Mergers in higher education institutions: a proposal of a novel conceptual model

Purpose – The macroeconomic changes as well as the challenges facing universities nowadays result in the transfer and adaptation of various concepts and organizational methods from enterprises to universities. One of such solutions is mergers. Even though there are a very large number of practical examples of university mergers in the world, at the same time there is a shortage of frameworks that would help manage mergers. The purpose of this paper is to present key areas of focus in HEIs’ consolidation processes and the creation of the conceptual model of the universities’ mergers. Design/methodology/approach – In this paper synthesis, the inductive approach for model development and case study description were used. The analysis and findings were based on the systematic literature review taken out from management and public policy areas. The new public management and public value governance approaches as well as strategic and process theories of mergers were applied. The descriptive approach to management was used as well. Findings – Formulation of a Conceptual Model of Universities’ Mergers and ten principles of effective management of universities’ mergers that cover the entire process, from planning, through implementation, to integration. Research limitations/implications – There is a need to verify the proposed inductive model of universities’ mergers through further qualitative and mixed-methods research studies. Practical implications – The paper offers a significant opportunity for practical application of the presented content, because it indicates how the know-how from one (business) sector can be valuable for the future of another sector (the over-fragmented sector of higher education). Originality/value – This study presents the key areas of focus in HEIs’ consolidation processes and proposes a novel Conceptual Model of Universities’ Mergers. It concludes with the principles of effective management of universities’ mergers.

limiting the accumulation of market power in markets not deemed to possess a natural monopolistic market structure. Such regulation serves the public interest in that accumulation of market power leads to higher equilibrium prices and restricted equilibrium output. In turn, these equilibrium changes unambiguously decrease the total economic surplus generated by the market (see e.g. Church and Ware, 2000) for an overview as to the relationship between market structure and surplus). On the other hand, M&A activity is desirable from the perspective of many firms in that market power increases the rate of economic profit in a market for which scale economies remain on the margin (Church and Ware, 2000). Hence, the activity and regulation of M&As constitutes a tug-of-war between private firms and the US Department of Justice.
The regulation of M&As is somewhat transparent. If a market's level of firm market share concentration (i.e. its HHI) is above 2,500, then the US Department of Justice considers that market to be highly concentrated and thus subject to being blocked or otherwise regulated by the Department of Justice. What does an HHI Index value above 2,500 indicate? Let us consider the index's equation, as well as an example on the topic. The HHI is calculated as follows: where HHI i,t represents the HHI index value for industry i at time t, s j,t is the (percentage) market concentration for firm j (of industry i) at time t, and P n j¼1 s 2 j;i;t represents the sum of squared market concentrations for all firms j in industry i at time t. Across industries, the HHI must fall in the range (0, 10,000], where a value near zero (of ε) indicates a perfectly competitive industry and a value of 10,000 indicates a textbook monopoly industry.
In the USA, the Justice Department monitors M&As closely to ensure that they do not "enhance market power" (US Department of Justice, 2019). Such an M&A activity is defined objectively as one in a market that already features an HHI greater than 2,500 and would further increase the HHI in that market by more than 200 additional units. To understand the types of market structure that this rule restricts, let us consider a simplified example in which firms are symmetric. In an industry of n symmetric firms, we can solve for the maximum number of firms such that the HHI remains greater than or equal to 2,500 (Table I).
Given symmetric firms, the market HHI is greater than 2,500 if the number of firms is less than 4. For the n ¼ 4 case, the HHI just equals 2,500. As such, we also know that HHI is greater than 2,500 for any four-firm market in which market concentrations are not symmetric. That is to say, the symmetric case strictly minimizes HHI for any n. Consider the symmetric case. As market shares must sum to 100, any departure from the symmetric case must involve a gain (gains) for some firms above a share of 1/n and an equal loss (losses) for
It is less tractable to summarize the conditions under which an M&A activity would raise the market HHI value by more than 200 points. For example, any industry with HHI greater than 2,500 may have a firm that is sufficiently small that an M&A activity involving this firm would not substantially raise HHI. However, we can summarize from the market HHI analysis that the Department of Justice pays particular (and less equivocal) attention to M&A activity in industries for which n ⩽ 4. These ground rules and regulatory effects will be relevant as we consider the market M&A and joint venture activity studied in this Special Issue.

A characterization of Special Issue articles 3.1 Market effects of M&A activities
Walia and Boudreaux (2019) examine the market effects of M&A activities. Specifically, they systematically review the literature on hospital M&A events. The authors find that, while somewhat mixed, the literature largely supports market effects from hospital M&A events that are consistent with economic theory (i.e. higher price, restricted output, accumulation of market power and lower costs). Thus, it appears that hospital M&A events lead to adverse market outcomes for consumers. They discuss the role of price cap regulation to diminish the price and output implications of hospital M&A events. The article raises the point that there is often economic (market surplus) cause to regulate proposed M&A events.

Firm and market value implications of M&A activities
Several papers in the present issue address topics of firm market value and the effect of M& activities upon this value. Tutuncu (2019) studies the effect of pre-acquisition earnings on financial performance of management buyouts. Studying 291 private firms in the UK and using cross-sectional discretionary accrual modeling, the authors find that management buyouts are preceded by earnings overstatement. Subsequently, then, there is found to be financial performance deterioration. As such, the authors find that estimates of value creation from M&A activity can be affected by this activity (e.g. if one looks only at immediate valuation consequences). Mohil et al. (2019) study the relationship between options availability and occurrence of informed trading prior to an M&A action. The authors examine 864 M&A announcement in India and find that option listing status increases the likelihood and extent of informed trading prior to an M&A event there. As such, there are potential market efficiency issues surrounding M&A activities in such cases. Ibrahimi and Meghouar (2019) use accounting variables to find sources of value creation or destruction during M&A events in France. They find that control of operating expense fluctuations throughout the event is a key indicator of value creation from the M&A event.
Wonder and Lending (2019) examines the effect of acquisitions upon the acquiring company's shareholder base. In a sample of 348 US acquisition events, they find statistical evidence of large increases to shareholder base for acquisitions completed at least partly in stock. These increases sustain over a four-year period. The magnitude of this effect is diminished in the case of cash acquisitions in the USA. Anagnostopoulos and Rizeq (2019) use a neural network approach to study M&A events in the US technology sector. They find that such an approach provides more explanatory power than a traditional regression in predicting valuation effects of M&A events in this sector. Aggarwal and Garg (2019) examines the effect of spin-off announcements upon the share price of the parent firm using an event study methodology. They find that spin-offs positively affect parent firm share price in the weeks following the announcement. This is potentially due to the adverse selection issues surrounding which divisions a parent firm would select to be spun-off (e.g. possibly those that do not complement other divisions of the firm).

Implementation of M&A activities
Two papers in this issue consider issues of implementation for M&A activities. Sulkowski et al.
(2019) examines mergers in higher education, presenting an inductive conceptual model of said activity. The model is meant to determine the value and optimal implementation of a potential merger activity (i.e. to render a more viable institution). The authors conclude that mergers often render greater institutional value if key principles are identified and (case) studied. Cheng (2019) studies the impact of the post-merger integration period on firm capital structure. The author finds that managers in the acquiring firm are forward looking when considering the financing of a potential M&A event, factoring in integration characteristics of the post-M&A firm.

Conclusion
This Special Issue addresses an important topic in managerial finance: that of M&A events. The topic is addressed using multiple empirical and theoretical methodologies. Moreover, key issues surrounding M&A events are considered. For example, the Special Issue addresses issues of value creation and destruction from M&A events, as well as issues of overall market (power) effects and of implementation issues surrounding these events. Collectively, these articles consider the issue of market value from M&A events from the often contrasting perspectives of shareholders, firm managers and consumers. As such, the Special Issue is able to collectively consider the myriad effects of M&A events upon societies.

Introduction
The 1990s and 2000s witnessed an increase in merger and acquisition (M&A) activity for private and non-profit hospitals. Cutler and Morton (2013) report that more than 60 percent of the US hospitals were part of a merged system of hospitals as of 2012 and that this value had increased by 7 percentage points since 2002. From 2007 to 2012, for example, 432 hospital M&A agreements were announced involving 835 hospitals. Industrial organization theory suggests that such an aggressive consolidation period within an industry is bound to have marked economic effects upon outcomes within this vital service market. Indeed, hospital M&As are often the subject of debate because the benefits of consolidation (such as increased efficiency and lower costs) are often weighed against the increased market concentration effect, as well as the potential for higher prices that stem from lower levels of competition. From a policy perspective, therefore, hospital M&As are often the focus of antitrust cases (Blackstone and Fuhr, 1993;Capps et al., 2002).
The purpose of this paper is to review the extant literature on hospital M&As. To do this, we review the literature on the outcomes of hospital M&As along several market dimensions, including costs, pricing, market power and efficiency. While our focus is on these considerations, several studies have given attention to alternative areas such as subsequent productivity (Ferrier and Valdmanis, 2004), welfare effects (Town et al., 2006), patient referrals (Nakamura et al., 2007), infant mortality and readmission for heart attack patients (Ho and Hamilton, 2000). and 11 of those results occurred in the past 5 years (2013-2018). These articles arose from a variety of disciplines, including economics, health care management and health policy. From this list, we reviewed 11 abstracts. We included any study that examined the outcomes of the merger or acquisition process along some metric. Common metrics were cost, price and quality, but we also found articles that examined which hospitals were more likely to be merged or acquired. Two studies examine how mergers affect costs (Harrison, 2011;Sinay, 1998). One study examines how (vertical) mergers affect patient referrals (Nakamura et al., 2007). One study examines how vertical mergers between hospitals and physician practices affect profitability. Two studies examine why certain hospitals are the targets of M&As (Noles et al., 2015;Dor and Friedman, 1994). There were also three review articles (Haas-Wilson and Vita, 2011;Brown, 1996;Brown et al., 2012), and one article on mergers of teaching hospitals (Kastor, 2001).
We had much better luck, however, by conducting an additional search that includes both hospital (singular) and mergers in the title. This search yielded 467 results according to Google Scholar or 141 results in the past 10 years and 74 results in the past 5 years. These studies tended to emphasize antitrust issues in relation to how M&As affect hospital market power, efficiency, productivity and pricing. The literature presenting new empirical evidence as to the market effect of hospital M&As in the USA was relatively sparse. In total, 8 articles evaluated the cost effect(s) of the US hospital M&As, 11 articles considered pricing effect(s) and 7 articles considered effects on efficiency and market power.
We excluded articles that discussed hospital mergers in other contexts, such as legal and ethical cases (Hochberg, 1996) or organizational culture ( Jones, 2000). Unpublished manuscripts were also excluded from our analysis.
Analysis of hospital M&A effect results from the literature Table I lists several categories of outcomes emphasized in hospital M&A studies. These categories include: provision cost, pricing and market power and efficiency. Of the eight  Table I. Articles by category of hospital M&As hospital efficiency or market power. Findings from several studies indicate that mergers can and do act to improve efficiency (Harris et al., 2000;Lynk, 1995b;Sinay, 1998;Groff et al., 2007). Hospital mergers can increase scale efficiencies, but they can often make it possible to consolidate two small clinical departments into one larger unit, which reduces the "peak load problem" of daily patient census (Lynk, 1995b).
However, Brooks and Jones (1997) found that mergers between hospitals are not driven directly by considerations of market power or efficiency as much as by the existence of specific merger opportunities in the hospitals' local markets. Kjekshus and Hagen (2007) found no effect of mergers on technical efficiency and a negative effect of 2-2.8 percent on cost efficiency. Other studies find that the effect is very sensitive to hospital ownership and governance and the structure of the market following the consolidation (Spang et al., 2009).

Policy implications
In this review, we find significant evidence of increased market power in the case of hospital M&As. As such, price cap regulation of hospitals is expected to lower prices, raise consumer surplus and raise overall market surplus in the market for hospital services. Figure 1 demonstrates the price and quantity of hospital care chosen by a local market monopolist hospital firm operating in a free market setting. Figure 2 demonstrates these allocations for the same hospital firm operating under price cap regulation.
In Figures 1 and 2, we have that P represents unit price of hospital services, Q represents quantity of hospital services, MC is the marginal cost of services, MR is marginal revenue from services and D represents the Demand function for hospital services in the local market. We also have the following equilibrium allocations: P n U P n R À Á represents equilibrium price in the unregulated (price cap regulated) market and Q n U Q n R À Á represents equilibrium quantity exchanged in the unregulated (price cap regulated) market. From the standard analysis of price cap regulation in Figures 1 and 2, it is clear that P n U 4P n R and Q n U oQ n R such that access to hospital services, consumer surplus and total market surplus are greater under price cap regulation. Hospital mergers

Introduction
Existing literature shows widespread earnings management before going private acquisitions (Perry and Williams, 1994;Wu, 1997;Fischer and Louis, 2008;Mao and Renneboog, 2015). The managerial self-interest inherent to MBO acquisitions provides sufficient incentives and forms an ideal setting for managers to exercise their discretion over accruals and understate earnings to pay an undervalued equity price. However, the current research has an exclusive focus on going private buyouts. This study investigates earnings management preceding private firm MBOs and follows up with post-acquisition performance investigation. The private firm choice is motivated by the fact that private firms are the largest source of buyouts worldwide (Strömberg, 2008). In the UK, private firms are subject to the same reporting requirements as public firms as they file annual reports with the registrar. EU fourth Directive Article 47(1) and 51(1), respectively, state that companies must make their annual reports publicly accessible and have their financial accounts audited. The fourth Directive also clarifies that member states can lighten publication requirements of annual accounts for small and medium sized companies and auditing exemptions can be introduced for small companies. In particular, private firms are exempt from publishing cash flow statement. These regulations enable us to utilise a fairly large sample of private firms for tests. Another motivating factor is the lack of event-based studies for private firms. Past research is either based on case studies (Howorth et al., 2004), or multi-country private firm populations (Burgstahler et al., 2006) absent a major corporate event.
Past research investigates relevance of an MBO acquisition to firm performance (Kaplan, 1989;Guo et al., 2011;Jelic and Wright, 2011), and provides us ample evidence on earnings management around equity issues and following performance (Rangan, 1998;Teoh et al., 1998a, b;Jo et al., 2007). This study examines earnings management on and post-acquisition performance of MBOs. The sample comprises 291 UK private firm MBO acquisitions between 2004 and 2012. Earnings management tests are performed by means of four discretionary accruals models. Main results show that private firm managers overstate earnings prior to buyout. The year preceding acquisition coincides with large positive changes in total accruals and earnings. The performance analysis is carried out with 254 MBOs for which the data requirements are met. Performance, measured by return on assets (ROA), peaks in the first pre-acquisition year and begins to deteriorate after the first post-acquisition year. MBO firms outperform industry in the five years examined; however, performance drops to industry levels when private equity (PE) backed buyouts are excluded. On the other hand, PE sponsors invest in firms with already high levels of profitability and no significant performance improvement is observed after PE investment. Spearman correlation test and cross-sectional tests of discretionary total accruals (DTA) show that earnings management is negatively associated with performance changes and discretionary accruals is a significant determinant of changes in performance.
This study makes mainly two contributions to the existing research. First, it extends case study evidence of Howorth et al. (2004) and adds to the buyout literature by providing the first empirical analysis of earnings management in private firm acquisitions. It adds to the PE literature by providing an analysis of earnings management in PE-backed and non-PE-backed MBOs. Such an analysis is important to shed light on the role of PE funds in their target investments. While there is a large body of literature on corporate governance mechanisms of venture capital (VC) and PE firms relatively few studies examine their role in earnings management practices of their portfolio companies. The findings of this study are complementary to the existing evidence on earnings management in VC-backed initial public offerings (IPOs) (Morsfield and Tan, 2006;Hochberg, 2012;Wongsunwai, 2013) and the role of buyout sponsors in reverse leveraged buyouts (Chou et al., 2006;Wang, 2010).
The second contribution is to the value creation debate surrounding buyouts. The negative relationship between post-acquisition performance and discretionary accruals is important in the sense that past studies might have overstated buyout performance due to their omission of earnings management factor and led to the erroneous conclusion that buyouts perform better following MBO deal. Taken in isolation, profitability ratios are not indicators of the real buyout performance and effects of earnings management must be controlled to draw conclusions. This study complements research on equity offerings, Rangan (1998) and Teoh et al. (1998b) in particular, and helps us develop a more insightful understanding of buyout performance.
The rest of the paper is organised as follows. Section 2 reviews the literature and develops hypotheses. Section 3 discusses data and methodology. Section 4 presents results of discretionary accruals estimations. Section 5 examines post-acquisition performance. Section 6 concludes the paper.
2. Literature and hypothesis development 2.1 Private firm management buyouts Schulze et al. (2003) argue that governance mechanisms designed for public companies do not work as planned when firms are private. Private companies are often owned by a few individual blockholders in contrast with public companies where equity is dispersed across a large number of investors. This ownership concentration allows managers and shareholders to establish more personal relationships (Fama and Jensen, 1983) and to use 1364 MF 45,10/11 private communication channels to exchange information (Burgstahler et al., 2006). Many private firms are run by families where the roles of managers and owners are not clearly separated. The unclarity of borders between managerial and ownership roles invalidates inferences based on the standard principal-agent relationship. However, different types of private firms might introduce different issues within the organisation. Howorth et al. (2004) report that information asymmetries can be strong in private nonfamily firms since they have separate ownership and management structures. In family firms, however, MBO acts as a viable solution to deal with succession issue, allowing the family to realise investment and maintain the independence of the company (Howorth et al., 2004;Scholes et al., 2008). In this case, the purchasing team is likely to have good relations with the family, which would minimise conflicts of interest. Their incentives to manage earnings for personal gain are then substantially reduced.
As many other private firm, MBOs typically have concentrated ownership and their managers are not under pressure of strict capital market scrutiny. In addition to other manipulative forces, however, MBO managers are likely to have additional incentives due to their acquisition of company. In private family firms where MBO acts as an acceptable succession tool (Howorth et al., 2004); the friendly nature of the deal would substantially mitigate incumbent managers' incentives to manage earnings to their benefit. On the other hand, non-family owner-managers, who typically have large ownership stakes to sell, could be sufficiently motivated and powerful to overstate earnings to at the expense of acquiring team. Note that unlike going private MBOs, none of the parties are interested in understating earnings. In general, it is conjectured that considerations related to family succession issues and managers' personal wealth will offset incentives for to understate earnings. Since this study does not differentiate between different types of private firms (e.g. family, non-family, lone founder), segregation of their earnings management practices and predicting direction of earnings management is not possible. The following hypothesis is proposed: H1. Managers of private firms do not understate earnings prior to management buyout.

PE sponsors
The literature presents us conflicting evidence regarding the role of PE in the process leading to the buyout and after the buyout transaction. While performance improvements are documented in the post-transaction firm (Kaplan, 1989;Jelic and Wright, 2011), the evidence related to their pre-buyout involvement is less positive. Acharya and Johnson (2010) show presence of large insider trading in PE-backed buyouts. Bargeron et al. (2008) find that PE funds pay lower acquisition premiums than other acquirers. They argue that this could be a reflection of their limited capacity to generate synergy gains and their limited time to extract returns; nonetheless it also raises the question of whether PE firms collaborate with target management in MBOs. With managers lacking the funds to acquire the total equity, the financial assistance of PE could benefit both parties in an undervalued acquisition.
On the other hand, PE firms are known to have a positive bias for better performing firms ( Jelic and Wright, 2011). The performance provisions related to PE funding might lead managers to seek better performance to attract PE investment. Fischer and Louis (2008) argue that managers' desire for personal gain might be offset by their need for external financing. Their incentives for earnings management are related to their financial independence and personal ability to finance MBO deal. If managers cannot finance transaction with their personal wealth, they might manage earnings upwards to show their firm as an attractive investment option. However, PE firms would detect earnings management practice if their screening skills allow them to extract managers' private information. Moreover, PE firms are 1365 Performance of management buyouts repeat market players who have serious reputation considerations. Collaborating with managers to understate earnings would taint their credibility in the case of detection by regulators. European private firms report audited financial statements, which would increase the probability of detection and reduce manipulative incentives related to going private. The evidence from IPOs suggests that effective monitoring and reputation concerns of VC firms constrain earnings management around IPO. For example, VC-backed IPO firms tend to show lower abnormal accruals and exhibit more conservative earnings management than comparable non-VC-backed IPOs (Morsfield and Tan, 2006;Hochberg, 2012). In the same vein, Wang (2010) finds that presence of buyout sponsors leads to improvement in discretionary accruals in reverse leveraged buyouts. One exception is Chou et al. (2006) who find that buyout sponsors engage in upwards earnings management prior to IPO exit. Wongsunwai (2013) and Brau and Johnson (2009) report negative association between earnings management and VC reputation. The opposing considerations related to managers incentives for personal gain, their need for external financing and reputation concerns of PE firms are expected to temper the incentives for earnings management in either direction. Therefore, the following hypothesis is proposed: H2. PE-backed MBOs do not exhibit earnings management prior to buyout.

Data and methodology 3.1 Sample selection
The sample of research is constructed by the following procedure. First, MBOs completed between 2004 and 2012 are identified from Thomson One Banker (TOB) database. Deals are selected based on three criteria: acquisition target must be registered in the UK, target must be a private company and deal must be led by an incumbent management team. The search results in 1,004 MBOs. TOB provides deal announcement and completion dates, firm industry and SIC codes, deal value where available, and deal synopsis that gives information on bidders, presence of PE investor and the origin of acquisition target. Secondary buyouts, divestments, going private buyouts and management buy-ins are identified from deal synopsis and excluded from sample. These buyouts are excluded to obtain a homogenous private firm sample since mixing with other types of buyouts might change managerial behaviour. Finally, financial firms are excluded following the standard practice in corporate finance research. A total of 144 firms are dropped in this step. For the remaining 860 firms, financial statements are collected from Fame database. After firms with missing data are dropped, the final sample contains 291 MBOs. Table I presents the distribution of MBO transactions and their deal values across years. The TOB population holds 860 deals, of which 291 are included in the cross-sectional tests. This number is larger than any of previous related studies. There is substantial deal clustering at the time referred to as the mega buyouts period. The highest number of deals is reported in 2006. Two of the sample years (2005,2006) collectively account for more than 40 per cent of the sample buyouts. Over 80 per cent of deals are completed between 2004 and 2008 and the number of buyouts drops after 2008. This pattern is consistent with the sharp fall in worldwide buyout activity around 2008 (Gilligan and Wright, 2010). However, the sample represents a good portion of the UK population, comprising 64 per cent of the aggregate transaction value where deal information is disclosed. In terms of total number of private firm buyouts, it represents 34 per cent of the MBO population.

Measuring accruals and model selection
This study follows the balance sheet approach where total accruals are computed as non-cash working capital minus depreciation expense. This definition is the same as in Perry and Williams (1994), Dechow et al. (1995) and Burgstahler et al. (2006). It is argued that current 1366 MF 45,10/11 accruals are more relevant when measuring year-to-year discretion since non-current portion of discretionary accruals may not reflect the recent accounting practices ( Jones, 1991;Teoh et al., 1998b). Kothari et al.'s (2005) performance modification is also useful to correct misspecifications in cases that companies are likely to exhibit extreme financial performance. Therefore, working capital accruals (WCA) and performance-adjusted accruals are also used in the tests in addition to total accruals. Note that cash flow-based models cannot be tested since private firms are not required to publish cash flow statements. Finally, original Jones time-series model is used; however, this results in a significant drop in the number of firm observations due to extensive data requirements.
Tests are conducted for the two years preceding buyout. For cross-sectional tests, portfolios of private industry firms matched on two-digit SIC codes are used to estimate parameters. The sample contains 52 unique two-digit industries. The top three two-digit industry groups with the largest number of MBOs (SIC code 50 (whole sale-durable goods), SIC code 73 (business services) and SIC code 87 (engineering, accounting, research, management and related services)) account for about one-third of the sample. In total, 225 unique regressions are performed to obtain parameters. In the DTA model, total accruals are a function of inverse lagged assets, revenues and tangible assets: This model is estimated with OLS regression on the portfolios of control firms. Obtained parameters serve to compute MBO discretionary accruals as follows: DTA jt ¼ TA jt Assets jt À1 À b1 jt 1 Assets jt À1 þb2 jt DREV jt Assets jt À1 þb3 jt PPE jt Assets jt À1 : (2) WCA are computed as non-cash working capital, and can be obtained by subtracting depreciation from total accruals. Since depreciation represents a long-term accrual, its removal leaves current accruals in the equation. PPE variable is dropped because it is associated with depreciation. The model used to estimate parameters is as follows:

1367
Performance of management buyouts Kothari et al. (2005) suggest two ways of performance adjustment. The first one involves matching each MBO on a firm with the same industry and nearest profitability. The second one is carried out by augmenting the original regression by an additional profitability (ROA) variable. This study uses augmented cross-sectional regression: Finally, DTA are re-estimated by Jones' (1991) pooled time-series model. This model utilises 101 MBOs and their estimation period ranges from 4 to 8 years.

Tests of earnings management 4.1 Cross-sectional and time-series tests
The results of discretionary accruals tests are presented in Table II. Mean and median discretionary accruals in year −1 are positive and significant, with the exception of median time-series accruals. Discretionary accruals in year −2 are generally negative, smaller in magnitude and insignificant. The drop in significance for the time-series model can be attributed to the short estimation windows used in the model as well as the low observation count in accruals tests due to data limitations. The results show strong upward management preceding private firm MBOs, indicating that private firm managers have stronger incentives to make upward adjustment than their potential wealth benefits from downward adjustment. This result is consistent with H1. A possible explanation for this pattern could be such that private firm owner-managers can reap higher benefits by overstating earnings, especially if they are selling their shares in the acquisition.
To examine the effect of PE involvement in earnings management, the sample is stratified based on PE sponsor status. Slightly less than half of the sample firms has PE sponsors. Table III displays the results of cross-sectional tests for PE-backed and non-PEbacked subsamples. Time-series tests are not presented due to insufficient number of observations and powerless test statistics resulting from short parameter estimation windows. Consistent with H2, PE-backed buyouts do not show significant earnings management in year −1. Only mean DTA and median discretionary WCA are weakly significant. The non-PE-backed sample, on the other hand, shows significant upwards management in all models, with the exception of performance-adjusted mean discretionary accruals. The results lend support for H2 that PE-backed buyouts do not engage in earnings management, highlighting the differences in motivations, PE firms' continuing involvement in the markets and their reputation concerns. The findings also support existing evidence from IPOs that PE firms constrain earnings management (Morsfield and Tan, 2006;Wongsunwai, 2013).

Additional accruals tests
There might be concerns associated with private company regulations that previous results might not be extrapolated. For example, EU member states might introduce disclosure and audit exemptions for small and medium sized firms. Since previous studies examine going private MBOs that do not face such exemptions, it might be useful to repeat the tests excluding these firms. In addition, there might be concerns about the large differences between the firm observations of control portfolios. Finally, a difference-in-differences (DID) estimation is used to address the potentially false hypothesis that discretionary accruals must be statistically equal to zero to show no earnings management. DID estimation enables us to control for the non-buyout firm discretionary accruals and show the margin of the sample relative to firms in the same year and industry. The results of the experiments are presented in Table IV Notes: DTA, WCA, PADJ, and TIMS stand for discretionary total accruals, working capital accruals, performance-adjusted accruals, and time-series accruals, respectively. Significance is tested by two-tailed t-test and Wilcoxon sign rank test. t and z values are in parentheses. *,**,***Significant at 10, 5 and 1 per cent levels, respectively Finally, propensity score matching (PSM) method is used to estimate earnings management. Contrary to dimension-to-dimension matching applied in the previous crosssectional models, PSM controls for multiple dimensions to select a matching firm with similar characteristics (Li and Prabhala, 2005). The procedure involves estimation of a probit regression to predict the likelihood of becoming a buyout target, where dependent variable is a dummy that equals 1 for sample MBOs and 0 for non-buyouts, and independent variables are sales growth, size (natural logarithm of assets), ROA and asset turnover (sales/assets). The regression is executed in each calendar year instead of pooling the data. After the balancing property of regression is satisfied, for each MBO firm a non-buyout with the nearest propensity score is selected with replication. The PSM discretionary accruals are calculated as MBO firm DTA minus PSM matched firm DTA. The results presented in Table V show that MBOs overstate earnings prior to transaction. The results related to the provision of PE backing also suggest that PE-backed MBOs engage in less earnings management relative to non-PE-backed MBOs. Overall, findings are consistent with prior estimations and private firm MBOs conclusively exhibit upwards earnings management.

Earnings management and MBO performance
The performance dimension is important to demonstrate that buyouts can create value and offer an above-market return to their investors. Prior buyout studies do not investigate the link between earnings management and performance. However the research on IPOs and seasoned equity offerings (SEO) documents that earnings management around equity issues has substantial impact on subsequent firm performance (Teoh et al., 1998a, b;Rangan, 1998;Li et al., 2006;Jo et al., 2007). The negative association between discretionary accruals and performance is well documented (Rangan, 1998;Teoh et al., 1998b). The income-increasing earnings management around share issues results in post-issue accrual reversals and leads to the deterioration of operating performance and returns in the following periods. The effects of earnings manipulation are also manifested in the subsequent delisting method, where IPOs associated with conservative earnings management are more likely to be merged or acquired, and IPOs associated with aggressive earnings management are more likely to delist involuntarily from markets (Li et al., 2006). In line with the studies above, earnings management is expected to be inversely related to subsequent buyout performance. Since accrual management effectively means shifting income from one period to another, a sample of firms inflating earnings should underperform in the following periods. Hence, it is hypothesized that: H3. Aggressive earnings management results in performance deterioration.
H4. Discretionary accruals are negatively associated with performance changes after buyout. Performance of management buyouts VC classification reveals clustering around three industry groups. Consumer, business and industrial, and business services industries account for over 73 per cent of the sample, compared to 65 per cent in Jelic and Wright (2011). This pattern of buyout concentration around business and service industries is consistent with the UK (Weir et al., 2015;Jelic, 2011) and worldwide market trends.

