Should banks be averse to elections? A GMM analysis of recent elections in Ghana

Abstract

Purpose

Actions of incumbent politicians and firms’ managers during election years have been cited as sources of many problems that afflict economies and business entities. Given the controversies surrounding the impact of elections on firms’ soundness, this paper poses a question of whether banks should be averse to elections. Specifically, this study aims to investigate the impact of elections on the profitability and efficiency of banks.

Design/methodology/approach

Based on the authors’ knowledge, this is maiden analysis in this context for Ghana where relatively advanced appropriate GMM technique has been used on annual data from 2012 to 2016.

Findings

This study reveals that banks make higher returns in election years. Additionally, the authors report that government’s economic policies in election years are detrimental to management efficiency, though insignificant.

Practical implications

From an emerging economy perspective, this study would guide policymakers in designing policies that respond to, or minimize, the impact of elections on bank performance. The result of this analysis would also substantiate the market reaction to the changes in the economic, political and financial conditions.

Originality/value

This analysis suggests that firms’ performances in an election year depend on policies and political institutions in place. The authors argue that Ghana, with its exemplary democratic credentials and strong institutions, living alongside a high perception of corruption, is different. The contribution to literature is, first, by limiting this work to the banking sector of Ghana and, second, by incorporating the behaviors of incumbent governments and individuals in the regression specification model.

Citation

Broni, M.Y., Hosen, M., Mohammed, H.N. and Tiamiyu, G. (2019), "Should banks be averse to elections? A GMM analysis of recent elections in Ghana", Journal of Economics, Finance and Administrative Science, Vol. 24 No. 47, pp. 47-65. https://doi.org/10.1108/JEFAS-03-2018-0029

1. Introduction

As far as the management of firms, nations’ economies and individuals’ finances are concerned, situations appear to run smoothly, until when an election year sets in. Governments’ inertia to intervene during the crisis, firms’ attempt to please incumbent politicians, and the behaviors of managers of corporate entities and individual voters especially, are some reasons for the high research interest to investigate how politics affects firms’ performances.

On one hand, it is postulated that managers of firms would take actions to ensure the re-election of an incumbent presidential candidate, even if their firms suffer the consequences of lower performances. On the other hand, it is believed that providing campaign financing, makes it possible for firms to obtain political favors and hence increase their performances. With regards to economic policy, the opportunistic model has it that, governments prefer the adoption of expansionary economic policies in periods close to elections. But the accountability model runs contrary to this and states that for fear of being punished by voters, who do not want to be saddled with huge debt burdens and taxes, governments would resort to contractionary policies instead in election years. These theoretical contradictions pave the way for further inquiry into this subject.

Again, another theory indicates that the government is likely to direct state-owned banks to fund certain projects for political expediency, even if such projects are of less significance to the public interest. Also postulated is the view that, the presence of strong institutions promotes better fiscal policies and hence prevents abuse of incumbency. As the theory is unable to provide clear-cut proof as to the course of action that managers of firms and the economy are likely to take, it is prudent to analyze results of empirical work with regards to the impact of election year on the performance of firms, and banks for that matter.

Empirically too, there seem not to exist any consensus in investigations that have been undertaken so far. For instance, Akitoby and Stratmann (2010) report on a negative impact of elections on financial markets. Agyei-Sasu and Anaman (2012) present a contrary finding. They find a positive relationship between return on equity (ROE) and election year.

This work is motivated firstly by the controversies prevailing both in theory and empirical studies, and hence requiring further inquiry into the relationship between an election year and firm performance. Secondly, this paper is inspired by the work of Lehrer (2018) who suggests that the impact of elections on firm performance is country-specific, depending largely on the state of institutional and political systems in place. Most importantly, when two banks (Capital bank and UT bank) collapse some few months after the 2016 major elections in Ghana (Osabutey, 2017), the most likely suspect for such a calamity would be no any other factor, then manipulations of incumbency, or the inaction of the government. As noted in many political models, job destructions and firm closures may not be announced or allowed in periods close to elections (Bertrand et al., 2007). Moreover, Haggard and Mo (2000) suggest that political uncertainty and party politics have the possibility of causing delays in the government’s response to problems in the financial sector.

Extending the knowledge in this field, this paper focuses specifically on the banking sector in Ghana, a country in sub-Saharan Africa, which is democratically young, but considered a beacon of democracy on the African continent. The use of recent data from 2012-2016, and covering two election years, is also an important contribution in this field. The main issue addressed in this paper is, given Ghana’s unique nature, would an election year affect the performances of its banks? Putting it differently, should banks in Ghana be averse to elections?

From the perspective of this paper, Ghana is unique in its own rights. Being an African country, its democratic credentials are not expected to merit any serious respect. Examples abound in many African countries including Togo, Burkina Faso, Cote D’ Ivoire, and Kenya just to mention a few, where elections have turned into violence resulting in the destruction of many valuable lives and properties. But the same cannot be said of Ghana, which has since the inception of the fourth republican constitution, emerge from every election stronger and more united.

Despite the successes chalked by Ghana in democracy, the processes involved cannot be said to have been managed without problems and apprehensions. In the 1992 elections, when the transition was made from an unconstitutional rule to democracy, the opposition parties disputed the presidential election results that put former president Rawlings as the first democratic leader in the fourth republic. The opposition parties boycotted the parliamentary elections which were to be held in about four weeks after the presidential one. The situation did not deteriorate into conflict. The New Patriotic Party (NPP), then the most prominent opposition party, wrote a book entitled “The stolen verdict” (Party, 1993), the nation learned its lessons, and amended the electoral laws that resulted in the present situation where both presidential and parliamentary elections are held on the same day. Obviously to avoid future boycotts.

