Investor sentiment and timely loss recognition

Hong Kim Duong (Department of Accounting and Legal Studies, Salisbury University, Salisbury, Maryland, USA)
Michael Schuldt (Department of Accounting and Legal Studies, Salisbury University, Salisbury, Maryland, USA)
Giorgio Gotti (Department of Accounting and Information Systems, The University of Texas at El Paso, El Paso, Texas, USA)

Review of Accounting and Finance

ISSN: 1475-7702

Publication date: 13 August 2018

Abstract

Purpose

The purpose of this paper is to investigate the impact of investor sentiment on timely loss recognition by examining a sample of firms for the period 1988-2015.

Design/methodology/approach

The authors use the accruals-based model of Ball and Shivakumar (2005) and a sentiment measure in their primary analysis. Supporting analyses include an extension of Simpson (2013) using an abnormal accruals analysis with subsamples of firms with bad news, the use of a Khan and Watts (2009) quarter firm-level measure of conservatism and an investigation of the monitoring role played by financial analysts.

Findings

The study finds that managers strategically report more losses in high sentiment periods than in low sentiment periods. This loss timing behavior results in an average 37.8 per cent increase in the acceleration of loss recognition. This study additionally finds a negative correlation between investor sentiment and abnormal accruals when managers are reporting bad news, and that a greater number of financial analysts following a firm curtails managers’ acceleration of loss recognition in high sentiment periods.

Originality/value

This study contributes to the corporate disclosure literature by showing that managers strategically recognize losses, and such behavior is more prevalent in high sentiment periods. Managers take advantage of prevailing investor sentiment to accelerate losses in high sentiment periods to mitigate market penalties from reporting bad news.

Keywords

Citation

Duong, H., Schuldt, M. and Gotti, G. (2018), "Investor sentiment and timely loss recognition", Review of Accounting and Finance, Vol. 17 No. 3, pp. 383-404. https://doi.org/10.1108/RAF-07-2016-0104

Download as .RIS

Publisher

:

Emerald Publishing Limited

Copyright © 2018, Emerald Publishing Limited


1. Introduction

Accounting researchers have long been interested in whether corporate financial reports present faithful information to capital providers. In the Conceptual Framework – Statement of Financial Accounting Concepts No. 8 published in 2010, the Financial Accounting Standard Board (FASB) has emphasized that faithfulness is an important qualitative characteristic of a useful financial report:

To be a perfectly faithful representation, a depiction would have three characteristics. It would be complete, neutral, and free from error.”[…] “A neutral depiction is without bias in the selection or presentation of financial information. A neutral depiction is not slanted, weighted, emphasized, deemphasized, or otherwise manipulated to increase the probability that financial information will be received favorably or unfavorably by users.

Conceptual Framework - Statement of Financial Accounting Concepts No. 8, Financial Accounting Standards Board (FASB) (2010, pp. 17-18).

However, corporate disclosure literature has documented that financial reports are not always faithful and neutral. Prior research has provided evidence that managers manipulate the depiction of financial information by strategically timing their firms’ earnings reports and/or reports of other financial information to increase the probability that the reports will be received more favorably, or in some cases less harshly, by investors.

Bad news tends to be reported when investors pay less attention, to minimize market penalties, such as stock price decreases. Patell and Wolfson (1982) provide early evidence that managers take advantage of investor inattention by releasing quarterly earnings bad news after the markets close. Subsequent papers have provided additional evidence of investor inattention, and found that Friday earnings announcements tend to have worse news than those released on other days (Penman, 1987; Damodaran, 1989; DellaVigna and Pollet, 2009; Bagnoli et al., 2006). Research has also found that managers voluntarily disclose bad news in a manner that minimizes the market penalty by strategically timing management forecast announcements during periods of lower market attention (Doyle and Magilke, 2015).

Investor sentiment is a phenomenon that biases investors’ expectations of future firm performance; sentiment influences how managers’ report earnings and other financial information. Several studies have documented such effects. Bergman and Roychowdhury (2008) show that in high sentiment periods when investors are optimistic and overvalue firms, managers reduce the horizon of their future earnings forecasts; further, when investors are pessimistic and undervalue firm future performance, managers increase their forecast horizon to “walk-up” investors’ beliefs. Brown et al. (2012) find that managers are more likely to report adjusted earnings (“pro forma” earnings), especially one that exceeds the Generally Accepted Accounting Principles (GAAP) earnings number, as investor sentiment increases. Mian and Sankaraguruswamy (2012) find that asymmetric market responses to earnings news vary with investor sentiment, and specifically, that stock price sensitivity to bad earnings news is lower in high sentiment periods than during low sentiment periods. Their findings are consistent with the notion that as sentiment increases, investors place a lower penalty on cash flow decreasing news embedded in earnings announcements. The inevitable question is whether managers take advantage of market sentiment to strategically report bad news. Our study aims to answer this research question and fill this gap in the extant literature.

We postulate that, in an attempt to mitigate market penalty on bad earnings news, managers strategically report more losses in high sentiment periods than in low sentiment periods. Managers are insiders and have better knowledge about their own firms’ performance than outside investors. If bad earnings news must be reported at some point in time, managers strategically report bad news when the market penalty is low. Examples of bad earnings news include bad debt expenses, asset write-offs and accrued liabilities. In an accrual-based earnings system, bad news is recorded in income through managerial estimation of income-decreasing accruals. Ball and Shivakumar (2005) develop an accruals-based model to capture the timely recognition of losses. According to their model, accruals are a source of timely loss recognition, and thus accruals are predicted to positively correlate with contemporaneous negative cash flows. To test our hypothesis, we introduce a measure of investor sentiment (SENT) as an interactive explanatory variable in the Ball and Shivakumar (2005) model. Following prior research, we use the investor sentiment index developed in Baker and Wurgler (2006) to compute the quarterly-average Investor Sentiment Index (SENT). If managers accelerate reporting losses in high sentiment periods, we expect to find a stronger positive correlation between accruals and negative cash flows in high sentiment periods. Consistent with our hypothesis, we find that the positive correlation between accruals and negative cash flows is significantly higher in high sentiment periods. On average, the loss recognition acceleration in high sentiment periods is 37.8 per cent higher than in low sentiment periods. The results are robust to controls for size (SIZE), growth (measured by the market-to-book ratio) (MTB), leverage (LEV) and return on equity (EARN).

Our study is related to Simpson (2013), which examines whether managers manage earnings through abnormal accruals in high investor sentiment periods. Simpson (2013) documents a positive relation between abnormal accruals and investor sentiment and concludes that managers use abnormal accruals to increase earnings and avoid negative earnings surprises when investor sentiment is relatively high and report more conservatively when investor sentiment is relatively low. Simpson (2013), however, does not examine the impact of investor sentiment on timely loss recognition, and our findings suggest that managers accelerate loss recognition in high investor sentiment periods. To reconcile the facially contradictory findings of our study and Simpson (2013), we partially replicate Simpson (2013) and investigate the relation between investor sentiment and abnormal accruals. Consistent with Simpson (2013), we find that investor sentiment is positively related to abnormal accruals in our sample. However, our study focuses on the timing and recognition of losses and its relation to investor sentiment; thus, we extend Simpson (2013) by examining the relation between abnormal accruals and investor sentiment in three subsamples of firms reporting bad news: firms reporting negative earnings; firms missing financial analysts’ forecasts and firms reporting negative operating cash flows. Our regression results for all three subsamples show that investor sentiment is negatively correlated with abnormal accruals; thus, conditional on reporting bad news, investor sentiment is negatively associated with abnormal accruals. Thus, our findings complement the findings in Simpson (2013); conditional on bad news reporting, managers accelerate timely loss recognition in high investor sentiment periods to avoid negative market reaction.