Net income performance of MBOs
Since results in the previous section show an upwards earnings management pattern, the expected income-decreasing effect of future accruals reversals will pull earnings down. To examine whether upwards or downwards earnings management introduce distinct performance outcomes, sample is stratified by the direction of earnings management. Sample firms are also ranked by magnitude of discretionary accruals into aggressive and conservative quartiles. Each quartile contains 64 firms. Finally, Spearman rank correlation and cross-sectional regression tests are performed to examine the link between earnings management and post-buyout performance. Note that expected sign of relation between performance and discretionary accruals is negative without regard to the direction of earnings management. Table VII reports net income performance in the six years around MBO transaction. Performance analysis is limited to three post-buyout years for mainly two reasons. First, most of the corporate governance and performance changes or improvements occur in the first three years following buyout (Guo et al., 2011). Second, buyout sponsors are more  27,36,38,48,50,73,78,87 30 11.82 Consumer 8,20,22,23,25,27,31,32,34,39,50,51,[54][55][56][57][58][59]73,76 58 22.83 Business and Industrial 7,[15][16][17]26,28,30,[32][33][34][35][37][38][39]42,45,50,51,79 likely to exit after the first three years. Average holding period in the UK for PE-backed buyouts is around 3.5 years (Nikoskelainen and Wright, 2007;Jelic, 2011). Hence, the first three years would suffer less from survivorship bias and offer a better representation of the performance. Top two panels display medians for the entire sample and bottom two panels present performance by PE sponsor status. Year 0 is the buyout year and year −1 is the first year before buyout where earnings management activity is observed. Performance is measured in two ways: raw operating performance which is calculated as net income divided by beginning assets and industry-adjusted net income performance. Reported statistics represent median performance for the relevant samples and years. Since scaling income by lagged assets can inflate performance, tests are repeated using net income scaled by current assets, only to find similar results. Observed patterns for the entire sample suggest improvements prior to buyout and slight deterioration after. Unadjusted performance rises from 7.7 to 8.9 per cent in year −1. Industry-adjusted performance observes a similar improvement from year −2 to −1. relative to buyout transaction year (year 0 is acquisition year). The unadjusted performance is measured as net income divided by beginning assets. Industry-adjusted performance is measured as net income divided by beginning assets minus industry median. Performance changes are computed as year-to-year changes in unadjusted and industry-adjusted net income divided by beginning assets. *,**,***Significant at 10, 5 and 1 per cent levels, respectively

Performance of management buyouts
Performance peaks in year 0 (10.5 per cent unadjusted and 2.9 per cent industry adjusted), then begins to decline in year 1 before eventually reverting to pre-buyout levels in year 3. Performance improvements before buyout are significant as shown by industry-adjusted change of 2.1 per cent. The percentage of firms with negative performance changes rises from 41 per cent in year −1 to 58 per cent in year 2. The post-buyout changes are significant at conventional levels. Note that despite decreasing levels of net income, MBOs continue to outperform industry peers in all post-buyout years. PE-sponsored and non-sponsored buyouts have substantial performance differences. Sponsored MBOs are more profitable throughout and about two times more profitable than non-sponsored MBOs in the post-buyout period. Unadjusted performance of PE-sponsored sample peaks at year 1 while non-PE sample performance monotonically declines following buyout. More importantly, industry-adjusted performance of PE-backed buyouts shows improvements and significantly outperforms industry from year −1 to year 3 while non-backed buyouts remain at the same industry levels. Mann−Whitney test statistics show that performance differences between medians are significant. It appears that better-than-industry performance is driven solely by PE-sponsored buyouts. The reported performance changes in the bottom panel exhibit a similar pattern. Industry-adjusted performance increases in the earnings management year by 1.9 per cent and 2.6 per cent and declines following buyout. The differences in performance changes between PE-backed and non-backed samples are not significant at conventional levels. In sum, PE sponsorship is associated with higher performance; however it does not prevent the deterioration of performance in post-buyout period.
The performance of MBOs stratified by the direction and magnitude of earnings management are reported in Table VIII. The results are based on DTA; however, similar patterns are observed in unreported tests based on WCA and performance-adjusted accruals. Panel A suggests that upwards earnings managers tend to be more profitable prior to buyout and less profitable afterwards. The opposite pattern is true for the downwards earnings managers. The downwards subsample registers significant performance improvements from MBOs with negative discretionary total accruals are classified as downwards and MBOs with positive discretionary total accruals are classified as upwards earnings managers. Conservative and aggressive quartiles contain MBOs with the smallest and largest absolute discretionary total accruals in year −1, respectively. *,**,***Significant at 10, 5 and 1 per cent levels, respectively pre-to post-buyout period. While they underperform industry by a median of 2.8 per cent two years before buyout, by the end of year 3 they outperform industry by a median of 3 per cent. Upwards subsample performance, on the other hand, declines to industry levels in year 3. The results presented in Panel B for aggressive and conservative earnings management quartiles are consistent with prior evidence that aggressive earnings managers subsequently experience performance deterioration (Teoh et al., 1998b). Unadjusted earnings for aggressive quartile MBOs are around 10 per cent of assets which is then reduced to just over 6 per cent in the year 3. In contrast, conservative quartile MBOs register higher earnings in all post-buyout years. More importantly, aggressive earnings managers fail to perform better than industry after buyout while conservative earnings managers outperform the industry in all post-buyout years. The performance differences between aggressive and conservative quartiles are significant in year 1 and year 2.
Overall, the performance analysis offers support to the proposition that earnings management prior to buyout influences performance after buyout. In the specific context of private firm MBOs, accrual reversals following upwards earnings management give result to the deterioration in performance. H3 is also supported. Upwards earnings managers outperform industry only for two years while aggressive earnings managers do not outperform the industry at all. The results so far imply a negative relationship between earnings management and performance. OLS regressions are conducted in the next sub-section to examine this relationship (H4).

Regression of post-buyout performance and discretionary accruals
Spearman rank correlations between discretionary accruals and performance changes (ΔROA) are reported in Table IX. Panels A and B display unadjusted and industry-adjusted correlations, respectively. Correlations in both panels are consistent, discretionary accruals are negatively correlated with performance changes in all years; with varying degrees of significance. The correlations with DTA are significant in all years, while correlations with performance-adjusted discretionary total accruals (PADJ) are only significant in year 1 and correlations with WCA are not significant. Therefore, it is concluded that only DTA predict the long-term buyout performance.
Prior research documents a negative relation between earnings management and performance in equity issues (Rangan, 1998;Teoh et al., 1998b;Jo et al., 2007). Consistent with this literature, H4 posits that earnings management is negatively associated with post-buyout performance changes. To examine this prediction, cross-sectional regressions are estimated in each of the three post-buyout years. The performance is modelled as a function of DTA and controls. Following single and multiple regression models are estimated:  Table X reports single regressions. The dependent variables are unadjusted and industry-adjusted changes in ROA relative to year −1. DTA is the common independent variable in all regressions. The first and last three columns display results for unadjusted and industry-adjusted ROA changes, respectively. Consistent with previous studies, the results demonstrate that DTA is negatively associated with post-buyout performance changes. In Panel B, several other variables are controlled following the prior literature. The estimated model is: where ΔROA t ¼ raw and industry-adjusted net income change in year t; t ¼ 1, 2, 3. DTA ¼ discretionary total accruals in year −1; PE ¼ dummy variable equals 1 if MBO is PE-backed, and 0 otherwise; SGRO ¼ percentage growth in sales from year −2 to −1, included following Rangan (1998); SIZE ¼ inflation adjusted log of total sales in year −1; g1, g2 and g3 are industry dummies for buyouts in high-tech industries defined as in Gompers et al. (2008). The model R 2 ranges from 7.68 to 16.7 per cent. The main variable of interest DTA maintains a negative sign and remains statistically significant in all regressions. Consistent with the findings in univariate performance analysis, PE is not significantly associated with performance changes. To assess the economic significance of results, the effect of one standard deviation change in DTA on dependent variable is calculated by multiplying its coefficient with its sample standard deviation following Rangan (1998) andTeoh et al. (1998b). The sample standard deviations and economic impact of DTA are presented in Table XI. The results show a consistent trend of increasing negative impact on earnings from year 1 to year 3. For example, accruals reversals are associated with a 2.66 per cent negative impact on raw performance changes in the first year following buyout, which rises to a cumulative 4.75 per cent impact in the third year. For industry-adjusted income, discretionary accruals are associated with 2.29 per cent decline in performance, rising to 4.35 per cent in the third year.
The implied economic impact of discretionary accruals in univariate and multivariate tests are consistent and quantitatively similar. These results are economically important in the sense that earnings management explains more than 4 per cent of the performance changes in the three years following MBO, which is the approximate improvement or deterioration in earnings reported by recent buyout studies (e.g. Boucly et al., 2011). Prior literature also documents that most of the improvements in earnings are limited to first two or three years subsequent to buyout (Kaplan, 1989;Smith, 1990;Opler, 1992). Therefore findings provide useful insights into post-transaction performance of buyouts. In unreported multivariate regressions with changes in ROA in year 0 as the dependent variable, DTA also carries a negative sign; however, which is not significant at conventional levels. Thus the coefficients in year 0 do not have economic importance.
In sum, the results show that upwards earnings management in the immediate year before MBO coincides with increases in net income. Consistent with an accrual reversals explanation, buyout transactions are followed by deterioration in performance in the three subsequent years. Earnings management proxied by DTA is a significant determinant of post-buyout performance changes. Given the fact that prior studies document performance improvements subsequent to going private buyouts (Kaplan, 1989;Boucly et al., 2011;Guo et al., 2011) andearnings understatement prior (Perry andWilliams, 1994), the findings suggest that income-decreasing earnings management may partly account for post-buyout performance improvements. The performance analysis suggests that pre-buyout earnings management can explain post-buyout performance changes. The tests show significant drops in performance following buyout in aggressive earnings managers and upwards earnings managers. Aggressive earnings managers fail to outperform industry peers while upwards earnings managers cease to outperform industry after second post-buyout year. The results of univariate and multivariate regressions indicate that performance drops can be explained by upwards earnings management prior to buyout. Earnings management proxy DTA are statistically and economically significant in all regressions. Consistent with the prior evidence from equity offerings, the results suggest that earnings management is a significant determinant of post-transaction performance.

1376
This research makes several contributions for practitioners and regulators. It shows that earnings management is not a generic phenomenon in MBOs and there is considerable heterogeneity with respect to the existence as well as direction of earnings management. The heterogeneity exists in the form of public and private firms, as well as PE-backed and non-backed acquisitions. The findings related to performance analysis underline challenges in value creation and difficulties in assessing it, as well as demonstrating that the classic agency view does not fully explain post-buyout performance in private firm buyouts.
This study does not attempt to explore earnings management practices in other types of buyouts. Therefore, a useful area for further research would be to examine earnings management in other types of buyouts, distinguishing between public and private firm governance structures (e.g. family firms, owner/manager-led firms, agent-led firms) and deal type (e.g. management buy-in, divestment, secondary buyout). Further research can regression, the results of which are reported in Table X. Impact is calculated by multiplying the DTA coefficient by the relevant sample standard deviation. **,***Significant at 5 and 1 per cent levels, respectively

Introduction
Mergers and acquisitions (M&As) are price-sensitive events and remain unknown in the public domain until officially announced. Only a selective number of traders possess the detailed information about the upcoming M&A announcement. Sometimes, these informed traders try to use this confidential information for their personal benefits. However, in regards to the possibility of abnormal gains on the parts of informed traders, the viewpoints of researchers are divided into two different forms of the efficient market theory. One class of researchers support the strong form of efficient market hypothesis, mentioning that informed traders are unable to access abnormal returns (ARs) around announcements of mergers and acquisitions. They explain that the ARs earned on the basis of confidential information are offset by the risk of indulging in illegal corporate transactions. These researchers opine that since risk-adjusted returns get negligible, informed traders refrain from altering their stock holdings before M&A announcements (Lorie and Niederhoffer, 1968;Jarrell and Poulsen, 1989;Agrawal and Jaffe, 1995).
The other researchers support the semi-strong form of efficient market hypothesis and explain that informed traders are able to earn significant abnormal stock returns on the basis of confidential information about the upcoming M&A announcement. They elaborate that there are specific settings when the risk of trading is less and returns are high. Hence, the informed traders take positions selectively under those venues where risk-adjusted returns are significantly positive (Keown and Pinkerton, 1981;Cornell and Sirri, 1992;Meulbroek, 1992;Agarwal and Singh, 2006;Agrawal and Nasser, 2012).
In association to the favorable venue viewpoint, informed trading has a strong liaison with the options market, as the risk in the options market is limited to the premium, leverage is high and the transaction cost is less (Amin and Lee, 1997;Black, 1975;Cox and Rubinstein, 1985;Du et al., 2018;Easley et al., 1998;Skinner, 1990;Stephan and Whaley, 1990;Geppert and Kamerschen, 2008;Ross, 1976;Manaster and Rendleman, 1982). So, the informed traders either completely move to the options market or migrate between stock and options market to earn ARs to utilize the confidential information they possess about the impending M&A announcement. This is significantly true for the companies with option listing status, otherwise informed traders do not have any choice but to adjust their traders only with the stock market transactions. If the informed traders displace from or migrate between the stock market to options market, then trading in the stocks would differ for optioned[1] and non-optioned companies ahead of M&A announcements. The main objective of the current study is to examine whether the option listing status of the acquiring company affects the information-based trading in the stock market prior to the M&A announcements.

Research question
It is evident from the above discussion that informed traders may behave differently with the involvement of options market due to lesser risk and higher profitable opportunities. Besides, there are some other external factors such as regulatory environment and crisis period when risk of insider trading gets affected. The informed traders become highly prevalent during turmoil periods (Netter and Mitchell, 1989;Abumustafa and Nusair, 2011;Cziraki, 2018), whereas the level of informed trading would be less if regulatory environment is strong (Atanasov et al., 2006;Podolski et al., 2013;Agrawal and Jaffe, 1995).
Further, the existing studies [2] concentrate on the developed economies, namely, the USA (Lorie and Niederhoffer, 1968;Keown and Pinkerton, 1981;Jarrell and Poulsen, 1989;Cornell and Sirri, 1992;Meulbroek, 1992;Arnold et al., 2006;Cao et al., 2005;Agrawal and Nasser, 2012) and the UK (Spyrou et al., 2011) where regulatory environment is stronger. On the contrary, the possibility of informed trading gets higher in emerging countries because the regulatory environment in their underlying capital market is under-developed (Klapper and Love, 2004;Miller et al., 2008). So, it would be ideal to get information on informed traders belonging to an emerging country like India where not only the possibility of informed trading is high but the options market of India is also growing continuously over the recent years as per the Securities and Exchange Board of India [3]. Additionally, the percentage of the M&As announced by optioned acquirers in India has shown an increasing trend throughout the years (Table AI and Figure A1).
So, this study is conducted to accomplish the following objectives: (1) To analyze and compare the effect of options availability on the level of informed trading occurring in the stock market of acquiring companies around M&A announcements in India.
(2) To find out the joint effect of option listing status and the external factors (crisis period and regulations) on the informed trading.
The empirical evidence demonstrates that stock market of optioned Indian acquirers follows a semi-strong form of efficiency prior to the M&A announcements and informed traders earn significant ARs under this setting. The results show that cumulative ARs are significantly high for optioned Indian acquirers in the study windows of −30 to −1 and −20 to −1 days, which reflect that informed traders act differently when underlying company has options availability. The informed traders hedge the risk by selling their positions or reducing the stock buying well before the M&A announcements when acquirers do not have options availability, whereas they take long positions in the stock market and hedge the risk and/or earn ARs by taking equivalent positions in an options contract. Hence, it can be concluded that option listing status affect the information-based trading occurring before the M&A announcements. The results also reflect that informed traders were more pervasive after the sub-prime crisis period than before the crisis period during M&A announcements of optioned

1383
Informed trading in M&As Indian acquirers. We also find that informed traders do not increase their stock purchases when regulatory reforms are introduced.
The paper contributes to the three important streams of finance: mergers and acquisitions, informed trading and options market. To the best of authors' knowledge, the current study brings into notice the changing dynamics of informed trading over different time periods under the alignment of M&A announcements and option listing status of acquiring companies. Besides, a different country-wide perspective is outlined through an emerging country like India, which is novel in this context.

Related literature and hypotheses development 2.1 Informed trading around M&A announcements
Lorie and Niederhoffer (1968) initially notified that insiders have an access to confidential information about the upcoming corporate announcements. They also mentioned that if insider trading occurs, it can be reflected through stock price movements. Jarrell and Poulsen (1989) and Agrawal and Jaffe (1995) showed that insiders do not indulge in illegal transactions ahead of acquisition announcements. Keown and Pinkerton (1981), Cornell and Sirri (1992) and Agarwal and Singh (2006) measured the insider trading through abnormal stock price behavior and found that non-public information is used ahead of the merger announcements. Meulbroek (1992) examined the illegal insider trading prosecuted by the authorities and showed that abnormal stock trading before the takeover announcements gets significantly higher on the day when insiders trade. Agrawal and Nasser (2012) described that insiders reduce their open transactions significantly to avoid the legal penalties and risk of losing reputation and job. They also mentioned that insiders do indulge in passive trading by reducing their stock sales, hence escaping losses. Thus, we see that the literature is indecisive about the occurrence of informed trading before the announcements of M&As. Some of the studies affirm the pervasiveness of informed traders around M&A announcements. Yet, few studies do not support the view that informed trading happens before the M&A announcements. The limitation of these studies is that most of them rely solely on reported/detected transaction of insiders, whereas insiders do not report all their transactions, especially when they are based on material non-public information. Also, these studies are restricted to developed economies, whereas regulatory regime is comparatively stronger than the regulatory dynamics of an emerging country like India. Based on these viewpoints, we develop the following hypothesis of the study: H1. The absolute ARs get significantly high prior to the M&A announcements of Indian acquiring companies.

Effect of option listing status
Spyrou et al. (2011) opined that informed trading has a strong association with the options market. They showed that only non-optioned target companies experience significant abnormal trading, whereas informed traders displace to options market ahead of M&A announcements. They also highlighted that the probability of informed trading may vary, but some informed traders always are always present before the announcements. Cao et al. (2005) compared the optioned and non-optioned target firms and found that informed traders migrate from stock market ahead of M&A announcements if underlying targets have options availability. Arnold et al. (2006) proposed that option listing status of the target company affect the demand for stocks before the tender offer announcements. They showed that the magnitude of ARs gets higher and sooner for optioned targets than non-optioned targets in the pre-announcement period. These studies show that the possibility and patterns of informed trading tend to change with the indulgence of options market. Also, the main concentration is on the target companies, whereas the M&A announcements affect the 1384 MF 45,10/11 acquiring companies differently. In order to extensively understand the dynamics of informed trading in the context of acquiring companies, we develop the following hypothesis: H2. The pre-announcement ARs for optioned and non-optioned acquirers differ significantly when the informed traders displace from or migrate between stock and options market.
Usually, informed traders have two ways to use the inside information, either to buy or sell stocks well before the M&A announcements. But the overall wealth creation for acquiring companies at the time of M&A announcements is either negligible or slightly negative (Mandelker, 1974;Asquith et al., 1983;Roll, 1986;Markides and Ittner, 1994;Morck and Yeung, 1991;Kiymaz, 2004;Betton et al., 2009;Banerjee et al., 2014;Wang, 2008;Andrade et al., 2001). So, insiders tend to either sell their existing positions in the shares or do passive trading by reducing their buying positions in the stock of acquiring companies. Thus, they can hedge the risk of losing money due to upcoming downturn at the time of M&A announcement. But this move happens when options are unavailable. These transactions of informed traders reduce the stock market returns in the pre-announcement period. In case of options availability, the informed traders do not have to necessarily sell their holdings. In fact, they can take long positions in the stock market and hedge the risk and/or earn ARs by taking equivalent positions in an options contract ( Jayaraman et al., 2001;Spyrou et al., 2011). This move increases the stock trading before the M&A announcements. Based on these viewpoints, we construct the following sub-hypothesis: H2a. The ARs prior to the M&A announcements of optioned acquirers get significantly higher than the non-optioned acquirers.

Crisis period and regulation effect
Abumustafa and Nusair (2011) found that insider trading increased significantly during and immediately after the financial meltdown of 2008. Similarly, Cziraki (2018) elaborated that insiders perceive the risk of crisis as per the exposure level to the crisis. He showed that if the risk exposure of the company is high, insiders start selling their holdings before the crisis. Thus, if the companies opt for M&As during the crisis period, it affects the risk level of the company. Regarding the regulatory impact, researchers like Ordu and Schweizer (2015), Healy and Palepu (2001) and Atanasov et al. (2006) explained that regulatory reforms bring down the level of informed trading. Learning from already happened real examples, the authorities improve the existing leakages in the current system that have been causing the insider trading. Hence, insiders indulge more in illegal transactions to compensate the risk increased due to the turmoil in the market. They also try to avoid the risk emerged from strong regulations, and thus reducing the prior announcement trading based on inside information. The followings hypotheses are developed based on these viewpoints: H3a. The insider trading gets higher around those M&A announcements that occur during/immediately after the crisis period.
H3b. The insider trading gets lower around those M&A announcements that occur after the regulatory reforms.

1385
Informed trading in M&As we start our analysis from 2004 due to various reasons. First, the options market of India was at its nascent stage in the initial years and the number of stock option contracts started growing continuously from 2004 onwards [4]. In India, the weekly options also started in the same year, which increased the depth of options market. Second, many pro-M&A regulatory changes such as leniency in regulation by the central bank of India for investment in mergers, introduction of automatic route, etc., happened from the year 2004 onwards (Nayyar, 2008;Banerjee et al., 2014).
Since 2004, the number of these options contracts has been continuously growing until the year 2008 when equity options market in India faced 39 percent drop in the number of contract traded. We take 2008 as the first cut-off year for the analysis to gauge the effect of global financial crisis.
The next cut-off year for the analysis is 2013 when the new Companies Act 2013 was passed, which brought more emphasis on corporate governance practices in India. The new act also brought in significant amendments in the regulations regarding the M&A activities in India, and it intended to make them more streamlined and transparent [5] , [6]. The time period 2013 onwards is considered to measure the effect of regulatory reforms.
After selecting all the M&As announced in the defined period, we screen them on the basis of four criteria[7]: all the deals get successfully completed until April 2017, the minimum deal value is given and at least $1m, if the same acquirer has more than one M&A announcement within 3-month period then we consider only the first one and there are required number of data points available for the analysis. We get a final sample size of 864 deals. The year-wise distribution of deals is presented in Panel A of Table I. There are 540/324 deals announced by acquirers without/with options availability.
The deal-specific attributes are detailed in Panel B of Table I. There are 646 acquisition announcements, and the remaining 218 announcements belong to mergers. The number of deals in which target firm shareholders are paid in cash is 630, whereas 159 deals have payment through stocks. There are 26 such deals for which payment is done by using a mixed mode of payment, and the mode of payment is not disclosed for 49 deals at the time of the announcement. The average deal value for all the M&A announcements is $162m. There are 128 deals in which the target company is Indian, and the remaining 736 M&A deals are related to cross-border.
The M&A announcement data are obtained from Bloomberg database and PROWESS[8], and cross-checked by consulting company-specific websites. The daily stock price and implied volatility (IV ) data are also taken from Bloomberg database. The options market data are taken from the official website of the NSE.

Event study
In order to conduct the empirical analysis, we employ one-factor market model of event study methodology [9]. We use the following three-step process for computing the ARs generated around M&A announcements. The deviation of AR from 0 in the pre-announcement period reflects the level of informed trading and shows market reaction on/after the first public announcement day. In the first step, we compute the expected return (ER it ) on a particular day t for a company i through the following equation, where the estimation period[10] for the model is t ¼ −160 to t ¼ −61 days relative to the actual announcement day (t ¼ 0): where α i refers to the level of return for the security i when R mt is 0 and β i shows the sensitivity of R it to the market wide factors. R mt is the return on market index [11], whereas ε it measures the effect of variables with respect to a security i.

1386
MF 45,10/11 In the second step, we calculate the daily ARs through the second equation and average of ARs for n companies (AARs) through the third equation: where AR it is abnormal return for a security i on day t and AAR t is the arithmetic mean of ARs on day t for N companies. In the last step, average ARs are summed up over different study windows to get cumulative average abnormal return (CAAR) for specific windows (Equation (4)): where CAAR(k, m) is the cumulative average abnormal return for N companies with respect to a specific window, and each window starts at time k and ends at time m. The CAARs are measured separately for study windows of −30 to −1, −20 to −1, −10 to −1 and −1 to +1 days. In order to get the significance of CAAR, we use the following test statistics (Equation (5)): where AAR is the average abnormal return of all the securities from day k to day m, and n is the number of days over which AAR s are summed up. δ AAR shows the standard deviation of AAR s over the estimation period (t ¼ −160 to t ¼ 61). We separately compute the CAAR for optioned acquirers (CAAR O ) and CAAR for non-optioned acquirers (CAAR NO ) with respect to all the study windows.
After analyzing the informed trading for optioned and non-optioned acquirers individually, the option listing effect for different study windows is also examined. Option listing effect measures the difference between the abnormal stock returns occurring before the M&A announcements in case when options market is available or not. For analyzing the significant difference in ARs of optioned and non-optioned acquirers, independent sample one-tail t-test is conducted.

Black-Scholes-Merton model
In the end, we verify if informed traders use combination of stocks and options for utilizing the confidential information about the upcoming M&A announcement. For this, we compare the stock market trading with the level of options market volatility through following process.
At first, we sort the sample of optioned acquirers into three terciles, named Q33, middle and Q66 for different study windows. Q33 represents top 33 percent and Q66 represents bottom 33 percent, thereby dealing with the highest positive (negative) cumulative abnormal stock returns in a specific study window. We exclude the middle tercile for the robustness.
Second, we get the daily IV[12] for the estimation and study window(s), as defined in Subsection 3.2. IV is propounded by Black and Scholes (1973) and Merton (1973). It cannot be observed directly and is derived by reverse engineering the following Black-Scholes-Merton model: where C denotes the call option's price and p is the put option's price. S o and X represent the spot and exercise price, respectively. T is expiry time and r shows risk-free rate of interest. N denotes cumulative standard normal distribution, ln is normal log and s is volatility.
Then we take average of implied volatility (AIV) for n number of deals underlying Q33 and Q66 terciles. In the end, we divide AIV of a specific study window to AIV of estimation window called implied volatility ratio (IVR). The IVR is computed separately for call and put options again with respect to Q33 as well as Q66. It is done to examine if informed traders migrate between stock and options market. If they develop strategies with combination of 1388 MF 45,10/11 stocks and options, then we expect to have high IVR for Q33 than Q66. For having a deeper understanding of whether call or put is preferred by informed traders, we also check the put/ call volume ratio with respect to the specific study windows for terciles Q33 and Q66.

Options availability and abnormal trading
In this section, we discuss the empirical evidence whether informed trading occurs well before the announcements of M&As and vary as per the option listing status of the acquiring company (Table I). First, we examine the abnormal stock trading for the entire study period . The results show that the level of CAAR NO is statistically insignificant for all the study windows. CAAR O gets insignificant in the shorter study window of −10 to −1 days, but CAAR O is significantly positive in the sub-periods −30 to −1 (1.15) and −20 to −1 (0.88) days. We can observe a continuous buildup in CAAR O , which starts 30 days prior to the first public announcement of the acquisition (Figure 1). These results support the conjecture that the abnormal stock trading increases significantly prior to the M&A announcements of Indian acquiring companies, but our finding is conditional on the availability of options. In case of options unavailability, we do not observe any significant presence of informed traders consistent with the findings of Lorie and Niederhoffer (1968), Jarrell and Poulsen (1989) and Agrawal and Nasser (2012).
Next, the results of the mean difference test (Panel B) show that CAARs for optioned companies are significantly higher than CAARs for non-optioned companies in the sub-periods −30 to −1 (3.15) and −20 to −1 (1.83). These results indicate that the pre-announcement ARs for optioned acquirers get higher than non-optioned acquirers. This finding is in alignment with the results of Arnold et al.  (Table II).