After the problems uncounted in the 1992 elections, all other elections (1996, 2000, 2004, 2008, 2012 and 2016) have passed, each with its own unique problems. But the most significant and notable is the 2012 elections where the National Democratic Congress’ (NDC) presidential candidate, then incumbent President John Mahama was declared the winner by the electoral commission. NPP’s presidential candidate then, now President of the Republic Nana Addo Dankwa Akuffo-Addo, together with Mahmudu Bawumia (now Vice President) and the then chairman of NPP, the Late Jake Obitsebi Lamptey, headed to the Supreme Court seeking for the annulment of the results, and to declare Nana Addo as the elected president. The proceedings of the court lasted for about eight months. But eventually, the pleas of the Petitioners were refused and the incumbent was pronounced as duly elected. Even this acrimony did not throw the country into conflict. The necessary reforms were made, and the nation moved on earning Ghana an enviable respect in democracy and electioneering management.

Ghana’s uniqueness is also seen from the perspective of its institutions. Undoubtedly, the nation is blessed with many strong institutions namely the judiciary, parliament, media, labor, and civil society organizations. Others include anti-corruption agencies such as the Commission for Human Rights and Administrative Justice (CHRAJ), and the Economic and Organized Crime Office (EOCO). Most of these organizations are required to operate independently without hindrance from the executive. The Military and Police, despite taking instructions from the President, do well to maintain a seemingly neutral posture during elections.

Unfortunately, in the mist of these strong institutions, is the prevalence of high levels of corruption in public and private institutions. For many years running, Ghana has been poorly rated on the Ghana Integrity Initiative’s Index of perception of corruption in the country. For instance, Ghana was rated 70th out of 176 countries, with a score of 43 per cent (see www.transparency.org/country/GHA). This is unusual as it is unthinkable that corruption could thrive in an environment surrounded by strong institutions. This again further supports the uniqueness of the country that merits an independent research of this kind.

It must be noted however that, the atmosphere preceding elections and aftermath of results’ declaration, is full of fear, uncertainty and anxiety. Experience has shown that violence normally breaks out, where some leading party personalities suffer attacks from their opponents. In this regard, the general populace does imagine what is likely to happen in the event that these clashes become aggravated. This may happen when results are rejected. Though the nation has never come to this level, the public is fully aware that a major disagreement in elections is not impossible.

Even though Ghana has earned the accolade of being a peaceful nation, in this paper’s view, individuals are more concerned about the possibility of losing their properties, if not lives, in the event that election violence does occur. As a result, the search for a safe haven to keep one’s resources will preoccupy the minds of individuals, especially years when an election contest is very keen. Arguably, banks would emerge as safest institutions because of their skills and resources. It is envisaged that many resources would be converted into cash and cash equivalents and deposited in banks. Hence banks would have more than enough resources to loan out in election years, to meet the high demand for loans (Cole, 2009).

This paper seeks to achieve the following objectives:

• to investigate how elections impact on banks’ profitability; and

• to investigate whether managers of banks do compromise on efficiency in election years.

Using annual data of twenty-nine banks, and Two-step General Method of Moments (GMM) estimation technique, the findings seem to indicate that Return on Asset (ROA) of banks is higher in election years. In addition, the results suggest that expansionary fiscal and monetary policies do lower banks’ efficiency, though not significantly. The rest of this paper is arranged as follows; Section 2 provides a review of relevant literature, Section 3 describes the variables and data used, Section 4 looks at methodology, Section 5 gives a presentation of empirical results and discussions, and Section 6 concludes and provides implications for policymaking.

2. Theoretical underpinning and literature review

This study looks at the impact of elections on financial sector performance. In this regard, many hypotheses have been formulated based on the behaviors of governments, firms and the voting public. For example, it is postulated that to help incumbent governments and politicians retain power, managers would increase hiring whiles postponing layoffs and plant closures in election years. Such practices are believed, would have negative consequences on the financial performance of firms. An alternate view sees campaign financing as a way of buying firm-specific political favors. This view may favor good financial outcomes for firms when considered alongside the Cartel theory which posits that, agenda setting in parliament is based on party lines. Hence policies are likely to favor firms that provide campaign financing especially, for legislators in party leadership and governing coalitions.

There are also the opportunistic and accountability models. Under the opportunistic model, it is thought that governments like to embark on expansionary fiscal and monetary policies during periods close to elections. Voters are assumed to base their decisions of who a good candidate is, by looking at recent economic conditions, and are likely to cast their ballots for the incumbent. However, under the accountability model, voters are concerned about higher taxes and debt burdens and are likely to punish the incumbent for adopting expansionary economic policies. Faced by political risk, incumbents are unlikely to adopt expansionary policies during election years even when they are necessary.

Again, another theory states that the government may direct state-owned banks to channel funds for projects that serve party political interest, thus neglecting public interest. In such scenarios, swing constituencies or constituencies where the incumbent is likely to lose, are flooded with many projects during election years in an attempt to brighten the chances of the incumbent. But the hypothesis that the existence of better institutions promotes better fiscal policy, may not allow an incumbent to channel resources arbitrary for political expediency.

All the theories listed above do not provide clear-cut direction about how government, firms, and the financial sector are likely to behave, and the possible outcomes of firms’ performances. Therefore, there is the need to refer to empirical work on financial sector performance in election years.

Given the importance of the relationship between government economic policy and firms’ performances, there has been a growing interest in research devoted to studying interactions between politics, democracy, elections and firms’ performances. How governments behavior in terms of economic policy, for example, depends on political election cycles. Government fiscal and monetary discipline may be abandoned at some point if such a path appears to be costly politically. Dreher (2003) finds that on the average, IMF programs with countries seem more likely to break down just before major elections. He studied 104 countries and also observes that the severity of program interruption is less probable in more developed democracies. The government has an overall responsibility in ensuring that the economic environment is conducive for businesses to thrive. How the government responds to economic distress has also been a subject for investigations. According to Haggard and Mo (2000), pressing political problems linked to elections in addition to disunity in the ruling party, were responsible for inhibiting a coherent government response to corporate and financial distress, and the deterioration of the external account that befell Korea in 1997.