We further check the robustness of this study’s empirical analysis by examining the influence of investor sentiment on firm-level conservatism. Following Khan and Watts (2009), we create a firm-quarter conservatism score (KW C_Score) and then regress this KW C_Score on our investor sentiment variable. The regression results report that investor sentiment is negatively associated with firm-level conservatism; in general, managers report less conservatively in high investor sentiment periods than in low investor sentiment periods. However, the regression results using a subsample of loss firms and a subsample of firms reporting negative operating cash flows show a positive association between investor sentiment and the KW C_Score. This supplemental analysis corroborates our prior results and suggests that during high investor sentiment periods, managers strategically accelerate loss recognition if their earnings or cash flows are negative.

In further robustness tests, we use the Michigan Consumer Sentiment Index as an alternative measure of investor sentiment and obtain empirical results consistent with our prior findings using the measure developed by Baker and Wurgler (2006). Additionally, we investigate the role of financial analysts in monitoring managers’ reporting behavior. Yu (2008) finds that firms with a larger financial analysts following evidence lower earnings management. We re-examine our prior findings with the Ball and Shivakumar (2005) model using two subsamples, one comprising the top quartile and the other the bottom quartile, based on the number of analysts following the firm. The interaction coefficient representing the association between accruals and negative cash flows is positive in both subsamples but only statistically significant in the bottom quartile subsample. Our findings complement Yu (2008), suggesting that financial analysts serve as external monitors to managers and that firms followed by a greater number of financial analysts are less likely to accelerate loss recognition in high investor sentiment periods.

The study contributes to the extant literature in several important aspects. First, our findings contribute to the corporate disclosure literature by showing that managers strategically recognize losses, and that this behavior is more prevalent during high investor sentiment periods. This finding is consistent with prior evidence that financial reports are not always neutral, and therefore, are not always a faithful presentation of corporate performance. Second, the study provides evidence regarding the consequences of investor behavioral biases by documenting that managers take advantage of a specific market inefficiency, investor sentiment, and strategically time loss recognition to mitigate market penalties. Third, our findings complement the findings of Simpson (2013) by showing that conditional on loss reporting, managers use abnormal accruals to accelerate loss recognition in high sentiment period to avoid negative market reaction. Finally, our study complements Yu (2008) by showing that financial analysts serve as external monitors to mitigate managers’ opportunistically reporting behaviors. This study should be of interest to shareholders, corporate directors, researchers, policy makers and others concerned with understanding the determinants of corporate earnings disclosure, including the associations between firm disclosures and market-wide, investor-driven expectations.

The remainder of the paper is organized as follows. In the next section, we review the relevant literature and develop our hypothesis. In Section III, we provide our research design and data. Section IV contains empirical results, and the results for robustness tests are described in Section V. We summarize and conclude our study in Section VI.

2. Related literature and hypothesis development

Timely loss recognition is considered an enduring and important element of financial reporting quality (Ball and Shivakumar, 2005). Timely loss recognition is central to the definition of conditional conservatism. Khan and Watts (2009) define this form of conservatism as “the asymmetric verification threshold for gains versus losses”, with such threshold being higher for the recognition of good news (gains) as opposed to bad news (losses). Thus, losses are generally recognized on a timelier basis than gains[1]. Timely loss recognition mitigates agency problems associated with managers’ actions; it incentivizes managers to quickly address and limit operating investments with expected economic losses, and assists lenders by providing timely information regarding contractual constraints, such as leverage, investment and dividend restrictions (Basu, 1997; Ball and Shivakumar, 2005).

In addition to agency considerations, other factors create incentives for managers to time their earnings-based disclosures, particularly disclosures of bad news. Managers hasten disclosure of bad news in earnings forecasts to avoid litigation and reputational costs (Skinner, 1994; Baginski et al., 2002). Managers delay good news and accelerate bad news in their earnings forecasts in the periods immediately preceding stock option grants to lower the exercise price of their options (Aboody and Kasnik, 2000). However, while managers may have incentives to accelerate the disclosure of bad news, in other instances, they face opposing incentives to withhold bad news. Managers often possess private information concerning firm operations and future prospects which may be used to evaluate their performance; good news disclosures can result in continued employment and more favorable compensation contracts, and bad news disclosures can result in salary and bonus reductions, as well as personal wealth loss through termination (Nagar, 1999). Disclosing bad news can negatively impact manager equity compensation arrangements, such as stock options and incentive stock grants (Bergstresser and Phillipon, 2006; Cheng and Warfield, 2005).

Kothari et al. (2009, p. 273) provide evidence that “managerial incentives to withhold bad news dominate managerial disclosure behavior” and that managers strategically withhold bad news up to a threshold where it becomes too costly or difficult to continue withholding. As disclosure of bad news can be costly, managers have incentives to avoid or minimize its impact. Managers can gamble on subsequent favorable events that could obfuscate the good news with the bad (Kothari et al., 2009; Graham et al., 2005). Alternatively, given an inevitable timely loss disclosure threshold, managers may use investors’ behavioral biases to manipulate the timing of bad news disclosures and minimize their cost.

The behavioral finance and accounting literature has provided evidence that investors’ behavioral biases can affect earnings disclosure decisions. These studies provide evidence that managers take advantage of investors’ behavior biases and manipulate earnings-related disclosures to increase the probability that positive financial information will be received more favorably, and that unfavorable news will have lessened impact. Several such biases have been examined. One such bias is the investor inattention/distraction effect, which documents that managers strategically time the release of negative information during high distraction/low attention periods, such as before weekends (i.e. on Fridays) and before national holidays (Niessnar, 2015; DellaVigna and Pollett, 2009; Bagnoli et al., 2006), and after trading hours (Doyle and Magilke, 2015; Genotte and Trueman, 1996; Patell and Wolfson, 1982).

Another more recently documented investor bias which has generated researcher attention is investor sentiment. Baker and Wurgler (2007, p. 1) defines investor sentiment as “a belief about future cash flows and investment risks that is not justified by the facts at hand”. Market sentiment has discernable and significant effects on the cross-section of future stock returns by causing regular patterns of mispricing (Baker and Wurgler, 2006). When sentiment is high, investors are more inclined to speculate and over-value stocks (by either over-estimating the magnitude of the cash flows or underestimating their risk); the reverse is true when sentiment is low. This misvaluation reverses in the future, thus creating a negative relation between current sentiment and future returns (Mian and Sankaraguruswamy, 2012). Two accepted proxies used in prior investor sentiment research are the monthly sentiment index developed by Baker and Wurgler (2006), and the Michigan Consumer Sentiment Index. The Baker and Wurgler index is a composite sentiment index based on the first principal component of five sentiment proxies[2]. The monthly Michigan Consumer Index is consumer confidence index published by the University of Michigan and Thomson Reuters based on at least 500 telephone interviews in the USA in which participants are asked questions about their outlooks on the economy[3].

Many recent finance studies have examined market sentiment and its relation to corporate events (e.g. equity issues, dividend payouts and mergers and acquisitions). However, only limited research has examined how managers strategically adjust their financial disclosures in response to varying market sentiment. Baker et al. (2007) suggest that managers will act to increase firms’ appeal to investors based on the prevailing sentiment. Such action could include modifying or timing financial disclosures. Several accounting studies have examined the relation between financial reporting decisions and market sentiment. Bergman and Roychowdhury (2008) demonstrate that managers react strategically by reducing the frequency of their longer-term earnings forecasts during high sentiment periods to maintain their firms’ prevailing optimistic valuations, and by increasing the frequency of such forecasts during low sentiment periods to boost investor optimism. Brown et al. (2012) provide evidence of a positive relation between managers’ propensity to issue pro-forma earnings disclosures (especially such disclosures which exceed the GAAP amount) and investor sentiment. Simpson (2013) extends this line of study and investigates the relation between managers’ use of earnings management and investor sentiment; she finds a positive association between managers’ use of income-increasing abnormal accruals and investor sentiment.