Crisis period and regulation effect
In order to examine the behavior of informed traders around crisis period and regulatory reforms, we report the empirical results in Table III. We examine the abnormal stock returns across different years. When we look at the results of CAAR NO for the pre-crisis period (Panel A: 2004(Panel A: -2008, none of value is statistically significant; however, CAAR O is insignificantly positive in all the study windows, except the sub-period −20 to −1 days in which it is significantly positive with 1.33 percent value.  The results of the time period immediately after/during the sub-prime crisis (Panel B: 2008-2012) reflect higher magnitude and consistently positive CAAR O throughout all the years. CAAR O is significantly high in the pre-announcement window periods of −30 to −1 and −20 to −1 days, with values 2.87 and 1.53 percent, respectively, and insignificantly positive in the study window of −10 to 1 (0.59 percent) days. Similar finding is illustrated by Netter and Mitchell (1989) who showed that insiders have an accurate idea about the company's health, and they can earn ARs around those corporate announcements that occur immediately after the crisis period.
When we look at the CAAR values during and after the year when Companies Act 2013 was introduced (Panel C: 2013-2017), we see that CAAR NO is significantly negative with values −3.16 and −1.93 percent in the sub-periods −30 to −1 and −20 to −1, respectively. CAAR O is significantly negative in the study window of −30 to −1 (0.98) days. The CAAR values for the remaining windows are not statistically significant. The significant negative CAAR value signifies the passive trading on the parts of informed traders, either in form of 75 Notes: Column 2 shows the number of deals with respect to parameters enlisted in Column 1. Columns 3-10 display the values of cumulative average abnormal returns and the respective t-statistics for their significance for sub-periods −30 to −1, −20 to −1, −10 to −1 and −1 to +1. *,**,***Significant at 1, 5 and 10 percent levels, respectively 88 Notes: Column 2 shows the number of deals with respect to parameters enlisted in Column 1. Columns 3-10 display the values of cumulative average abnormal returns and the respective t-statistics for their significance for sub-periods −30 to −1, −20 to −1, −10 to −1 and −1 to +1. *,***Significant at 1 and 10 percent levels, respectively  Agrawal and Nasser (2012). But, we do not find any trace of increase in purchases of informed traders. It shows clearly that regulatory reforms played a noteworthy effect on the informed trading in India in the context of mergers and acquisitions, and informed traders did not indulge in long positions of stock for developing options strategies.
In the unreported results[13], we compare the CAAR values of time periods 2008-2012 and 2004-2007 and 2013-2017 and 2008-2012 individually in order to know the crisis and regulation effect, respectively. We employ independent sample one-tail t-test and find that CAAR O values belonging to the time period 2008-2012 are significantly higher than CAAR O values of the time period 2004-2007 for the study window of −30 to −1 days. The results show that informed trader got more pronounced in the stock market of optioned acquirers prior to those M&A announcements that occurred during/immediately after the crisis period. Regarding the regulatory effect, the CAAR O values belonging to the time period 2013-2017 get lower than CAAR O values of the time period 2008-2012 for the study window of −30 to −1 and −10 to −1 days. However, this inference does hold for the M&A announcements belonging to the acquirers without option listing status in case of crisis as well as regulatory effect.
The CAAR values in the sub-period −1 to +1 are presented to capture the reaction of the market when M&As are announced publicly. We see that neither CAAR O nor CAAR NO deviates significantly from 0. These results are consistent with the prior literature which show that acquiring firm shareholders' gains are negligible at the time of M&A announcements. We also find that unlike informed traders, the regular market participants do not consider the availability of options market substantial to act differently to the announcements.

Robustness check
In what follows, we conduct the robustness check by comparing abnormal stock returns with the options market variables to assure that the level of informed trading has a liaison with the options availability. Our indicative analysis demonstrates that informed traders get more prevalent when acquiring companies have availability of options. In this section, we verify if it occurs possibly because informed traders migrate between stock and options market and develop strategies by jointly using tools of stock and options market. If it is true, then call and/or put IVR should be higher when abnormal stock trading is higher. Table IV shows the result of call and put IVR along with put/call ratio with respect to Q33 and Q66. The migration effect is analyzed for two sub-periods −30 to −1 and −20 to −1, because these study windows reflect highly significant informed trading in the preliminary analysis. The results show that call (1.02) and put (1.04) IVR is higher for Q33 than the call (0.96) and put (0.96) IVR in the study window of −30 to −1 days. The IVR for both call as well put is higher in case of Q33 than Q 66 in the study window of −20 to −1 days also. If we

1391
Informed trading in M&As compare the IVR between the study windows, all the IVR values are higher in the sub-period −20 to −1 than the sub-period −30 to −1. If we look back at the values of CAAR O for these sub-periods, we see that the CAAR O is higher in the sub-period −20 to −1 than the sub-period −30 to −1. It affirms that informed traders migrate between stock and options market for utilizing the material non-public information about the upcoming M&A announcement.
When we compare the put/call volume ratio of Q33 and Q66, we see that this ratio is smaller for Q33 than Q66 in both the study windows. It exhibits that call options are used more than put options by informed traders. The supportive results are found by researchers like Jayaraman et al. (2001), Cao et al. (2005) and Augustin et al. (2015) who mentioned that informed traders select more/only call options for using confidential information about upcoming M&A announcements.

Conclusion and implications
On the basis of the above discussion, it can be determined that informed traders use the confidential information about the upcoming M&A announcements, especially when options market tools are available to be used jointly with the stocks. Another noteworthy outcome is that magnitude and possibility of informed trading in the stock market of optioned Indian acquirers increase significantly during and immediately after the crisis period. The reason being that informed traders perceive the turbulence in the economy as quite risky and they end up hedging the risk or earning ARs on the basis of unpublished price-sensitive information (UPSI) about the acquisition announcements. The underlying opportunities are wider when underlying acquiring company has options market alongside, so informed traders get more prevalent under such condition. The results also indicate that informed traders do not get involved in stock purchases when new regulatory reforms are introduced. This study brings into notice that option listing status is a special condition wherein the likelihood of informed trading is higher and has implications for analysts, regular traders and regulatory authorities. The empirical evidence of the study provides a base to gauge the informed trading more cautiously as per availability of options, along with the other considerations like crisis and regulatory impact.

Limitations and future scope
The limitation of the study is that informed trading cannot be directly and completely traced until the informed traders report all their transactions based on UPSI, which is unlikely in the real-life scenario. Another limitation of the study is that options strategies are not identified.
Future study can be conducted for the identification of precise options strategies adopted by informed traders prior to the M&A announcements. The effect of options availability on the informed trading can also be examined for other corporate announcements. 8. This is a database software originated by Centre for Monitoring Indian Economy (CMIE).
9. The market model is widely used by researchers to compute the abnormal returns occurring around M&A announcements ( 11. Here, the market index employed is NIFTY 500. 12. We consider implied volatility of one-month options contracts with absolute delta value of 50. It is due to the reason that researchers like Arnold et al. (2006), Wang (2008) and Chan et al. (2015) explain that informed traders prefer short term at the money contracts.
13. These results are available from authors on request.   49 Notes: Column 2 shows the number of deals with respect to parameters enlisted in Column 1. Columns 3-10 display the values of cumulative average abnormal returns and the respective t-statistics for their significance for sub-periods −30 to −1, −20 to −1, −10 to −1 and −1 to +1. *,**,***Significant at 1, 5 and 10 percent levels, respectively Findings -To create value after a horizontal M&A, it is necessary to concentrate on turnover and the restructuring of charges without neglecting the control of debt capacity. To avoid destroying value after a horizontal M&A, it is necessary to concentrate on the control of debt capacity and restructuring of charges in order to reduce financial charges and financial risk. Horizontal M&A also create value through the reduction of investment costs and through tax optimization.
Research limitations/implications -This paper is different from other contributions in that the majority of existing literature concerning the sources of value creation in M&A has been based on abnormal returns or microeconomic data. This paper analyzes accounting data that are likely to be influenced over the long term by corporate decision making. These kinds of decisions influence the firm's value as well as the long-term gains that industrial investors may hope to obtain. Originality/value -This study makes a significant contribution to the existing literature insofar as it seeks to divide the sources of value creation into three categories: sales synergy, cost synergies and hybrid synergies. To the best of the authors' knowledge, this is also the first study to provide explanations from companies' accounting data, which can lead managers to a greater vision of post-merger strategy management, reinforcing the mechanism for value creation.
Keywords Mergers, Value creation, Value destruction Paper type Research paper

Introduction
Horizontal mergers and acquisitions (M&A) occur when two competitors combine, often with the key benefit of an increase in market power for the newly joined entity (Gaughan, 2010). Those which allow an increase in shareholder equity, fixed assets and turnover by increasing the company's size also result in increased expenditure to maintain this size. Consequently, the new entity faces several difficulties: management of the integration process, the elimination of duplicates (in human resources and equipment), the renegotiation of contracts with business and financial partners and the control leakage problem. The latter issue, which is a potentially expensive problem, can bring the restructuring of the group into question. It results from the increased size of the new entity size's growth and from the multiplication of its services. It can only be resolved through effective, short and rapid restructuring. Academic research has contributed to understanding this phenomenon through studies of motivations and performance. Thus, while there has been a significant amount of research on M&A, more is required to help us understand the paradox of value creation and destruction and to provide recommendations that will help strategists enhance their performance in terms of M&A strategy (Hitt et al., 2009).
Eger (1983), Rappaport (1986), Porter (1987), Ansoff (1984), Seth (1990), Healy et al. (1992), Mueller (1992, Perdreau (1998), Creswell (2002 and others have explained the determiners of value creation in M&A by proposing several hypotheses without empirical contributions. Among these proposals, we wish to highlight market power, asset restructuring, tax optimization, increased debt capacity, the co-insurance effect, financial risk reduction, economies of scale, reduction of investment costs, etc. No empirical studies have analyzed the sources of value creation and destruction. Malatesta (1983) and Bradley et al. (1988) confirmed that no empirically obvious facts exist because of the methodology used in these cases, which does not offer the possibility of isolating the link between various sources of advantages.
In this paper, we aim to fill this gap, using these theoretical contributions as a basis from which to propose and empirically test accounting variables which are considered to have the capacity to influence the value of the new entity. We measured the sources of value creation from the perspective of industrial investors, operating to ensure business continuity. The value measured here is therefore the result of the efficiency of operational centers and functional departments. Strategic or operational synergies due to horizontal M&A are thus generated as a result of strategic decisions, rethinking the redeployment of tangible and intangible assets. This value is therefore a long-term balance sheet measure.
Using a sample of 90 French companies involved in a horizontal M&A between 2005 and 2014, we identified a value-creating group and a value-destroying group in order to highlight those accounting indicators which are sources of value creation and destruction in horizontal M&A. Our study makes a significant contribution to the existing literature insofar as it seeks to divide the sources of value creation into three categories: sales synergy, cost synergies and hybrid synergies. To the best of our knowledge, this is also the first study to provide explanations from companies' accounting data, which can lead managers to a greater vision of post-merger strategy management, reinforcing the mechanism for value creation.

Value creation and destruction in M&A: a general overview
Value creation is one of the most common motives in M&A (Meghouar, 2016). It is defined as the gain in efficiency ensuing from advantages in terms of costs (Morck and Yeung, 2002) or as the capacity to make a combined company more profitable than two individual companies (Brealey and Myers, 2002). Koller et al. (2015) found that companies which dedicated themselves to value creation were healthier and stronger.
In order to measure performance in horizontal M&A, Capron (1999) distinguished two types of sources of performance: cost synergies obtained via asset divestiture and revenue synergies obtained through the redeployment of resources. According to Dimopoulos and Sacchetto (2017), there are two possible sources of synergies: a reduction in the fixed cost of production, or an increase in marginal productivity. The presence of synergies thus means that mergers are potentially profitable for the coalition of merged and merging firms (Ding et al., 2013). For example, in air transport, companies involved in horizontal M&A enjoyed increased scale returns ( Johnston and Ozment, 2013); demonstrating the presence of marginal and fixed cost savings (Chow and Fung, 2012); positive effects on productivity have also been identified (Schosser and Wittmer, 2015). In the same way, Tao et al. (2017) found evidence which was consistent with improved productive efficiency and increased buying power as sources of gains in horizontal mergers.
However, a horizontal M&A tends to reduce innovation incentives in the absence of efficiencies (Motta and Tarantino, 2017) and has an 18-19 percent greater probability of being challenged in industries which are subject to the market concentration hurdle hypothesis (Tao et al., 2017). Other studies indicate a destruction of value for shareholders 1399 Value creation and destruction in horizontal M&A (Kürsten, 2008) and a reduction of economic performance following a merger, or at best a small improvement of this performance (Trautwein, 1990). Most recent studies estimate these failure rates at around 45-50 percent (Schoenberg, 2006). However, it is important to note that there is no consensus concerning M&A performance (Ibrahimi and Taghzouti, 2014) and the criteria used to measure this (Zollo and Meier, 2008). This paper is different from other contributions in that the majority of existing literature concerning the sources of value creation in M&A has been based on abnormal returns or microeconomic data. Our paper analyzes accounting data that are likely to be influenced over the long term by corporate decision making. These kinds of decisions influence the firm's value as well as the long-term gains that industrial investors may hope to obtain.

Characteristics of the empirical framework 3.1 Sample characteristics
We selected French companies involved in a merger or acquisition. Our sample respects the following criteria: the transaction was either a horizontal merger[1] or a majority horizontal acquisition (more than 50 percent of capital acquired) realized between two competitors selling the same product or serving the same market; the smaller of the two firms had to be at least one-third of the size of the other (see Footnote 1), measured on the basis of annual turnover; the turnover for each company for the year of the deal was in excess of €1m; and the deal was made at least three years previously, making it possible to observe its effects [2].
Using the "Thomson One Banker" database, we selected all French companies having realized a merger or a majority acquisition between 2005 and 2014. We then deleted operations involving banks and insurance companies, and those companies for which data are not available. This left us with a final sample of 90 forms. These 90 firms were split into two groups: Group 1 (value creating) contains 47 companies presenting increased net cashflow three years after the operation, and Group 2 (value destroying) contains 43 companies presenting reduced net cash-flow three years after the operation.

Variables
In our study, value creation and its sources are only measured by accounting indicators (Barney, 1988;Harrison et al., 1991). Our literature review allowed us to include eight sources of value creation (or destruction) put forward by various authors. Our choice was made on the basis of the operability and importance of variables in the literature. We consider that the indicators of value creation can also indicate value destruction (see Table I and its discussion for more explanations).

Sign of β Revenue synergy
Costs synergy Hybrid synergy Situation ΔTv ΔOpC ΔFA ΔWC ΔFiC ΔTax ΔDebt ΔFinRisk If the β is positive, the average variation of its variable must be positive and less than the average variation of revenue synergy Value destruction (the variable to explain is negative) + or − if average variation is negative and lower than the average variation of cost synergy variables If the β is positive, the average variation of its variable must be positive and greater than the average variation of revenue synergy  (1997) and Laux (1999), used cash-flow to distinguish healthy companies from failing ones. In the same way, the free cash-flow theory  suggests that there is a connection between cash-flow and takeovers. Since then, other studies have shown a positive correlation between cash-flow and takeover risk (Oprea, 2008) and between cash-flow and diversification (Doukas and Kan, 2004). Finally, whilst some researchers have mobilized ROA or ROE to measure M&A performance, Barney (1986Barney ( , 1988 recommends the use of cash-flow to measure the long-term performance of M&A. However, the means of measuring cash-flow differs from one study to the next. Researchers may use net cash-flow, gross cash-flow, free cash-flow or operating cash-flow. Germain and Trebucq (2004) studied the relation between commissionership and mergers, providing an analysis of gross operating cash-flow and net cash-flow after financial elements and corporate tax. In pecking order theory, on the other hand, Myers (2001) chose to discuss net cash-flow. In our study we have chosen to focus on net cash-flow because it represents a strategic quantity including the money supply currently available to the company. Additionally, in our regressions, we have considered the impact of interest and tax on value creation, two elements which are contained within net cash-flow contain these two elements. Finally, measurement of net cash-flow involves the addition of allocations for provision and amortization to net income, and the deduction of the variation in working capital and Capex.

Sources of value creation and destruction
• Turnover (Tv): by studying the evolution of turnover, we can easily analyze its financial challenges. This value represents the market share attributed to every company. A merger or an acquisition creates value thanks to an increase in market power (Porter, 1987) by reducing competition in the market and by raising barriers to entry (Mueller, 1992;Porter, 1987). Thanks to its new competitive position and its new market power, a merged firm can increase the sale price of its products (Porter, 1987) and/or increase the quantities sold.
• Operating cost (OpC): while turnover growth reflects a key element in company performance, a more rapid increase in operating expenses can reduce this performance, potentially to the point of destruction. Combination operations are often accompanied by a rationalization of charges to reduce wage costs via job cuts and to reduce raw material purchasing costs by leveraging the new entity's greater bargaining power. These two costs represent the greatest part of operating costs, which are equal to the difference between turnover and operating profit. Improved exploitation, economies of scale (Rappaport, 1986), operational synergies (Seth, 1990) and an increased productivity (Healy et al., 1992;Rahman and Limmack, 2004;Schosser and Wittmer, 2015) are thus sources of value creation in horizontal M&A.
• Working capital (WC): profiting from its increased size, the new entity may force business partners to bear part of its costs in order to reduce its working capital (e.g. by imposing tighter delivery deadlines, reducing payment terms for customers and increasing payment terms for suppliers). This bargaining power aspect was put forward by Porter (1987) and Mueller (1992). Put simply, working capital comprises uncollected turnover, purchases which have yet to be paid for, sold production or untransformed raw materials. If this value is not controlled, it can lead to a lack of liquidity. It is equal to the amount of inventory and trade receivables less the amount of trade payables. If working capital increases faster than turnover, then value is destroyed. Its reduction, however, is a source of value creation.

1401
Value creation and destruction in horizontal M&A • Investment cost of fixed assets (FA): the main motivations for horizontal M&A include rapid access to new technologies and rapid acquisition of finalized investments (Creswell, 2002). The reduction in investment costs (tangible assets) is a source of value creation, allowing companies which did not adapt prior to making these investments to avoid heavy expenses; in the case of acquisitions, the charge is supported by both merged companies. M&A thus represents an internal market free from material means which promotes better use of assets and reduces the cost of investments (Rahman and Limmack, 2004).
• Financial charges (FiC): financial charges represent the remuneration of financial debts contracted by the firm. They depend on its level of debts and on interest rates. Therefore, financial restructuring is often envisaged in order to reduce the weight of financial charges and to encourage financial partners to lower their interest rates. Large size, which often goes hand-in-hand with multibankarity, increases the weight of financial charges (Dietsch and Golitin-Boubakari, 2002) which allow tax savings to be made (Modigliani and Miller, 1963). However, they can easily become a source of value destruction if they soak up any value created by the exploitation of the company.
• Profit tax (Tax): given that a part of profit returns to the state, it is easy to see the need to use tax potential effectively in the context of company combinations to create value (Rappaport, 1986). According to Hayn (1989) and Shih (1994), who believed that certain takeover bids were made exclusively for reasons of tax savings, this tax reduction in M&A results from four elements: non-taxation of increases in asset value from revaluation following a merger or an acquisition of total assets; increased depreciation following revaluation of the increase of fixed assets; the possibility of profiting from deferred tax advantages held by one of the companies involved; and/or the transferability of tax deductions based on past losses made by either company.
• Debt capacity (Debt): empirical studies show that there is a link between mergers and debt (Sverdlove, 2015) and that the level of debts is positively correlated with the firm's size and its tangible assets (Rajan and Zingales, 1995) thanks to the guarantees which they represent. This value is measured using the financial debts/shareholder equity ratio (Watts and Zimmerman, 1990). M&A favors the increase of debt capacity (Perdreau, 1998) by allowing the financing of investments, enabling new purchases, solving cash-flow problems and to funding growth. Once the optimal level is exceeded, however, this increase in debt capacity becomes a source of value destruction.
• Financial risk: in cases of increased financial risk, the companies may find themselves in a difficult situation due to the reduced confidence of financial and commercial partners. M&A allow the reduction of financial risk (Rappaport, 1986) and of illiquidity risk (Eger, 1983). Inspired by Wruck (1990) and John (1993), who connect financial risk to lack of payment, we define financial risk as a lack of liquidity or a temporary insolvency, making the firm unable to face terms and financial commitments. The ratio used to measure liquidity (Liquid) is the relationship between short-term assets (except stocks) and short-term liabilities (Michalopoulos et al., 1993); the ratio measuring solvency (Solv) is the relationship between total assets and total debts (Altman et al., 1994;Michalopoulos et al., 1993).

Methodology
3.3.1 Models. In our model, we chose to use the variation of indicators rather than their value in order to reveal the determinants of value creation or destruction and to neutralize the effect of inflation. For example, in a case where cash-flow passes from €200 to 240 1402 MF 45,10/11 (20 percent variation), we would only be interested in the 20 percent increase, rather than the actual value of €240. Working along the same lines, we calculated the variation of indicators explaining this increase. Value creation or destruction can be stopped by a significant variation of one or a combination of several variables. We have used regression analysis to understand the relationship between variables. This approach allows us to produce a model of the relationship between two or more variables and to demonstrate the characteristics of this relationship. Our multiple linear regression is calculated in the following way: where Y is the dependent variable (the variation of net cash-flow); a the constant; b 1 , …, b n regression coefficients (β); and X 1 , …, X n are the independent variables.
However, studies based on regressions analysis are often faced with a problem of multicolinearity. In these cases, other regressions have been proposed to avoid colinearity. Examples include stepwise regression (Foucart, 2006), which is perfectly suited to the purposes of our study (Tenenhaus, 1996). It consists of introducing variables one by one. At each stage, the model including all the selected independent variables is recalculated using the standard method. Variables with a contribution probability which fails to meet the fixed threshold (5 percent) are then gradually eliminated, one by one. Our regression models are as follows.
Model 1: Model 2: Where Δ is the variation. 3.3.2 Expected sign of β. In regression analysis, the β sign (b n ) can help us to identify the influence of each variable. However, given that the same variables can be responsible for value creation in one group and value destruction in the other, our β analysis is more complex. There are several possible situations, involving three fields.
We divided the sources of value creation and value destruction into three fields by paying specific attention to the β of significant variables. The first field corresponds to revenue synergy, which consists exclusively of the increase in turnover. This is the only source of cash input. In cases of value creation, this β must be positive. However, in cases of value destruction where the dependent variable is negative, the β must be also positive. If the β is negative, the average variation must be negative (reduction of revenue synergies) and less than the average variation of the cost synergy variables, meaning that cost synergies must be greater than revenue synergy.
The second field corresponds to cost synergies (decreased of expenses). It consists of the variation of operating expenses, of investment costs in fixed assets, the working capital, financial charges and tax on the profit. Their reduction (negative β) indicates value creation. On the other hand, their increase (positive β) reflects a destruction of value, unless their average variation is positive but less than revenue synergy.

Value creation and destruction in horizontal M&A
In practice, cost synergies often increase at the same time as revenue synergy. For example, increased turnover implies an increase in the quantities sold, thus requiring an increase in purchases of raw materials and in man-hours. We thus wish to know whether revenue synergy increased more than cost synergy.
In cases of value destruction and on the condition that the dependent variable is negative, we suppose that the sign of the β for costs synergies must be negative, or positive if the average variation is positive and greater than the average variation of revenue synergy.
The third field is a hybrid field which does not correspond to a source of cash input or to an opportunity for savings, but which is intended to reflect the image of the company in the eyes of financial and commercial partners. An increase in debt capacity (positive β) is taken to reflect the creation of value, and a reduction in this capacity (positive β as the dependent variable is negative) reflects value destruction. On the other hand, if the average variation of this variable is negative in cases of value destruction, then the β may be negative.
In terms of financial risk, an increase in the indicators in question (ΔLiquid and ΔSolv) reflects reduced financial risk. In the case of value creation, the β must be positive. On the other hand, if the dependent variable is negative, the β must be positive, or negative if the average variation is negative. Table II shows the number of operations and their percentage per year. Table III gives the number of operations per sector and their total turnover. Several interesting elements are immediately apparent from these tables. A total of 70 percent of the operations in question took place in 2009, 2010 and 2011. In terms of sectors of activity, 23 percent of the total sample was industry related, 20 percent high tech and almost 16 percent was connected with consumer goods.

Results and comments 4.1 Analysis of sectors of activity
The three sectors alone thus make up 59 percent of our sample. The rest is made up of companies from other sectors. Note that in the year of the operation, these 90 companies had a total turnover of €349,608m. However, the distribution of turnover by sector is not the same as the distribution of the number of companies. The energy sector, for example, is represented by only three operations, but these operations make up 20 percent of total turnover.

Analysis of accounting data
This analysis shows the characteristics of each group before and after the merger or acquisition (Tables IV and V ).
Comparing the two groups, the averages of the elements representing size (total assets, shareholder equity and turnover) are higher in the value-creating group than in the valuedestroying group. This shows that the groups in the first group were larger than those in the second group, which seems to indicate that large firms far better than smaller companies in horizontal M&A.
The second notable feature is that the evolution of the means in the two groups is very different. Three years after the operation, almost all of the figures for the value-creating group are significantly higher, particularly in the case of turnover (up from €5,164m to 5,960m), the operational result (from €18m to 222m) and net profit (from €17m to 244m). In the value-destroying group, however, almost all figures had decreased three years after operations, notably turnover (down from €4,013m to 3,757), the operating result (from €384m to 242m) and net profit (from €191m to 97m).
This clear difference between the two groups supports the choice of net cash-flow as a means of measuring value creation, highlighting a first group for which the general situation

1405
Value creation and destruction in horizontal M&A improved, and a second group where the general situation deteriorated. Out of the 90 firms included in our sample, over half (52 percent; 47 firms) of the merged companies increased their net cash-flow, while the remainder experienced a reduction in this value. In other words, three years after the deal, almost half of horizontal M&A resulted in value destruction [3]. Table VI provides a more detailed comparison of the performance of the two groups. Three years after merger operations, the performance of the value-creating group improved. The proportion of turnover taken up by OWC was down to 18.59 percent. Operational profitability increased by 965.71 percent and net profitability went up by over 1,000 percent. In the second group, performance was lower three years after the merger, with operational and net profitability decreasing by 32.5 and 45.7 percent, respectively.
In the year of the merger, the mean OWC represented 7.21 percent of turnover in the first group, increasing to 24 percent for the second group. Logically, this suggests that larger companies can exert more pressure on their trading partners. However, during the same year, the second group enjoyed better overall performance than the first group (Group 1: interest cost/total debts ¼ 2.69 percent, operating result/total assets ¼ 9.57 percent, net results/turnover ¼ 4.77 percent; Group 2: 2.73, 0.35 and 0.34 percent, respectively). This shows that companies in the value-destroying group generated positive results prior to the merger or acquisition, and that their poor performance three years later is due to the combination operation. Table VII shows descriptive statistics for variables for the value-creating group used in our analysis.

Descriptive analysis of the value-creating group
In this value-creating group, net cash-flow increased by 82 percent on average, with a 76 percent average increase in turnover. This shows that the increase in net cash-flow is not exclusively due to increased turnover and may also result from increased control of other costs. The lowest observed variation in turnover (2.13 percent) supports this hypothesis: the company in question increased its net cash-flow, as it forms part of the value-creating group. The highest average, 76 percent, concerns the variation in turnover and solvency. For the value-creating group, two figures fell: taxation (−10 percent) and fixed assets (−8 percent). This may reflect the firms' efforts to optimize their fiscal potential and reduce their investment expenditure for fixed assets. However, lower indebtedness (−6 percent) does not result in lower interest expense, as we see from our sample. Value-creating companies may prefer to contract short-term debts with higher interest rates, leading to increased interest expense. Table VIII shows data for companies in the value-destroying group.

Descriptive analysis of the value-destroying group
The largest variation in this sample of 43 companies, with an average reduction in net cash-flow of 95 percent three years after merging, is in solvency (−91.45 percent). Since assets increased, the reduction in solvency is due exclusively to the excessive increase in total indebtedness [4]. This observation is validated by regression analysis. Note that the smallest reduction corresponds to the variation in tax (−2.97 percent). This change is a logical consequence of the lower profits obtained by the companies in question after combination.
We see that the average changes in the cost of fixed assets (+20.33 percent), operating expenses (+15.93 percent) and the OWC (+17.19 percent) are greater than the evolution in turnover (+9.84 percent). This imbalance in variations demonstrates an inability to master the three costs in question, leading to a reduction in average net cash-flow (−95 percent).
Finally, we have assumed that an increase in debt capacity constitutes a source of value creation, in that companies are thus able to access cash at any point. This is the reason for the positive variation in the liquidity ratio (11.84 percent). However, debt capacity also increased in value-destroying firms (21.26 percent): our findings do not support the above hypothesis, which has also been used by other authors. The increase in interest expenses (2.30 percent) reflects increasing financial debts. From this perspective, the results of regressions may prove interesting, as we also presumed that an increase in debt capacity would create value as long as a firm's debts remained below the optimum value-creation threshold.

Sources of value creation
Tables IX-XI show colinearity statistics (tolerance must be above to 0.33 and the VIF less than 3). It is not possible to analyze the significance of variables using the stepwise regression method, as, by its nature, it only tests those variables which are significant. However, replacing the liquidity variable by a solvency variable in the second model did not change the results of the first model. We have therefore created a single Turnover is the first relevant variable for the increase of net cash-flow. It is significant in the two models. After combination, the merged firms concentrate their efforts on increasing their sales. However, given the lack of data, we cannot confirm whether this increase results from the quantity of sales or from increased prices, which would have provided an indication of bargaining power with regard to customers. Horizontal M&A favor an increase in turnover, as all of the regressions show it to be explanatory factors for value creation. This growth in turnover can result from the horizontal expansion of the market or from the creation of monopolies thanks to industrial concentration. Operating charges are significant, in second position behind turnover. Two stepwise regression models confirm this, showing the strong significant power of the influence of operating costs on the variation of net cash-flow. In France, workforce strikes are a common occurrence in the wake of merger attempts. Staff knows that optimization and redundancies are likely to follow combinations. Furthermore, in cases of financial difficulty, management teams prefer to pass the economic risk onto their employees rather than onto the shareholders, who are responsible for their appointment and remuneration. Managers take advantage of combination operations to eliminate the least profitable elements and rationalize the operational management of their group. The main objective of these changes is to lower the cost price of products in order to favor the increase of operating income.
Profit tax comes in third position. In both models, the tax variable is more important than the reduction in investment costs. Companies may then redirect tax savings toward other investments or accelerate debt repayment. This factor highlights the determination of merged companies to reduce their tax bill. These principles are easily respected when it comes to value creation.
The reduction of fixed asset investment costs is fourth in the list. This variable is significantly different from 0 in two models. Without these fixed assets, the company cannot exist, and they must be continually renewed in order to ensure its continued survival. Horizontal M&A contribute to this process, providing rapid and easy access to new technologies and expensive investments and reducing the need for expenditure in obtaining these elements.
However, our two models show that the variation of working capital has no significant effect. This may be explained by the fact that the selected three-year period is not sufficient to observe the influence of this new power, or imply that the combination operation does not really permit new groups to impose conditions on their business partners. Finally, we should note that debt, financial risk and financial charges have no significant explanatory power according to our regressions. Companies do not, or cannot, immediately prioritize these three variables in the wake of a merger.