Furthermore, incumbent governments may decide to exert pressure on institutions and firms to undertake certain activities that provide political advantage to incumbents over their opponents, especially when elections are close. Bertrand et al. (2007) indicate that firms with politically connected CEOs display higher levels of job and plant creation, and lower levels of plant destruction in election years. They examined publicly traded firms from 1987 to 2002. In addition, the report of lower profits in politically connected firms due to the high wage bills, as they use more in keenly political contested areas. Using a panel data of 98 developing countries over the period 1976-2004, Ha and Kang (2015) report that governments tightened fiscal and monetary policies during the crisis, however, the extent of tightening was considerably moderated by the presence of large political constraint, strong leftist partisan power in government, and upcoming legislative or presidential elections. Moreover, the independence of the central bank is normally tested during periods close to elections, when incumbent politicians seek to pressure them into adopting the loose monetary policy. Ranking the independence of central banks using 52 countries from 1972-2001, Alpanda and Honig (2010) findings show that countries with more independent central banks have lower average inflation rates over both election and non-election years.

Questions of why corporate bodies and individuals provide campaign contributions to politicians and political parties have also engaged the attention of many researchers. Using elections data of 1998 and 2002 in Brazil, Claessens et al. (2006) report that Brazilian firms that provided contributions to elected deputies, experienced higher stock returns than those that did not. In their view, the results show that contribution leads to the shaping of policy on the firm-specific basis. Additionally, they observed that contributing firms had more bank financing after the elections compared to the firms that did not contribute. On the contrary, Lehrer (2018) observed that political connections had little impact on cumulative abnormal returns, when he researched on 413 firms listed on the Tel Aviv stock exchange, in the week of the 2015 elections. Though he admits that some specific sectors responded significantly to the election outcome, he concludes that the market’s lack of responsiveness to political connections after the 2015 elections could indicate that, political connections do not carry financial value in the Israeli stock market. However, Akitoby and Stratmann (2010), report that financial markets’ view of election years is negative. They studied a panel of emerging market countries and assert that financial markets value democratic institutions more than economic and fiscal outcomes shaped by those institutions.

Literature also indicates that depending on the type of ownership, firms may experience different impact from election cycles. For instance, Baum et al. (2010) report that even though there exist differences in bank assets, liabilities, and performance across different stages of parliamentary elections, the behavior of government-owned banks is not different from foreign-owned and private banks before, during and after parliamentary election cycles. Their study covered 86 Turkish banks in the period of 1963-2005. In addition, they indicate that government-owned banks underperform foreign-owned and private banks. Chen and Liu (2013) provide results that support that of Baum et al. (2010). Using a panel data of financial institutions in Taiwan covering 1994-2009, Chen and Liu (2013) report that during election years, private financial institutions earned higher ROA and loan growth, than the government- and foreign-owned ones. According to their findings, government-owned institutions are not affected by current elections, whiles foreign-owned are significantly affected. Also, Muttakin et al. (2015) find that politically connected family firms outperform family firms that are not politically connected. Checking through their robustness test, it is evident that the relationship between elections and performance was found to be insignificant. Investigating firms listed on the Dhaka stock exchange over the period of 2005-2009, they also indicate that non-family firms with political connections demonstrate lower firm performance than non-family firms without political connections. In comparing Islamic and conventional banks, Bitar et al. (2017) show that Islamic banks underperform their conventional counterparts in more democratic political systems after analyzing data of 729 banks (139 Islamic), from 33 developing countries over the period 1999-2013.

Of significance to firms’ performance also, is the political ideology of the ruling party. Hence firms should be interested in which party thrives in elections. According to Belo et al. (2013), firms with high government exposure, experience higher cash flows, and stock returns when Democrats are in power in the USA, but the opposite pattern holds true during Republican presidencies. They based their findings on Industry level data, from July 1955-December 2009.

Finally, we present the findings of Agyei-Sasu and Anaman (2012) who studied all the 22 listed firms on the Ghana stock exchange over the period 1999-2009 and report that the impact of democratic political transition on ROE is mixed. Their results show that ROE increases in an election year, but declines in a transition year after an election.

Contradictions in theory and empirical findings provide justification for a further empirical inquiry into how election years impact on the performance of firms, especially those in the financial sector. Firstly, whether incumbent politicians or governments will pursue the opportunistic or accountability model in election years is not known, mainly because the ability to implement any of such policies depends on how developed democratic institutions are. Secondly, Akitoby and Stratmann (2010), asserts that financial markets value elections negatively as opposed to Agyei-Sasu and Anaman (2012), who observed that ROE increased in an election year. Such contradictions in empirical findings seem to lend credence to the view that, the impact of election year on firm’s performance is country-specific, based on the prevailing institutional and political system (Lehrer, 2018). Finally, the motivation for this study is also rooted in the fact that, it is focused on the impact of election year on the performance of the financial sector in Ghana. This paper is different from that of Agyei-Sasu and Anaman (2012) in that, they studied only firms in various sectors listed on the GSE.

3. Description of variables and data

This paper tries to find out how banks’ profitability and efficiency are affected in election years. In consistent with the literature, return on asset (ROA) and ROE being proxies for profitability, and cost to income ratio (CTI) for efficiency, represent dependent variables. Election Year, one of the main independent variables is represented by a dummy variable EY. In an election year, EY has a value 1, and in other years, a value 0. The data cover only two election years, 2012 and 2016.

Going by literature and theory, the danger faced by firms and the economy in an election year is the actions or inactions of incumbent governments, whose main objective is to retain political power, totally ignoring the financial health of firms and the economy (Haggard and Mo, 2000). Therefore, both fiscal and monetary policies are likely to impact significantly on firms in election years. Budget Balance/Deficit (BD) and Money Supply/M2 (M) have been used as proxies for fiscal and monetary policies respectively.

Changing attitudes in election years is not limited to only the incumbent government or the central bank. Individuals and member banks also respond to coming elections. Given the acrimonious nature in which electioneering campaigns are conducted in Ghana, the issue of safety of property is of major concern to individuals. The uncertainty of the acceptability of the verdict that would be announced by the electoral commission is a major concern to many. As characteristic of many African countries, such disagreements are normally greeted with anarchy, vandalism, and violence, accompanied by theft and, loss of lives and properties. This paper takes the position that individuals’ major concern will be, to search for an appropriate place to lodge their assets for safekeeping. As banks possess the expertise in this direction, individuals most likely would convert most of their assets into cash and deposited in banks, at least until after the elections are completely done with. Hence banks’ performances in election years would be greatly impacted by deposits. For this reason, deposit to asset ratio (DA), reflecting individuals’ behavior is included as an independent variable.