This study further extends this literature stream by focusing on managers’ disclosures of negative news, and whether such bad news disclosures are strategically timed with market sentiment. Consistent with the opportunistic managerial sentiment view set forth in Brown et al. (2012) and reiterated in Simpson (2013), we assume managers are cognizant of, and can identify instances of high and low sentiment for their firm and exploit such biases for their benefit. Further, we note that prior literature suggests a relation between investor sentiment and bad news reporting. Mian and Sankaraguruswamy (2012) examine the impact of sentiment on the sensitivity of stock prices to earnings surprises and finds that bad news generates a smaller stock price reduction during high sentiment periods than low sentiment periods. Managers have a generally strong aversion to disclosing bad news due to its attendant consequences (Kothari et al., 2009) and may not always have the ability to either obscure the bad news with serendipitously timed good news, or time-alter the bad news through the use of earnings management techniques. Given an inevitable required bad news disclosure, we postulate that managers will time to report more losses in high sentiment periods than in low sentiment periods.

This study frames this postulation in terms of an accrual-based earnings system and an accrual model. Dechow et al. (1998) demonstrates that working capital accruals make earnings timelier than operating cash flows and a better forecast of future cash flows. Ball and Shivakumar (2005) extend Dechow et al. (1998) by noting a second role for accruals – the timely recognition of gains and losses. Bad news – economic losses such as bad debt expenses, asset write-downs or other accrued charges “are more likely to be recognized on a timely basis as unrealized (i.e. non-cash) accrued charges against income” (Ball and Shivakumar, 2005, p. 95). If managers must report losses via accruals, they will seek to time their use of accruals to minimize any market penalty. Thus, we hypothesize that the positive correlation between accruals and contemporaneous negative cash flows is stronger in high investor sentiment periods. Our hypothesis is stated in an alternative form as follows:

H1.

The positive correlation between current accruals and negative cash flows is stronger in high investor sentiment periods than in low investor sentiment periods, ceteris paribus.

3. Method and data sample

3.1 Method

We start with the accrual-based timely loss recognition model of Ball and Shivakumar (2005) to examine the relation between accruals and operating cash flows, specified as follows:

(1) ACCRt=β0+β1DCFOt+β2CFOt+β3DCFOt×CFOt+vt

The dependent variable, ACCR, is the firm total accruals, computed as the difference between earnings and operating cash flows, CFO is the firm operating cash flows, DCFO is a dummy variable equal to 1 when CFO is negative and 0 otherwise, and subscript t corresponds to firm year. Both accruals and operating cash flows are scaled by the beginning book value of total assets. The role of accruals in the Dechow et al. (1998) model is to reduce the noisiness in operating cash flows and create a smoother earnings variable; the implication is that accruals and cash flows are contemporaneously negatively correlated (Dechow et al., 1998). Ball and Shivakumar (2005) outline a second role of accruals when developing their accruals-based test of loss recognition. The timeliness of gain or loss recognition is based on expected, but unrealized, cash flows, and therefore is accomplished through accruals. “It follows that timely recognition of economic gains or losses is a source of positive correlation between accruals and current period cash flows, thereby attenuating the negative correlation predicted by the Dechow et al. (1998) model” (Ball and Shivakumar, 2005, p. 93). In their model, the positive correlation between cash flows and accruals is greater in the case of losses; thus, timely loss recognition mitigates the negative correlation between accruals and operating cash flows. The model’s implications are that β2 < 0 (as per Dechow et al., 1998) and β3 > 0 (timely loss recognition for negative cash flows).

To examine our hypothesis, we modify the annual Ball and Shivakumar (2005) by converting the model to a firm-quarter basis (i.e. subscript t thus correspond to the fiscal quarter) and extend the model by introducing SENT as an interactive variable. The extended model is specified as follows:

(2) ACCRt=β0+β1DCFOt+β2CFOt+β3DCFOt×CFOt+β4SENTt+β5SENTt×DCFOt+β6SENTt×CFOt+β7SENTt×DCFOt×CFOt+vt
where SENT is a quarterly-average investor sentiment index (SENT) computed by averaging the monthly investor sentiment index from Baker and Wurgler (2006) for each quarter. For example, for firm A with fiscal quarter ended March 31, 2010, SENT is computed as the average of the investor sentiment indices in January, February and March of 2010. The extended model allows an asymmetric relation between accruals and cash flow levels that differs between high and low sentiment periods. Consistent with our hypothesis, we expect to find a positive β7.

3.2 Data sample

Our sample consists of all non-financial, non-utility and non-government US firms from January 1988 to October 2015 from the merged the Center for Research in Security Prices (CRSP) and Compustat North America databases. Financial firms (SIC codes 6000-6999), utilities (SIC codes 4900-4999) and government entities (SIC codes greater than 8999) are excluded from the sample to avoid corporate decisions dictated by specific regulations. We restrict our sample to December-fiscal year end firms to avoid overlapping quarterly observations and further limit our sample to include only firms with ordinary common shares with CRSP code 10 and 11. We further exclude firms with missing data on total assets and cash flows from operations. Investor sentiment indices are obtained from Jeffrey Wurgler’s website[4]. We limit the sample period from January 1988 to October 2015 for two reasons. First, the SFAS 95 became effective in 1987 requiring firms to report cash flows. Hribar and Collins (2002) show that cash flows estimated from changes in balance sheet accounts are subject to biases; thus, we use cash flows from the cash flow statement as reported by firms from 1988. Second, the investor sentiment indices are available between January 1965 and October 2015, so our sample period ends on October 2015. To mitigate the influence of outliers and data coding errors, we winsorize all continuous variables at the 1st and 99th percentiles. The final sample has 169,929 firm-quarter observations. Details regarding variable construction and sources are presented in the Appendix Table AI.

4. Empirical results

4.1 Descriptive statistics and univariate test results

Table I reports the descriptive statistics for our sample, including the mean, median, standard deviation (StdDev) and first (Q1) and third (Q3) quartile of the variables used in the main tests.

Table II shows the correlation matrix for the variables used in our main tests for the period beginning January 1988 and ending October 2015. The upper (lower) right triangle presents the Pearson (Spearman) pairwise correlation coefficients. The Pearson and Spearman correlation coefficients between CFO and ACCR are significantly negative (−0.869 and −0.888, respectively). The correlation coefficients between DCFO and ACCR are significantly positive (0.594 and 0.644, respectively).

4.2 Multivariate test results

The multivariate results of our main regression model, Model (2) are reported in Table III.

As a robustness check and further enhancement to our equation (2) baseline model, we control for SIZE, MTB and LEV, and their interactions with our main variables. Khan and Watts (2009) specifically use these firm characteristics to capture variation in firm-level conservatism[5]. Further, following Basu (1997) and Dechow (1994), we include EARN and its interactions with our main variables as an additional firm characteristic and performance measure[6]. Consistent with the results reported in Dechow (1994) and Ball and Shivakumar (2005), our regression results across all six regressions in Table III show that the association between accruals and cash flows is significantly negative (β2 < 0) and that economic losses receive timelier recognition than gains (β3 > 0). The coefficient on our test variable, SENT × DCFO × CFO (β7) is significantly positive in the regression results reported in Columns (2) to (6) of Table III, documenting a significant, enhanced relation between accruals and negative cash flows in high sentiment periods. In the regression results reported in Column 2 of Table III, β7 is 0.035 and β3is 0.084. The implication is that during high investor sentiment periods, on average, managers accelerate loss recognition by 42 per cent (0.035/0.084). Similarly, when controlling for SIZE (Column 3), managers increase loss recognition by 21 per cent in high sentiment periods. When controlling for MTB, LEV and EARN, loss recognition acceleration increases by 27, 45 and 54 per cent, respectively. Thus, loss recognition acceleration, as reported across the regression results in Columns 2 to 6 of Table III, increases on average by 37.8 per cent in periods of higher sentiment. The results are consistent with our prediction and provide support to our hypothesis: managers take advantage of investor’s biases in belief about firm future performance and strategically time reporting more losses during high investor sentiment periods to minimize negative market reaction and penalties.