Sources of value destruction
This value-destroying group consists of 43 companies which presented reduced net cashflow three years after a merger or acquisition. This variable is therefore negative for all companies. Our objective here is to identify the parameters which were responsible for this decrease. We carried out two regressions using four explanatory variables.
The first model highlights the influence of debts. This result is confirmed by the second model which shows a significant variation at a level of 5 percent. These debts, generally contracted to finance the acquisition, are the main cause of value destruction following combination operations. Although mergers favor increased debt capacity by giving

1409
Value creation and destruction in horizontal M&A constant access to cash, our models indicate that this increase in debt capacity is at the root of all problems for the value-destroying group. These companies appear to have exceeded the optimal-level debt for value creation. The influence of the operating expense variable is also confirmed by the first model. Firms which do not restructure their operational services and do not rationalize their expenses following a combination destroy value. While turnover is not addressed by the regressions, this is because it followed the normal evolution of the merger. However, its evolution was absorbed by the uncontrolled increase of operating costs and other charges. Following a combination operation, companies must be able to handle all elements at the same time from the outset, restructuring the group, improving productivity, paying off debts, making profitable investments and more. To honor its terms, a firm must generate enough cash-flow, but nothing is guaranteed after the merger.
Moreover, the solvency ratio is also very significant. This ratio indicates that the companies in this group experienced financial difficulties. It seems that, further to the merger or acquisition, these companies were confronted with problems of solvency because of excessive debts. There is, in fact, a vicious circle. If firms go into debt in order to finance a merger or acquisition then fail to generate profit from the operation, then they must increase their debts to finance continued operations, as their available cash has been absorbed by an unprofitable investment. This can degenerate into a spiral of increased debt, reduced profitability and solvency.
Finally, there is also a highly significant variation in financial charges, strongly linked to the previous point. The increase in these charges destroys value. From our findings, we believe that the main problem in cases of value destruction is an excessive increase in debts.
These regressions demonstrate that investment costs do not explain value destruction, as horizontals M&A facilitate the transfer of tangible and intangible assets and do not initially require new investments. The same can be said of working capital which does not explain the variation in net cash-flow. We feel that our chosen period of three years post-merger is not sufficient to clearly show the effects on bargaining power with customers and suppliers. Financial logic indicates that increased debt will result in decreased tax on profits. Our model highlights this finding, showing that tax has no significant effect in cases of value destruction.

Discussion
We have seen that, following M&A, French value-creating companies concentrate their efforts on increasing turnover, restructuring operational charges, reducing investment costs and tax optimization. Our group of value-destroying French companies failed to recognize the dangers of excessive debts and of failing to restructure operational charges. These findings are illustrated in Figure 1. As tax savings and reductions in investment costs are temporary and only have an impact in the first years following an M&A, we may include the other elements in our two main factors: turnover growth and restructuring of operational charges. Both of these factors must be present to create value for shareholders.
In short, an M&A without turnover growth is doomed to failure. All M&A aim, directly or indirectly, to increase turnover growth. This forms the basis for all concentrations. However, without the restructuring of operational charges, attempts to increase turnover are also doomed to failure.
For all industrial companies, if we eliminate exceptional cash income such as that resulting from the sale of fixed assets and financial instruments, the only source of regular income lies in sales of goods and raw materials. All other sources of value creation relate to savings made by spending less. To ensure value creation in a horizontal M&A, turnover growth must be accompanied by charge restructuring. The difference between these two variables may be used to pay off debts, pay taxes, remunerate shareholders and build up a financial reserve. This combination of factors therefore facilitates increased liquidity and solvency.
The destructive role of excessive debt should not be underestimated. However, we consider that companies with sufficient liquidity will avoid excessive indebtedness. Nevertheless, in reality, all variables are closely linked and can have a direct or indirect influence on value creation. To illustrate these statements, Figure 2 shows the interconnection and crossing of factors.
If turnover growth is accompanied by efficient restructuring of charges, value creation will occur. If, on the other hand, the charge restructuring process is inefficient, the result may be value creation or value destruction. In the case of value creation, turnover growth favors the reduction of debt, resulting in reduced financial expenses. Finally, in the same scenario, turnover growth, debt reduction and the reduction of financial expenses also reduce financial risk, enabling value creation. Value creation and destruction in horizontal M&A In cases of a reduction in sales and of non-efficient load restructuring following a horizontal M&A, value will be destroyed. However, if reduced turnover is accompanied by efficient charge restructuring, then value creation is still possible. In cases where value is created, there are two possibilities: the first case of value creation as described above and the second case of value destruction. In this last scenario, a reduction in sales is likely to lead to an increase in debt and, consequently, an increase in financial charges. Financial risk therefore increases and the company's value falls.
As we have identified excessive debt as the primary culprit in cases of value destruction, we feel that simply controlling this factor would not be sufficient to create value. Debt control needs to be accompanied by an increase in sales and by charge restructuring. It would be foolish to concentrate on a single parameter to the detriment of the others.
Finally, contrary to Porter (1987) and Gao et al. (2017), finding that concentration in the aviation sector resulted in increased the bargaining power, our model does not indicate that working capital has an effect on value creation. As we indicated previously, this may be due to the fact that French companies do not exert power over their customers and suppliers; that this power is not sufficient to create value; that our chosen measurement period is too short to reflect this change; or that our variables do not permit us to measure this power, given that Porter and Gao et al. based their results on prices from a single sector. However, it seems clear that increased size results in greater bargaining power when negotiating with suppliers, customers and subcontractors.

Conclusion
The objective of this empirical study is to gain an understanding the sources of value creation and destruction in horizontal M&A. We tested eight accounting indicators grouped into revenue, cost and hybrid synergies which are likely to influence the value of the new entity. This work guides managers in decision making during post-merger phases. Using this evidence, our research provides a valuable complement to current studies based on economic variables and abnormal returns.
Our conclusions are as follows. First, to create value after a horizontal M&A, it is necessary to concentrate on turnover and the restructuring of charges without neglecting the control of debt capacity. To avoid destroying value after a horizontal M&A, it is necessary to concentrate on the control of debt capacity and restructuring of charges in order to reduce financial charges and financial risk. Horizontal M&A also create value through the reduction of investment costs and through tax optimization.

Notes
1. The desired effect is that integration will significantly influence the overall structure of the new entity by pushing it to initiate strategic and financial restructuring. This suggests that these companies have premises, machines, employees, etc., which may be pooled in order to realize synergies. This is the object of our study.

Introduction
Corporate acquisitions entail many challenges and benefits. While the introduction and integration of a mass of new employees are extensively discussed, the addition of a large group of new shareholders appears to generate less interest. Although new shareholders require fewer adjustments for management than does melding groups of employees, increasing the number of shareholders potentially offers significant costs and benefits. This paper measures the change in number of shareholders following an acquisition and examines the drivers of that change. It will assess whether this restructuring of the shareholder base is sustained in the years following the acquisition. Previous studies have generally found a large shareholder base to be associated with favorable outcomes. Research has shown that an increase in the number of a firm's owners may increase its stock's liquidity (Lipson and Mortal, 2007;Li, 2006); decrease risk (Foerster and Karolyi, 1999); and decrease cost of capital (Bodnaruk and Ostberg, 2013). Other studies have shown an increase in market value after controlling for liquidity enhancements from broadening the shareholder base (Kadlec and McConnell, 1994;Foerster and Karolyi, 1999). A potential cost of increasing shareholder base is that more dispersed shareholders may reduce monitoring and increase agency costs (Rozeff, 1982). An acquisition's impact on the shareholder base is, therefore, an important issue.
This paper finds that American firms completing major acquisitions experience large increases in the shareholder base, the number of shareholders of record. Not surprisingly, the broadening of the shareholder base is mainly for purchases paid for with at least some stock; cash acquirers experience median decreases in the number of shareholders. Another intuitive result found is when the target firm has more shareholders prior to the acquisition, the acquirer's shareholder base experiences larger increases after the acquisition. This increase in shareholder base is sustained four years following the acquisition, indicating target shareholders appear to be passive investors. The relationship between shareholder base and the relative number of shareholders of the target in a stock acquisition is maintained after controlling for other factors in both single and multistage regression analysis.
When the target company's shareholders receive stock following the acquisition, they might be expected to sell the stock immediately, because their new stock investment was not made purposefully. According to Merton (1987), investors hold only stocks about which they have information. Becoming the shareholder of a new firm via an acquisition could be the push that overcomes an initial cost of becoming informed about a stock. In fact, a striking finding of this paper is that four years following an all-stock acquisition, the mean (median) number of shareholders is 230 percent (39 percent) higher than prior to the acquisition. While a large increase in shareholders could result from investor interest due to the news associated with a large acquisition, the increase in shareholder base is seen only for stock acquisitions. In multivariate regressions, the ratio of target shareholders in acquiring shareholders predicts growth in shareholders for four years following the acquisition. This paper is the first to reveal evidence for a very passive target shareholder base in the years following a stock acquisition. While other papers have shown increases in shareholder base reduce risk, this paper finds no clear evidence of reduced risk related to the change in shareholders.
The next section discusses the existing literature in greater detail. Following that are descriptions of the data, methods, and results and then a conclusion.

Literature
The research on shareholder base largely is motivated by Merton's (1987) model of an incompletely informed asset market. Merton assumes that investors are mean-variance maximizers who are rational but are aware of only a subset of the stocks available in the market, consistent with a fixed cost of learning about each firm. With incomplete information, imperfect risk-sharing arises and there is an additional risk premium for holding the stock. Therefore, the price of the stock is reduced due to the low number of investors. Likewise, equilibrium firm stock value rises as its number of informed investors rises. This establishes a potential justification for a firm to devote resources to increase its visibility to investors and thereby reduce its unfamiliarity discount, even if the steps taken do not impact beliefs about the firm's stock among aware investors.
One possible effort a company may undertake to increase investor awareness and resultant participation is to list stock on an exchange. Kadlec and McConnell (1994) find that when OTC stocks are listed on the NYSE, the stock price jumps and the number of shareholders increases. This stock price increase is correlated with increases in the number of shareholders. Foerster and Karolyi (1999) find similar results for foreign firms who begin listing their stock on US exchanges as ADRs. This listing is also correlated with declines in betas. These declines are weakly related to increases in the shareholder base, implying another potential channel for the increased value with increases in shareholder base.

1417
Acquisitions and the shareholder base There are additional reasons firms may want to increase shareholder base. Bodnaruk and Ostberg (2013) infer from Merton (1987) a wedge between internal and external financing costs and suggest that firms with small shareholder bases might limit dividend payouts and maintain larger cash reserves. Bodnaruk and Ostberg confirm these relations in the data, finding lower dividends on firms with low excess shareholder base relative to a fitted value based on factors likely to covary with the number of shareholders. They also find that firms with a small shareholder base are less likely to engage in potentially base-reducing repurchases. Bodnaruk and Ostberg (2013) also mention an additional theoretical reason firm value might be enhanced by increases in the shareholder base: a larger base and increased liquidity trading might reduce information asymmetry among investors (Holmstrom and Tirole, 1993). A stock's liquidity may be favorably affected by an increase in the base, and greater liquidity reduces the cost of financing, as theoretically demonstrated by Amihud and Mendelson (1986) and empirically supported by Kadlec and McConnell (1994). Kadlec and McConnell (1994) also demonstrate that liquidity is not the sole reason for value effects deriving from the exchange listing.
Shareholder base changes, nevertheless, are associated with changes in liquidity. Lipson and Mortal (2007) find that mergers with larger increases in the number of shareholders ( from pre-announcement to post-completion) result changes in liquidity of the stock. Using a decomposition of the bid-ask spread based on Madhavan et al. (1997), they find that increases in the number of shareholders following an acquisition result in larger declines in the portion of the spread attributable to adverse selection component, while the remainder of the spread (the order-processing component) is independent of the change in shareholder base. Li's (2006) merger study finds that stock bidders experience greater liquidity improvements in the weeks surrounding mergers than do cash bidders, a result she notes is consistent with stock mergers generating greater increases in the shareholder base.
Deviating from Merton's (1987) model of unaware investors, Baker et al. (2007) develop a model showing a downward sloping demand curve for individual stocks based on heterogeneity of beliefs among investors and restrictions on short-selling. With this difference-of-opinion model, stockholders of a target in an acquisition should unload their shares following the news of stock acquisition. These investors do not want to exchange their current stock investment for shares in the acquiring firm that are not optimal holdings based on their beliefs. Given only these elements of the model, the investors would immediately sell their stock; however, Baker et al. additionally posit that some investors will react slowly to changes, due perhaps to lack of attention. When a potential target is largely held by these inertial investors, paying for acquisition with stock is more appealing than going to the market with a seasoned equity offering and facing the downward sloping demand curve. Baker et al. (2007) test this model through an examination of brokerage data on individual investors. They find that only a small amount of acquisition-induced selling occurs prior to merger completion. Lipson and Mortal (2007) provide additional evidence of inertial investors. They find that shareholder base expands by 26 percent in the fiscal year immediately following completion of an acquisition; this is on all acquisitions, regardless of mode of payment.
This paper examines how the number of shareholders of the acquiring firm changes up to four years following the acquisition. This paper extends the study of Baker et al. (2007) by examining the long-term impact on the company's shareholder base rather than short-term actions of individual stockholders. It extends Lipson and Mortal (2007), who assess shareholder base in the short-term and do not distinguish between modes of payment. This paper examines the differential effect on the shareholder base of stock and cash acquisitions. While it is intuitive that placing acquirer shares in the hands of target 1418 MF 45,10/11 shareholders will result in larger effects on the acquirer's shareholder base than would paying in cash, this paper examines this as an empirical question. It also examines whether this larger shareholder base is a long-term change. The present study demonstrates the importance of the relative number of shareholders and controls for other factors that may lead to changes in shareholder base using regression analysis and instrumental variable analysis on the decision for a cash or stock acquisition. Furthermore, it complements Lipson and Mortal's examination of liquidity effects of mergers by addressing merger-induced changes in beta and equity value via the shareholder base channel, and it alerts researchers to pitfalls in the use of Compustat shareholder data.

Data
This paper investigates changes in the number of acquirer's shareholders following an acquisition. The acquisitions in this study are completed between 1993 and 2013 and are obtained from the Mergerstat collection on LexisNexis Academic. Additional restrictions on the sample are both buyer and target must be based in the USA, and the deal type must be "acquisition" or "acquisition tender offer." Both the buyer and target must be publicly traded and the buyer must seek 100 percent ownership in the target. The sample is limited to large acquisitions; the final purchase price must exceed $500m. To reduce confounding effects, the acquirer must not have completed another acquisition of at least half the market value of the relevant deal within five years of the closing (completion) date.
Company-specific data are obtained from Compustat, so an additional requirement is that both parties to an acquisition appear in Compustat data. Acquisition characteristics including deal price, acquirer market value 30-days pre-announcement, buyer and target stock exchanges, and the announcement and closing dates are collected from the Mergerstat data. Announcement returns and beta are gathered using CRSP return data; there must be 245 trading days before acquisition announcement and 245 days after the acquisition completion. Acquisition announcement returns and pre-and post-acquisition betas are calculated using Eventus.
The above restrictions lead to 348 acquisitions whose the completion years are shown in Panel A of Table I The variable of primary interest is the number of shareholders. Corporate reporting of this variable is not always consistent in Compustat or 10-K reports. A large majority of firms filing 10-Ks report the number of registered shareholders, or equivalently shareholders of record, excluding the beneficial shareholders whose stock is held in the name of a broker or bank[1]. Some companies report both figures, and occasionally a firm that reported one figure will switch to the other (or switch from reporting both to reporting only one) in a later year. Beneficial shareholders are in some cases orders of magnitude larger than registered shareholders. Compustat shareholder data do not indicate whether a figure is the registered shareholders, the beneficial shareholders, or the sum of the two, though it typically sums the two when both are reported. To limit inconsistencies, shareholder data that appear inconsistent are further investigated with reference to either 10-K filings or Mergent Online (which often describes the type of shareholder data reported). Data are hand checked from the above sources when the acquirer's number of shareholders 1419 Acquisitions and the shareholder base decreases more than 66 percent or increases more than 300 percent over the years of the sample. Another trigger for hand checking the data is if the target-to-buyer ratio (shratio) is above 5 or below 0.025. The number of shareholders is corrected based on the new information when possible; however, occasionally data are eliminated from the sample when a consistent reporting method across time or between buyer and target is not available. The inconsistency in corporate reporting of number of shareholders has likely been a source of noise in other papers. Some authors have mentioned supplementing Compustat when data were missing, but this is apparently the first paper that corrects the inconsistent Compustat shareholder data. Lastly, firms often report the shareholder figure, whether beneficial or record holders, as of a date sometime after the fiscal year end for which the rest of the 10-K data applies, but the gap is usually limited to one or two months.

Methodology
This paper hypothesizes that acquirers experience larger increases in the shareholder base if the acquisitions are paid for with stock rather than cash and if the ratio between the target's shareholder base to that of the acquirer's is higher. However, a target with a large shareholder base might not make any difference in a cash merger, particularly if the inertial shareholder story of Baker et al. (2007) is the sole reason for increases in the acquirer base. If target investors in a stock acquisition are totally passive and 100 percent do not sell their new acquiring firm stock, the share ratio will equal the change in shareholders (see example below). Therefore, another key hypothesis is that the ratio of shareholder bases (shratio in this paper) interacts with the method of payment in driving changes in the base.
The first hypothesis will initially be examined through a comparison of the changes in acquirer shareholder base over time across all-cash acquisitions, partial-stock acquisitions and all-stock acquisitions. Then a multivariate regression test will simultaneously assess all three hypotheses (the effects of the payment type, the shareholder base ratio and their interaction) while controlling for other factors that may influence changes in shareholder base. The multivariate analysis regresses the change in the number of registered shareholders (chgshr) on several independent variables, including method of payment (allstock), target size (reltargetval) and shareholder ratio: chgshr À1;t ¼ b 0 þb 1 allstockþb 2 shratio þb 3 relt argetval þb 4 allstock_shratio þ b 5 exchsameþb 6 sicequal þb 7 divsame þb 9 revgrowdif f þb 10 advchg À1;0 þb 11 chginvpr À1;t þ b 11 chgmktcap À1;t þb 12 dumrep 1;t þb 13 dumissue 1;t þe; (1) where year −1, the benchmark year for calculating changes in the shareholder base as well as the year used for shratio, is the fiscal year end just prior to the announcement [2], and year 0 is the year end immediately following the closing of the acquisition. Use of the pre-announcement number of shareholders ensures that the analysis captures the entire change in the shareholder base, including changes in shareholders based on the news of the acquisition (variables are defined in the Table AI).
An example from the sample shows the importance of the ratio of shares as compared to the changes in shareholders for the acquiring firm. Westwood One, Inc. announced on June 2, 1999, that it would buy Metro Networks Inc. in an all-stock deal. Just prior to the announcement, Westwood One had 4,220 registered shareholders, while Metro Networks had 1,801. That implies a shratio of 1,801/4,220 ¼ 0.427. Assume either that the existing Westwood One shareholders did not sell their shares or that a new shareholder replaced any selling shareholder. If all Metro Networks shareholders held their new shares in Westwood One shares after the deal closing and through Westwood One's year end on December 31, the shareholder base would rise to 4,220 + 1,801 ¼ 6,021. This would give the value of (6,021/1,801)−1 ¼ 0.427 for chgshr −1,0 , exactly the same as the shratio. That is, the shareholder base would have risen by exactly the ratio of target to acquirer shareholders in this unlikely scenario. Furthermore, if investors are passive and do not make changes in shareholdings, the acquirer in a cash acquisition will not be affected by the number of shareholders in the target firm. Therefore, shratio is not expected to affect the changes in shareholders given the presence of the interaction term allstock_shratio in the regressions.
The regression includes the relative size of the target (reltargetval) to control for size. This variable is similar to a variable Martin (1996) used to explain the choice of merger payment based on risk-sharing considerations, but it appears in regression (1) as an attempt to address the possible conflation of firm size and shareholder base.
The paper also includes indicator variables on the similarities between the acquirer and the target firms. Exchsame, sicequal, divsame and revgrowdiff are included with the rationale that shareholders of the target may be more willing to hold shares in the acquirer if those shares have similar features and therefore perhaps have similar return distributions or perceived pricing errors. Jeon et al. (2010) claim that because target shareholders receiving stock in an acquirer are more likely to liquidate those holdings if the dividend payment status differs, buyers will be more likely to extend stock offers if their dividend policy resembles that of the target [3].

1421
Acquisitions and the shareholder base The tests include other variables shown to affect shareholder base in Grullon et al. (2004). Advchg −1,0 is motivated by Grullon et al. (2004), who find that the number of shareholders, the number of institutional investors and stock liquidity are positively correlated with advertising expense. Other control variables are inverse stock price and market capitalization. Grullon et al. (2004) hypothesize transaction costs and minimum block sizes may lead to stock ownership clientele and market capitalization affects investor visibility.
The last two variables in (1) appear only in the regressions for longer term changes. Repurchases (dumrepur 1,t ) and stock issuance (dumissue 1,t ) are controls for an acquirer's actions that could affect the shareholder base independent of the acquisition (though those actions might themselves be influenced by the fact that the acquisition took place).
Initial tests of Equation (1) employ ordinary least squares. However, heteroscedasticity is likely a concern. In particular, while the linear model could apply as given even for acquisitions with very large values of shratio (so that Winsorizing or eliminating large observations would be inappropriate), such acquisitions may be associated with larger error variance. Weighted least squares tests will be performed to avoid results that are driven by outliers.
Another confounding factor is that an acquirer's choice of payment in an acquisition may partly be based on the expectation of changes in the shareholder base, or factors that drive those changes may also influence the mode of payment decision. Therefore, the study will also consider a specification with instrumental variables for the regressors allstock and allstock_shratio. The former is a binary variable, while the latter is bounded at zero, so with the hope of gaining some efficiency of estimation over a purely linear two-stage approach, the study first will use instruments to fit a probit regression to allstock and Poisson regression to the interaction variable. The fitted values and the same instruments will be used in the first stage in a two-stage least squares estimation. This three-stage approach follows Adams et al. (2009) and avoids the inconsistency of a non-linear first stage in a two-stage process. The instruments used are similar to variables that in previous research have been found to be statistically significant determinants of the mode of payment. The Probit and Poisson regressions are as follows: where X i is the vector of the variables included in regression (1) and Z i consists of instrumental variables defined in the Table AI. The instrumental variables in Z i are those found in the literature to affect the method of acquisition payment. Measures of growth opportunities for the acquirer are equity book to market (Bktomkt) and revenue growth (revgrow bu ). These growth opportunities encourage the flexibility afforded by equity rather than debt financing (Martin, 1996;Jung et al., 1996). Buylev and targetlev are acquirer and target leverage prior to announcement. Higher existing leverage has been associated with lower unused debt capacity ( Jeon et al., 2010). The final independent instrument is the dummy variable post_2001; fewer stock acquisitions are expected in 2002 or later due to accounting rule changes (De Bodt et al., 2018). The instrumental variables are not expected to influence shareholder base except through the merger payment channel. All instruments are interacted with shratio in an additional five variables because the dependent variable in the Poisson regression, allstock_shratio, is itself an interaction variable.

Shareholder changes through time
The changes in the number of shareholders for acquirers differ dramatically between stock and cash acquisitions. Figure 1 documents the median change in number of shareholders for acquiring firms based on whether the deal is all stock, some stock or all cash. This change is the percentage change from the year prior to the announcement of the acquisition (year −1). Prior to the acquisition announcement, all-stock acquisitions show slight increases in median number of shareholders while cash and partial-stock acquirers lost shareholders prior to the acquisition. The difference in the changes in shareholders is amplified in the years following the acquisition. Median stock and partial-stock acquisitions gain shareholders the year of and the year following the acquisition [4]. The difference in sample medians based on the method of payment suggests that the shareholder base increases immediately following stock acquisitions is specifically associated with stock acquisitions rather than reflecting a general trend of increasing investor popularity or of changing the measurement of registered shareholders. While the number of shareholders later declines for these stock acquirers, four years following the acquisition the median cumulative increase in shareholders is 12 percent for partial-stock acquisitions and 33 percent for purchases using exclusively stock. The changes in shareholders appear dramatically different based on payment type; tests of differences in means and medians along with regressions provide further analysis.
The acquirer's number of shareholders and changes in shareholders by the form of payment are shown in detail in Table II. The 79 cash acquisitions have a median of 7,730 shareholders the year prior to the announcement (−1), whereas the median is 4,760 for the 157 stock acquisitions and 6,330 for partial-stock, partial cash acquisitions. After the acquisition completion (0), the median shareholder base goes up for the full sample and for acquirers using at least some stock. As seen in Figure 1, a greater number of shareholders are sustained for the stock and combined cash and stock offerings. The mean number of shareholders is much larger than median, showing skewness in the distribution.  shareholders, but more than half of acquirers have decreased numbers of shareholders as seen by the negative medians in all years. The mean and median change in shareholders for cash acquisitions is significantly lower than those with any stock payment in all years following the acquisition. While the results in year zero are expected, the four-year results support long-term passive target investors in stock acquisitions.

Multivariate results
To control for both firm and acquisition characteristics, multivariate regression is necessary. Regression results based on Equation (1) appear in Table III to examine whether changes in shareholders found above are attributable to the stock acquisition or other factors. In the OLS regression (A), somestock and somest_shratio are significantly positive, indicating a stock or partial-stock acquisition increases the shareholder base and a greater ratio of target shareholders to acquirer shareholders in the stock acquisition also increases shareholder base. When the indicator variable is allstock (column B), partial-stock acquisitions are identified as zero and allstock is negatively related to the change in shareholders. This indicates an all-stock acquisition does not lead to increases in shareholders if the number of target shareholders is very small as compared to the acquirer shareholders. The positive significant allstock_shratio indicates that a large ratio of target to acquirer shareholders increases the shareholder base in all-stock acquisitions. This is evidence for passive target shareholders; large numbers of shareholders are gained in the acquiring firm due to the target shareholders not selling their newly acquired stock. The insignificant coefficient on shratio in column (A) indicates that in all-cash acquisitions, targets with more relatively more shareholders are not associated with larger increases in the acquirer's shareholder base; the high significance of shratio in (B) is associated with partial-stock acquisitions, for which the interaction term equals zero in that specification. The R 2 value for the allstock specification (B) is much higher than that for the somestock specification (A), so all subsequent regressions will use allstock and allstock_shratio.
With heteroscedasticity likely, robust standard errors are reported. 171.63% 13.38% 20.14%*** −18.66%* 170.52% 32.60% 148.43% 11.50% Notes: Panel A reports the mean and median number of shareholders in thousands prior to and following the acquisition by type of financing. Panel B reports mean and median changes in shareholders from the year prior to the acquisition announcement by type of financing. *,***Indicate Cash acquisitions' levels of change in shareholders is significantly (at the 1 percent level) less than offerings with some stock in z-tests and signed rank tests, respectively  Notes: This table presents determinants of the change in the number of acquiring firm shareholders from before announcement to after completion. It also includes firststage regressions for the method of payment. Robust standard errors are in parentheses. *,**,***Significant at 10, 5 and 1 percent levels, respectively Heteroskedasticity also motivates the use of weighted least squares (WLS), which is more efficient than OLS given weights that align with the inverse error variance. Weighted least squares results are shown in column (C). In an unreported OLS regression with all the controls, the squared fitted errors were found to be to be strongly related to shratio (more than to the square of shratio). The weights in column (C) and other WLS regressions were based on that estimated variance dependence. For all-stock acquisitions, the fitted change in the shareholder base can be calculated using the estimated coefficients on allstock, shratio, and allstock_shratio together with the constant term, −0.0212 + (0.216+0.411) × shratio + 0.215, which simplifies to 0.194 + 0.627 × shratio. An all-stock acquirer's shareholder base is projected to increase by 19 percent plus 63 percent of the target's shareholder base on average. This is an economically significant change in the number of shareholders. The coefficients have implications for the Westwood One -Metro Networks acquisition. The shratio of 0.427 in the Westwood One -Metro Networks acquisition was similar to the median in the data. The coefficients from the previous paragraph indicate that the fitted proportional change in Westwood One's shareholdings is 0.194 + 0.627 × 0.427 ¼ 46 percent. In fact, Westwood One reported 5,368 shareholders after the acquisition for an increase of 27 percent, somewhat lower than the fitted value but nevertheless a major impact.
Adding control variables in column (D) of Table III result in little change on the coefficients of interest. The interactive term allstock_shratio remains positively related to the change in number of shareholders. The only control variable significant at the 5 percent level is divsame (both target and acquirer either pay or do not pay dividends); however, the negative coefficient indicates the number of shareholders decreases when dividend policy is the same, which is opposite expectations. Change in market capitalization and change in inverse share price have the expected positive coefficients at the 10 percent significance level. Advertising generates no effect in contrast to results in Jeon et al. (2010) and Grullon et al. (2004). This result is not dependent on the functional form; whether using change or scaling by assets, the results remain insignificant.
To control for endogeneity, instrumental variables are estimated for allstock (Probit) and allstock_shratio (Poisson); these first stage results are displayed in Columns (E) and (F). The explanatory variables discussed previously are included, as are their interactions with shratio. The explanatory variable chgmktcap −1,0 has been removed from these regressions because a change in the acquirer's equity value also could be viewed as endogenous. None of the instruments are statistically significant determinants of allstock itself, though joint significance is high. An unreported version of the Probit finds significance on a tender offer dummy variable, but presumably the tender offer decision is also endogenous. The coefficient on post-2001 is negative and significant only for allstock_shratio. Three of the interacted instruments also are significant determinants of allstock_shratio.
The main regression (using the stages derived in E and F) is in column (G) of Table III. The multistage process yields similar results to those in (D) for the key variables allstock, shratio, and allstock_shratio. Allstock_shratio is positively related to the change in acquirer shareholders in the year of acquisition completion. This relationship is not due to factors influencing the choice of a stock acquisition. Coefficients of the control variables are also comparable, though the relative market value of the target is a significantly positive driver of shareholder base changes in the three-stage model. The Hansen J. statistic for the test of overidentifying restrictions has a p-value of 0.58, so the instruments appear to be legitimate. At the same time, the Durbin-Wu-Hausman test of endogeneity of allstock and allstock_shratio fails to reject exogeneity with a p-value of 0.31, suggesting there is a chance the one-stage approach was sufficient. Regardless, the results are very similar across the two specifications. The results in Table III are Table III does not establish the persistence of these effects over time. That matter  is addressed by Table IV, which shows cumulative changes in number of shareholders through years 1, 2 and 4 are driven by the same factors that drive the initial change. In an all-stock acquisition, the target firm's shareholder base increases the acquirer's shareholder base at year 2 more than for cash or somestock acquisitions as seen by the 0.447 coefficient of allstock_shratio. The impact of allstock_shratio decreases by year 4, but it is still positive and significant at the 10 percent level. The share ratio is positive and significant in all regressions, providing some evidence that regardless of payment mode, the number of target shareholders positively affects the change in acquirer shareholders.
The dummy variables for repurchases and stock issuance, dumrepur and dumissue, following the acquisition do not significantly affect the number of shareholders. Untabulated regressions show the channel for the Table IV regressors' collective impact on future changes is through the initial jump: the year 0 change in number of shareholders explains more than 90 percent of the variation in cumulative changes through year 3. This is unsurprising but confirms that the shareholder base for acquirers does not dramatically fluctuate. Overall, the results in Table IV show evidence that target investors are passive in the long-term; most do not appear to sell the stock of the acquiring firm although they had no intention of buying this stock. This passivity supports an argument of Baker et al. (2007): consider a company facing downward sloping demand curves for its shares and intending to conduct an acquisition. If it further knows that it needs at some point to issue stock, then issuing that stock to passive target shareholders will be less costly than would doing a cash acquisition followed by a seasoned stock offering.