Also, banks may take advantage of the government’s quest to embark on fiscal expansion, and the increased demand for loans by parliamentary candidates, to grow their loan portfolios in election years. Thus, the view of this paper is that the quantum of loan growth or asset growth would be higher in election years. It is for this reason that, total asset growth (TAG) is included among the main independent variables. Also, interaction variables formed between the election year dummy and each of the four main independent variables mentioned above, have been utilized namely EDA (EY*DA), ETAG (EY*TAG), EDB (EY*DB) and EM (EY*M).

Finally, macroeconomic variables to control for inflation (INF), and economic activity using GDP per capita (GDPC) have been used. Bank-specific controls used are a total liability to total asset ratio (LA), and natural logarithms of total assets. All data cover the periods 2012-2016 and involves twenty (29) banks. Thompson Reuters Datastream is the source of all macroeconomic variables used in this study, whiles the bank-specific variables were obtained from FitchConnect database. Further description of variables is provided in Table I.

4. Methodology

The panel data in this study has a number of units (groups), greater than the time series. Twenty-nine banks are covered over a time period of 5 years span (2012-2016). In this case, the appropriate technique for estimating the model should be a panel technique. Due to the correlation between error term and lagged dependent variable, results of OLS estimation of panel techniques provides results that are inconsistent. To overcome this problem, General Method of Moments (GMM) is used to estimate coefficients that would be consistent and efficient. GMM uses instruments to eliminate the correlation between the error term and the lagged dependent variable.

Between Differenced and System GMM, the latter is preferred when the number of units (N) is relatively larger than the time series (T). It is for this reason that this paper resorted to the use of System GMM for estimations. The use of two-step System GMM is important in eliminating the effects of heteroscedasticity and serial autocorrelation (Davidson and MacKinnon, 2004). However, the two-step uses residuals which are downward bias, from the one-step estimation. To make the two-step System GMM estimates more efficient, the method developed by Windmeijer (2005) for finite sample correction is applied. This makes the results of the two-step System GMM more robust than the one-step (Roodman, 2009).

When Windmeijer (2005) method is applied STATA is unable to provide a Sargan test to check whether overidentifying restrictions are valid or not. The availability of a robust version of the Arellano and Bond (1991) AR(1) and AR(2) for serial autocorrelation, together with the number of instruments generated by STATA is enough to indicate whether instruments are well specified. For efficient estimates, the number of instruments should be less than the number of units, and there should be no serial autocorrelation especially for AR(2). According to Yalta and Yalta (2012), the failure to reject the null for Sargan test and AR(2), shows that the instruments used are valid.

We specify the model for the GMM estimation in this paper as shown below;

(1) Yit=αi + β(Yt-1 + EYt) + Ɵ(xit + EY*Xit) + ƞCit + ϵit,

Where α_i represents bank-specific effect, β, Ɵ and ƞ are coefficients to be estimated, ϵ_it is the error term. X_it is a matrix of the key independent variables consisting DA, TAG, BD, and M. EY_t is election year dummy. Similarly, C_it represents a matrix of bank-specific and macroeconomic control variables namely LA, LTA, INF, and GDPC. Finally, the three performance dependent variables ROA, ROE, and CTI, are represented by Y_it.

5. Empirical results and discussions

In Table II, descriptive statistics are provided. Going by the results, there is an indication that the profitability of banks is highly volatile as shown in the reported standard deviations of ROA and ROE. This situation should be a matter of great worry to shareholders and stakeholders in the banking industry.

The statistics also reveal that the government is restraining itself from using expansionary fiscal policy to gain political advantage, being aware of the responsibility hypothesis. This might seem to suggest that voters in Ghana are matured enough, capable of reading between the lines in distinguishing genuine projects from politically motivated ones. Supporting this argument lies in the minimum values of BD and EBD which indicates that, the biggest deficit in non-election years (−11.6 per cent) is higher than that of election years (−5.0 per cent). Figure 1 provides further evidence that in both elections’ years 2012 and 2016, government deficit was relatively normal. An examination of M and EM, however, tells a different story. The maximum value for money supply over the period under review is 27 per cent, this is found to have occurred in an election year. As illustrated in Figure 2, money supply fell in 2013 after the 2012 elections, and since then, it has been on the ascendancy. Therefore, one may not be wrong in suggesting that preference is given to expansionary monetary policy in periods close to elections, then an expansionary fiscal policy.

The descriptive statistics further provides credence to the suspicion of this paper in connection with the banking behavior of individuals in election years. We postulated that deposits would rise during such periods. The figures indicate that maximum deposit to asset ratio, occurred in an election year as shown in the maximum value of EDA. A closer look at the data reveals the highest deposit to asset ratio for most of the banks, were recorded in 2012. Additionally, the maximum value of ETAG is indicative that growth in banks’ total asset is highest in election years.

Correlation between all important variables is presented in Table III. Some of the key independent variables are highly correlated, and where there exists co-linearity, some variables are dropped when put together in a regression as the independent variable. The election year variable EY is highly correlated with its interaction variables EDA, EBD and EM. Correlation values of 0.8 are normally considered to be high. According to the matrix, EY is negatively correlated with both ROA and ROE, but positively correlated with CTI. Even though these correlations are not significant, they imply that banks experience low returns and high cost in election years. However, correlations are not enough for making judgments without controlling for certain important variables.

In this study, our main objective is to find out how banks’ profitability and management efficiency are affected by coming elections. Through two-step GMM estimations, results are provided in Tables IV-VI. Table IV presents the GMM results with ROA as the dependent variable. There are eight columns in this table. Column 1 is a regression with the main independent bank-level and economic-policy variables. In columns 2-4, interaction variables are included in the regression. In the last four columns, bank-level and macroeconomic controls are added to check for robustness.