5. Additional analyses

In this section, we examine the robustness of our prior findings by first comparing and contrasting our findings with Simpson (2013), and then re-examining the relation between investor sentiment and conservatism using the conservatism score (KW C_Score) developed by Khan and Watts (2009). Next, we use an alternative measure of investor sentiment, the Michigan Consumer Sentiment Index, to address possible measurement errors associated with our main explanatory variable. Finally, we examine our hypothesis in contexts where firms are under increased scrutiny of financial analysts.

5.1 Investor sentiment and abnormal accruals

Simpson (2013) documents a positive relation between investor sentiment (measured by the Michigan Consumer Sentiment Index) and abnormal accruals and concludes that managers strategically time to use abnormal accruals to boost earnings to meet or beat analyst forecasts in high sentiment periods, and use abnormal accruals more conservatively during low sentiment periods. Simpson’s findings facially appear to contradict ours, as our study finds a positive association between accruals and timely loss recognition in high sentiment periods. However, Simpson’s study does not specifically address the potential association between abnormal accruals and sentiment under conditions where managers may have differing earnings management incentives (i.e. during periods of bad news). Further, our main analysis focuses on total accruals, as opposed to abnormal accruals. Thus, we partially replicate Simpson (2013) and consider the potential association between abnormal accruals and investor sentiment under circumstances involving disclosure of bad news. Our study’s results suggest a negative relation between investor sentiment and abnormal accruals, conditional on loss recognition. We examine this suggestion using a subsample of firms with negative earnings (the “loss firm subsample”), a subsample of firms that miss financial analysts’ forecasts (the “miss forecast subsample”) and a subsample of firms with negative operating cash flows (the “negative cash flows subsample”).

Following Simpson (2013), we estimate signed firm-quarterly abnormal accruals by using the Dechow et al. (1998)[7] model:

(3) ACCRit=α0+α1ΔREVit+α2PPEit+α3CFOit+eit
where ΔREVit is change in revenues, PPEit is property, plant and equipment and CFOit is the cash flows from operation. All variables and the intercept are scaled by the firm’s average total assets for that quarter. Abnormal accruals (AbACCR) are the residuals from regression Model (3). We regress abnormal accruals on investor sentiment and firm characteristics such as size (SIZE), market-to-book (MTB), leverage (LEV), return on equity (EARN), lags of abnormal accruals and macroeconomic factors such as industrial production growth (INDPRO), inflation (INFL) and growth in real gross domestic product (GDPGR). Additionally, we include the firm’s operating cycle (OPCYCLE) as a control variable, as Dechow (1994) finds that the length of a firm’s operating cycle is an important determinant in the size of the change in a firm’s working capital requirements. We include industry fixed effects, year fixed effects and dummies for Quarter 1 to Quarter 4. The estimation results are reported in Table IV.

Our regression results reported in Columns 1 and 3 are consistent with the results documented in Simpson (2013). The coefficient on SENT is positively significant (p-value < 0.01) in the regression using the total sample (Column 1, Table IV) and in the regression using a subsample of firms that meet or beat financial analysts’ forecasts (Column 3, Table IV). The results indicate that managers’ time reporting of higher abnormal accruals in high investor sentiment periods to boost earnings to meet or beat financial analysts’ earnings forecasts. However, the findings do not hold in the loss firms subsample (Column 2), the miss forecast subsample (Column 4) or the negative cash flows subsample (Column 5). The coefficients on SENT are significantly negative and equal −0.006, −0.002 and −0.009, respectively. The findings suggest that managers accelerate the reporting of losses using abnormal accruals in high sentiment periods when their firm earnings or operating cash flows are negative or when they miss analyst earnings forecasts. These findings supplement and support our main analysis, and further contribute to the extant literature by extending the findings reported in Simpson (2013) by showing a negative association between abnormal accruals and earnings in high sentiment periods when managers are compelled to recognize losses.

5.2 Investor sentiment and firm-quarterly conservatism score

Khan and Watts (2009) proposed a firm-year measure of accounting conservatism, noting that other measures either obscure the cross-section variation in conservatism of individual firms in an industry, or obscure the timing of conservatism of individual firm financial reports. Their methodology provides a simple, parsimonious method for estimating a firm-year measure of conservatism; further, the measure’s properties are consistent with conservatism metrics confirmed in the prior accounting literature (Khan and Watts, 2009). Their measure appears uniquely suitable for our supplemental analysis, as such measure can readily be used as a dependent variable in an empirical analysis involving investor sentiment. Following Khan and Watts (2009), we control for MTB, SIZE and LEV to avoid spurious associations between our sentiment variable and the firm conservatism measure.

Khan and Watts (2009) derive a year firm-level measure of conservatism. We follow the Khan and Watts (2009) methodology and create a quarter firm-level measure of conservatism. The Khan and Watts (2009) model is based on the Basu (1997) model of asymmetric recognition of gains and losses. In the Khan and Watts model, each coefficient in the Basu model is expressed as a linear function of firm size, market-to-book ratio and leverage as follows:

(4) EARNi=β1+β2Di+Ri(μ1+μ2SIZEi+μ3MTBi+μ4LEVi)+DiRi(λ1+λ2SIZEi+λ3MTB+λ4LEVi)+(δ1SIZEi+δ2MTBi+δ3LEVi+δ4DtSIZEi+δ5DtMTBi+δ6DtSIZEi)+εi
where EARN is earnings, R is monthly returns aggregated by quarter, D is a dummy variable that takes the value of 1 if R is negative and 0 otherwise, SIZE is the natural logarithm of market capitalization, MTB is the market-to-book ratio, and LEV is leverage, defined as the sum of long-term and short-term debts deflated by total assets. The KW C_Score equals (λ1 + λ2SIZEi + λ3MTBi + λ4LEVi). This is a firm-quarter measure of the incremental timeliness of loss recognition versus gain recognition. The KW G_Score measures the timeliness of gain recognition and equals (µ1 + µ2SIZEi + µ3MTBi + µ4LEVi). Following Khan and Watts (2009), we exclude from our sample firms with a price per share less than $1, and firms with negative book value of equity. We truncate observations in the top and bottom percentiles of EARN, SIZE, MTB and LEV. These restrictions leave us a sample of 47,547 firm-quarter observations. We then regress KW C_Score and then KW G_Score[8] on SENT and additionally control for firm characteristics and macroeconomic factors in our total sample, a loss firm subsample and a subsample of firms with negative operating cash flows. The regression results are reported in Table V.

With respect to the regression using KW C_Score as the dependent variable, the coefficient on SENT is significantly negative (coefficient = −0.002; p-value < 0.01) in the first regression using the full sample, indicating that firm conservatism is reduced during high investor sentiment periods. However, the coefficients on SENT are significantly positive in the regression using only a loss firm subsample (coefficient = 0.010; p-value < 0.01) and the regression using only firms with negative operating cash flows (coefficient = 0.013; p-value < 0.01), suggesting when faced with negative earnings or cash flows, managers time accelerating losses and report more conservatively in high investor sentiment periods. These results are consistent with our main analysis empirical findings.

We do not find significant results regarding the impact of investor sentiment on the timeliness of gain recognition (i.e. KW G_Score). The coefficients of SENT in Columns (4, 5 and 6) of Table V are insignificant.

5.3 Investor sentiment – Michigan Consumer Sentiment Index

Prior research on investor sentiment also uses the Michigan Consumer Sentiment Index to measure the market overall sentiment level (Bergman and Roychowdhury, 2008; Simpson, 2013). As an additional robustness check, we re-estimate our Model (2) using the Michigan Consumer Sentiment Index as an alternative measure for the overall market sentiment. The estimation results using this additional sentiment measure are reported in Table VI. We document results consistent with our previous findings[9]. The coefficient of the variable SENT × DCFO × CFO is positive and significant (coefficient = 0.011, p-value < 0.01).