Impacts on value and risk
In light of the empirical literature on the shareholder base, the enduring impact of the shareholder base ratio on stock acquirers' number of shareholders suggests that the interaction term allstock_shratio may affect acquiring firm value and risk. This research offers an initial investigation of that matter. Abnormal returns for the acquirer are calculated for various intervals, using an estimation period from trading day −301 (relative to the announcement) to −46 were available and requiring at least 200 days. Pre-acquisition betas stem from the same

1427
Acquisitions and the shareholder base estimation period and employ the CRSP equal-weighted index as the market portfolio. The symmetric period after acquisition completion (+46, +301) supplies post-acquisition betas. The change in beta is calculated as the post-acquisition beta minus the pre-acquisition beta. For the purpose of regressions, these changes are winsorized at the 2 percent and 98 percent levels of −1.3 and +1.49. Column (A) of Table V presents a regression of the change in acquirer beta on the same explanatory variables used in the chgshr regressions. The coefficient on shratio by itself is positive, but the larger negative coefficient on allstock_shratio indicates that for all-stock acquisitions in the present sample, higher values of the shareholder base ratio lead to greater reductions (or smaller increases) in market risk. The only other variable with a statistically significant impact on the change in beta is exchsame, which is associated with reductions in beta. The rationale behind a link between allstock_shratio and beta is, of course, that the former is a predictor of increases in the shareholder base and that shareholder base increases have been shown to be negatively associated with beta changes around ADR listings (Foerster and Karolyi, 1999.) Then the direct relationship between changes in the base following acquisitions and changes in betas ought to be more clearly negative than the indirect relation, portrayed in Column (A), between mere predictors of shareholder base changes and beta changes. This logic, though, is called into question by Column (B), which shows the coefficient on chgshr −1,0 in a single-variable regression is small in magnitude (relative to the sum of the Column (A) coefficients on shratio and allstock_shratio) and statistically insignificant. Augmenting chgshr −1,0 with the controls from column (A) does not appreciably change the coefficient or significance of chgshr −1,0 , as shown in Column (C). Finally, Column (D) indicates that shareholder base changes over the longer period from year −1 to year +1 likewise have a limited connection to changes in beta, leaving a puzzle regarding the nature of an effect by allstock_shratio.
If shareholder base increases are associated with firm value increases for any reason, whether because of reduced betas, reduced risk premiums, increased liquidity, or another channel, then the market reaction to an initial acquisition announcement might be influenced by observable predictors of the shareholder base change. Regression results for cumulative abnormal returns (CARs) at the time of announcement appear in Table VI, with figures presented for the (−1, 0) 2-day period relative to announcement, for the (+1, +30) 30-day period, and for the combined period, (−1, +30). Payment in stock is predictably

1428
MF 45,10/11 associated with negative returns for all periods. (See Liu and Wu, 2014, for discussion of theory and evidence on announcement returns for stock acquisitions.) Somewhat surprisingly, shratio has a highly significant negative coefficient for the overall announcement period return (Column C). That result indicates that cash or partial-stock acquisitions trigger more negative reactions when the target has relatively more shareholders, even controlling for relative target relative value and even though shratio is positively associated with shareholder base changes in Table III. The coefficient on allstock_shratio, though positive and larger than the coefficient on shratio (suggesting increases to the shareholder base would be value-enhancing), is insignificant.

Conclusion
The paper shows that the number of shareholders rises following major acquisitions paid for at least partly with stock. The larger shareholder base persists up to four years following the acquisition. For all-stock acquisitions, every three additional target shareholders translate into roughly two additional acquirer shareholders the year immediately following the acquisition. Four years after the acquisition, these three target shareholders still imply one additional acquirer shareholder. For acquisitions paid with at least some stock, median initial (year 0) increases in the shareholder base are over 30 percent and are approximately the same four years later. These results appear consistent with Merton's (1987) market in which investors hold only stocks about which they have information.
The implication for managers acting on behalf of shareholders is that one way to increase the number of shareholders is to conduct a stock acquisition of another company with many shareholders of its own. In a company where managers believe the stock is undervalued due to investor inattention, a stock acquisition can improve attention and increase shareholder base. This is not to imply that companies should make inappropriate acquisitions. Instead, the evidence shows if a company is making an acquisition, using stock can help increase investor attention. While a cash acquisition may increase news about the company, on average, a cash acquisition will not cause increases in shareholder base.
This paper examines whether this larger base affects the risk of the firm. This paper's tests on changes in beta provide limited evidence that among stock acquisitions, purchases of targets with large shareholder bases are more likely to reduce beta. If increased shareholder base is beneficial, the shareholder ratio should be positively related to

1429
Acquisitions and the shareholder base announcement returns; this paper does not find this. This could be because the net effects are small so the market fails to recognize them, or because the data set is insufficiently large to reliably assess the returns. Future research ideas are to study other effects of increased shareholders due to a stock acquisition. Additional shareholders lead to less monitoring and greater agency costs; this increase may be value reducing for shareholders. Testing the effect of the base changes on long-term returns may then be appropriate for future research. It is possible that market interest in an acquiring company based on publicity would increase the number of shareholders with any payment types, but cash acquisitions do not exhibit such an effect in this paper's data. A possible extension of this paper, however, would be to connect shareholder base changes to media attention or analyst reactions surrounding the acquisition. In addition, an analysis of the differential consequence of changes in registered versus beneficial shareholders may affect the implications of a stock acquisition. Given the lack of data of American firms reporting beneficial shareholders, this analysis would probably need to use another country's data set.
The message of this paper is simple. Acquisitions using stock may have effects unrelated to the combination of the two operations. These stock acquisitions on average increase the number of shareholders, potentially bringing costs and benefits to the acquirer.

Notes
1. Sweden and Hong Kong are among the countries where companies must report the total number of shareholders. Bodnaruk and Ostberg (2009) used Swedish data in a study finding stock returns were negatively correlated with the size of the shareholder base.

Corresponding author
Claire Lending can be contacted at: claire.lending@wwu.edu (Revenue −1 )/(revenue −3 ) -1 revgrowdiff abs (revgrow of targetrevgrow of acquirer) advchg −1,t (advertising expense of acquirer 0 )/(advertising expense of acquirer −1 ) -1 chginvpr −1,t (1/acquirer's market price per share 0 ) -(1/acquirer's market price per share −1 ) dumrepur 1,t Indicator variable equals 1 if the buyer reports positive amounts of stock repurchases between years 1 and t, and is zero otherwise dumissue 1,t Indicator variable equals 1 if the buyer reports positive amounts of stock repurchases between years 1 and t, and is zero otherwise bktomkt Acquirer's buyer's book-to market equity ratio revgrow buy Acquirer's revenue growth from year −3 to year −1 buylev Acquirer total debt (short-term debt + long-term debt) divided by total assets targetlev Target total debt (short-term debt + long-term debt) divided by total assets post_2001 Equals 1 if the announcement year is 2002 or later 1. Introduction Shareholder theory posits that management has a mandate to maximize shareholder wealth through decisions that add value to investments and stimulate growth. For any large company, growth through mergers and acquisitions (M&As henceforth) is often a key part of corporate growth strategy. Growth largely drives value creation and M&As can offer a course to growth when esoteric opportunities are restricted through projected financial, strategic and operational synergies achieved at a fair price. Numerous studies though have shown that M&As more often than not destroy value rather than create it. More than 50 percent of all M&As lead to a decline in relative total shareholder return after one year. Hence, effective target identification must be built on the foundation of a credible strategy that identifies the most promising market segments for growth, assesses whether organic or acquisitive growth is the best way forward and defines the commercial and financial hurdles for potential deals. It is thus crucial for companies' upper management to utilize credible and proven methodologies and models to ensure that target identification is based on sound background research. For example, as early as 20 years ago, researchers (Kaastra and Boyd, 1996;Rojas, 1996) argued for the systematic application of neural networks (NNs) as a method to deal with the problem of non-linearity in financial transactions. Rojas (1996) argues that where there is abundance of data but less theoretical understanding (e.g. behavioral patterns that are not easily identified through established linear methods or continuously keep changing) NNs "can discover statistical regularities and keep adjusting parameters even in a changing environment" (p. 247).
When valuing a firm three major sources of valuation inputs are considered: current financial statements; firm's past history; and peer group comparisons. While for most firms such crucial information is ready-made, for technology firms such vital sources might be absent. Their financial statements do not include much information about growth prospects either. Most technology firms have limited or no past history. They also possess unique businesses and/or products therefore leading to no directly visible peers or competitors (Damodaran, 2001): "As more and more technology firms get listed on financial markets, often at very early stages in their life cycles, traditional valuation methods and metrics often seem ill suited to them" (p. 19). Daniel et al. (1998) demonstrated that investors tend to be overconfident when examining unclear information and that mispricing is stronger for stocks whose value is closely tied to their growth. The very fact that technology firms more often than not exhibit unconventional growth patterns makes them difficult to evaluate and can lead to their stocks being massively mis-valued (most of the time overvalued) and therefore increasing M&A activity (Rhodes- Kropf and Viswanathan, 2004;Jovanovic and Rousseau, 2001). While there are idiosyncratic motives for undertaking M&A-led growth strategies, there are also substantial economy-wide factors which cause waves of global M&A activity such as responses to globalization forces and increases in competition, de-regulation and the associated economic reforms and liberalization, block/regional economic integration (i.e. the EU). As such, target firm identification has become a great research interest area both to the business world and academia alike. The three latest M&A waves (namely, M&As waves 5, 6 and 7) make the case in point:  Acquisitions and Alliances, 2016). It is also owed to the system-wide steps taken by central banks after the 2008 credit crisis, such as keeping near-zero interest rate and the quantitative easing procedures which supplied equity and bonds markets with enough liquidity.
With an increasing amount of M&As in the technology sector, it is crucial to identify targets before announcement date, as this can be significantly beneficial for investors, target and acquiring firms. From that comes the motivation of a reliable takeover predication model. Standard models have so far been relatively indeterminate in the past, and as a result have not had highly reliable estimations regarding the scale of an outcome or a conclusion on the directional relationships of the variables (Betton et al., 2008;Routledge et al., 2013;Eckbo, 2014). Hence, we switch our approach to an altered empirical exploration model where this is also tested. We elaborate further on our motivation below where by implication we discuss our reasons for utilizing NNs as opposed to the traditional regression techniques.
Having introduced our study topic, we discuss the relevant empirical evidence on the characteristics of M&A deals in the USA, valuation challenges and the technology sector in Section 2 that follows. In Section 3 we discuss methodological issues where determinant variables and NNs are compared with the traditional statistical techniques of discriminant analysis (DA) and logistic regression (LR) with regard to the identification of potential takeover targets. Section 4 discusses our methodology and Section 5 presents the results of our analysis. The conclusions of the study are presented in Section 6.

Motivation summary
Financial time series have some characteristics that make them hard to reliably forecast, especially when a traditional statistical method is employed. Such characteristics are as follows (Motiwalla and Wahab, 2000;Thawornwong and Enke, 2004;Versace et al., 2004): (1) non-stationarity of data, where due to different business and economic cycles, the statistical properties of financial data change randomly over time, which also introduces; (2) non-linearity of data, where the relationship between the financial and economical independent variables and the desired dependent variable may not be linear. Intensified by; and (3) noisiness through daily variations in financial time series.
On the other hand, NNs are more flexible and adaptable computing methods that provide the ability to potentially capture the patterns among variables more effectively. Hence, the use of NNs to forecast financial time series as an alternative is justified by some of the in-build qualities they posses. Such characteristics make them reasonably well suited for use in the financial forecasting domain (Hussain et al., 2007;Lin et al., 2006;Lam, 2004;Eakins and Stansell, 2003): (1) Their non-linearity. NNs can capture nonlinear relations between element (input or independent variables) and response (output or dependent variables).
(2) Their data-driven nature. No prior explicit relational assumptions on the model are made or modeled between inputs and outputs.
(3) Their generalizability. Once trained, NNs can produce relatable results even when the data structure has changed or when they are faced new input patterns.
(4) Their assumption neutrality. Dissimilar to traditional statistical techniques, NNs do not employ pre-constructed assumptions on the input data distribution.
Yet, as with any forecasting tool, the robustness of an NN application outcome can equally be questioned. This is addressed in our results and discussion sections at the end of this exposition.

1435
Confining value from neural networks

Literature review: mergers and acquisitions in the US technology sector
The USA is well known as the most preferred international investment destination measured by foreign direct investment (FDI) flows. In 1989, the FDI position in the USA (FDIUS) exceeded $400bn (Harris and Ravenscraft, 1991), while in 2014 for example it totaled $2.4 trillion with an average annual growth of 8.9 percent (Organization for International Investment, 2016; Bureau of Economic Analysis, 2017). It is ranked as the world's top market for five years consecutively (Laudicina and Peterson, 2017). Figure 1 illustrates M&As as the most exercised type of investment in the USA in volume vs expansions and new establishments. Rossi and Volpin (2003) suggested the role of the legal system as a factor affecting cross-borders M&A volume. The rationale being that countries with mature legal systems are better able to cope with economic changes, absorb shocks and provide shareholder protection thus improving the liquidity of the market as a whole (Eden and Dobson, 2005;Beck et al., 2003). Harris and Ravenscraft (1991) claimed that FDIUS increases when the dollar is weaker compared to the investor's home currency. Servaes and Zenner (1994) [1999][2000] was at the peak with 371 and 261 tech initial public offerings (IPOs), respectively, and in 1996 with 274 IPOs (Ritter, 2017). The growth of IPOs during the 1990s was fueled by venture capitalists excessively funding start-ups as funding rose from $3bn in 1990 to $60bn in 1999 (Lowenstein, 2004); furthermore, 57 percent of these tech firms going public were less than five years old and in some cases even less than two years old (Westenberg, 2009); institutional investors bought stocks with thin fundamentals as they purchased more than 63.  (Westenberg, 2000). Its growth has been substantial since the 1990s measured by the number of technology IPOs as indicated by Figure 3.

Valuation of technology firms
In an M&A transaction it is as crucial for the acquirer to determine a fair value of synergies for the target, as it is for the target to come to a value for itself. It is also important for the shareholders of both firms to justify the acquisition price (Petitt and Ferris, 2013). The dynamics of the technology sector are characterized by rapidly evolving firms which  Confining value from neural networks operate under high levels of uncertainty and risk (Lev and Zarowin, 1999). This, combined with the lack of positive cash flows (Aydin, 2015), makes their valuation very challenging as also demonstrated by Bakshi and Chenb (2005), where they demonstrate the potential for significant mispricing and departures from fair values. The complexity of valuing technology firms can be attributed to reasons such as: (1) Tech firms are often young ones, very dependent on innovation and require huge amounts of upfront investments in intangible assets. Chandra et al. (2011) state: "[…] this arises from the uncertain nature of long-run industry prospects as well as competition among firms for market share through first-mover advantages, creation of entry barriers and establishment of property rights in new technology" (p. 8) which leads to the second point.
(2) The value of many firms in the technology sector usually comes from intangible assets. These assets however do not always appear on the firm's financial statements due to the lack of accounting standards to accommodate such intangibles, such as innovation, customer satisfaction and human capital, resulting in complexities when it comes to perform an equity valuation (Chan et al., 2001).
(3) Tech firm value is directly dependent on growth; consequently, most of the value will originate from future customers or products not from current operations. That makes it challenging for investors to measure firm's beta (risk).
(4) The value of a technology is only known after it is commercialized to the market (Park and Park, 2004).
There are various methods to value firms; they are however categorized into three mainstream methods (Hodges, 2007). Table I shows the pros and cons as listed by Anadol et al. (2014). DCFs as a method boast a huge limitation in that firms in the technology sector more often than not either do not pay dividends (even in cases they pay dividends these are often very volatile) or instead choose stock buybacks therefore using this method can undervalue the firm (Palepu, 2003). Any valuation method can be misleading as it does not for example incorporate intangibles, yet typically, acquisition premiums achieve more than 50 percent above market value (DeAngelo, 1990). Also, multiple bidding offers can be significantly   (Bradley, 1980). Even hedge funds in the USA hire technology consultants to provide expert insights about tech firms as they are hard to value from a financial standalone perspective (Benou and Madura, 2005).

Takeover prediction techniques
Various researchers have studied the possibility of predicting acquisition targets through statistical aggregation and the associated distress signals (i.e. bankruptcy) using publicly available information of firms and then applying different statistical models on them. It is important to mention that the methodologies used to predict bankruptcy and predict takeover targets are very similar (DA and LR); therefore, we shall consider both broad approaches below.
3.1 Traditional analytical techniques 3.1.1 Regression models. Ohlson (1980) utilized LR analysis in order to examine the relationship between binary or ordinal response probability and explanatory variables. He was the first to point out weaknesses in Altman's (1968) model and highlighted the importance of using data from firms' financial statements directly as they will indicate whether the firm filed for bankruptcy before or after releasing them which will help the researcher avoid the "back-casting" issue (i.e. applying the model to firm's data after being bankrupt). This model produced an accuracy prediction rate of 96 percent with a cut-off point of 0.5. The binary LR, a nonlinear model, is one of the predictions' techniques where the dependent variable is a binary or dummy variable. Very few assumptions are required in such model in comparison to other similar dependence techniques such as DA. Harris et al. (1982) used a probit model where, for example, the dependent variable can take only two values (acquired or not-acquired), in order to produce a probability of a firm to be acquired or not as well as what are the characteristics that affected this probability. Dietrich and Sorensen (1984) used LR model to predict acquisition likelihood. Palepu (1986) used a binomial logit probability model with nine independent variables; his model suggested a good fit of success in predicting a high number of targets. It, however, predicted a high number of non-targets as targets; therefore, it was not sufficient to use this model to gain abnormal returns. Barnes (1990) used multiple discriminant models with five chosen industry-related ratios to increase the predictability of his model. While the previous studies, as above, have shown prediction power between 60 and 90 percent, Palepu (1986) argued, however, that these findings are overstated and suffer a biased estimate due to two main flaws in such methodologies: state-based sampling for model estimation and prediction testing; and using pre-determined, arbitrary, optimal cut-off probability. Furthermore,

1439
Confining value from neural networks Powell (1997) argued that the characteristics of hostile and friendly takeovers differ; therefore, using binomial models (treating hostile and friendly takeovers in the same group) will cause misleading results. Cudd and Duggal (2000) in their study used Palepu's (1986) factors but they added an industry dispersion factor to account for different industries which improved the accuracy of the said model. In addition, they also found the dummy variable "industry disturbance" to be significant; therefore, indicating that a takeover in the same industry in the past 12 months will increase the probability of takeover for the remaining firms in that industry.
3.1.2 Discriminant analysis (DA). It allows the researcher to pair two or more firms (or groups of firms) and compare their differences with respect to several variables simultaneously. Depending on how variables behave (i.e. jointly or independently of one another), DA models can be further applied into two sub-categories namely univariate or multivariate models; multivariate discriminant analysis (MDA) models consider simultaneously an entire portfolio of characteristics common to the firms and their interaction, whereas univariate models are limited to only one characteristic at a time. As a technique, DA does very well provide that the variables in every group follow a multivariate normal distribution and the covariance matrices for every group are equal. As early as 1971, Simkowitz and Monroe suggested that target firms tend to be usually smaller, with lower P/Es and dividend payout ratio and lower equity growth. Most importantly, they further observed that non-financial characteristics appeared to be as important as financial. Their MDA in-sample results predict 83 percent of the targets and 72 percent of the non-targets, while the holdout results are slightly worse predicting 64 percent of the targets and 61 percent of the non-target.

Machine learning (ML) techniques
A differentiated methodological approach used by researchers is the use of NNs, ML and data mining to predict bankruptcy or takeover targets. Sharda and Odom (1990) compared the use of both NNs and MDA models in bankruptcy predictions. In their study, the researchers utilized the same ratios used by Altman (1968) and after executing both models their findings suggest that NNs seem to outperform MDA based on different holdout samples with an accuracy level ranging from 77.78 to 81.48 percent. In another study, Tsai and Wu (2008) studied the effect of including multiple NN classifiers in bankruptcy prediction and credit scoring where it was found that single NN classifiers outperformed multiple NN classifiers in both credit scoring and bankruptcy prediction. Hongjiu et al. (2007) used self-organized mapping with Hopfield NN to cluster data and their model showed accuracy predictions of 80.69 percent for targets and 63.11 percent for non-targets. Their paper also suggests the importance of including non-financial factors to improve the predictability power. Iturriaga and Sanz (2015) used multilayer perceptron (MLP) to predict bankruptcy of US banks with a 96 percent success rate.
The evidence regarding method and model fit is far from conclusive though. Coats and Fant (1993), for example, confirmed that NN outperformed MDA in their sample 80 percent of the time. Numerous other studies have supported the use of NNs in outperforming LR in predicting bankruptcy (see e.g. Tam and Kiang, 1992;Jo and Han, 1996;Maher and Sen, 1997;Fan and Palaniswami, 2000;Tseng and Hu, 2010). Branch et al. (2008) utilized both NN and LR to predict whether a takeover attempt will succeed or not with the authors concluding that "[…] neural network model outperforms logistic regression in predicting failed takeover attempts and performs as well as logistic regression in predicting successful takeover attempts" (p. 1186). Salchenberger et al. (1992) compared NN with LR to test healthy and failed thrift institutions and concluded that NN achieved higher accuracy. In all of the above studies a common results' attribution emerges: NNs seem to possess a higher 1440 MF 45,10/11 flexibility and ability to address nonlinearities. This echoes Zhang's et al. (1999) statement that NNs can potentially be robust and can provide more reliable estimations when applied on different samples only once the optimal architecture is found.
On the other hand, Altman et al. (1994) reported that both MDA and NN performed almost the same when trying to predict Italian firms suggesting that contextual and structural considerations as well as firm-characteristics' variables are also important. Equally, Olson et al. (2012) used LR, NNs, support vector machines and decision trees to predict bankruptcy. They demonstrated that different data with different models present different results. There is trade-off between model accuracy and transparency and transportability. In a sense, in order to increase model transportability (i.e. applying it to new data sets and observations) the accuracy level will decrease. Table II, by Barnes (1998), summarizes other research showing the prediction rates of various methods for North America and the UK.

Takeover determinant variables
The main takeover relevant metrics/ratios that have been introduced by the financial literature to identify a takeover target are discussed below: • Inefficient management.
This hypothesis states that managers who fail to maximize their shareholders' wealth and firm's value shall be replaced in accordance with the market for corporate control theory. Therefore, incompetent management increases the probability of their firms to be taken over . Investors will seek to replace the management by purchasing a controlling stake in the firm due to the share prices being below their true value, and target managers will typically get replaced if the bid succeeds (Agrawal and Walkling, 1994). This hypothesis can be measured by EBITDA margin ROE, ROCE, ROA and/or asset turnover: • Undervalued firms.
This hypothesis suggests that firms with low market value compared to book value are targets since they represent a "cheap buy" (Powell, 1997;Palepu, 1986). It utilizes market to book and price to earnings ratios where a bidder will bid for an overvalued firm if it was still less overvalued than the bidder (Dong et al., 2006 Bartley and Boardman (1996) USA MDA 65 -Notes: DA, discriminant analysis; GLS, generalized least squares; MDA, multiple discriminant analysis; -, not reported Source: Barnes (1998)  Firm size plays a significant role in takeover probability, the bigger the size the lower the probability of it being taken over (Palepu, 1986), which explains why usually bigger firms acquire smaller ones (Levine and Aaronovitch, 1981). It has been shown that size is a significant factor (Powell, 1997) as measured by market capitalization and total assets: • Leverage, liquidity and growth. Powell and Yawson (2007) debated that many takeovers occur as a way to rescue the target firm from a certain bankruptcy due to high debt and poor performance. Therefore, firms with low growth and high leverage are more likely to be classified as targets and measured by debt to equity, current ratio and growth in revenues. While low liquidity does not singlehandedly affect the takeover likelihood, when coupled with growth and leverage it can have a significant effect (Palepu, 1986;Cremers et al., 2008) (Table III).

Sample, methodology and data
Our study required three generic groups of data. M&A transactions records, number of public firms in the technology sector from the year 2000 to 2016 and the relevant financial ratios for the same period. All sample data were gathered from Bloomberg. We defined a technology firm as a type of business entity that focuses primarily on the manufacturing and development of technology. This also includes the dissemination of information via high-tech companies. It also includes information technology companies as subsets of technology companies as provided by the NAICS coding system where we placed several restrictions and criteria for selecting our sample. We observed that the number of public technology firms in the USA has been declining over the last 17 years as shown in Figure 4. It is also important to mention that the decrease in public firms is not only affecting the technology sector, as it is affecting the whole US stock market. Since 1996 the number of public firms in the USA has decreased by 50 percent, as a result of: firms being delisted, acquired or bankrupt; and less IPO activities, where firms remain private due to available capital provided by venture capital and private equity firms (Mauboussin et al., 2017).
We pose certain sample restriction criteria for the purposes of our study. First, our study period covers the last 17 years where M&A transactions announced between the year 2000 and 2016 are included; second, we eliminate private firms where the target is a publicly traded company and having its domicile in the USA; third, we screen only technology firm  Figure 5 illustrates the acquirers' industries by number of deals. More than 80 percent of the M&As were completed. A total of 93 percent were classified as friendly takeovers with 3 percent representing hostile takeovers. The rest are classified as unsolicited/unsolicited-to-friendly. Technology firms were 53 percent of the acquirers' transactions. Financial firms came in second at 19 percent. Figure 6 shows the number and value of deals in the technology sector in the USA. The total dollar value of these transactions for our period of study reached $1.025 trillion. Most acquirers in our sample came from the USA with 88 percent and the remaining came from Europe with 10 percent.

Data sets
Our study sample consists of two data sets, targets and non-targets. The target-group data set includes firms which got acquired or received a bid to be acquired within our study period. The non-target-group data set includes firms which did not get acquired or received a bid to be acquired during the same period. The number of firms in our target data set reached 846. Due to data pre-processing and omitted values this number was brought down to 415. We followed Palepu (1986) in choosing pre-determined ratios for the purposes of   Confining value from neural networks consistency and comparability but also in order to avoid the statistical overfitting issue (see also further support in Section 4.3.1).
From this sample, 102 firms (24.5 percent) did not provide for a meaningful P/E ratio, and a further 87 firms (21 percent) did not have information on liquidity ratios. This further resulted in 189 firms been dropped from the sample producing a final 226 usable observations. The non-target data set reached 2,340 firms.