ROA_t−1 has a significantly positive impact on ROA indicating the presence of momentum in determining banks’ performance. This also justifies the use of GMM in the estimation. The results show that EY has a positive impact on ROA as reported in columns 1, 6, 7 and 8, making the results robust. ROA is higher in an election year than other years by a range of between 4.5-5.5 per cent.

According to the estimations, DA is negatively related to ROA though this impact is not significant. From the results, the other main bank-level variable TAG is found to have a positive and significant impact on ROA. These results are robust across all columns of Table IV. A 1 per cent increase in the TAG would cause an increase of about 0.025 per cent in ROA on the average.

Turning to the main economic policy variables, BD (budget balance or deficit) and ROA have a significantly negative association. An increase in BD indicates a contractionary fiscal policy, as the deficit is reduced. The results show that a 1 per cent increase in BD causes a 1.0-1.5 per cent decrease in ROA. On the other hand, the impact of M on ROA is positive but not significant in all the columns.

The interaction variable EDA shows a positive impact, which is not significant, on ROA. It stands to reason that deposits are more useful in election years given that in other years, the impact of DA on profitability is negative. According to the results in column 5, the impact of DA on ROA is only significant in an election year. Similarly, the relationship between ETAG and ROA is positive but, also not significant.

Considering the other interaction variables formed between the economic policy variables and election year dummy, EM and EBD have no significant impact on ROA. An interpretation of these results will suggest that fiscal and monetary policies have no impact on ROA in election years. With regards to EM, its impact on ROA is positive, indicating that expansionary monetary policy in an election year results in higher ROA. Similarly, the negative coefficient of EBD only points to the fact that, a contractionary fiscal policy in election year causes lower ROA. The opposite will mean that pursuing a budget deficit in an election year, produces higher ROA. Finally, control variables LA, LTA, INF, and GDPC show no significant impact on ROA. The inclusion of these controls in the regression as shown in columns 5-8 does not seem to alter the impact of EY on ROA and thus indicating its robustness.

Moreover, Table V is similar to Table IV except that the dependent variable is CTI. To investigate whether management efficiency is compromised, the cost component of operations must be analyzed. The results indicate that the impacts of all variables on CTI are not significant. However, it is important to report the signs of the relationships between the independent and dependent variables. Thus, EY, DA, TAG, and M show negative relationships with CTI. It follows from these results that in election years, the cost to income ratio is lower than other years. Also, an increase in DA, TAG and M lower CTI. Conversely, the last of the key variables BD show a positive relationship with CTI. Hence, an increase in BD (contractionary fiscal policy), results in an increase in CTI.

Again, three of the interaction variables, namely, EDA, ETAG and EBD, are negatively related to CTI, whiles EM show a positive association with the dependent variable. Accordingly, increases in DA, TAG, and BD in an election year, cause reductions in CTI. An increase in M in an election year, however, produces an increase also in CTI. However, turning to the control variables, LA and GDPC show positive, whereas LTA and INF show negative relationships with CTI. Columns 5-8 shows the robustness of the insignificant relationship and impact between EY and CTI.

Table VI provides robustness check for election year impact on banks’ profitability using ROE as a dependent variable. Similar to the results of Table IV, the impacts of EY, EDA and ETAG on ROE are positive, except that the relationship between EY and ROE is not significant. Inference can be made from the positive relationship between EY and ROE to the effect that, there is robustness in the findings that banks realize higher returns in election years.

However, there is a reversal of results in Table VI compared with table IV, when the interactive variables EM and EBD are considered. EM exhibits a negative impact on ROE, whereas its impact on ROA is positive. Again, the impact of EBD on ROE is positive, but its impact on ROA is negative. Despite these disparities, both Tables IV and VI indicate that the relationship between these interactive variables (i.e. EM and EBD), and the two profitability variables (ROA and ROE) are not significant. Generally, most of the variables show no significant relationship with ROE, except ROE_t-1 and TAG. Results from Table VI show that an increase in the TAG of 1 per cent would result in the increase in ROE of between 0.2 and 0.4 per cent.

In terms of profitability, the results seem to suggest that banks realize higher returns during election years. This finding supports Agyei-Sasu and Anaman (2012) when they found that, returns increase in election years. There may be several reasons for this phenomenon. Elections are very expensive especially for presidential and parliamentary candidates. Considering the belief that voters would cast their ballots based on recent memories of events, candidates are tempted to delay many projects for execution in periods close to elections. Candidates are confronted by two scenarios; deploy their own funds if they have the capacity, or resort to borrowing from local or external sources. When candidates deploy their own resources to undertake development projects, the economy is stimulated resulting in increased incomes. Bank deposits are likely to increase especially from individuals who may now have higher cash surpluses because their incomes have gone up. Hence banks are well positioned to provide more loans to their clients and therefore make more profits.

Those candidates that require extra resources may need to fall on banks to fill the deficit. Considering the number of candidates and volume of projects that are normally executed, banks would be expected to provide more loans, both in numbers and amounts in election years than any other period. In this case, banks’ interest incomes would be expected to surge.

Incumbent presidential candidates are privileged to fall on economic policy tools to ensure the accomplishment of projects that would win them electoral victories. It is a known secret that incumbent presidential candidates may direct the central bank to pursue expansionary monetary policy (Ha and Kang, 2015). Another avenue for an incumbent president is to undertake expansionary fiscal policy by running a deficit budget that may be detrimental to the economy. In both cases, however, banks stand to benefit. Increased monetary supply provides banks with excess reserves that allow them to create more loans for higher returns. On the other hand, if the government decides to undertake an expansionary fiscal policy, it would do so by borrowing from the banking system. Even if the government decides to source for foreign funds, injecting such funds in the local economy would indirectly benefit the domestic banking system, as banks act as intermediaries (Agbloyor et al., 2013).