5.4 The monitoring role of financial analysts

Our findings support the prediction that during high sentiment periods, managers time reporting more transitory losses. The motivation of this timing behavior is to mitigate market punishment for the reporting of bad news. Yu (2008) postulates that financial analysts play a monitoring role in mitigating managers’ opportunistic reporting behavior. His findings indicate that firms with a larger analyst following have significantly lower earnings management. Dechow et al. (2010) also find that firms with a larger financial analysts following are less likely to manage their earnings to protect their reputation. This implies that the presence of financial analysts could reduce managers’ inclination to time losses. Thus, we investigate whether the timing behavior of managers to report losses in high sentiment periods differs between firms with differing levels of financial analyst following.

We partition our sample into quartiles based on the number of analysts following the firm. The quartile partition results show that firms in the bottom quartile have on average 1.4 analysts, and firms in the top quartile on average have 13.6 analysts. We then re-estimate the regression Model (2) for the top and bottom quartile subsamples. The results are presented in Table VII.

The coefficient on SENT × DCFO × CFO is still significantly positive in the regression using the bottom quartile subsample (coefficient = 0.057; p-value < 0.01), which supports our prior main analysis results[10]. Interestingly, the coefficient on SENT × DCFO × CFO is positive but insignificant in the regression using the top quartile subsample (coefficient = 0.045; p-value = 0.36). These results suggest that financial analysts serve as market monitors and mitigate the strategic loss recognition behavior of managers in high investor sentiment periods.

6. Summary and conclusions

Prior research has investigated the effects of sentiment on various corporate activities, such as capital investment decisions, equity issuances and dividend policy. More recent research has focused on the relation between investor sentiment and corporate disclosure and has emphasized managerial discretionary disclosures. The recent disclosure literature has found that managers reduce the frequency of their longer-term earnings forecasts during high sentiment periods (Bergman and Roychowdhury, 2008), and that managers’ disclosure of pro-forma earnings metrics in press releases increases with the level of investor sentiment (Brown et al., 2012). Simpson (2013) extended this line of research and found that managers use positive abnormal accruals to overstate earnings during periods of high sentiment.

This study further extends this literature stream by focusing on conservatism and timely loss recognition, and its association with investor sentiment. We use a modified version of the Ball and Shivakumar (2005) accruals-based test of timely loss recognition and find that managers take advantage of investor sentiment to strategically accelerate loss recognition in high sentiment periods. When facing the inevitable prospect of bad news disclosure, managers increase loss recognition, on average, by 37.8 per cent to mitigate the negative market reactions to loss disclosure. These findings are robust to alternative econometric approaches and alternate sentiment and conservatism variable measurements, as well as controls for firm characteristics and for macroeconomic factors.

Our findings should inform policy makers, investors, directors and researchers concerned with how investor sentiment influences managerial reporting decisions. Regulators should be cognizant of market sentiment, and particularly discerning when examining the content and timing of reporting disclosures in periods of stock market exuberance “bubbles”. Similarly, external auditors and boards, particularly audit committee board members, should consider closer scrutiny of management disclosures during periods of high and low sentiment and be sensitive to the opportunistic timing of such disclosures in quarterly periods with unaudited financial data. In all such reporting circumstances, existing literature (Simpson, 2013; Hurwitz, 2018) highlights the need for increased attention from these monitors for firms with heightened sensitivity to market sentiment, notably small and younger firms, and firms with greater information asymmetry. Our study additionally cautions that opportunistic managerial behavior is not merely limited to earnings overstatement, but can including opportunistic timing of bad news disclosures. Future extensions of this research can include investigations of audit quality metrics and their relation to investor sentiment, as well as the reaction of regulators to loss disclosures in high versus low sentiment periods. Additionally, an examination of institutional investors’ reactions to loss disclosures in periods of high versus loss sentiment may provide further insights into the impact of sentiment on managerial as well as investor behavior.

Descriptive statistics

Variable Mean SD Q1 Median Q3
SIZE 6.25 1.80 4.95 6.14 7.40
MTB 1.75 1.89 0.63 1.12 2.10
LEV 0.22 0.21 0.03 0.19 0.35
EARN 0.03 0.04 0.01 0.03 0.05
CFO 0.03 0.10 −0.02 0.04 0.08
ACCR −0.03 0.08 −0.07 −0.03 0.01
DCFO 0.26 0.44 0.00 0.00 1.00
SENT −0.02 0.40 −0.18 −0.06 0.14
Notes:

This table shows descriptive statistics for 169,929 firm-quarter observations between January 1988 and October 2015. The mean, median, standard deviation (StdDev) and first (Q1) and third (Q3) quartiles are reported. SIZE is natural logarithm of market capitalization. MTB is the market-to-book ratio. LEV is leverage, defined as the sum of long-term and short-term debts deflated by total assets. EARN is the difference between revenue and the sum of cost of goods sold and selling, general and administrative expenses, scaled by market capitalization. CFO is cash flow from operations, deflated by total assets. ACCR is total accruals, computed as the difference between net income and operating cash flows, scaled by total assets. DCFO is a dummy variable taking the value of 1 if CFO is negative, and 0 otherwise. SENT is the average of three monthly Investor Sentiment Indices from Baker and Wurgler (2006) prior to the firm fiscal quarter end date. Investor Sentiment is a composite index of sentiment that is based on the common variation in five underlying proxies for sentiment: the closed-end fund discount, the number and average first-day returns on IPOs, the equity share in new issues and the dividend premium. Details of variable computation and data sources are included in the Appendix

Correlation matrix (Pearson top and Spearman bottom)

Variable SIZE MTB LEV EARN CFO ACCR DCFO SENT
SIZE 0.075 (0.000) 0.100 (0.000) 0.006 (0.013) 0.224 (0.000) −0.171 (0.000) −0.244 (0.000) −0.100 (0.000)
MTB 0.132 (0.000) −0.334 (0.000) −0.403 (0.241) −0.215 (0.000) 0.130 (0.000) 0.181 (0.000) 0.099 (0.000)
LEV 0.140 (0.000) −0.518 (0.000) 0.368 (0.000) 0.006 (0.013) −0.023 (0.000) −0.046 (0.000) 0.007 (0.944)
EARN 0.042 (0.000) −0.619 (0.000) 0.436 (0.000) 0.333 (0.100) −0.179 (0.000) −0.307 (0.109) −0.034 (0.840)
CFO 0.233 (0.000) 0.021 (0.000) −0.015 (0.000) 0.310 (0.000) −0.869 (0.000) −0.684 (0.000) −0.078 (0.000)
ACCR −0.177 (0.000) 0.027 (0.000) −0.023 (0.000) −0.185 (0.000) −0.888 (0.000) 0.594 (0.000) 0.055 (0.000)
DCFO −0.248 (0.000) 0.112 (0.000) −0.070 (0.000) −0.344 (0.000) −0.760 (0.000) 0.644 (0.000) 0.091 (0.000)
SENT 0.133 (0.000) 0.088 (0.000) −0.005 (0.029) −0.062 (0.001) −0.086 (0.00) 0.074 (0.000) 0.091 (0.000)
Notes:

This table presents the correlation coefficients for 169,929 firm-quarter observations between January 1988 and October 2015. The upper (lower) right triangle of the matrix presents Pearson (Spearman) correlations. The variables are the same as presented and reported in Table I; p-values are in parentheses. Details of variable computation and data sources are included in the Appendix