Modeling
We apply two distinct methods in order to account for the different predictive accuracy of the two categories (target and non-target). A traditional statistical technique as well as an ML, predictive analytics technique, the MLP, has been used to model M&A activity at the developed capital markets and to predict potential targets.
4.2.1 Model 1: multilayer perceptron (MLP) model analysis method. Over the last decade, a renewed growing interest in NNs as a tool for data analysis has been observed. To a certain extent, the attractiveness of artificial NNs vis-à-vis other statistical methods may have also been partially caused by human issues that merit some mention: often there is a shortcoming of statisticians to clearly communicate their methodologies and algorithms to non-statisticians. A large amount of the extant statistical knowledge raises a hurdle for potential investors of their methods. NNs, on the other hand, are in a mid-embryonic phase, meaning that the current knowledge is thinner compared to statistical techniques. Artificial neural networks (ANNs) originate from the biological human brain neurons. It is a network of nodes connected with each other through a weighted connection (Roiger, 2016), which can be greatly beneficial for complex non-linear relationships between variables (Hyndman and Athanasopoulos, 2013). ANN has been used in many industries such as telecommunications, industrials, banking, airlines and healthcare, and has been successfully showcased by Widrow et al. (1994). An example/representation of this model is shown in Figure 7.
The nodes in the input layers are passive nodes as they only pass the data from the input layer to the hidden layer. In the hidden layer, a weight (W n ) will be generated for each input node.
For the first iteration it is a randomly generated number based on Gaussian distribution. Then each input (X n ) will be multiplied by its weight (W n ) to produce a weighted input (XW n ). The summation of these weighted inputs goes into the activation function to produce an output between 0 and 1). The output number gets transferred to the output layer, there they get multiplied again with another set of randomly generated weights to produce the  Table IV shows the financial ratios used in this study which have been used by a number of influential research papers (Ohlson, 1980;Palepu, 1986;Powell and Yawson, 2007). EBITDA and ROA for most technology firms in our sample were non-existent hence had to be dropped as candidate variables as they would limit the sample to less than 100 firms. The number of nodes in this layer simply equals the number of independent variables, in our case six for each instance.
Hidden layer. This layer will receive the nodes sent from the input layer. It will generate a weight for each connection between any node in the input layer and any node in the hidden layer. Then it will multiply each node with its weight as shown in the following equation (summation of weighted inputs): where b is the bias node; X 1 the financial ratio (Ex: ROE for the first instance); W 1 the weight associated with X 1 (randomly generated number between 0 and 1). The net input function z will go into a non-linear activation function (sigmoid function). It will act as a smooth thresholding function to determine the relationship between inputs and outputs. Our sigmoid function performs better for negative variables and classifiers (Zhang et al., 1998) based on the following equation (sigmoid activation function): With differentiation φ' (z) ¼ φ (z) (1−φ(z)), in updating the curve. The cost function used in the study was sum of squared errors using an optimization gradient descent method with the following parameters: initial learning rate ¼ 0.4 and momentum ¼ 0.9. The cost function C is given as follows: Output layer. The output value will then be multiplied with its connection weights again and the final value will go into the output layer. Hidden layers adjust the weightings on those inputs until they reach the optimization stage, that is, the error of the NN is minimized. An interpretation of this is that the hidden layers extract salient features in the input data which have predictive power with respect to the outputs. This is the discussed feature extraction function and it is parallel to the function of statistical techniques such as principal component analysis. This layer consists of a binary node [4]; it will receive the value from the hidden layer, indicating the data set which the firm is predicted to be in. One indicates a target, zero indicates a non-target. The final output value will be compared with the desirable target value. This whole process is called standard forward propagation. The final model architecture is shown in Figure 8. Validation. We used the cross-validation method that involves dividing the data records into three sets: training data set: data records that are used to train the model; testing data set: records that are used to observe the error rate while training in order to further tweak the model; and holdout data set: this set of records is used to assess the model's final error rate and performance. Validation is used to measure the performance and the generalization ability of this model (Kaastra and Boyd, 1996). While there is no standardized rate of division in the literature, some researchers (Hammerstrom, 1993) recommend using the 70/30 ratio. In our study, our data are randomly divided into three groups as follows: 70 percent for training; 20 percent for testing; and 10 percent as a holdout. We first clustered the data into years (Sample 1). Then it was clustered into target and non-target firms. Next, the records were randomly sorted, and the analysis was performed on three sets: • Sample 1: all records of all years, randomly sorted.
However, once this approach was finalized we discovered that this would potentially create a considerable overtraining issue because then the requirement would be to repeat the steps above 17 times (17 years) with the same companies appearing on all data sets. We then took a second-sample approach.
Data itemization in our study could potentially suffer from unreliability owed to sample limitations where the data available were not enough to train different networks on different subsets of the data. Consistent with Srivastava et al. (2014), Dekel et al. (2010 and Hinton and Salakhutdinov (2006), at this stage we only performed the training once in order not to fall into the overfitting and overtraining where the network would just memorize the outcome and not learn thus making it only usable in our specific data set. The data were fed to the network in at once but it used the data ten times (learning epochs ¼ 10) to update the weights. Following the above authors' prior work on data size and data diversity considerations, we performed the experiment based on three trials and then took the average of these trials as shown in the analysis. 4.2.2 Model 2: logistic regression (LR) model. LR method was used as the nature of our study is to forecast takeover targets. Therefore, the output is always binary (i.e. target, non-target) so it is important to use a technique that can classify a data instance into two classes by predicting the probability of an input being in a certain class. We convert the values of our independent variable from a string format to a numerical format assigning the following codes: 0 ¼ Non-Target, 1 ¼ Target. The LR model starts with no predictive variables and only includes the intercept (constant) and measures the prediction power of this model using −2 log likelihood.
It then adds one predictive variable per step and calculates −2 log likelihoods again to measure if the new variable improved the prediction accuracy for all predictive variables. The model allows us to calculate the odds of an input (firm) to be acquired or not using Odds ¼ e a þ b i X À Á , where a is the intercept (constant), b is the predictive variable added in step i and X is the independent variable (Acquisition Status, 0 or 1).
Next, we convert the odds to probabilities using (Probabilities ¼ (Odds)/(1+Odds)). Based on the probabilities result for each input the model classifies them into target or non-target based on a threshold (0.5), any input with a probability equals 0.5 or more will be classified as target, anything less than that will be classified as non-target. Based on this classification, the model produces a classification showing the number of cases correctly classified vs incorrect classifications in order to produce an overall prediction accuracy.
Predictive variables. The same inputs from Model 1 are utilized; therefore, maintaining consistency in our predictive variables (independent variables), namely, return on equity, price to earnings, market capitalization, debt to equity, current ratio, rate of change of annual revenues. Our model is based on three traditional empirical formulae as proposed by Swaminathan and Rogers (1990) formulation of the LR procedures where.
The Odds function is as follows: The probability using Odds is as follows: 1447

Confining value from neural networks
The LR equation is as follows: where Odds is the ratio of probability occurring divided by the probability of it not occurring; P i the probability of firm i being taken over; β 0 the intercept; Z i the weighted sum of the predictive variables; and β n the coefficients for the financial ratio X n .

Model I: MLP
The results of our sample include a total number 226 technology firms, 50 percent target firms and 50 percent non-target firms. Table V shows the prediction percentages for training, testing and holdout data sets based on three trials. As the table shows 70 percent of the data are reserved for training, 20 percent for testing and the final 10 percent for our final holdout sample. We apply a standard feedforward propagation NN with a single hidden layer in our sample in order to identify potential takeover targets (and hence, e.g., the possibility to yield positive abnormal returns from investing in these targets stocks).

1448
MF 45,10/11 standard binary regression technique. Our sample had an out-of-sample overall average prediction accuracy of 71.4 percent, with an average of 28.6 percent of incorrect predictions. The attempt rate (i.e. trials) at three trials showed improvement in our holdout sample for correctly identifying the target companies with a 87.5 percent accuracy prediction rate. Yet it has to be recognized that it also correctly identified non-targets only 50 percent of the time giving an overall prediction accuracy rate of 71.4 percent at trial three. Our NN model attempts to provide a tool that can adaptively sift through noise and identify patterns in complicated financial relationships where non-linearity might pose problems. Using six inputs considered to be the most relevant, and having only four hidden nodes, our sample gets around the issue of having a relatively small data set. Adaptability also lies in the recognition of not adding too many nodes which could lead to mode overfitting.
Our results support that such an approach can potentially provide meaningful explanation regarding dependent and independent variables compared to a traditional regression model. We turn to this in Tables VII and VIII.  Table VIII illustrates the three steps taken by our regression model when adding new predictive variables to the model and the accuracy of correct predictions on each step. The model was able to increase the accuracy with each step, albeit marginally, reaching an overall accuracy of 61.9 percent. This model correctly identified the target companies with a 66.4 percent accuracy prediction rate and it also correctly identified non-targets only

1449
Confining value from neural networks 57.5 percent of the time. Comparatively, the first model achieves a higher accuracy overall over Model 2 providing some support for the utilization of NNs.
It has to be said though that the training, testing and holdout results differ from each other in each of the three trials for the 50/50 samples. We suggest that it is due to the random number generator where the network starts with a random initial numbers to start with and then keeps updating the weights accordingly; this is important in order to create a global optimum solution. In the first instance we actually had an average 6 percent change from one step to the next but a variability of 16.5 percent in-between the steps. Compared to the second model the step difference is 1 percent with a variability of 14.5 percent in-between the steps. The observations drawn are: the regression model is static throughout the sample and trials whereas the NN model shows evolution and adaptability; there are large swings in variable values where, for example, the RoE, D/E and liquidity swing wildlydeep in negative and high up in positive territoryfrom year to year; and the number of observations is relatively limited where the holdout sample is strictly anecdotal data since it covers only a limited number of observations, hence, the expressive power of the network is potentially not enough to capture the target function. One alternative would be to add more layers or more hidden units in fully connected layers. So while it is helpful to test different methods, and provide for greater accuracy, it does not by itself, conclusively, determine which method is best owed to data limitations. In addition, an examination of the variables also provides some interesting insights. Figure 9 shows the importance of each variable fed into the model in terms of characterizing its output. While the variable importance analysis shows the input effects on the output, it can also be clearly seen that the three variables mentioned above account for over 80 percent of the effects on output.
The variable importance analysis showed a great importance for ROE, D/E ratio and liquidity. These are consistent with the inefficient management, leverage and liquidity takeover hypotheses but the direction of the relationship between the independent and dependent variables is not clear. These three variables also showed the greatest volatility throughout our sample period.

Conclusions and limitations
This paper examines the use of an NN method for pricing mergers. With over 50 percent of mergers failing, it is critical for acquiring firms to identify the characteristics of a target prior to a merger that will provide synergies once the merger is complete. The NN model presented in this paper simplistic as it may be at this stage shows overall improvements on the accuracy of predicting merger targets over linear regression results. ANN has outperformed logistic models in both senses of discrimination and calibration, although from the arbitrary standpoint of accuracy (cut-off point 0.5), logistic models can be superior to ANN models. The fact is that in some applications NNs fit better than other models such as linear regression and this usually occurs when there are nonlinearities involved though it is important to evaluate other aspects. For example, a linear regression model will have less parameters to estimate compared to an NN for the same set of input variables. Hence, an NN will require a larger data set for its calibration and subsequent optimization in order to get the required benefit of generalization, applicability and nonlinear mapping. In the absence of critically enough data, despite existing nonlinearities involved, a linear regression model may indeed be better calibrated. Improvements as per the accuracy of target prediction can translate into significant savings in offering prices for target companies. Reliable predictions can improve the quality of decisions and business strategy in target determination and fair price decisions. NN methods permit the use of an expanding number of prospective venture opportunities with the added benefit that as market changes are introduced and more dynamic analysis is eventually involved new and more inputs can be loaded onto the model with less resource devotion. Having said that it is also important to identify that NN methods do not provide for a fuller analysis of significance for each of the autonomous variables in the model as traditional regression methods do.
Using a different activation function and a "deeper" network with more hidden layers could potentially account for how each successive layer uses the output from the previous layer as input. It could also further show how the algorithm self learns from multiple levels of representations that correspond to different levels of abstraction (i.e. the levels form the hierarchy of concepts above). The quantity of data at our disposal though is relatively limited for more hidden layers to be involved in terms of describing potential causal connections between input and output. The transfer function is the calculated derivative sigmoid function utilized; we see this as important when calculating the weight updates in the network based on the amount of data and the computational load of our simulation. Finally, while the extra layers could potentially help in learning features indeed the authors felt that with such a sample introducing Rectified Linear Unit (ReLU) we may also run the risk of naively training a "deeper" NN. As argued above, the possibility of added layers of abstraction could also show rare dependencies modeling in the training data.
It could arguably have also been interesting to investigate how model performance is influenced by using different activation functions (e.g. utilizing the so called ReLU method instead of Sigmoid) or also involve in the analysis a higher number of hidden nodes. This is another area for research where, traditionally, ML evaluation works best in producing an extrapolative model. The trade-off though, of creating a flexible, nonparametric predictive model on the other hand, is that causal interpretations can potentially be lost. Equally, linear regression is a relatively inflexible approach yet it is less complicated in its interpretation. Flexible constructs avoid the assumptions of a particular functional form for a model, but they also require a larger number of observations and are more complicated and challenging to interpret. In addition, it can also be supported that NNs with different initializations produce different signals for a certain feature. As seen above, our NN with a certain initialization produced better signals in some cases and incorrect signals in some other.
Our results are also consistent with 20 years of research and some seminal papers that date as back as the 1990s until today (see e.g. Sen and Gibbs, 1994;Sinha and Richardson, 1998;Fescioglu-Unver and Tanyeri, 2013;Spangler et al., 2015;Tkáč and Verner, 2016). Such studies indicate that although NNs map the data satisfactorily, it is still questionable whether they predict merger targets significantly better than LR. This strongly suggests 1451 Confining value from neural networks that the financial models used to predict mergers are relatively inadequate. Firms should approach the development of merger prediction models cautiously and identify other factors that are more likely to predict mergers. NNs give the best overall results for the largest multiple classification cases. There is substantial room for improvement in overall performance for all techniques. The results indicate that data mining methods and data proportions and characteristics have a significant impact on classification accuracy. Zhu et al. (2001), for example, state that within data mining methods, rough sets provide better accuracy, followed by NNs and inductive learning.
The generalization breadth of this study is limited within a specific sector (technology) in a specific country (USA) covering a specific period (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). One of the most important limitations was data collection, as we had to omit approximately 50 percent of the initial sample due to unavailable data on firms and their financial ratios. The takeover determinants were chosen from previous studies done by other researchers that showed statistical significance; this may affect the results of this analysis as the sample size, sector and study period are different. Further research can be done to extend this model and improve the accuracy of it by including for example: technology firms' specific ratios, this will allow the model to study technology firms not just from a financial but from an operational perspective too; social profiling, social media and softer social variables not captured or modeled by standardized techniques, where these can be leveraged in order to discover opportunities or create maps for those interested audiences (Beese, 2015). Monitoring social media impression of the firm or its management might give an indication of its takeover probability [5].
Old may be, but this echoes also Kuo and Reitsch's (1995) early research in the managerial forecasting problem; many managers value the "softer" features of neural nets, particularly when standard regression models tend to emphasize the causal interpretations (more the why) of the problem and not the solution.

Introduction
Corporate restructuring denotes the modifications in ownership, business composition, assets composition and the structure of alliances in order to augment the shareholder value. Therefore, corporate restructuring may comprise restructuring of ownership, business and assets. A company may carry out restructuring of the ownership through mergers and acquisitions, strategic alliances, leveraged buy-outs, joint ventures, spin-offs, buy back of shares, etc. (Daru, 2016). Whereas merger and acquisition remain the centrepiece of corporate restructuring, in the recent decades, India has also observed large number of corporates trimming their size by spinning off one or more divisions. While mergers are said to create synergies (Cigola and Modesti, 2008) spin-off are also expected to add value to shareholder's wealth due to increased focus. Spin-off, also known as demerger, or split-off, results in reduction in the size of the parent company. In the process, the parent company gets divided into two or more separate legal entities. The new companies created by the parent company are usually formed as a subsidiary of the parent company. Shareholders of the parent company receive pro rata distribution of equity shares in the newly formed companies. The new company formed as a result of spin-off is termed as resulting company in our study. The key outcome is that the resulting company becomes a separate decision-making firm and its control gets segregated of control from parent company ( Johnson and Klein, 1996). Spin-offs are the mirror image of mergers (Hite and Owers, 1983). Mergers create value from positive synergy and spin-offs create value by excluding divergent divisions that have adverse synergy. While the acquirer look for a gain due to expected increase in efficiency of operations (Ravenscraft, 1991), the parent company in spin-off also envisages gain due to more concentrated approach. Spin-off is a process of restructuring the business for focusing on the core business of the organisation. Spin-offs help firms in redeploying resources, particularly financial resources to fund other initiatives, which add more value to the firm (Montgomery and Thomas, 1988). The main objective of spin-off is achieving sustainability through effective management of resources of a particular unit of an entity (Basak, 2017).
Sometimes spin-off is used interchangeably to divestitures, but there is a fine line of difference between the two. As per Vijh (1994), divestitures are consummated by selling an unsolicited subsidiary or division to another company or a group of investors for cash. This divestiture form is ordinarily known as a sell-off. Divestitures generates value by selling off the assets which have negative present value. (Nguyen et al., 2013) In many cases, however, the parent company renounces the control of a subsidiary by simply allocating the subsidiary shares to current stockholders. This latter divestiture form is known as a spin-off and has been gaining popularity in recent years. Undervalued firms are more likely to spin-offs and overvalued firms are more likely to sell-off. Spin-offs are mainly transacted during investor optimism (Alexandros and Simonyan, 2015). After the spin-off takes place the parent company and resulting entity are still owned by the same shareholders (Burgelman, 1984).
The rest of the paper is organised as follows. The second section describes the literature review on the spin-offs. The third section describes the data and research design. The fourth section reports the main findings and finally the fifth section concludes the paper.

Literature review
The rationale behind this paper is to unlock the potential of business organisations to create and preserve shareholder value (Schipper and Smith, 1983). Spin-off becomes a driving force to increase value in the cut-throat competitive environment, both economic and financial (Ghosh, 2014). While merger and acquisition is the only remedy for number of companies to create value and to grow, the others feel separating or segmenting the ownership might become more advantageous (Coyne and Wright, 1986). To them corporate break-ups seem to be more effective and attractive to enjoy improved operating performance and better information flow to the investors because of separate financial disclosures. The review of literature suggests that sometimes spin-off also takes place as a result of any earlier merger and acquisitions transaction (Kaplan and Weisbach, 1992). If the acquisition is not giving the desired results, then the company may think of spinning off the division. Investors become more equipped and the companies can raise additional equity funds by boosting their valuations up and unlocking hidden shareholders' value as a sequel of spin-off (Miles and Rosenfield, 1983).
A subset of the reconfiguration literature argues that divestiture is an important element of reconfiguration activities, either as an independent mode of reconfiguration (Berry, 2010) or as a proactive tool used in combination with acquisitions or alliances (Capron et al., 2001;Villalonga and McGahan, 2005). Divestiture activity includes selling all (full divestiture) or portions of (partial divestiture) existing business units, including liquidating substantial sets of assets (Duhaime and Grant, 1984;Kose and Ofek, 1995). Full and partial divestitures include sell-offs to other firms and spin-offs of units into independent operation. Divesting full units allows firms to undertake extensive immediate corporate change, whereas divesting partial units allows more fine-grained adaptation (Montgomery and Thomas, 1988). Greater divestiture activity in a given period means divesting assets with greater financial value and/or undertaking more divestiture events. Reconfiguration via divestiture can arise from both weakness and strength.

1459
Restructuring through spin-off The literature available on the effects of spin-offs usually moves around the companies from the USA. A few studies from Japan, India and Australia are also available but the quantum is much less as compared to the studies from the USA. Rosenfeld (1984) studied 35 spin-offs during a period of 1963-1981. He has considered the event window period of −30 to +30 days and has calculated the average abnormal returns (AARs) and cumulative average annual returns. He has further analysed the returns over different event window periods, to examine closely the impact of spin-off announcement on the stock prices of parent firm. His results show that both spin-off and sell-off announcements tend to have a positive influence on the stock prices of the divesting firms. Schipper and Smith (1983) also investigated the effect of voluntary corporate spin-off announcements on shareholder wealth. They found a significantly positive share price reaction for a sample of 93 spin-off announcements between 1963and 1981. Chai et al. (2018 studied the impact on spin-off announcements in Australia. They found that Australian spin-offs are associated with a positive impact of spin-off announcement. Few studies have also been carried out on Indian data long-run excess stock performance for up to 36 months after the spin-off. Few Indian studies have also been carried out to examine the impact on spin-off announcement on the share price of parent firm. We have tried to summarise some of the studies on the effects of spin-off announcements in Table I. Most of the studies in Table I have been conducted on the data from the USA. Sample size varies from 35 to 146 spin-offs. Almost all the studies have tried to find out the impact of spin-off announcements of the share prices of the parent company by applying event study methodology. The results of all the studies show a positive impact on the share prices of the parent company. Different event windows have been taken by different researchers but the basic methodology used is the same. Few studies in Table I are from Australia and the UK. These are handful of studies with sample sizes varying from 103 to 223. The methodology adopted by them is also similar and they also confirm the results obtained from the USA. Table I also shows that Indian studies on spin-off are very few and the sample size is also too low to authenticate the study. The review of literature suggests that the overall impact of spin-off announcement has been positive on the stock returns of the parent company. During the review of literature, it was found that abnormal returns are between 2.4 and 4.3 per cent as shown in different time periods and geographical locations. Veld and Veld-Merkoulova (2008) performed meta-analysis using a sample of 26 firms and found that spin-offs result in positive wealth creation for the shareholders. Galai and Masulis (1976) claimed that there could be transference of wealth between bondholders and stockholders because of the spin-off. The assets of the spin-off firm no longer are collateral to the bondholders of the parent firm and hence the value of debt of the parent should be reduced. However, Schipper and Smith (1983) and Hite and Owers (1983) both examined the transfer of wealth hypothesis and found no empirical support for it. However, Parrino (1997) concluded that wealth transfer takes place from bondholders to shareholders after spin-off. In his study on Marriott Corporation he has documented that total value of the firm has decreased following the spin-off announcements.
It has also been observed in literature that in some studies spin-offs have been divided into two parts, namely focus-increasing and non-focus-increasing spin-offs. Both kinds of spin-offs have been analysed separately. Daley et al. (1997), Krishnaswami and Subramaniam (1999) and Desai and Jain (1999) concluded that the abnormal returns for the focus-increasing spin-offs are higher than for the non-focus-increasing spin-offs (Boreiko and Murgia, 2016). Abnormal long-run stock returns and operating performance are observed for spin-off firms only and mostly for internally grown business units and nonfocusing subsidiaries. Our study has not categorised the spin-offs into focus increasing and non-focus increasing. The cumulative average abnormal return is significantly positive for the completed spin-offs Krishnaswami and Subramaniam (1999) 1979 Positive abnormal returns were found around the announcement of spin-offs Daley et al. (1997) 1975-1994 The results indicate significant excess returns around the announcement of cross-industry spin-offs ( focus-increasing spin-offs) Johnson and Klein (1996)  Spin-off announcements produce an abnormal positive stock returns. He supported the conclusion by calculating CAAR also Hite and Owers (1983) 1963-1981 For 59% of the sample cumulative excess returns are showing positive impact. For the rest 41% the impact is negative. When tested for different observation periods, the announcement was found to have significantly positive stock price reactions. The overall indication is that stockholders experience positive wealth gains Miles and Rosenfield (1983) 1962-1980 USA 55 Spin-off announcements creates an increasing impact on shareholder wealth Schipper and Smith (1983) 1963-1981 USA 93 A significant positive share price reaction is documented Table I. Literature Review on the impact of spin-off on the stock prices of parent firm As per our review of literature, it is indicative that the Indian data of spin-offs have been explored quite less. Whereas a large number of studies on US data substantiate the conclusion that spin-off announcements have positive impact on stock prices, at the same time we could find very few studies to substantiate the same conclusion for Indian data. Recent study by Vyas et al. (2015) analysed the impact of spin-off announcement on the stock prices. They took data of 51 spin-off announcements during a period of 2012-2014. Their result is also in consonance with the US studies that spin-off announcements lead to a positive impact on the stock prices of the parent company. Apart from this study, the other studies on Indian data have used either a very small sample size or they are in the form of case studies discussing a single spin-off announcement in detail. Bendre and Apte (2017) studied 24 spin-offs from India announced during 2012-2017. They concluded that 60 per cent of the companies have shown positive impact on their share prices as a result of spin-off announcement. Ghosh (2014) studied 20 spin-offs from India. He studied the movement of share prices before and after spinoff and found that price of the parent company's share has seen an increase thereby increasing the shareholder's wealth. Singh and Bhowal (2009) examined five spin-offs that took place in 2006 and concluded that demerger/spin-off leads to positive abnormal returns. India has witnessed an increase in spin-off cases in the last two decades. Probably, with the growing business of the corporate houses, the need arises to reduce the size of the organisation so that the focused approach can be followed. The eventual purpose of the business is to maximise the market value of owners' equity (Cummins and Xie, 2009). Our paper attempts to find if this purpose is achieved after spin-off.

Data and research methodology Data
This study has been done to analyse the impact of spin-off announcement on the shares price of Indian companies. We selected 249 companies that were listed on Bombay Stock Exchange and went for a spin-off during the period 2010-2011 to 2015-2016. Out of these, 104 companies were eliminated as they were trading at a price below Rs20. Data for very low priced securities were too scattered as it is seen that they are not traded daily. A total of 11 companies were eliminated as they had got delisted due to some reasons. Data of 19 companies were not available as they had got suspended from the stock exchange due to penal reasons. For 23 companies, exact date of spin-off announcement could not be verified due to non-availability of the data. Data for 16 companies were too scattered. Although these companies were trading above Rs20, they were not getting traded daily so the daily movement of prices could not be captured. After elimination of all the above companies, we were finally left with 76 companies which announced their spin-off decision during the period 2010-2011 to 2015-2016. The date of announcement has been taken to be the date when the company intimated the spin-off decision to the stock exchange after it was duly approved in the shareholder's meeting. Daily data of stock prices of individual securities have been extracted from the Bombay Stock Exchange. For the movement of market, BSE 500 index has been considered.

Methodology
Finance literature substantiate that even study methodology has been extensively used in evaluating the impact of a certain event on the stock prices. The methodology adopted in our paper is similar to the one adopted by Rosenfeld (1984). Event study methodology has been used to examine short-term stock price reaction to the announcements of spin-offs.
Following are the steps that comprise the mechanism of event study.
Defining the event and the date of announcement. The foremost step in the event study methodology is to describe the date on which the event is first declared to the public. Day 0 is stated as the day on which the spin-off announcement gets published in any newspaper.

MF 45,10/11
To define Day 0 is the most important part in any event study. As a part of the process, spinoffs are approved in the board meeting first before further processing. The day on which Bombay Stock Exchange is intimated about the said board approval has been identified as the announcement day in our study. These dates are carefully verified (manually) from the archive section of the corporate announcements of the Bombay Stock Exchange.
Estimation period. The period prior to the event is defined as the estimation period. It is essential that there is no overlap between estimation window and event window. This ratifies that the normal return estimations are not impacted by the event-related returns. Estimation window of 256 (−290, −35) days is considered in the study.
The market model is used to assess the expected returns. Regression of a stock's returns against a market index is done to calculate the expected returns. S&P BSE 500 is used as a market index. The strategic issue in event studies is to determine the part of the price movement that is essentially affected by the event under study. It is imperative to measure the effect of the specific event on stock returns. This generates the concept of ARs. The AR is the difference between the actual return and the expected return on a particular day. The AR of the jth stock (AR jt ) is achieved by reducing the expected returns in the non-occurrence of the event E(R jt ), from the actual return during the event period (R jt ) as per the following equation: The market model relates the return of a security to the return of the market portfolio as per the following equation: where, t ¼ -290, …, -35, α j is a constant term for the jth stock, β j is the beta of the jth stock, R mt is the market returns, and ε jt is an error term. The parameters of the model will be assessed by using the time-series data from the estimation period before every particular announcement. The estimated parameters will be used in the calculation of ARs for each day in the event window. These will then be compared with the actual returns during the event period. The daily excess return, that is, the AR of firm j for the day t(AR jt ) is estimated from actual returns across the event period and the estimated coefficients from the estimation period as per the following equation: where t ¼ -35, …, +35.

1463
Restructuring through spin-off The average abnormal return (AAR t ) for each day in the event window is calculated as per the following equation: where N is the number of companies. The cumulative average adjusted return for day t, CAAR t , is defined by: To take into account any cross-sectional dependence of the abnormal returns over the observation period, t-tests are performed using the crude adjustment method suggested by Brown and Warner (Rosenfeld, 1984). Table II denotes the AARs over the event window. AARs have been shown for all the 70 days in the event window. Along with AAR, CAAR has also been worked out for each day in the event window. It can be seen from the table that the abnormal returns being 0.905 are the highest on Day 0, i.e. on the day of announcement of the spin-off. The return on Day 0 has a high t-value (Brown and Warner) of 2.77 which is significant at the 1 per cent level. Apart from Day 0, Day −1 also shows high abnormal returns. Day −1 has return of 0.104, with t-value of 2. It is significant at the 5 per cent level. This signifies that the information about spin-off starts impacting the stock price of the parent company even before it is officially announced. This could be due to the information leakage before it officially strikes the market. As per Table II, the market's response to the spin-off continues on Day +1 as well. The returns on Day 1 are 0.048 with a high t-value of 2.22, which makes it significant at the 5 per cent level. It is clearly reflected here that the spinoff announcement has a significantly positive impact on the stock prices of the parent company during Days −1, 0 and +1. The CAAR is maximum on Day +1 signifying that the returns to the shareholders would be highest on Day +1. The result of day-wise analysis also gets reinstated in Table III where the AARs on different intervals are being shown. The AAR of 0.857 is maximum for the interval (0, +1) with quite high t-value of 3.32 which makes it significant at the 1 per cent level. The interval (−1, +0) and (−1, +1) also shows a positive AAR of 0.74 and 0.751, respectively. Both are significant at the 5 per cent level. The results of day-wise analysis and interval-wise analysis verify each other. CAAR for the interval (0, +1) is the highest. This makes the interval to be highest return giving interval in the entire event window period. Table II shows that the AAR starts getting positive almost from Day −9 onwards. Although the AARs from Day −9 to Day −3 are not significantly positive, we can say that the stock prices of the parent company start getting impacted before the official announcement of the spin-off. The similar results are also obtained in Table III. The AARs for the interval (−10, −5) and (−5, 0) are positive but not significantly positive. After Day +1, the stock prices of the parent company do not seem to be impacted by the spin-off announcement. Apart from this, Day −23 and Day −16 show a significant decline in AARs. Day +21 shows significantly positive AARs. The results obtained in Tables II and III are related to each other rather they verify each other. Table III shows that AARs are not significant for any interval before the official announcement of the spin-off except the two intervals, namely (−1, 0), (−1, +1).

