Dividend policy in Indonesian banking sector during COVID-19 pandemic period

Abstract

This research examines the effect of the crisis caused by the COVID-19 pandemic on the dividend policy in banking companies in Indonesia with the focus on the 2014 to 2020 period. The samples were selected using the purposive sampling method, and the data were analyzed statistically using dynamic panel data regression with two estimation techniques which include the first difference generalized method of moments (FD-GMM) and the system generalized method of moments (SYS-GMM). Moreover, the predictor of the crisis caused by the COVID-19 pandemic was proxied by GDP growth and annual inflation rate, and this led to the estimation of four parameters to obtain comprehensive results. The model specification test also found only three-parameter estimates considered feasible for use in the research and the results showed that banks tend to distribute higher dividends during the period of the crisis caused by the pandemic. It was also discovered that profitability and dividends from the previous year have a significant positive effect on dividend policy. Furthermore, banking liquidity was found to also have a positive effect, while financial leverage, investment opportunity, and bank size have a negative effect on dividend policy. These findings are expected to serve as a reference for investors in making business decisions to achieve optimal returns and to allow company management, especially those in the banking sector, to formulate optimal dividend policies during a crisis.

Keywords:

• dividend policy
• Indonesian banking sector
• COVID-19 pandemic
• Indonesia stock exchange
• dynamic panel data regression

JEL Classifications:

• G21
• G30
• G35

1. Introduction

COVID-19 pandemic was a global occurrence as indicated by the spread of the virus first discovered in Wuhan, the People’s Republic of China, to almost all countries in the world (Mas-Coma et al., 2020). This led the government of Indonesia to implement a Large-Scale Restriction Policy (PSBB) towards suppressing the spread of the virus by limiting human activities outside the home with the focus on the education sector, offices, religious activities, events in public facilities, socio-cultural activities, as well as the movement of people and goods using different modes of transportation. The restrictions on the movement of people and goods have led to disrupted production and business chains, subsequently paralyzing economic activities. Banks, as intermediaries responsible for aggregating and channeling funds, have also been affected by the hindered circulation and allocation of resources due to the crisis conditions (Basse et al., 2014).

A bank is one of the financial institutions playing an important role in the economy through the provision of credit and other banking services to the industry and trade sector. Meanwhile, the COVID-19 pandemic caused a global financial crisis (Altig et al., 2020) with a significant effect on the banking sector, especially in Indonesia (Hasan, 2020). It is also important to note that the Large-Scale Social Restrictions Policy paralyzes the economy with further effect on the banking sector (Herwany et al., 2021). This created vulnerabilities which eventually caused several actions such as confiscation, bad loans, and other banking problems. Moreover, the business activities of banks were declined during the period with significant influence on their performance (Goodell, 2020).

The economic crisis caused by the pandemic also reflects in the country’s gross domestic product (GDP) as indicated by −2.07% recorded year-on-year in 2020 and this implies the market value of all goods and services produced in Indonesia was experiencing negative growth (Rahman, 2013). The trend was also observed in the inflation rate of 1.68% reported in 2020 which is the lowest in the last 5 years compared to 2.72% in 2019, 3.13% in 2018, 3.61% in 2017, and 3.02% in 2016 as presented by the data obtained from the Central Statistics Agency of Indonesia. This shows a decrease in the money supply in 2020, thereby, increasing the tendency of people to collect their money in banks considering the uncertainties caused by the pandemic (Blyth & Lonergan, 2014).

Share prices of companies generally fell in Indonesia due to negative sentiment over the pandemic and this was confirmed by the fact the Composite Stock Price Index experienced its lowest fall at 4,194.94 throughout 2020 as indicated by the data obtained from the Indonesian Stock Exchange (IDX). It was also discovered that the investment behavior of investors changed based on the information received (Herwany et al., 2021) in relation to the pandemic, thereby, leading to the rearrangement of their investment portfolios to obtain optimal returns (Zainuri et al., 2021).