Investor sentiment and timely loss recognition

Dependent variable: ACCR
Variable Predicted sign (1) (2) (3) (4) (5) (6)
DCFO −0.002*** (3.07) 0.000 (0.00) 0.033*** (15.87) 0.007*** (7.76) −0.001 (−0.87) −0.002*** (−2.53)
CFO −0.786*** (−95.27) −0.753*** (−95.55) −0.605*** (−56.54) −0.785*** (105.23) −0.736*** (−74.69) −0.772*** (−122.02)
DCFO × CFO + 0.112*** (9.78) 0.084*** (8.35) 0.152*** (27.21) 0.101*** (10.49) 0.080*** (7.38) 0.061*** (6.50)
SENT 0.002*** (6.19) 0.003*** (8.29) 0.001*** (3.99) 0.002*** (6.16) 0.002*** (5.18)
SENT × DCFO −0.008*** (−8.28) −0.008*** (−10.79) −0.005*** (−5.50) −0.008*** (−8.31) −0.007*** (−7.65)
SENT × CFO −0.029*** (−6.74) −0.031*** (−8.15) −0.022*** (−5.70) −0.030*** (−6.51) −0.025*** (−6.10)
SENT × DCFO × CFO + 0.035*** (3.72) 0.032*** (4.70) 0.027*** (3.26) 0.036*** (3.94) 0.033*** (4.10)
SIZE 0.004*** (25.79)
SIZE × DCFO −0.005*** (−14.31)
SIZE × CFO −0.030*** (−21.06)
SIZE × DCFO × CFO +/− 0.012*** (4.71)
MTB 0.002*** (4.37)
MTB × DCFO −0.006*** (−9.37)
MTB × CFO −0.004*** (−2.57)
MTB × DCFO × CFO +/− 0.001 (0.48)
LEV −0.001 (0.28)
LEV × DCFO 0.002 (0.61)
LEV × CFO −0.152*** (−4.15)
LEV × DCFO × CFO + 0.134*** (3.51)
EARN 0.159*** (16.03)
EARN × DCFO 0.177*** (7.87)
EARN × CFO −0.005 (−1.29)
EARN × DCFO × CFO +/− −0.021 (−0.15)
Industry fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Quarters 1 to 4 dummies Yes Yes Yes Yes Yes Yes
Adjusted R2 76.30% 77.04% 77.72% 78.09% 77.21% 78.64%
N 169,929 169,929 169,929 169,929 169,929 169,929
Notes:

This table presents pooled regression results of regressing Accruals (ACCR) on operating cash flows (CFO), Investor Sentiment (SENT), a Dummy variable (DCFO) and their interactions on a sample of 169,929 firm-quarter observations between January 1988 and October 2015. The variables are the same as presented and reported in Table I. Standard errors are clustered by firm; t-statistics are reported in parentheses. Statistical significance at the 10, 5, and 1% level is denoted by *, ** and

***

, respectively. Details of variable computation and data sources are included in the Appendix

Investor sentiment and abnormal accruals reporting

Variable Dependent variable: AbACCR
Total sample Negative earnings subsample Meet or beat analyst forecasts subsample Miss analyst forecasts subsample Negative cash flows subsample
(1) (2) (3) (4) (5)
SENT 0.001*** (3.50) −0.006*** (−3.54) 0.003*** (2.84) −0.002** (−2.12) −0.009*** (−3.57)
SIZE −0.000 (−0.19) −0.006*** (6.38) −0.002*** (−6.96) −0.001 (−0.25) −0.006*** (−5.04)
MTB 0.000 (0.01) −0.002*** (−3.73) 0.004 (1.32) −0.001 (−1.41) −0.002** (−2.22)
LEV −0.009*** (−6.57) 0.001 (0.16) 0.007*** (2.91) −0.009*** (−3.08) −0.005 (−0.64)
EARN 0.072*** (8.88) 0.443*** (10.11) 0.027** (1.91) 0.211*** (15.10) 0.480*** (8.19)
OPCYCLE 0.000 (0.88) 0.000 (1.44) −0.000 (−0.10) 0.000 (1.13) 0.000* (1.66)
INDPRO 0.001*** (4.84) −0.001 (−0.76) −0.002*** (−3.38) −0.002*** (−2.87) −0.006* (−2.34)
INFL −0.0001*** (−4.33) −0.0001 (−1.49) 0.000 (0.22) 0.000 (0.13) 0.000 (0.34)
GDPGR −0.0002*** (−2.87) 0.0003 (0.80) 0.0003** (2.08) 0.0001*** (5.62) 0.0006 (1.01)
AbACCRt-1 0.371*** (69.09) 0.368*** (22.54) 0.394*** (41.86) 0.414*** (35.03) 0.377*** (16.77)
AbACCRt-2 0.057*** (12.32) 0.052*** (3.38) 0.062*** (7.71) 0.070*** (6.90) 0.066*** (3.02)
AbACCRt-3 0.016*** (3.68) 0.031** (2.27) 0.024*** (3.00) 0.015 (1.48) 0.041* (1.88)
AbACCRt-4 0.078*** (14.92) 0.023 (1.58) 0.080*** (9.82) 0.075*** (7.59) 0.023 (1.15)
Industry fixed effects Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes
Quarters 1 to 4 dummies Yes Yes Yes Yes Yes
Adjusted R2 18.67% 14.13% 17.34% 17.73% 16.62%
N 110,049 13,595 49,832 32,084 6,524
Notes:

This table presents pooled regression results of regressing abnormal accruals (AbACCR) on Investor Sentiment (SENT), firm characteristics and macroeconomic activities: growth in industrial production (INDPRO), inflation measured as the change in seasonally adjusted consumer price index (INFL), and gross domestic product growth (GDPGR) on a sample of 169,929 firm-quarter observations between January 1988 and October 2015. AbACCR is abnormal accruals measured as the residuals of the regression using the Dechow et al (1998) model. OPCYCLE is operating cycle, defined as the sum of account receivable turnover and inventory turnover. SIZE, MTB, LEV, EARN and SENT are the same as presented and reported in Table I. Standard errors are clustered by firm; t-statistics are reported in parentheses. Statistical significance at the 10, 5, and 1% level is denoted by

*

,

**

and

***

, respectively. Details of variable computation and data sources are included in the Appendix

Investor sentiment and conservatism

Variable Dependent variable KW C_Score Dependent variable KW G_Score
Total sample Loss firms subsample Negative cash flows subsample Total sample Loss firms subsample Negative cash flows subsample
(1) (2) (3) (4) (5) (6)
SENT −0.002*** (−3.36) 0.010*** (2.86) 0.013** (2.15) 0.0001 (0.18) −0.0002 (−0.16) −0.002 (−1.14)
SIZE 0.001*** (6.08) −0.052*** (−16.61) −0.050*** (−13.71) −0.012*** (16.42) −0.011*** (−17.90) −0.010*** (−11.94)
MTB 0.002*** (12.94) 0.011*** (8.79) 0.011*** (7.73) −0.005 (−19.96) −0.005*** (−10.12) −0.005*** (−9.40)
LEV −0.004*** (−3.57) 0.032** (2.12) 0.038*** (2.58) 0.026*** (16.96) 0.038*** (7.77) 0.045*** (8.10)
EARN −0.005 (−0.66) 0.388*** (2.41) 0.331* (1.86) 0.014** (1.99) −0.048* (−1.69) −0.081** (−2.06)
StdEARN 0.071*** (6.58) 0.260 (1.44) 0.059 (0.25) 0.042*** (2.97) 0.099** (2.13) 0.007 (0.11)
OPCYCLE −0.000 (−0.77) −0.000* (−1.95) −0.000** (−2.09) 0.000 (0.31) 0.000 (0.10) 0.000 (0.25)
INDPRO 0.0002*** (7.84) 0.003*** (5.77) 0.003*** (4.58) 0.0002*** (5.49) 0.0005*** (3.78) 0.0005*** (3.12)
INFL −0.0001*** (−7.98) −0.001*** (−3.69) −0.001*** (−3.47) 0.0004*** (30.95) 0.0003*** (5.64) 0.0003*** (4.87)
GDPGR −0.0014 (−15.04) 0.027*** (18.17) 0.023*** (12.68) 0.002*** (27.06) 0.001*** (4.53) 0.002*** (4.09)
Industry fixed effects Yes Yes Yes Yes Yes Yes
Year fixed effects Yes Yes Yes Yes Yes Yes
Quarter 1 to 4 dummy Yes Yes Yes Yes Yes Yes
Adjusted R2 28.53% 23.14% 20.27% 33.99% 26.42% 25.08%
N 47,457 5,034 3,169 47,457 5,034 3,169
Notes:

This table presents pooled regression results of regressing Khan and Watts (2009) C_Score (columns 1, 2 and 3) and Khan and Watts (2009) G_Score (Columns 4, 5 and 6) on Investor Sentiment (SENT) and firm characteristics on a sample of 47,457 firm-quarter observations (Columns 1 and 4) and a subsample of loss firms of 5,034 firm-quarter observations (Columns 2 and 5) and a subsample of negative cash flows firms of 3,169 firm-quarter observations (Columns 3 and 6) between January 1988 and October 2015. StdEARN is standard deviation of earnings. All other variables are the same as presented and reported in Table I; Standard errors are clustered by firm. t-statistics are reported in parentheses. Statistical significance at the 10, 5, and 1% level is denoted by

*

,

**

and

***

, respectively. Details of variable computation and data sources are included in the Appendix

Investor sentiment and timely loss recognition – alternative measure of investor sentiment

Variable Dependent variable: ACCR
Predicted sign Coefficient t-stat
Intercept −0.012*** −2.85
DCFO −0.035*** −7.41
CFO −0.892*** −32.57
DCFO × CFO + 0.179*** 7.28
SENT_Michigan 0.001*** 3.71
SENT_Michigan × DCFO 0.002*** 5.45
SENT_Michigan × CFO 0.008*** 3.42
SENT_Michigan × DCFO × CFO + 0.011*** 3.21
Firm fixed effects Yes
Year fixed effects Yes
Quarters 1 to 4 dummies Yes
Adjusted R2 54.7%
N 169,929
Notes:

This table presents pooled regression results of regressing accruals (ACCR) on operating cash flows (CFO), investor sentiment (SENT_Michigan), a dummy variable (DCFO) and their interactions on a sample of 169,929 firm-quarter observations between January 1988 and October 2015. SENT_Michigan is the average of three monthly Michigan Consumer Sentiment Indices from University of Michigan prior to the firm fiscal quarter end date. All other variables are the same as presented and reported in Table I. Standard errors are clustered by firm; t-statistics are reported in parentheses. Statistical significance at the 10, 5, and 1% level is denoted by *, ** and ***, respectively. Details of variable computation and data sources are included in the Appendix

Investor sentiment and timely loss recognition – monitoring role of financial analysts

Dependent variable: ACCR
(1) (2)
Top quartile Bottom quartile
Min of number of analysts: 8 Min of number of analysts: 1
Max of number of analysts: 47 Max of number of analysts: 2
Mean of number analysts = 13.6 Mean of number analysts = 1.4
Median of number analysts = 12 Median of number analysts = 1.0
Variable Predicted sign Coefficient t-stat Coefficient t-stat
DCFO −0.003** (−2.05) 0.003*** (2.80)
CFO −0.811*** (−113.3) −0.763*** (−87.0)
DCFO × CFO + 0.190*** (6.84) 0.070*** (8.24)
SENT 0.001* (1.95) 0.004*** (6.45)
SENT × DCFO −0.005*** (−2.43) −0.011*** (−6.59)
SENT × CFO −0.010* (−1.78) −0.046*** (−6.27)
SENT × DCFO × CFO + 0.045 (0.91) 0.057*** (4.81)
Industry fixed effects Yes Yes
Year fixed effects Yes Yes
Quarter 1 to 4 dummies Yes Yes
Adjusted R2 82.18% 81.59%
N 19,589 42,048
Notes:

This table presents pooled regression results of regressing Accruals (ACCR) on operating cash flows (CFO), Investor Sentiment (SENT), a dummy variable (DCFO) on a sample of 19,589 firm-quarter observations (Top quartile) and on a sample of 42,048 firm-quarter observations (Bottom quartile) between January 1988 and October 2015. All variables are the same as presented and reported in Table I. Standard errors are clustered by firm; t-statistics are reported in parentheses. Statistical significance at the 10, 5, and 1% level is denoted by

*

,

**

and

***

, respectively. Details of variable computation and data sources are included in the Appendix

Variable definition, computation and data sources

Variable Computation Data source
AbACCR Abnormal accruals, measured as the residuals of the regression using Dechow et al. (1998) model Compustat North America, Fundamental Quarterly
ACCR (niqoancfy)/atq Compustat North America, Fundamental Quarterly
(Net Income – Operating Cash Flows)/Total Assets
EARN (revtq – cogsq − xsgaq)/(cshoq * prccq) Compustat North America, Fundamental Quarterly
(Total Revenue – Cost of Goods Sold – Administrative and General Expenses)/(Common Shares Outstanding * Share Price)
CFO oancfy/atq Compustat North America, Fundamental Quarterly
(Operating Cash Flows/ Total Assets)
DCFO Equal 1 if CFO < 0, and 0 otherwise
SIZE Log(cshoq * prccq) Compustat North America, Fundamental Quarterly
Log(Common Shares Outstanding * Share Price)
MTB (cshoq * prccq)/atq Compustat North America, Fundamental Quarterly
(Common Shares Outstanding * Share Price)/Total Assets
LEV (dlttq + dlcq)/atq Compustat North America, Fundamental Quarterly
(Long-term Debts + Short-term Debts)/Total Assets
OPCYCLE [(rect + rectt−1)/2]/[revt/360] + [(invt + invtt−1)/2]/[cogs/360] Compustat North America, Fundamental Annually
(average account receivable/sales/360) + (average inventory/cost of goods sold/360)
SENT 1/3(SENTtm + SENTtm-1 + SENTtm-2) www.stern.nyu.edu/∼jwurgler
where tm corresponds to month, and SENTtm is the monthly sentiment index developed by Baker and Wurgler (2006)
SENTtm is the first component of a Principal Component Analysis of five underlying proxies for market sentiment: the closed-end fund discount, the number and average first-day returns on IPOs, the equity share in new issues, and the dividend premium
SENT_Michigan 1/3(SENT_Michigantm + SENT_Michigantm-1 + SENT_Michigantm-2)
where tm corresponds to month, and SENT_Michigantm is the monthly sentiment index. This index is constructed monthly based on at least 500 telephone interviews in the US in which participants are asked questions about their outlooks on the economy. Three main focuses of the interview questions are (1) how consumers view the prospects of their own financial situation; (2) how they view the prospects of the economy in short term; and (3) how they view the prospects of the economy in long term
The Federal Reserve bank of St. Louis, Economic Research
https://research.stlouisfed.org/fred2/series/UMCSENT/downloaddata
INDPRO Growth in industrial production Federal Reserve Bank of St. Louis
INFL Inflation, measured as the change in seasonally adjusted consumer index Federal Reserve Bank of St. Louis
GDPGR Gross domestic product growth Federal Reserve Bank of St. Louis
KW C_score A measure of firm-quarter conservatism based on Khan and Watts (2009) model. Details about the derivation are in Section 5.2 Compustat North America, Fundamental Quarterly
KW G_Score A measure of firm-quarter good news timely recognition based on Khan and Watts (2009) model. Details about the derivation are in Section 5.3 Compustat North America, Fundamental Quarterly

Notes

1.

Ball and Shivakumar (2005) clarify that conditional conservatism involves timely loss recognition, which reflects a contemporaneous economic loss being reflected in accounting income, as contrasted with unconditional conservatism, which reflects an accounting bias toward low book values of stockholders’ equity (i.e. low book values of assets among possible alternatives, high values for liabilities, deferral of revenues and acceleration of expenses [Watts and Zimmerman, 1986]).

2.

The five proxies are as follows: the dividend premium, which is the difference in average market-to-book ratios between dividend payers and non-dividend payers; the closed-end fund discount; the number of initial public offerings (IPOs); the average first-day returns on IPOs; and the share of equity issues in total equity and debt issues (Baker and Wurgler, 2006).

3.

The three main focuses of the interview questions are: how consumers view the prospects of their own financial situation; how they view the prospects of the economy in short term; and how they view the prospects of the economy in long term (https://research.stlouisfed.org/fred2/series/UMCSENT/downloaddata).