Actions and inactions of individuals may also well explain the higher profitability of banks in election years. We envisaged that given the acrimonious nature of elections in Ghana, citizens are unsure about what would happen after election results are announced. Even if they are sure about the likely winning presidential candidate, the acceptability of the results by the losing candidate cannot be ascertained. The most worrying aspect of the outcome of elections, however, is not about winning or the acceptability of results, the concern is the violence and vandalism that would be meted out to lives and properties. Therefore, the concerns of the majority of citizens are how to protect their properties from post-election violence. This paper considers that individuals would find banks as safest places to keep their properties, and hence are likely to convert most of their assets into cash and deposit same into banks. The desire to resort to banks for safekeeping of their cash at least until after the elections by individuals provides resources for banks to make more loans and profit from them. As mentioned earlier, the descriptive statistics seem to support this view. The fact that the maximum value of deposit to income ratio occurred in an election year (maximum of EDA), confirms our assertion that bank deposits are likely to be higher in election years.

As can be seen from the results of Tables III and V, the impacts of deposits (DA) on ROA and ROE respectively in ordinary years are negative. In election years, however, these impacts are positive, though insignificant. This phenomenon reflects the reality that, deposits are more useful and perhaps easily available in election years. Additionally, coefficients of ETAG are positive implying that increasing bank total assets would be more profitable in election years as against ordinary periods.

For fear of losing elections, governments have a difficulty in using contractionary fiscal policy (Ha and Kang, 2015). This implies that in most cases, expansionary policies would be adopted without regard to consequences that firms and the economy may suffer. The results of this study show that the impacts of money supply and budget deficit/surplus in election years, on profitability (both ROA and ROE) are not significant. An indication that banks may not suffer from the carelessness of the actions of government and the central bank. Alternatively, it is an indication that government and the central bank have always acted responsibly, even in election years. Furthermore, the active roles played by the media, civil society groups, political parties in opposition, economic and financial analysts, anti-corruption campaigners and the judiciary, might have put government actions in check to curtail any abuses, politically or economically (Akitoby and Stratmann, 2010).

Generally, all the results of Table V show no significant relationship between the independent variables and CTI. But we may derive certain vital information from them. For instance, the results indicate a reduction in cost to income ratio (CTI) of banks in election years. This contradicts other evidence which argues that banks especially the state-owned, may be forced to provide funding for projects, and to political cronies at higher costs or low-interest rates, resulting in inefficiency and misallocation of resources (Baum et al., 2010). Given the situation that individuals would prefer keeping their cash in banks for fear of violence and theft, banks obtain extra deposits at no extra cost but lend such funds to election candidates, who virtually are in serious need, at higher interest rates (Cole, 2009). Moreover, borrowers are readily available during such periods, and banks need not incur high costs providing them with financing. Thus, the processes of obtaining funds and lending them out in election years were of lower cost in Ghana as far as the periods 2012 and 2016 were concerned. The result may seem to suggest that, the one main factor that differentiates the situation in Ghana from developed democracies, is the fear of uncertainty that surrounds the declaration of election results and its aftermath. This fear provides an avenue for banks to obtain liquidity at virtually no or least cost.

Furthermore, the results show that the impact of money supply increases cost to banks in election years, while reducing cost in ordinary periods. This phenomenon is explained by the interaction between demand and supply of money. Money supply in excess of its demand, would bring down interest rates and therefore becomes a cost to banks. Perhaps in election years, money supply is too excessive to the extent that, it affects the efficiency of banks. Another cost that banks would have to incur, emanates from competing with rivals to attract customers (election candidates), in the form of increasing advertisement for example.

Finally, Table V indicates that contractionary fiscal policy in election years reduces CTI compared to an increase in other years. The opposite means that expansionary fiscal policy increases CTI and hence is detrimental to banks’ efficiency. Though this impact is not significant, it supports the assertion that, banks’ management may be willing to support incumbent candidates by providing funding without regards to efficiency. To sum up, it seems though that economic policy, fiscal and monetary, are potential threats to banks’ management efficiency, even if not significant. It should be noted, however, that as far as this study is concerned, incumbent government preference for economic policy, is monetary as stated earlier.

6. Conclusion and policy implications

In this paper, we set out to contribute to the ongoing debate concerning the relationship between democratic elections and firms’ performances. Considering the importance of the banking sector in every economy, we limited our study to the banking sector of Ghana, arguing that Ghana is different and hence merit such an investigation. This paper wishes to address the question of whether banks should be averse to coming elections. The main objectives of this paper are:

• to find out how banks’ profitability is affected by elections; and

• to investigate whether management of banks compromises efficiency to give political advantage to an incumbent government.

Literature is not conclusive on the kind of impact elections exert on the performance of the banking sector, and in theory, it is not clear what type of economic policy an incumbent government may pursue in times close to elections. These shortcomings are primarily the motivation for this research. But the outcome of such a study would also depend on policies and level of development of political institutions. Ghana arguably possesses such strong institutions, except that perception of the existence of corruption in the country, is likewise strong. Apart from corruption, the one other factor that makes Ghana appear different from elite democratic countries like USA, UK, Germany, and France, is the high probability that elections results may be rejected, and the country thrown into a state anarchy and lawlessness.

Taking into consideration Ghana’s uniqueness, we postulate that individuals for fear of violence after the declaration of election results would increase their deposits with banks to ensure the safety of their wealth. Therefore, we further contribute to the literature by including individuals’ banking behavior in our model.

Using two-step GMM estimations, and annual data of twenty-nine (29) banks from 2012-2016, the descriptive statistics appear to indicate that, the expansionary monetary policy is preferred to expansionary fiscal policy in election years. The choice of policy seems to be based on the fact that voters in Ghana have matured, and are likely to vote out the incumbent who will overburden them with high tax and debt levels, as alluded by the responsibility theory of elections. Analysis of regression estimations have culminated in the findings listed below:

• ROA is higher in years when elections are held compared to other years.

• Fiscal and monetary policies in an election year are detrimental to banks’ management efficiency, though not significant.

• Increase in bank deposit to asset ratio leads to an increase in ROA only in election years. Again, this impact is not significant.