The companies in the banking sector carefully considered their dividend policy during the crisis considering the fact that their values were previously set based on normal economic conditions. It is important to be pointed out that these companies need to maintain a good financial position to keep business running while fulfilling obligations to their shareholders. This led to the examination of the determinants of banking dividend policy in Indonesia during the economic crisis caused by COVID-19 (Abdulkadir et al., 2015). Reddemann et al. (2010) showed a reduction in the dividend rates during crisis and the same was also reported by Hauser (2013) for the dividend policy of companies during the 2008 and 2009 economic crisis. Abdulkadir et al. (2015) also found a reduction in the dividends distributed to shareholders during the economic downturn in Nigeria, while Krieger et al. (2021) reported that 1,400 companies in the USA paid, 213 withheld, and 93 did not pay dividends during the COVID-19 pandemic. Moreover, Renitia et al. (2020) descriptively found that companies in Indonesia have different behavioral variations with some discovered to have cut or even eliminate the dividends distributed to their shareholders, while several others distributed at a higher rate than the previous period.

Therefore, the main motivation of this research is to examine the banking dividend policies in Indonesia during the COVID-19 pandemic crisis period. This study aims to establish causal relationship between crisis variables and dividend policies. The macroeconomic indicators utilized to gauge the crisis conditions encompass the growth of Gross Domestic Product. Additionally, a Inflation Rate is employed to differentiate as macroeconomics variable associated with the pandemic. Moreover, this research also investigates other predictors that partially influence banking dividend policies.

Several predictors were also tested and used as the postulates to ensure a comprehensive assessment of the complex empirical models and those associated with the dividend policy are profitability, financial leverage, investment opportunity, company size, banking liquidity, and the previous year’s dividend (Kullab et al., 2022; Lim, 2016; Sharma & Bakshi, 2019; Singla & Samanta, 2018). Hence, the hypotheses of this research entail the presumed impact of the COVID-19 pandemic crisis on dividend policies, as well as the presumed influence of profitability, financial leverage, investment opportunity, company size, banking liquidity, and the previous year’s dividend on dividend policy.

All the banking companies listed on IDX distributing dividends were the focus of this research, and a dynamic panel data regression was applied as the analytical tool to produce robust parameter estimates, overcome problems related to endogeneity, and test the previous year’s dividend (Arellano & Bond, 1991; Li, 2016). This study hopefully will contribute in providing an empirical research in the pandemic era by employing appropriate method, which is dynamic panel data regression to produce robust parameter estimates for the banking sector in Indonesia in terms of the dividend policy’s determinants.

This research presents a significant contribution to the existing literature by introducing novel findings. Firstly, it examines the dividend policy of banks in Indonesia during the 2020 pandemic crisis. The causal analysis is conducted using two approaches, namely GDP growth and annual inflation rate. Secondly, the study employs a sophisticated analytical tool, namely dynamic panel data regression. This method is considered the most suitable estimation technique for testing causality using regression models with panel data structure. Specifically, the estimation method utilizes the SYS-GMM approach for the two-estimator technique. The results obtained differ from general expectations. In crisis conditions, companies tend to reduce or even eliminate dividend payouts to ensure survival. However, the findings indicate the opposite behavior, as banks in Indonesia tend to distribute dividends positively during crises. This raises suspicions that banks distribute dividends positively during crises to provide positive signals to the market, particularly when the stock market conditions are weak.

This study is presented in several sections. Firstly, the introduction section provides the background information, previous research studies, research gap, motivation and objectives of the study, a brief overview of the planned research, and its systematic presentation. Secondly, the literature review presents the grand theory that underlies the study, leading up to the development of hypotheses. Thirdly, the research method section presents the type of research, selected samples, definitions and measurements of variables, as well as the statistical analysis instruments. Fourthly, the results and discussion section presents the statistical tests and comprehensive discussions of the research hypotheses. Finally, the conclusion section summarizes the findings, limitations, and provides suggestions for further research.