5.

Firm size (SIZE) could be positively or negatively related to accruals and conservatism. Larger firms tend to have a richer information environment and thus decreased demand for conservatism; however, larger firms are likely to have more segments, divisions and complexity, which induces a higher demand for conservatism. Firms with a high MTB ratio (growth firms) tend to have greater information asymmetry, higher agency costs and a greater demand for litigation conservatism; however, such firms also tend to be unregulated, implying a lower demand for conservatism. More levered firms (LEV) tend to have higher contracting demand for conservatism.

6.

Firms with a higher return on equity (EARN) may have a higher taxation demand for conservatism due to higher profitability; however, high ROE firms may have reduced agency conflicts between shareholders and lenders, suggesting a lower contracting demand for conservatism.

7.

We obtain similar results when using Kothari et al. (2005) performance matched model to estimate abnormal accruals.

8.

We would like to thank an anonymous reviewer for this suggestion.

9.

Although not presented in Table VI, we re-estimated the regression in Table VI by controlling for SIZE, MTB, LEV and EARN and their interactions with CFO and DCFO and obtained robust results. We would like to thank the editor for this suggestion.

10.

Although not presented in Table VII, our results remain unchanged when including SIZE, MTB, LEV and EARN as control variables.

Appendix

Table AI

References

Aboody, D. and Kasnik, R. (2000), “CEO stock option awards and the timing of corporate voluntary disclosures”, Journal of Accounting and Economics, Vol. 29 No. 1, pp. 73-100.

Baginski, S.P., Hassell, J.M. and Kimbrough, M.D. (2002), “The effect of legal environment on voluntary disclosure: evidence from management earnings forecasts issued in US and Canadian markets”, The Accounting Review, Vol. 77 No. 1, pp. 25-50.

Bagnoli, M., Clement, M. and Watts, S.G. (2006), Around-the-clock media coverage and the timing of earnings announcements. McCombs Research Paper Series No. ACC-02-06, also available at: http://ssrn.com/abstract/570247

Baker, M., Coval, J. and Stein, J.C. (2007), “Corporate financing decisions when investors take the path of least resistance”, Journal of Financial Economics, Vol. 84 No. 2, pp. 266-298.

Baker, M. and Wurgler, J. (2006), “Investor sentiment and the cross‐section of stock returns”, The Journal of Finance, Vol. 61 No. 4, pp. 1645-1680.

Baker, M. and Wurgler, J. (2007), “Investor sentiment in the stock market”, The Journal of Economic Perspectives, Vol. 21 No. 2, pp. 129-151.

Ball, R. and Shivakumar, L. (2005), “Earnings quality in UK private firms: comparative loss recognition timeliness”, Journal of Accounting and Economics, Vol. 39 No. 1, pp. 83-128.

Basu, S. (1997), “The conservatism principle and the asymmetric timeliness of earnings 1”, Journal of Accounting and Economics, Vol. 24 No. 1, pp. 3-37.

Bergman, N.K. and Roychowdhury, S. (2008), “Investor sentiment and corporate disclosure”, Journal of Accounting Research, Vol. 46 No. 5, pp. 1057-1083.

Bergstresser, D. and Phillipon, T. (2006), “CEO incentives and earnings management”, Journal of Financial Economics, Vol. 80 No. 3, pp. 511-529.

Brown, N.C., Christensen, T.E., Elliott, W.B. and Mergenthaler, R.D. (2012), “Investor sentiment and pro forma earnings disclosures”, Journal of Accounting Research, Vol. 50 No. 1, pp. 1-40.

Cheng, Q. and Warfield, T. (2005), “Equity incentives and earnings management”, The Accounting Review, Vol. 80 No. 2, pp. 441-476.

Damodaran, A. (1989), “The weekend effect in information releases: a study of earnings and dividend announcements”, Review of Financial Studies, Vol. 2 No. 4, pp. 607-623.

Dechow, P., Ge, W. and Schrand, C. (2010), “Understanding earnings quality: a review of the proxies, their determinants and their consequences”, Journal of Accounting and Economics, Vol. 50 Nos 2/3, pp. 344-401.

Dechow, P.M. (1994), “Accounting earnings and cash flows as measures of firm performance: the role of accounting accruals”, Journal of Accounting and Economics, Vol. 18 No. 1, pp. 3-42.

Dechow, P.M., Kothari, S.P. and Watts, R.L. (1998), “The relation between earnings and cash flows”, Journal of Accounting and Economics, Vol. 25 No. 2, pp. 133-168.

DellaVigna, S. and Pollet, J.M. (2009), “Investor inattention and Friday earnings announcements”, The Journal of Finance, Vol. 64 No. 2, pp. 709-749.

Doyle, J.T. and Magilke, M.J. (2015), “The strategic timing of management forecasts”, Working paper, Utah State University and Claremont McKenna College.

Financial Accounting Standards Board (FASB) (2010), “Conceptual Framework for Financial Reporting. Statement of Financial Accounting Concepts No. 8”.

Genotte, G. and Trueman, B. (1996), “The strategic timing of corporate disclosures”, Review of Financial Studies, Vol. 9 No. 2, pp. 665-690.

Graham, J., Harvey, C. and Rajgopal, S. (2005), “The economic implications of corporate financial reporting”, Journal of Accounting and Economics, Vol. 40 Nos 1/3, pp. 3-73.

Hribar, P. and Collins, D.W. (2002), “Errors in estimating accruals: implications for empirical research”, Journal of Accounting Research, Vol. 40 No. 1, pp. 105-134.

Hurwitz, H. (2018), “Investor sentiment and management earnings forecast bias”, Journal of Business Finance and Accounting, Vol. 45 Nos 1/2, pp. 166-183, available at: https://ssrn.com/abstract=2899603

Khan, M. and Watts, R.L. (2009), “Estimation and empirical properties of a firm-year measure of accounting conservatism”, Journal of Accounting and Economics, Vol. 48 Nos 2/3, pp. 132-150.

Kothari, S.P., Leone, A.J. and Wasley, C.E. (2005), “Performance matched discretionary accrual measures”, Journal of Accounting and Economics, Vol. 39 No. 1, pp. 163-197.

Kothari, S.P., Shu, S. and Wysocki, P.D. (2009), “Do managers withhold bad news?”, Journal of Accounting Research, Vol. 47 No. 1, pp. 241-276.

Mian, G.M. and Sankaraguruswamy, S. (2012), “Investor sentiment and stock market response to earnings news”, The Accounting Review, Vol. 87 No. 4, pp. 1357-1384.

Nagar, V. (1999), “The role of manager’s human capital in discretionary disclosure”, Journal of Accounting Research, Vol. 37 No. 3, pp. 167-181.

Niessnar, M. (2015), “Strategic disclosure timing and insider trading”, Working paper, Yale School of Management, available at: http://ssrn.com/abstract=2439040

Patell, J.M. and Wolfson, M.A. (1982), “Good news, bad news and the intraday timing of corporate disclosures”, The Accounting Review, Vol. 57 No. 3, pp. 509-527.

Penman, S.H. (1987), “The distribution of earnings news over time and seasonalities in aggregate stock returns”, Journal of Financial Economics, Vol. 18 No. 2, pp. 199-228.

Simpson, A. (2013), “Does investor sentiment affect earnings management?”, Journal of Business Finance & Accounting, Vol. 40 Nos 7/8, pp. 869-900.

Skinner, D.J. (1994), “Why firms voluntarily disclose bad news”, Journal of Accounting Research, Vol. 32 No. 1, pp. 38-60.

Watts, R.L. and Zimmerman, J.L. (1986), “Positive accounting theory”, Prentice-Hall Inc.

Yu, F.F. (2008), “Analyst coverage and earnings management”, Journal of Financial Economics, Vol. 88 No. 2, pp. 245-271.

Corresponding author

Hong Kim Duong can be contacted at: hkduong@salisbury.edu