Results of this paper seem to suggest that, banks in Ghana should not be averse to elections as returns are higher, and efficiency not significantly compromised during these periods. However, government and the central bank must endeavor to ensure that, their respective fiscal and monetary policies do not endanger the management efficiency of banks. Civil society, the media, judiciary, parliament, and donor organizations, both domestic and foreign, should continue to monitor the activities of incumbent election candidates. This is the only way the integrity of the financial and banking sectors can be guaranteed. Boards and management of banks must put in place measures, to uphold their independence and integrity, and resist attempts of manipulation by those that will power, especially in election years. Finally, as the incumbent has resorted to the use of expansionary monetary policy during election years, we wish to call on parliament and all stakeholders, to work towards the promulgation of laws that guarantee the independence of the central bank.

Figures

Figure 1.

Budget deficit

Figure 2.

Money supply (M)

Table I.

Definitions of variables

Name	Representation	Measure	Unit
Return on Asset	ROA	Profitability	%
ROE	ROE	Profitability	%
Cost to Income ratio	CTI	Efficiency	%
Election Year	EY		Dummy
Deposit to Asset ratio	DA	Liquidity risk	%
Total Asset Growth	TAG	Asset growth	%
Budget Deficit	BD	Fiscal policy	% of GDP
Money supply M2	M	Monetary policy	% of GDP
DA in an Election year	EDA
TAG in an Election year	ETAG
BD in Election year	EBD
M in Election year	EM
Total (Liabilities/Assets)	LA	Financial leverage	%
Total Assets	LTA	Bank size	Natural log
Inflation	INF	Price changes	%
GDP per capita	GDPC	Economic activity	Natural log

Source: The author

Table II.

Descriptive statistics

Variable	Mean	Std. deviation	Minimum	Maximum
ROA	3.4156	2.6866	−7.3000	8.6400
ROE	31.7791	26.0902	−60.6500	86.1000
CTI	57.4513	17.7749	13.5300	101.4400
EY	0.4000	0.4915	0	1
BD	−8.2840	3.0295	−11.5500	−4.2900
M	24.8386	1.4213	22.7557	26.6241
DA	64.4995	17.2883	2.3200	87.5100
TAG	32.37516	27.28491	−17.06	184.06
EBD	−1.8660	2.3056	−5.0400	0
EM	10.0800	12.4211	0	26.6241
EDA	23.9757	33.8532	0	87.5100
ETAG	12.0974	24.4627	−0.2200	184.0600
LA	85.1010	4.9011	62.8000	94.1001
LTA	7.0223	1.5816	2.8367	10.8145
INF	14.1760	3.2722	9.1600	17.4700
GDPC	7.3175	0.0902	7.1959	7.4518

Source: The author

Table III.

Correlation matrix of variables

	roa	roe	cti	ey	bd	m	da
roa	1.0000
roe	0.8948*	1.0000
cti	−0.7431*	−0.7464*	1.0000
ey	−0.0386	−0.0375	0.0226	1.0000
bd	−0.0573	−0.0529	0.0280	0.9788*	1.0000
m	−0.1717	−0.1356	−0.0000	0.2083*	0.2872*	1.0000
da	−0.1327	−0.1092	0.3969*	0.0798	0.0987	0.0225	1.0000
tag	0.2528*	0.2360*	−0.2196*	0.0558	0.0330	−0.1614	−0.1861*
ebd	0.0521	0.0478	−0.0204	−0.9947*	−0.9654*	−0.2728*	−0.0727
em	−0.0481	−0.0447	0.0211	0.9973*	0.9705*	0.2540*	0.0749
eda	−0.0572	−0.0485	0.0884	0.9293*	0.9146*	0.1064	0.2869*
etag	0.0846	0.0446	−0.0640	0.6629*	0.6600*	0.0081	−0.0015
la	−0.1715	0.0815	0.0096	0.0221	0.0298	0.1555	0.0239
inf	−0.1554	−0.1191	−0.0155	−0.2156*	−0.1751*	0.8773*	−0.0377
gdpc	0.0959	0.0738	0.0008	0.2625*	0.1042	−0.7455*	−0.0298
lta	0.0911	0.2106*	−0.5194*	0.0025	0.0151	0.3246*	−0.2803*
	tag	ebd	em	eda	etag	la	inf
tag	1.0000
ebd	−0.0402	1.0000
em	0.0449	−0.9995*	1.0000
eda	0.0303	−0.9182*	0.9225*	1.0000
etag	0.5314*	−0.6443*	0.6505*	0.5988*	1.0000
la	0.0257	−0.0338	0.0303	0.0308	−0.0656	1.0000
inf	−0.1913*	0.1313	−0.1564	−0.3230*	−0.3151*	0.1418	1.0000
gdpc	0.1061	−0.2457*	0.2510*	0.2312*	0.1781*	−0.1114	−0.7000*
lta	0.0813	−0.0337	0.0244	−0.0563	0.0215	0.2705*	0.3350*
	gdpc	lta
gdpc	1.0000
lta	−0.1995*	1.0000

Source: The author

Table IV.