2. Literature Review

COVID-19 pandemic threatens several companies, especially those in the banking sector, due to the economic downturn and systemic impact caused by the restriction of movement for people and goods. The bank, which is an intermediary institution, was affected by the crisis (Goodell, 2020) and companies were required to adjust their dividend policy to maintain business stability (Altig et al., 2020). This is in line with the pecking order theory (Hasan, 2021; Hasan & Islam, 2022) which recommends that companies prioritize internal funding sources such as retained earnings in their order of priority scale in the context of the risk level (Damodaran, 2015; Myers, 1984). This shows that business organizations need to focus on their level of dividends and retained earnings to ensure stability (Hartono & Raya, 2022; Tinungki et al., 2022).

The grand theory of this research is adapted from the Pecking Order Theory. The theory suggests that companies tend to prioritize internal financing over external financing. Internal financing is considered less risky and incurs lower cost of capital, compared to external sources of financing, such as issuing stocks and bonds, which pose higher risks and entail cost of capital more. In crisis conditions, ideally, banking companies would reduce or even eliminate dividend payouts to ensure their survival during the crisis. This is also due to the uncertain end date of the pandemic until the end of 2020, which further supports the need for companies to prioritize capital preservation (Basse et al., 2014; Damodaran, 2015; Myers, 1984).

2.1. Impact of the COVID-19 pandemic crisis on dividend policy

The pandemic caused economic uncertainty in 2020 with a significant impact on the profitability of companies due to the disruption of the industrial production chain. It also affected the non-current business cycle such as the banks serving as intermediaries for these trade transactions (Goodell, 2020; Renitia et al., 2020). The crisis influenced certain macroeconomic factors and an important indicator used to determine its effect is the Gross Domestic Product growth which is the proxy for economic development opportunities (Hartono et al., 2023; Tinungki et al., 2022). The banks experienced a decline in cash flow due to a reduction in banking transaction activities because of the pandemic, thereby, leading to a reduction in the level of dividends distributed to shareholders in order to maintain business continuity (Abdulkadir et al., 2015; Goodell, 2020; Krieger et al., 2021).

Abdulkadir et al. (2015) and Krieger et al. (2021) found that companies tend to reduce dividend rates during the crisis in line with the decline in their profitability. Ong et al. (2018) and Romus et al. (2020) also showed the positive influence of Gross Domestic Product on dividend policy and this indicates that it can be used to measure the impact of the economic crisis caused by the COVID-19 pandemic on dividend policy. This, therefore, led to the formulation of the following first hypothesis:

Hypothesis 1 (H1):

Gross Domestic Product (GDP) growth has a positive effect on dividend policy.

The banking sector is usually affected by certain macroeconomic variables such as the inflation rate. This is observed from the low money supply associated with the reduction in banking transactions during the pandemic which tends to lower the inflation rate. Moreover, the low transactions also have the ability to reduce profitability and, subsequently, the dividend rate. This implies that a lower inflation rate can reduce the level of dividends to be distributed to shareholders (Basse et al., 2014; Kosasih et al., 2021). Basse and Reddemann (2011), Kosasih et al. (2021), and Mirbagherijam (2014) also showed that the inflation rate has a positive effect on dividend policy and this led to the formulation of the second hypothesis as follows:

Hypothesis 2 (H2):

The inflation rate has a positive effect on dividend policy.

2.2. Impact of profitability on dividend policy

Profitability has a positive effect on dividend policy due to the fact that a higher level of bank profitability usually leads to a higher tendency to distribute dividends (Kullab et al., 2022; Maladjian & El Khoury, 2014; Silalahi et al., 2021). Banking transaction was observed to be low during the crisis, and this led to a reduction in profitability from interest received on loans and interest expense paid to customers and the trend was even higher due to lower money supply and other declining incomes (Hauser, 2013; Lestari et al., 2021). It was also reported by Kullab et al. (2022), Marfo-Yiadom and Agyei (2011), and Silalahi et al. (2021) that profitability has a positive effect on dividend policy, and this led to the formulation of the following third hypothesis:

Hypothesis 3 (H3):

Profitability has a positive effect on dividend policy.