1465
Restructuring through spin-off spin-off announcement, AAR is significantly positive for the interval (0, +1). After that AAR is not significant for any other interval except (+30, +35). For this interval the AAR is significantly negative at the 1 per cent level. The CAAR is maximum for the interval (0, +5) as the returns for the interval (0, +5) are also positive, although not significantly positive.

Conclusion
The results obtained in our analysis are consistent with the findings of Rosenfeld (1984), Veld and Veld-Merkoulova (2004) and Cooney et al. (2004). The results for the complete sample of 72 spin-offs show that spin-off announcements have a positive effect on shareholder wealth and such announcement increases shareholder value. The spin-off announcement has a significantly positive impact on the stock prices of the parent company.
The impact starts coming a day before the official announcement and remains till one day after the official announcement. The CAAR is highest on Day +1. The pre-announcement period seems to be affected more in terms of impact on share prices as compared to the postannouncement period. The shareholders of the parent company gain as a result of the spinoff announcement. The study of spin-off announcements on Indian data has substantiated the results in the USA, Australia and the UK.

Introduction
In today's reality, characterized by increased cross-border competition in both educational services and research outputs, universities are under strong pressure to both compete and collaborate. Higher education institutions (HEIs) compete for financial resources, talented students, high-quality lecturers, brilliant researches, good reputation and status, as well as high scores in international rankings. The increasingly competitive global market is the important driver of institutional mergers which become an important strategy of many HEIs. Other strategies, as underlined by Harman and Harman (2008, p. 99), include informal collaboration; joint business ventures; strategic alliances; regional, national and international networks and consortia; as well as cross-institutional mergers of academic and/or service departments. This study concentrates on strategic management of mergers in HEIs. The consolidation of universities is a major theoretical and practical challenge. However, despite a very large number of practical examples of university mergers worldwide, at the same time there is a shortage of frameworks that would help manage mergers in the HEIs context. This paper is an attempt to respond to these needs and grow the body of knowledge in this area. One of the concepts giving the theoretical basis to the topic of strategic mergers of universities concerns the theory of social identity, other refer to strategic and process theories of mergers and acquisitions (M&A) (Cai, 2006;Cartwright and Schoenberg, 2006;Gleibs et al., 2013). In the public sector, the basis for analyzing the concept of consolidation is the theories of "new public management" (NPM) and "public value management" (Bryson et al., 2017;Hartley et al., 2017). According to the trends of "new public management," university management can be treated as a complex process, similar to organizing the work of an enterprise (Dunleavy and Hood, 1994;Dunleavy et al., 2006). Mergers of public universities can illustrate the logic of this approach. Currently, in the process of university merger research, the higher education sector goes through the induction stage, where hundreds of case studies and a few comparative studies have been gathered that draw a complex picture of the mergers' practices and can serve as a source of guidance. However, there is a need for inductive synthesis of the sources of information and creation of a conceptual model that will help to guide the management processes.
The paper is organized as follows: Section 1 highlights context, reasons and strategic goals of mergers in HEIs. The following section refers to the areas of strategic focus during mergers. In Section 3, we present the proposal of the Conceptual Model of Universities' Mergers that is followed by an example of a merger that resulted in establishing the Université Grenoble Alpes. The study finishes with the conclusions and the proposal of ten principles of effective mergers' management at universities.

Mergers in higher education institutionscontext, reasons and strategic goals
Strategic mergers are formal combinations of two or more organizations into a single organization deliberately planned, so as to more effectively meet external challenges and opportunities (Harman and Harman, 2003). In relation to higher education, strategic mergers are described as strategies of "merging colleges for mutual growth" (Martin and Samels, 2002). In terms of getting two institutions together, the following terms are used: M&A, consolidation processes, takeovers, fusions, buy-outs and marriage. Despite the fact that these terms should not always be treated as synonyms, they are often used interchangeably. In a merger, one company takes over another, including all assets and liabilities. In a consolidation, two or more companies merge to form one new, larger company. All of each company's assets and liabilities then become the property of the new organization. Mergers and consolidations are ways in which companies can merge, following essentially the same process; therefore in our study, we will use the terms "mergers" and "consolidation" in relation to universities' mergers interchangeably.
The diagnosis of trends concerning changes in higher education has been developed on the basis of many studies and is widely described in the literature on this topic. One significant, clear trend is toward the development of larger and stronger "producers" of 1470 MF 45,10/11 educational services and research. Moreover, there are several tendencies observable in the HEIs sector that prejudge the strategic changes in universities: (1) High and increasing diversity of universities; the educational sector consists of organizations that differ substantially in their founding structure, activity, quality, specialization and size.
(2) Internationalization resulting in the increasing mobility of students, researchers, programs and entire institutions.
(3) Privatization and commercialization of education on a global scale, where higher education becomes a service coming from the sphere of "private goods," and science is an intellectual product.
(4) The development of the "entrepreneurial university" model. (5) The reduction of the state's participation in subsidizing or even co-financing universities.
New globally competitive higher education environment dictates strong incentives toward competition between institutions, but at the same time makes many of them decide to cooperate, following different types of partnerships: from a very informal cooperation between researchers, through alliances, consortia, affiliations and federations to full scale mergers.
In the case of HEIs, the implementation of the mergers' plans should lead to fulfillment of the mission and achievement of strategic goals related to the improvement of research and education or/and to the implementation of the universities' third mission (Di Berardino and Corsi, 2018;Zomer and Benneworth, 2011). In the second half of the twentieth century, there was a departure from the traditional formation of the Humboldtian university vs the entrepreneurial university (Clark, 1998). It is more and more visible that universities, especially those private ones, however also public HEIs, use economic logic and solutions developed in the field of business management. Competition, commercialization of research and cooperation with the socio-economic environment are becoming increasingly important. Internationalization and globalization of universities is growing, complex cooperation networks are being formed, and universities are competing for the best researchers and students through international cooperation. Universities begin to resemble business units and therefore they face similar competitive challenges. The university is transforming into an economic market organization that follows the concept of "new public management" (de Boer et al., 2007;Hood, 1995;Sułkowski, 2016).
Entrepreneurship tendencies in the university culture are reflected, inter alia, in the orientation on innovation, in scientific activities carried out in cooperation with the industry, in the application of organizational solutions of "quasi-business" and "quasi-corporate" type, as well as in the pursuit of generating revenues from educational and scientific activities. In addition, "entrepreneurial universities" implement a market mission and create competitive strategies, use accountability and governance methods, and make decisions using a managerial model (management and supervisory authorities have the power), not a collegiate one (based on an academic staff ). Mergers may be treated as a manifestation of the development of entrepreneurial university formation and academic entrepreneurship, both in relation to public and private HEIs. Private universities, through merger and consolidation processes, develop economies of scale and improve organizational methods, which lead to more effective market operation and fulfillment of their mission (Rudden, 2010). Moreover, in private HEIs, e.g. in the USA, mergers have been commonly used by individual institutions to deal with threats of closure, declining enrollments or even bankruptcy (Harman and Harman, 2003). Public universities recognizing that there are too many too small institutions try to obtain through mergers a "critical mass" in scientific,

1471
Mergers in higher education institutions educational and operational activities (Aula and Tienari, 2011;Tirronen and Nokkala, 2009). Generally, e.g. European universities are having hard time competing with their American counterparts, because they are relatively small and poorly funded. Lang (2003), while analyzing the reasons of mergers among public universities, underlines that governments want new programs at relatively low marginal costs. Moreover, mergers can reduce sunk costs of previous investments as the facilities may be utilized more efficiently. Some studies have confirmed also the financial drives of many mergers in HEIs (Eastman and Lang, 2001) pointing out that universities' mergers can result in significant economies of scale (Brinkman and Leslie, 1986;Lang, 2003;Lang, 2002;Sears, 1983).
Following Pinheiro et al. (2017) and Sułkowski (2017), it is possible to indicate several strategic goals, concerning universities' mergers: (1) increase of the effectiveness and efficiency of the universities' operations (Pinheiro and Stensaker, 2014); (2) limitation of the higher education system fragmentation (concentration); (3) expansion of students' access to the education network; (4) strengthening the autonomy, responsibility and accountability of the university; (5) creation of larger universities, growing scale of scientific, educational and operational activities, gaining the economies of scale and "critical mass" (Aula and Tienari, 2011;Docampo, Egret and Cram, 2015;Tirronen and Nokkala, 2009); (6) optimization of operating costs (Harman and Harman, 2003); (7) strengthening the competitiveness of a particular university at the national level (Goedegebuure and Meek, 1994;Harman and Harman, 2003); (8) support for university competitiveness at the international level (Harman and Meek, 2002); (9) strengthening the competitiveness and visibility of the entire country and the national education system at the international level (Docampo et al., 2015); (10) meeting the needs of different stakeholders, in particular students and employers in a more efficient way; (11) implementation of an effective strategic management mechanisms; (12) restructuring and rationalization of university management; (13) change of the competitive model to oligopolistic or even monopolistic in the case of private universities; (14) diversification of the educational offer; and (15) market expansion (mainly in case of private universities).
All the stakeholders of consolidating universities could benefit from their successful merger as it means a stronger institution that is in a position to compete better in today's global economy and become more effective and efficient.

Areas of strategic focus during mergers
In consolidation processes, strategic management plays a key role (Pinheiro and Stensaker, 2014). First of all, the decision about a merger itself should be preceded by a strategic analysis of the organization and the environment, which is the premise for making the decision on the merger. There should be consultations with various stakeholders and due diligence groups. The strategic objectives of the merger, which will be the basis for the 1472 MF 45,10/11 preparation of the strategic plan, should be set (planning stage). The adoption of the strategic plan for the merger is related to the transition to the process of strategic coordination (implementation stage). Strategic management at this stage consists of drawing conclusions from due diligence as well as participation in negotiations and conclusion of contracts. Institutionalization of the mergerin the form of signing agreements and validating the decisions taken on the consolidation of entitiescloses the implementation stage and constitutes the transition to the integration stage. Strategic management at the integration stage is associated with (Sułkowski, 2017): • supervising the correctness of the merger and implementing the strategic plan; • corrections to the strategic plan related to unforeseen situations; • strategic controlling of the merger process; • managing the work of management teams and the integration team; • coordination of central unit activities; and • conflict resolution and organizational and public communication.
Undoubtedly, decisions on mergers belong to the strategic ones. They are complex management processes that require a long-term implementation plan, consistent with the strategic plan for the development of the entire organization. The degree of complexity and difficulty of running consolidation processes depends on the number of factors: institutional characteristics, the type of consolidationif it is voluntary one or a compulsory (take-over), the profile of the HEIs involved, number of partners and cultural context, just to mention a few. The actual strategic success of the merger is not just the implementation of the university merger itself, but also the effects it brings. In order to achieve them, universities must consider and deal with several challenges of consolidation processes. In this paper, we indicate five areas of strategic focus during mergers: academic due diligence, appropriate selection of methods and tools in restructuring, project management during mergers, academic leadership, and finally university brand management and marketing activities of universities in the merger process.

Academic due diligence
Due diligence means in-depth analysis, examination and verification of previous information, thanks to which a potential buyer or merger partner can make an assessment (Sułkowski, 2017, p. 186). Such verification usually takes the form of a written document that presents the actual situation of the organization and pays special attention to current and future possible risks that may occur after the merger. The general characteristics of due diligence should meet the requirements of credibility, accountability, validity, accuracy, transparency, completeness and clarity. The scope of due diligence is wide, because this comprehensive analysis includes financial, legal, infrastructural, technological, organizational, intellectual resources, human resources and organizational culture analyses.
The due diligence in universities should focus on the most important goals, potential synergic effects of consolidated institutions as well as difficulties/barriers in the following areas: • management (strategies, structures and organizational processes); • material resources (campuses and laboratories); • human resources (scientific and didactic staff, administration, students and graduates); • financial resources (endowment, cash flow, costs and revenues); the merger itself. In the case of public universities, restructuring is rarely the most important goal of a merger, but it is often implemented as a consequence of the consolidation process. In the mergers of private universities, restructuring is usually one of the key objectives, because the possibility of creating added value through the merger requires a deep reengineering of the whole institution. Identification of areas for restructuring takes place at the stage of analysis and strategies creation (due diligence). Such identification should lead to: (1) analysis of organizational strategy, structure and culture; (2) definition of measures and indicators, the measurement of the current status and planned targets (after merger and restructuring); (3) choosing the methods of restructuring during the merger and after its implementation; and (4) creation of a merger plan taking into account restructuring processes.
Depending on the area of change, three types of restructuring can be distinguished: operational, financial and concerning the ownership. Each of them focuses on different aspects that are presented in Table II.
The university restructuring methods used in the merger processes relate appropriately to all processes and functional areas of the organization, namely finance and accounting, as well as the management of: people, quality, information, marketing, infrastructure and other aspects of operations. The typology presented in Table III does not cover all restructuring methods, tools and approaches that can be applied at universities, but it constitutes a list of options to consider in the merger processes.
In the mergers of universities, various restructuring methods and techniques are used; however, as Sułkowski (2017, p. 206) indicates, there should not be introduced too many complex methods of university restructuring (e.g. reengineering, Six Sigma or lean management) simultaneously. Before planning the merger, it is necessary to reflect and select the appropriate mix of useful methods, techniques and tools to be used in organizational changes.

Project management in university merger processes
Mergers can be operationalized as inter-organizational projects that lead to the improvement of processes: research, education and cooperation with the environment. Mergers of universities refer to deliberate organizational change, with a framed plan, time restrictions and budget limits. Projects have become means of implementing the organizational changes in HEIs. Effective project management in universities results in an increased competitiveness and value. The composition of the projects involved in the merger may Type of restructuring Description Operational restructuring Concerns changes in the core business of the enterprise and in the case of the university sector it may relate to, among others, marketing activities of the university, human resources (academic and non-academic staff ), technological and property resources of the university, as well as the organization and management processes implemented at the university Financial restructuring Focuses on financial aspects related to indebtedness, liquidity and efficiency of using capital

Ownership restructuring
Begins with changes in the structure of equity of the university and may then include further areas of activity Source: Authors' own study

1475
Mergers in higher education institutions vary depending on the strategy and mission of the particular university. University mergers should lead to synergy that provides the opportunity to improve the core processes (Patterson, 1999). In the management process of a merger, a mix of various methods, tools and techniques listed in Table III can be applied. Apart from that a three-level typology of merger projects in HEIs can be built according to the parameters: project duration, range of changes, degree of complexity, project effects and scope of changes. The typology of the universities' merger projects is presented in Table IV.

Academic leadership during the merger
Leadership plays a key role during a merger process. Harman and Harman (2003, p. 29) state: "sensitivity to human and cultural factors and effective leadership are the most important factors for achieving success in university merger." Academic leadership is to encourage members of the organization to act together, leading to the realization of the goals of this institution. The concepts of academic leadership are derived from a rich theory, and are developed on the basis of organization and management, psychology and sociology. Effective leadership that is crucial for the success of the merger may be analyzed from the perspective of various theoretical schools. They embrace mainly: theory of attributes, situational theories, management style concepts, critical trend regarding the school of leadership, the school of neocharismatic and transformation leadership and the team management school. In the context of university mergers, there are four main concepts of leaders that seem to be the most relevant. Their main characteristics are described in Table V 3.5 University brand management and marketing activities of universities in mergers Building reputation through intensive communication, marketing and internal branding (employer branding) is gaining importance in the academic world. Branding and image are notions related to reputation, they are also associated with culture and organizational identity (Aula and Tienari, 2011). All these areas of the organization's activity are subject to profound transformations in the processes of a merger. During mergers, the universities must use the concepts of marketing communication, brand management and organizational identity. The international and national recognition of universities for students and other external stakeholders depends to high extent on the university brand. The name of the university creates its image by providing a message that reflects the identity of the university. It creates trust, loyalty and reputation of the institutions. More and more often, students who make the decision about choosing a university take into account the value of the brand, e.g. the benefits that may result from receiving a graduation diploma from that particular institution. Following Aaker (1991), while verifying the value of the university brand, five basic elements described in Table VI should be considered.
The search for sophisticated concepts and marketing tools by universities is increasingly noticeable in order to build an appropriate brand image of the university and to attract customers, especially during or after merger. However, the marketing activities of universities concerning mergers are not limited to brand management. Contemporary marketing activities at universities focus not only on the regular promotion of the educational offer, but also on relational activitiesbuilding the image of the university. Especially, the concept of relationship marketing that has been transferred from other market sectors has become very popular among professionals performing marketing activities at universities (e.g. Plewa et al., 2005).

The Conceptual Model of Universities' Mergers
The systematic literature review summarized in the previous sections provided an input to form a map of the concepts related to mergers in the higher education sector, thus allowing us to formulate a conceptual model, which essentially represents an "integrated" way of looking at the topic of universities' mergers (Liehr and Smith, 1999). Miles and Huberman (1994) defined a conceptual model as a visual or written product, one that "explains, either graphically or in narrative form, the main things to be studiedthe key factors, concepts, or variablesand the presumed relationships among them" (p. 18). The proposed Conceptual Model of Universities' Mergers is an attempt to build a simplified representation of the The leader sets the direction of change and allows for quick, efficient and effective operations. This approach is based on the mutual trust of the leader and team members. It is also based on values and shaping of collective and individual identities. The leader focuses on transforming the perception and interpretation of reality (sensemaking and sensegiving), which allows the group to believe in the sense and value of the change that is taking place There is a threat concerning the creation of illusions and mistakes. Therefore, the perspective of transformational leadership should be balanced; its task is to combine a positive attitude to change with a pragmatic view of the process implementation Situational school of leadership The basis of research in this approach is searching for the conditions of effective leadership in university mergers processes that is affected by a particular situation/ conditions. It means that each case is to be analyzed on an individual basis The benchmarks are precious. In the literature on the subject of university mergers, case studies are the most numerous, although at the same time there is no comparative analysis. In the induction process, large number of data on effective leadership is gathered, however it is still difficult to build a universal theory now Team leadership Team leadership in complex organizations such as universities plays a key role. Effective mergers of universities can only be carried out by committed, competent and flexible employee teams. The role of leaders is important because it involves making strategic decisions; however, with a significantly high level of organizational complexity of merger processes, the codecision makers are mostly team members. Team leadership focuses on self-learning teams that by cooperating with each other and with external entities develop a consolidation project Not always it is possible to build a team that meets all the requirements of an effective and efficient cooperation. Behavioral aspects make crucial role in the works of teams Critical perspective of leadership The processes of mergers at universities create changes that may have a negative impact on the objectivity of leaders. They may fall into the traps of autocracy, narcissism, oppressive treatment of employees, manipulation of people and treating the scientific and educational mission as a smokescreen Critical analysis allows leaders to combine ethical solutions with pragmatism necessary to implement organizational changes (Aasen and Stensaker, 2007) Source: Authors' own study   Figure 1 assumes that mergers between universities, just like in business, do not easily succumb to managerial control and project management, which enable full implementation of the objectives.

Elements Description
Relationships with the brand The reactions related to the association of the brand against the background of other brands are analyzed. Universities create relationships with graduates, students, employees and stakeholders, which leads to strengthening the university's brand Other assets related to the brand These are added features that the brand offers, for example certificates, accreditations, signed cooperation agreements. The name, logo and brand are strengthened by accreditations and certificates Loyalty It is manifested in the attachment to the university of students and employees that shape ties with the university. Increasing number of universities implement loyalty programs among students and graduates Brand awareness It involves: brand recognition, the number of associations with a given brand (in case of universities, these may embrace features like: very good quality of education, a recognized diploma in the labor market, good study conditions, qualified scientific and teaching staff ), but also for example a well-known sports team Perceived value of the brand The perception of the university as an institution that offers products and services of a corresponding quality is measured. Very often, the perception of the quality of education at a given university is a determinant of choosing this particular institution Source: Authors' own study based on Aaker (1991)

Mergers in higher education institutions
The effectiveness of the implementation of mergers between universities is conditioned by a number of supporting and inhibiting factors.
Supporting factors include: • relative homogeneity of the merging organizations (similar type of activity, similar founding or ownership structure); • origin of the merging organizations (lack of far-reaching cultural and legal differences if the universities come from one country); and • benefits of mergers that may occur for the institutions involved.
Inhibiting factors include: • conservatism of traditional academic cultures; • a strong ethos of academic professions often oriented toward maintaining the status quo; and • the dominance of the stakeholder model favoring the maintenance of compromise between groups of influence.
The merger process takes place at three levels: • Level I: change of university organization systems. The merging universities integrate or/and restructure their organizational systems. Strategic areas, structures and processes as well as cultural aspects of the merged organizations require alignment.
• Level II: influence of the closer environment on the merger's process. The dynamics of the consolidation process are directly impacted by: − competitors present in the country and in the world or new emerging competitors and institutions trying to fill the market niche by offering alternative educational offers; − cooperation networks consisting of national and international entities cooperating with universities; − internal stakeholders, i.e. academic staff, students and university administration; − external stakeholders, i.e. ministries, central and local government, employers; and − public policy, reflected in law, financing of universities and central and local programs.
• Level III: influence of the further environment on the merger's process. In this respect, the significance of the following key variables needs to be taken into account: − Economic factors, such as: living standards and unemployment rate. They have a significant impact on the motivation to start a merger. One can risk a statement that the deterioration of the country's economic situation should become a catalyst for a wave of mergers, especially in dispersed higher education systems. − Demographic variables related to fertility and the flow of human capital. They form the basis for assessing the demand for higher education at the national level. − Social factors related to the level and structure of scholarisation in a given country. Social patterns have a significant impact on decisions if to study and what educational direction to choose. It is wherein worth mentioning that the transformation from an elite to an egalitarian higher education system is becoming a global megatrend. The waves of systemic mergers, carried out, among others, in China, EU countries and the USA in the 1990s, were designed to better adapt to the mass and even universal model of education (Mao et al., 2009). − Cultural context that has a significant impact on consolidation processes, although the assessment of its impact is very difficult. Culture not only shapes the organizations themselves and the behavior of people in organizations, but also affects the 1480 MF 45,10/11 functioning of the entire education system and even the dynamics of the consolidation process. In countries with a high level of social capital, with developed civic culture, university mergers have a greater chance of success, although there are also examples that contradict this thesis. Successful centralist mergers in China prove that even in a society with a relatively low civic culture, it is possible to effectively implement university consolidation (Cai and Yang, 2016, pp. 71-85). On the other hand, some unsuccessful mergers in the UK and Australia prove that culture is only one of the variables in the complex mosaic of influence factors (Martin, 1994, pp. 83-91). − Scientific and technological environment: it is connected with the development of science requiring the reorganization of research units. The general tendency is to group together scientific units, which leads to the creation of a "critical mass" that allows to implement ambitious research projects and to develop renown scientific schools. New technologies emerge in the cooperation of the university with the industry and through the creation of spin-offs. New communication and network technologies also have a direct impact on conducting research (e.g. methods and laboratories) and education (e.g. on-line and distant education were motivational factors when merging even faraway campuses and schools). − Global variables related to the internationalization of science and higher education and the development of global competition. One of the key mechanisms to drive the wave of strategic mergers in the public university sector that is sweeping through the world is global rankings. The globalization is dominated by the English-speaking countries due to the international role of English and the scientific and educational position of universities. In many countries, mergers are carried out to promote the internationalization of universities by: opening joint, dual and double degree programs, as well as English-language programs, attracting foreign students and strengthening their academic exchange.
The growing numbers of publications, research and cooperation projects convince that mergers may lead to the implementation of many strategic goals and may affect both private and public universities. The proposed Conceptual Model of Universities' Mergers sheds some more light on this complex phenomenon. Understanding the context for the universities' mergers, realizing supporting and hindering factors, processes, structures and variables playing roles in the whole process can help plan future mergers more effectively.

The use of the conceptual model: a case of the Université Grenoble Alpes
The proposed Conceptual Model of Universities' Mergers can constitute a useful framework for analysis of the merger that resulted on January 1, 2016 in reuniting of three universities in France: Joseph Fourier University, Pierre Mendès-France University and Stendhal University, forming the Université Grenoble Alpes. The split was made in 1970s and the XXI brought a strategic decision to reunite the three from four institutions that previously constituted the University of Grenoble (except from Polytechnic Institute of Grenoble). Table I presents the key elements of the proposed model and its relevance to the merger of the three universities in Grenoble. The data for the description of the Université Grenoble Alpes were collected from May 2016 till mid-2017 in France through observations, interviews and collection of documents. The follow-up study visit to Grenoble was organized in May 2017 (Table VII). In summary, the Université Grenoble Alpes currently educates over 45,000 students and employs 5,500 employees in over 80 organizational units. The merger brought first results in the form of: intensification of scientific activities, improvement of organizational and managerial efficiency and generation of savings from consolidated processes and structures. Much effort was required from the employees to adapt to the new situation, which means that the results of scientific and educational activities should improve year by year. The interviews indicated As a result of strategic analyses, due diligence, research and negotiations, the following were agreed: the strategy and stages of the merger, the structure of the consolidated university, the new name and the authorities of the consolidated university Strategies In the restructuring process, a new strategy was created which was focused on cooperation with the society, innovation as well as internationalization and development of high-quality research and education within a comprehensive university. The mission underlined the growing role of international interdependence, innovation and interdisciplinary research. It also confirmed that the heart of the university's activity was the combination of education and research Structures The university's rector was chosen (Lise Dumasy, the rector of Université Stendhal for three tenures) as well as Vice-rectors who represented the merging universities New organizational structure was created. In total, 24 units such as departments, schools and institutes were distinguished in the organizational structure after consolidation. As part of the matrix structure, 6 large disciplinary research units were established The university developed also a number of major improvement projects, which in turn led to a faster development of research. EQUIPEX enabled furnishing of laboratories in order to allow to undertake the most innovative research. IDEFI focused on the innovative education of students and researchers. LABEX allowed to establish and develop scientific cooperation with the best world centers and researchers. Infrastructure was also being developed, using public-private partnerships. A center of creativity and innovation was built to serve interdisciplinary research and education focused on innovation. Other examples were the health education and research center as well as the buildings of the law and social sciences departments Deeper structural changes were introduced also in university-wide service units that cover various functional areas, such as finance and accounting, human resources, international cooperation, education and university life, research and innovation, information systems, logistics, cooperation with the environment and others Processes The merger was preceded not only by a long period of close and formalized cooperation, but also by a six-year planning and preparation process for merger implementation at the strategic and operational level In the consolidation process, a new information system was implemented, covering not only the university management, but also the entire scientific output of employees and units, international relations and education quality management. New websites for the university and all units were also created, which was coupled (continued ) A deepened specialization between research and teaching staff was introduced. Evaluation and remuneration and development systems rewarding higher productivity (performance based systems) were introduced. The positions, salaries and development opportunities of employees from different disciplines were differentiated. New branding and new identity were gradually being developed Influence of the closer environment on the merger's process A wide consultation process was carried out, followed by communication, both among employees as well as students and other stakeholder groups

Competitors
Each merging university had different focus in their areas of studies so they saw themselves as complementary rather than competitive entities. The merger was seen as strengthening of all universities by creating an entity that could become a stronger regional, national and international player and competitor Cooperation network The merger of the universities into the Université Grenoble Alpes offered a new path that other entities and cooperation networks started considering as a strategic option Internal stakeholders The staff participated in the preparation of the merger through systematic meetings in the framework of inter-university integration teams for several years The Polytechnic Institute of Grenoble, although initially discussed the merger, retreated in the course, mainly due to social resistance. The Polytechnic staff did not know if the merger would bring them sufficient benefits to compensate for the partial loss of independence External stakeholders Negotiation and integration teams were established where representatives of all universities as well as central and local authorities and external stakeholders participated Public policy The ministry and local authorities favored consolidation and actively supported it. The merger required some specific laws and regulations that were prepared. The merger process was co-financed under the Ministry's programs Influence of the further environment on the merger's process Economic The conviction that the merger can lead to higher economic rationality and efficiency Demographic There was an opinion that merger offers new developmental opportunities for employees. At the same time there was a fear for human resources reduction. The fact that the merging universities were complimentary in the areas of study limited the lay-offs scope Social The concept of creating the Université Grenoble Alpes involved the assumption that also other universities from the Rhone-Alpes region can be involved in the merger. The discussions were held, however, in the end no more than the three institutions decided to participate in the consolidation process. The public entities, authorities at local and national levels as well as employers saw the merger as an opportunities-generating undertaking for the region. New entity was expected to be followed by the creation of new identity among the stakeholders Cultural Merger was perceived as a mixture of opportunities and threats. New branding contributed to developing new identity of the institution and its stakeholders Scientific and technological One of the consolidation motifs in France was to build strong links between universities and enterprises, which was supposed to fuel economic, scientific and technological development. There was an expectation that Grenoble merger could result in the "Silicon Valley" type of solution Global The beginning of the 21st century in France brought "Shanghai shock," which was associated with the poor positions of French universities in international rankings. Achieving the "critical mass" in science through merger was to be a springboard to becoming a world-class academic institution with high international recognition Table VII. 1483 Mergers in higher education institutions improvement of consolidated university management through: more effective strategic management, real emphasis on international cooperation and cooperation with the society, effective marketing communication and more advanced financial management and accounting. Employees mention faster and more efficient operation of administrative units compared to the situation before the merger. UGA implements mechanisms that dynamize scientific activity, which has improved its position in national and international rankings in the last two years. There is also a gradual increase in the number of English-speaking students and programs, which favors the internationalization of the university. The UGA educational offer was expanded, and at the same time unified and modernized.