Two-Step System GMM estimation with ROA as dependent variable

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
ROAt−1	0.6036*** (0.000)	0.6531*** (0.000)	0.6531*** (0.000)	0.6531*** (0.000)	0.6713*** (0.000)	0.6657*** (0.000)	0.6657*** (0.000)	0.6657*** (0.000)
EY	4.4703* (0.061)	1.4240 (0.713)			1.5589 (0.572)	5.4210* (0.055)	5.5392* (0.060)	5.3808* (0.053)
DA	0.0057 (0.868)	−0.0088 (0.888)	−0.0088 (0.888)	−0.0088 (0.888)	−0.0537 (0.271)	−0.0159 (0.768)	−0.0159 (0.768)	−0.0159 (0.768)
TAG	0.0324*** (0.000)	0.0287** (0.034)	0.0287** (0.034)	0.0287** (0.034)	0.0223*** (0.004)	0.02297** (0.029)	0.0230** (0.029)	0.0230** (0.029)
BD	−0.9534* (0.052)	−1.2555** (0.025)	−1.2555** (0.025)	−1.2555** (0.025)	−1.5137** (0.002)	−1.2240** (0.025)	−1.2320** (0.027)	−1.1431** (0.018)
M	0.3665 (0.195)	0.5396 (0.109)	0.5396 (0.109)	0.5396 (0.109)	0.6333 (0.163)	0.1827 (0.635)
EDA		0.0528 (0.343)	0.0528 (0.343)	0.0528 (0.343)	0.0754 (0.013)
ETAG		0.0433 (0.429)	0.0433 (0.429)	0.0433 (0.429)	0.0366 (0.366)
EM			0.0535 (0.713)
EBD				−0.2825 (0.713)
LA					0.0209 (0.884)	0.0144 (0.945)	0.0144 (0.945)	0.0144 (0.945)
LTA					−0.3060 (0.831)	1.3336 (0.278)	1.3336 (0.278)	1.3336 (0.278)
INF							0.1094 (0.635)
GDPC								−1.7427 (0.635)
Observations	98	98	98	98	98	98	98	98
AR(2)	0.2842	0.4303	0.4303	0.4303	0.4903	0.2113	0.2113	0.2113

Notes:

****, ** and * represent significant level at 1%, 5% and 10%, respectively; p-values are in parenthesis

Source: The author

Table V.

Two-step system GMM estimation with CTI as dependent variable

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
CTIt−1	0.4168 (0.153)	0.5001 (0.168)	0.5001 (0.168)	0.5001 (0.168)	0.3077 (0.191)	0.2760 (0.157)	0.2760 (0.157)	0.2760 (0.157)
EY	−2.0670 (0.937)	8.8386 (0.728)			−23.988 (0.383)	−23.231 (0.369)	−23.538 (0.375)	−23.127 (0.367)
DA	−0.0523 (0.801)	−0.0581 (0.794)	−0.0581 (0.794)	−0.0581 (0.794)	−0.0909 (0.744)	−0.1015 (0.721)	−0.1015 (0.721)	−0.1015 (0.721)
TAG	−0.0877 (0.106)	−0.0783 (0.294)	−0.0783 (0.294)	−0.0783 (0.294)	−0.0525 (0.549)	−0.0453 (0.574)	−0.0453 (0.574)	−0.0453 (0.574)
BD	0.6297 (0.900)	2.6913 (0.444)	2.6913 (0.444)	2.6913 (0.444)	5.9911 (0.147)	5.7451 (0.307)	5.7659 (0.309)	5.5352 (0.283)
M	0.6360 (0.765)	−0.0634 (0.968)	−0.0634 (0.968)	−0.0634 (0.968)	−0.5414 (0.665)	−0.4743 (0.751)
EDA		−0.2885 (0.228)	−0.2885 (0.228)	−0.2885 (0.228)	−0.0320 (0.835)
ETAG		−0.0749 (0.871)	−0.0749 (0.871)	−0.0749 (0.871)	0.0520 (0.909)
EM			0.3320 (0.728)
EBD				−1.7537 (0.728)
LA					0.6578 (0.407)	0.6112 (0.359)	0.6112 (0.359)	0.6112 (0.359)
LTA					−10.136 (0.226)	−10.921 (0.181)	−10.921 (0.181)	−10.921 (0.181)
INF							−0.2839 (0.751)
GDPC								4.5235 (0.751)
Observations	100	100	100	100	100	100	100	100
AR(2)	0.4249	0.9110	0.9110	0.9110	0.4338	0.2981	0.2981	0.2981

Notes:

****, ** and * represent significant level at 1%, 5% and 10%, respectively; p-values are in parenthesis

Source: The author

Table VI.

Two-step system GMM estimation with ROE as dependent variable

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
ROEt−1	0.7133*** (0.000)	0.7680*** (0.000)	0.7680*** (0.000)	0.7680*** (0.000)	0.8071*** (0.000)	0.8392*** (0.002)	0.8392*** (0.002)	0.8392*** (0.002)
EY	28.671 (0.368)	−13.504 (0.695)			0.5237 (0.985)	43.4455 (0.201)	43.0444 (0.230)	43.5821 (0.191)
DA	−0.0327 (0.918)	−0.2245 (0.482)	−0.2245 (0.482)	−0.2245 (0.482)	−0.3235 (0.533)	0.0372 (0.951)	0.0372 (0.951)	0.0372 (0.951)
TAG	0.3611*** (0.000)	0.2564*** (0.003)	0.2564*** (0.003)	0.2564*** (0.003)	0.2056** (0.022)	0.2382** (0.027)	0.2382** (0.027)	0.2382** (0.027)
BD	−5.1752 (0.431)	−7.8949 (0.191)	−7.8949 (0.191)	−7.8949 (0.191)	−11.291** (0.031)	−9.6175 (0.180)	−9.5906 (0.189)	−9.8922 (0.102)
M	0.4358 (0.890)	3.1275 (0.257)	3.1275 (0.257)	3.1275 (0.257)	2.8548 (0.424)	−0.6203 (0.893)
EDA		0.6000 (0.167)	0.6000 (0.167)	0.6000 (0.167)	0.6306** (0.048)
ETAG		0.4739 (0.199)	0.4739 (0.199)	0.4739 (0.199)	0.4614 (0.133)
EM			−0.5072 (0.695)
EBD				2.6793 (0.695)
LA					0.3069 (0.747)	0.7281 (0.346)	0.7281 (0.346)	0.7281 (0.346)
LTA					4.8584 (0.551)	16.0099 (0.183)	16.0099 (0.183)	16.0099 (0.183)
INF							−0.3713 (0.893)
GDPC								5.9165 (0.893)
Observations	98	98	98	98	98	98	98	98
AR(2)	0.9790	0.8036	0.8036	0.8036	0.8180	0.6922	0.6922	0.6922

Notes:

****, ** and * represent significant level at 1%, 5% and 10%, respectively; p-values are in parenthesis

Source: The author