Summary
Merger processes have produced positive results in many countries in the form of: increased effectiveness in conducting research (obtained grants, publications and implementations), higher recognition of universities (positions in international rankings) or optimization of the universities activity costs. At the same time, some negative effects of mergers may appear. They may be related to lower than expected effectiveness, resistance of the academic community, increase in the degree of universities bureaucratization and the weakening of academic culture. Moreover, frequently, consolidation processes do not fully achieve their goals, many of merger attempts finish as failures. The conclusion is that consolidations between universities can give positive results; however, the merger process should be effectively managed.
Based on the analysis of the literature and observations gathered during the case study data gathering the following ten principles of effective management of the universities' mergers may be proposed: (1) analysis of the potential synergy effect and complementarity of the university in line with the properly conducted due diligence process; (2) verification whether the merging organizations match in their identity and whether the change will bring status benefits; (3) flexible and data-based analysis and strategic planning of the consolidation process, including controlling ("milestones," operational plans); (4) communication and commitment of the main merger stakeholders who should be aware of potential benefits; (5) taking into account the influence of culture and human capital management, enabling the satisfaction of the staff, students and other stakeholders from the consolidation process; (6) transformational leadership that implements change and emphasizes benefits as well as identity change; (7) effective management of the brand, PR and marketing communication processes, both inside and outside the organization; (8) implementation of restructuring processes of consolidation project management methods, structural changes and the use of management concepts, experiments and research on the consolidation of universities; (9) consideration of key areas of transformation, including systems: strategy, people management, IT, marketing, as well as research and education-related processes at all stages of implemented changes; and (10) development of the vision and concept of an entrepreneurial, flexible, innovative and competitive university.

1484
MF 45,10/11 These principles cover the entire process of a merger: from planning, through implementation, to integration. The proper application of these ten principles might contribute to more effective management of university mergers and a greater success of institutions that decide to take this strategic decision. The proposed Conceptual Model of Universities' Mergers offers a framework for better understanding of the merger context and its variables. However, it is also important to mention the limitations related to the wider applicability of the model. Mergers belong to complex organizational processes. They constitute a radical change which the entire organization is subjected to in a relatively short time. The processes accompanying mergers are multidimensional and entangled in cultural and social factors that cannot be fully controlled causing that the trajectory of revolutionary cultural change happening in universities becomes partly indeterministic. Therefore, the created model needs to be viewed with these limitations in mind.

Empirical research on mergers'
leverage dynamics and post-merger integration duration Yao Cheng BLCU Youth Talent Development Programme, Business School, Beijing Language and Culture University, Beijing, China

Abstract
Purpose -The purpose of this paper is to examine the effects of the post-merger integration duration on acquiring firms' leverage behavior before and after a merger, using a dynamic model in which full merger benefits cannot be consumed at the instant of a merger, but rather after a pre-specified post-merger integration period. Design/methodology/approach -This paper presents a dynamic model and empirical tests that describe the impact of the post-merger integration period on the capital structure dynamics of the acquiring and target firms prior to a merger and during the post-merger integration period. By incorporating costs associated with the post-merger integration period, the model can provide new implications for the leverage behavior around the merger.
Findings -Empirical tests support the model implications by showing that the longer the expected postmerger integration process, the less likely the acquirer will structure the financing of the combined firm in a manner that increases firm leverage. Since integration takes time to complete, an acquirer tends to retain financial flexibility during the integration process by assuming lower levels of debt when determining the capital structure of the merged entity.
Originality/value -The model generates new implications related to acquiring firms' leverage dynamics along with the method of payment choice. The analysis of the duration of the post-merger integration period extends both the theoretical and empirical literature that tacitly assumes that the merger-related synergy is realized immediately at the merger date. This is the first model in the literature that assumes that both the acquiring and the target firms can change their capital structure overtime, which allows us to analyze both the financing structure and the merger timing. Previous empirical studies also ignore the integration period in the analysis of the method of payment choice and leverage behavior around mergers. The model in this paper can be extended along a number of dimensions. Keywords Mergers' leverage, Payment method, Post-merger integration duration Paper type Research paper 1. Introduction

Research objectives
The integration of two merging firms takes time to complete. We refer to this time lag between the initiation of the merger and its completion as the "post-merger integration duration" (PMID). This means that the synergy gains from the merger cannot be captured instantly at the merger date but rather only after the firms go through an integration/ transition period. This period is often associated with temporarily higher costs and an elevated uncertainty about the merger success. Only after the post-merger integration is complete, can the newly merged firm fully enjoy the merger benefits. This merger integration period receives a great deal of attention among practitioners, but is largely ignored in both the theoretical and empirical literature on mergers. Annual reports of acquiring firms frequently discuss the challenges and difficulties firms may face during the integration period, such as possible problems in maintaining key employees, consolidating and rationalizing corporate infrastructures and eliminating redundant processes. There have been numerous reports of culture clashes, confusion and internal disruptions leading to declines in employee and customer satisfaction and loss of profitability. For these reasons, companies that expect a longer post-merger integration period may face temporarily higher expenses spread over a longer period coupled with higher operating risk. In turn, this expectation of long integration duration should directly affect an acquirer's decision when to merge, how to pay and most importantly the capital structure of the newly merged firm.

Content
This paper examines the effects of the PMID on acquiring firms' leverage behavior before and after a merger, using a dynamic model in which full merger benefits cannot be consumed at the instant of a merger, but rather after a pre-specified post-merger integration period. The model is built in continuous time with an infinite horizon framework that describes the leverage dynamics of two firms: the acquirer and the target. Each firm continuously generates earnings which depend on the price of its own product and its respective fixed production costs and taxes, which depend on the level of debt. The model implies that because earnings do not increase instantly upon the merger, there is no immediate need for additional tax shields, which can explain why firms that expect a longer integration period tend to use less debt. It should be stressed that the presence of the PMID is the necessary condition for optimal ratios being lower for the periods immediately following the merger.
The empirical section of the paper documents evidence that supports the model implications that the duration of the post-merger integration process significantly affects the leverage behavior of the merged firm before and after the year of merger. Starting with the universe of mergers that took place between 1999 and 2017, we select mergers in which we can gather data on the expected time for merger-related gains to materialize. Using this method, we come up with a sample of 3,120 mergers in which we can create the variable for the expected integration duration. For the sample we create a variable that measures the PMID. Specifically, for each merger, we manually read each 10-K filing, 8-K filing and merger-related proxy statements on Edgar Online and news stories from FACTIVA and search for information about the estimated timeline of cost savings and/or revenue enhancements. Because the model endogenizes both the capital structure decisions and the merger timing, it can also offer a rationale for several time-series observations around mergers.

Innovation points of the research
The model generates new implications related to acquiring firms' leverage dynamics along with the method of payment choice. The analysis of the duration of the post-merger integration period extends both the theoretical and empirical literature that tacitly assumes that the merger-related synergy is realized immediately at the merger date.
This is the first model in the literature that assumes that both the acquiring and the target firms can change their capital structure overtime, which allows us to analyze both the financing structure and the merger timing. For example, the related model in Morellec and Zhdanov's (2008) study presents the interaction between financial leverage and bidding contest. In their model, capital structure plays a role of a commitment device, which only determines the outcome of the acquisition contest, where the merger-related profits are realized immediately at the instant of merger. By contrast, our model predicts that the leverage of the winning bidder should be below the industry average and that acquirers should lever up after the takeover. These implications resonate with the implications of our model but they are based on different mechanisms.

Post-merger integration duration
Previous empirical studies also ignore the integration period in the analysis of the method of payment choice and leverage behavior around mergers. For example, Martin (2016) examines the motives underlying the method of payment in acquisitions and finds that the likelihood of stock financing increases with the acquirer's growth opportunities and higher pre-acquisition market and acquiring firm stock returns. In our tests, we control for the factors mentioned above and demonstrate that the expected integration duration is not subsumed by those variables implying that it has its own power in explaining the choice of leverage and merger financing method.
Compared with previous studies (e.g. Samitas and Kenourgios, 2007;Samitas et al., 2008;Huang et al., 2015), the model in this paper can be extended along a number of dimensions. First, we can assume that the initial bid for the target firm may not necessarily result in a successful merger due to possible bidding contests or uncertainty with respect to shareholders' approval. Also, one can endogenize the premium that the acquirer pays to the target firm. These types of extensions, while interesting, will not change the main implications of the paper.

Research framework
The rest of the paper is organized as follows. In Section 2, we describe the model which offers new implications as well as economically plausible explanations to several stylized facts about the observed capital structures of mergers. In Section 3, we calculate and demonstrate valuation of target firm, merged firm and acquiring firm. In Section 4, we describe the base case parameters of the model. In Section 5, we test several predictions derived from the model. Section 6 concludes the model and empirical evidence on mergers' leverage and post-merger duration.

Description of the model 2.1 The earnings of the acquiring firm and the target firm
The acquirer and the target firm continuously generate earnings by selling products whose unit market price, p 1 and p 2 , respectively, evolve through time in the manner described by the following stochastic process: dp 1 p ¼ gÀa 1 ð Þdt þd 1 dW 1 ; (1) where W 1 and W 2 are the Wiener processes under the risk-neutral measure Q; δ 1 and δ 2 the instantaneous volatility coefficients; r the risk-free rate, which is assumed to be constant; and α 1 and α 2 ( W0) the convenience yields. The two Weiner processes are correlated, where corr (W 1 , W 2 ) ¼ ρ, and the instantaneous correlation ρ is assumed to be constant. The firm's instantaneous net earnings before interest and taxes (EBIT) are assumed to equal p 1 −c 1 , and p 2 −c 2 , respectively, for the acquirer and the target firm, where c 1 and c 2 are constants (⩾0) that describe the continuous fixed production costs.

The net earnings of the merged firm
As soon as the acquirer, which currently sells its product at price p 1 , merges with the target firm, which sells its product at p 2 , the merged firm begins selling both products of two firms and generate earnings of p 1 +p 2 . The level of the combined production cost of the merged firm will depend on the length of the post-merger "integration period" T, over which the cost c 3 (τ) will gradually decline from c 3 (T), at the initiation of the merger, to the level of c 3 (0), at the end of the integration period, where τ(0 oτ⩽T) is remaining time until the integration 1490 MF 45,10/11 period is over. The combined production cost is assumed to be described by c 3 t ð Þ ¼ cUe yt=T , 0 oτ⩽T. At the initiation of the merger, τ ¼ T, the combined production cost is assumed to be higher than the total production costs of the two firms, i.e., c 3 T ð Þ ¼ cUe yT 4 C 1 þ C 2 . When the integration period is over, τ ¼ 0, the production costs decline to the level below the their combined pre-merger level, i.e., c 3 0 ð Þ ¼ coC 1 þC 2 , and will stay at that level thereafter. Thus, during the integration period, the merged firm's instantaneous net EBIT is p 1 +p 2 −c 3 (τ), if 0 oτ⩽T, and p 1 þp 2 Àc thereafter.

Corporate taxes and dividends
For all firms, the net earnings after debt payments are taxed continuously at a constant corporate rate λ, and periodic debt coupon payments are tax deductible. The firms use their earnings to meet debt obligations and pay taxes, with any residual being paid out as a dividend. The firm's instantaneous tax obligation equals (λ)·[p 1 −d 1 ], (λ)·[p 2 −d 2 ], and (λ)·[p 1 +p 2 −c 3 (τ)−d 3 ] for the acquirer, the target and the merged firm, respectively, where d 1 , d 2 and d 3 are the respective coupon payments. Thus, the firms instantaneous after tax dividends are the following:

The debt structure and recapitalization
Similar to the assumptions in the studies of Fischer et al. (1989), Leland (1998) and Titman and Tsyplakov (2015), we assume that the acquiring and the target firms and the merged firm issue perpetual coupon debt with a periodic coupon payment d 1 , d 2 and d 3 , respectively. Following Leland (1998) and Tsyplakov (2008), we assume that firms can increase their debt ratio but not to decrease it. We assume that the firms can instantly increase its debt ratio by simultaneously repurchasing its total outstanding debt at its market value D i and issuing new debt with greater face value and a greater periodic coupon. For accounting purposes, we assume that the face value of the debt is F 1 ¼ (d 1 /r), and F 2 ¼ (d 2 /r), respectively, F 3 ¼ (d 3 /r), which are the values of perpetual debt with a periodic coupon payment d 1 , d 2 and d 3 discounted at the risk-free rate. Specifically, when the firm changes its debt level by replacing old debt that has a face value of F i ¼ (d i /r), with new debt that has a coupon b d i , and face value b F i ¼ ð b d i =rÞ, the firm has to pay transaction costs of: where C debt   Table I reports the decision to merge/not to merge of the acquirer as a function of its earnings and the earnings of the target firm, for three different levels of its own debt payments d 1 ¼ 120, d 1 ¼ 160 and d 1 ¼ 180, as well for the target firm's debt payment of d 2 ¼ 0, d 2 ¼ 40, d 2 ¼ 80, d 2 ¼ 120 and d 2 ¼ 160. For the variable values that are outside this range, the acquirer will not start the merger.
In Table I, we hold the earnings of the target firm constant ( p 2 ¼ 800). The results in the table imply that the acquirer does not initiate the merger if the target firm's debt is above 120 for any level of the target's earnings p 1 . This is because the acquirer is less willing to guarantee the target's debt if it is of a relatively large size. For example, if the earnings of the target firm are p 2 ¼ 800, the optimal leverage is 30 percent (d 2 ¼ 251). At such a level of debt, the acquiring firm will not initiate the merger. If the target firm's leverage is 17 percent or lower (debt payment d 2 o85) which is significantly below its optimal level, then the acquirer would initiate the merger. This implication of the model suggests that firms that are underleveraged are more likely become a target of acquisitions. In other words, the leverage of the target firm observed at the instant of the merger tends to be lower than the optimal leverage of similar firms. This is due to the optimal timing on the part of the acquirer that will optimally choose (or wait for) to acquire the target firm that has a lower leverage.

4.2.3
The optimal leverage choice of the acquiring firm at the merger time. The optimal leverage decision of the acquiring firm is complex and it depends on the value of the merger option. Table II reports the optimal leverage of the acquirer and an otherwise identical firm that has no option to merge, as a function and its earnings p 1 , keeping the earnings ( p 2 ¼ 800) and the debt level (d 2 ¼ 0) of the target firm constant. For example, if the otherwise identical firm has no option to merge and its earnings are p 1 ¼ 1,200, the optimal debt level (i.e. the leverage to which the firm would instantly recapitalize to) corresponds to 42 percent. For comparison, the acquirer's optimal leverage is about 20 percent. This result demonstrates that, in periods prior to an acquisition, an acquiring firm will optimally choose lower leverage ratios relative to its peers. This lower leverage of the acquirer tends to be interpreted in the related literature as being driven by the managerial empire-building incentives. The model shows that, due to an anticipated delay in capturing merger-related gains, an acquiring firm should rationally preposition itself by choosing lower leverage immediately before initiating a merger.

Empirical tests
In the following sections, we test several predictions derived from the model with a particular focus on the effect of integration duration on the leverage decision of the merging firms. We exclude 121 cases in which we do not have enough information to calculate one-year market-adjusted pre-merger acquirer abnormal returns and another six cases in which sales are missing in the probit regression. In the Tobit model, we exclude 9 cases in which the inverse Mill's ratios are missing, and 12 cases in which we cannot calculate the pre-merger leverage deviation.

Sample description
The sample of mergers comes from the Securities Data Company's (SDC) US Mergers and Acquisition Database. We select domestic mergers with announcement dates between 1999 and 2017[1]. We only consider mergers in which acquirers end up with full control of the target firms and we also require that the acquirers control less than 50 percent of the target before the merger announcement. We further require that the merger is completed; the acquirer and the target are not in financial (SIC 6000-6999) and regulatory (4900-4999) industries; the acquirer is a public firm so that we have necessary Compustat data to compute relative size of target to acquirer market size; the deal value relative to the market             Table I.

1498
MF 45,10/11 value of the acquirer is at least 18 percent [2], to eliminate the influence of very small acquirers and very small deals, we require that acquirers' CPI-adjusted market value of assets are larger than $10m, and the transaction value of deals in our sample is at least $19.5m. Our requirements leave us with a final sample of 3,527 successful mergers between 1999 and 2017. We obtain managers' discussions about the merger deal from several data sources including annual 10-K filings, 8-K filings and merger-related proxy statements on Edgar Online and news stories from FACTIVA. In these data sources, we search for the information about managerial expectations as to when merger-related gains are expected to materialize in order to create the PMID variable. For each deal, we conduct a keyword search through the entire text of annual 10-K filings, 8-K filings and merger-related proxy statements from the fiscal year of merger announcement to the fiscal year of merger completion and identify all sections of text in which the integration process and expected merger gains are discussed. Using this approach, we are able to construct the PMID variable in 420 mergers (13.5 percent) from our initial sample. PMID ranges from 0.4 to 19 years with sample median of 2.7 years. There is some clustering with respect to the PMID variable.
Panel A of Table III reports the number of mergers in each PMID group. We find 155 cases in which PMID ¼ 2 years. Therefore, we split the PMID subsample into three relatively even groups of PMIDo2 years, PMID ¼ 2 years and PMIDW2 years, as shown in the table. Panel B of Table III shows that horizontal mergers consist of 53 percent of the PMID sample, and vertical mergers and diversifying mergers are 20 and 27 percent, respectively. Panel C of Table III describes the industry distribution for mergers in which we are able to create PMID. For each one-digit Standard Industrial Classification (SIC) code, we report the number of mergers during the entire sample period. While firms in the manufacturing industry have more observations with PMID reported, our sample is not dominated by any particular industry. Observations in each PMID group are distributed relatively evenly across industries.

Empirical findings
In this section we first report summary statistics on acquirer and target firm characteristics for the whole sample and subsamples in which we are able to create PMID. Next, we examine the association between PMID and market leverage at the year of merger.

Post-merger integration duration
Third, we examine the impact of PMID on the choice of method of payment. Fourth, we verify our results by examining the magnitude of the change in leverage around the merger that is attributable to PMID. Finally, we examine the leverage dynamics during the post-merger integration period. 5.2.1 Acquirer and target firm characteristics. Table IV provides median values for selected acquirer and target firm characteristics for the initial sample and subsamples across three PMID groups.
No. of obs.
Fraction of sample Panel A shows that the market value of acquirers paying with equity is smaller than that of acquirers paying with cash or paying with a mix of cash and equity. The higher ratios of market-to-book assets for acquirers paying with equity suggest that these acquirers have larger growth opportunities, consistent with the studies of Martin (2016) and Harford (2015). In addition, acquirers paying with equity tend to be less profitable and hold more cash. Furthermore, Panel A also shows that market leverage is much lower for equity deals. Given that acquirers paying with equity tend to have larger investment opportunities, it is likely that they hold larger cash balances and are less leveraged (i.e. Titman and Wessels, 2015). Panel A also indicates that the market value of acquirers for the PMID sample is larger than that of the sample of acquirers in which we are not able to construct PMID. This is not surprising as there is likely much more media/news coverage of larger firms. However, the market-to-book ratio of acquirers in which we can create PMID is slightly lower than that of acquirers in which we cannot construct PMID. Moreover, acquirers that have PMID are likely to be more profitable, hold less cash and have higher leverage. They are in general more similar to acquirers paying with cash or with mix of cash and equity. Panel A also reports acquirer firm leverage deviations. The pre-merger leverage deviation for all deals is negative; and more negative for stock acquirers. Firms in which we are able to create PMID are on average less underleveraged than firms in which we cannot create PMID. Specifically, the median leverage deviation is −0.06 and −0.14 for the PMID sample and non-PMID sample, respectively.
Panel B presents summary statistics regarding target firm characteristics. Particularly, market size and relative size of target to acquirer are both larger for deals in which we can construct PMID. Additionally, target market leverage is higher in deals in the PMID sample.
Panel C provides information about acquirer characteristics across different PMID groups. The median market value of acquirers' assets increases monotonically as PMID becomes longer. For example, the median market value of acquirers in PMID o2 years group is approximately $950m, while the median market value of acquirers in PMID W2 years group is approximately $5bn. In addition, the market-to-book ratio increases monotonically across PMID groups, but the differences are only marginally significant. Furthermore, Panel B shows that market leverage, defined as total book debt scaled by total market assets is 0.27 for the PMIDo2 years group, 0.39 for the PMID ¼ 2 years group and 0.40 for the PMID W2 years group. Moreover, the firms with shorter PMID tend to have a larger leverage deficit prior to the merger. 5.2.2 Market leverage at the year of merger. In this section, we examine whether PMID is a factor in determining the leverage behavior at the year of the merger. We follow the capital structure literature and regress firm leverage on PMID and a number of factors that have been documented to impact capital structure. It is again worth noting that since we are collecting our data and creating PMID from electronic financial filings as well as news searches on FACTIVA we are undoubtedly bound to have larger firms in our PMID sample as it is more likely that news coverage is of large established firms as opposed to smaller ones. In light of this, we study the impact of PMID on leverage determination by implementing a two-stage Heckman selection model to control for the propensity to find managements' statements of synergy description and thus our ability to create the PMID variable for larger firms and larger deals. Table V reports the results of the two-stage Heckman selection model. In the first stage, we estimate a probit model to explain whether acquirers report PMID or not. The dependent variable is a dummy variable set equal to 1 if the acquirer reports PMID, and 0 otherwise. Since acquirer managers tend to discuss PMID in news stories when the merger is announced, we include pre-announcement fiscal year-end acquirer size and market-to-book ratio in the regression. To control for the possibility that 1501 Post-merger integration duration acquirers are more likely to discuss PMID when they conduct acquisitions of a larger target, we include the relative size of the target to the acquirer as an independent variable. We also create a "merger type" dummy variable to control for the likelihood that the propensity to report PMID is influenced by whether the proposed merger is horizontal, vertical or diversifying. Finally, we include pre-announcement acquirer abnormal return, change in the index of leading economic indicators, specifically the change in S&P 500 index or the change in Moody's BAA-rated bond yield during the 12 months prior to the month of merger announcement, public status of the target and announcement year dummies. In the second stage, we use an OLS model to predict market leverage at the year of merger completion. The primary explanatory variable of interest in the regression is PMID and our hypothesis is suggests that market leverage will be negatively associated with PMID. To control for acquirer size, we include the natural logarithm of CPI-adjusted sales. Additionally, we also control for the level of pre-merger leverage level by including the acquirer and target value-weighted market leverage ratio at one year prior to the merger. We further control for the market-to-book ratio, profitability, tangibility, R&D, a dividend dummy variable set equal to 1 if the firm pays a cash dividend, and two dummy variables set equal to 1 if the firm has a net operating loss carry forward or investment tax credit. Except for the level of pre-merger leverage, all other variables in the regression are contemporaneous. Finally, we control for industry effects by including the firms' two-digit SIC code. The result in Table V shows that PMID is significantly negatively associated with the leverage of the combined entity at the year of merger. Furthermore, the magnitude of the coefficient on PMID indicates that this relation is not only statistically significant, but also economically significant. The coefficient on this variable suggest that if the post-merger integration is expected to take one additional year to complete, the market leverage at merger completion is reduced by approximately 2.92 percent. The results in Table V also show a significant relation between leverage and other relevant control variables. As expected, a significant positive relation exists between the pre-merger acquirer-target valueweighted market leverage and the leverage at the year of merger completion. 5.2.3 Fraction mergers paid for with equity. Table VI presents the results of a two-sided Tobit model to explain the fraction of the deal paid for with equity. The dependent variable in the Tobit model is truncated at 0 and 1. To control for the probability that we are able to find merger-related news articles (managerial statements) to construct PMID for certain deals as opposed to others, we also include the inverse Mill's ratio estimated from the previous probit model in the regression as a regressor. Our intuition suggests that the fraction of a deal paid for with equity should be positively related to PMID. That is to say, the leverage of the newly merged firm will be lower if acquirers use more equity and less cash (debt) to finance a deal.
To control for acquirer size we include the natural logarithm of CPI-adjusted sales (in 1999 dollars). Additional control variables include the pre-merger market value-weighted acquirer and target market leverage and pre-merger market leverage deviation. Following Martin (2016), we control for market timing on the part of acquirer management (i.e. acquirer management using overvalued shares as currency in order to purchase the target) by including the market-adjusted 12-month pre-merger acquirer stock return. Specifically, we control for the change in the Standard & Poor's 500 index or the change in the yield on Moody's BAA-rated bonds. In addition, we include the relative size of the target to the bidder since the larger size of the target could increase the bidder's likelihood to pay for the target with equity. Furthermore, we control for the public status of the target as this could influence the method of payment because of more information asymmetry regarding the valuation of private or subsidiary targets as these targets face a higher "liquidity discount" on the M&A market (i.e. Berkovitch and Narayanan, 2010). Finally, we also include acquirers' pre-merger cash balance in the regression.

Post-merger integration duration
We implement our analysis on our initial sample and our PMID sample. The coefficient of pre-merger year leverage deviation is positively and significantly associated with the fraction of the deal paid for with equity. However, this variable loses statistical power in our PMID sample. Consistent with our intuition, the results from show a positive association between PMID and the fraction of the merger paid for with equity. This is consistent with our hypothesis that acquirer managers are less likely to conduct leverage increasing financing of the merger if they expect post-merger integration takes time to complete. 5.2.4 Change in market leverage resulting from mergers. Our previous results demonstrate a negative relation between market leverage after the merger completion and PMID, and also a positive relation between PMID and the fraction of the deal paid for with equity. In this section, we further assure our results by analyzing the change in market leverage around the year of merger.
Table VII presents results of regressions that explain the change in leverage from year t−1 to year t (i.e. one year before merger to the year of the merger). Other than the variables included in the regression of Table VI, we also include the change in optimal leverage in each specification. To show that PMID has additional power in explaining the leverage change before and after a merger, we include the merger-induced change leverage in the regression. In line with our hypothesis, we expect a negative relation between PMID and the actual change in market leverage from year t−1 to year t. 5.2.5 Leverage dynamics during the post-merger integration period. We define the remaining PMID (RPMID) as PMID minus the number of years elapsed after a merger; therefore, RPMID ranges from 0 to initial PMID. We apply a firm fixed effects model to analyze the extent to which our PMID and RPMID variables affect the leverage dynamics during the integration period, controlling for a number of conventional variables.
Table VIII reports the regression results. As expected, market leverage is negatively and significantly associated with our PMID and RPMID variables. The magnitude of the coefficients implies that the relationship is not only statistically significant, but also economically significant. A one year increase in PMID leads to about a 6.9 percent decrease in a firm's leverage during the integration period, and a one year increase of the RPMID results in about a 2.5 percent decrease in a firm's market leverage during that period.
Whole sample PMID sample PMID sample (1) (2)   Therefore, the evidence in this section supports our hypothesis that merged firms tend to maintain a higher degree of financial flexibility by keeping a lower level of leverage during the integration process if merger-related synergies take more time to materialize. 5.2.6 Robustness of the results. Our results remain quantitatively similar if we include year dummies or if we use three-digit SIC code industries or if the PMID is 1 or 3 years. We also implement the study on our PMID sample without the Heckman two-stage specification. The coefficient estimation of PMID is also statistically and economically significant. We construct the fraction of the deal paid for with equity using SDC deal transaction reports. Furthermore, our results are similar if we include pre-merger market leverage of the acquirer instead of weighted average leverage of the acquirer and the target.

Conclusion
This paper provides a model and empirical evidence that the leverage behavior around mergers and during the post-merger integration period is affected by the acquirer's expectation about the length of the PMID. The model offers new implications as well as economically plausible explanations to several stylized facts about the observed capital structures of mergers. The model substantiates the argument that due to anticipated delays in capturing merger-related gains, an acquiring firm should rationally preposition itself by choosing lower leverage immediately before initiating the merger and should keep a lower degree of leverage during the post-merger integration period.
Empirical tests support the model implications by showing that the longer the expected post-merger integration process, the less likely the acquirer will structure the financing of the combined firm in a manner that increases firm leverage. Since integration takes time to complete, an acquirer tends to retain financial flexibility during the integration process by assuming lower levels of debt when determining the capital structure of the merged entity. We document that the market leverage of a newly merged firm is negatively associated with the length of the integration period. Our results also suggest that, other things being equal, acquires are more likely to finance the deal with equity when they expect a longer integration period. Finally, we show that the duration of the integration period can help explain leverage dynamics during the post-merger integration period. Overall, the model indicates that the PMID is negatively associated with the market leverage of newly merged firms at the time of merger completion and during the integration period. Further, acquirer managers are more likely to use equity to finance a merger when the integration duration is likely to be lengthy.