2.3. Impact of financial leverage on dividend policy

Companies financing their business with debt are usually charged at the cost of debt. The payment of interest and principal debt normally reduce the income needed to guarantee dividend payments and this means the higher debt is expected to have a negative impact on dividend policy (Marfo-Yiadom & Agyei, 2011; Ranajee et al., 2018; Singla & Samanta, 2018). The research by Naveenan et al. (2021), Sharma and Bakshi (2019), and Wahjudi (2020) showed that financial leverage affects dividend policy and this was used to formulate the fourth hypothesis as follows:

Hypothesis 4 (H4):

Financial leverage has a negative effect on dividend policy.

2.4. Impact of investment opportunity on dividend policy

Investment opportunity was explained by the pecking order theory to serve as the logical thinking related to the prioritization of internal funding, in the form of retained earnings over net income, by a company to reduce dividend rates on net income earned. The common proxy normally used to measure investment opportunity is the market price to book value ratio in which a higher market share price indicates a higher investment opportunity. This is relevant because the market share price is a future cash flow (Athari et al., 2016; Damodaran, 2015; Jovković et al., 2021; Nadeem et al., 2018). Some previous researches by Nadeem et al. (2018), Patra et al. (2012), and Rizqia et al. (2013) showed that investment opportunity has a negative effect on dividend policy and this led to the formulation of the fifth hypothesis as follows:

Hypothesis 5 (H5):

Investment opportunity has a negative effect on dividend policy.

2.5. Impact of bank size on dividend policy

The size of a bank reflects its ability to generate profit such that a bigger bank size usually indicates more profitability and the capability to distribute higher dividends. It was discovered that the COVID-19 pandemic caused a decrease in the level of profitability and this led to a reduction in the dividend rate (Goodell, 2020; Imran et al., 2013; Kullab et al., 2022). Another report by Imran et al. (2013), Kosasih et al. (2021), and Kullab et al. (2022) also showed that bank size has a positive effect on dividend policy and this was used to formulate the sixth hypothesis as follows:

Hypothesis 6 (H6):

Bank size has a positive effect on dividend policy.

2.6. Impact of banking liquidity on dividend policy

Banking liquidity is generally proxied by the loan-to-deposit ratio which is the ratio of loans to customer deposits in banks such that a higher level of loans to creditors against customer deposits usually leads to a reduction in the liquidity level and this can cause a decline in the level of dividends (Rahma & Syarif, 2019; Ramadani & Jumono, 2020; Yusuf & Muhammed, 2015). Moreover, Rahma and Syarif (2019) and Yusuf and Muhammed (2015) found that banking liquidity has a negative effect on dividend policy and this was adopted to formulate the seventh hypothesis as follows:

Hypothesis 7 (H7):

Banking liquidity has a negative effect on dividend policy.

2.7. Impact of Previous Year’s dividend on dividend policy

The previous year’s dividend normally provides a positive signal to the market according to the dividend signaling theory. It is a reference to determine the current period’s dividend rate due to its positive influence (Alzomaia & Al-Khadhiri, 2013; Hartono & Matusin, 2020; Sharma & Bakshi, 2019). A previous research by Alzomaia and Al-Khadhiri (2013), Hartono and Matusin (2020), and Maladjian and El Khoury (2014) showed that the previous year’s dividend has a positive effect on dividend policy and this was the foundation for the formulation of the eighth hypothesis as follows:

Hypothesis 8 (H8).

The previous year’s dividend has a positive effect on dividend policy.

3. Research method

The banking companies listed on the Indonesia Stock Exchange (IDX) were used as samples in the research with the focus on 2011 to 2020 with 2020 marked as the economic crisis period due to the COVID-19 pandemic (Altig et al., 2020; Arianto, 2021). Secondary data were obtained from the financial statements of the banking companies accessed from www.idx.co.id with those to be used selected using a purposive sampling technique through several specified criteria (Sekaran & Bougie, 2016) such as the absence of delisting and initial public offerings during the research period, the existence of complete financial reports, and the evidence showing the distribution of dividends at least once between 2011 and 2020. It is important to note that conventional banks were used because state-owned companies usually have different dividend policy behavior associated with political economy issues, so we excluded the state-owned banks (Lai et al., 2020). The population of this study are 19 conventional banks which are listed on the IDX. It was discovered that 19 out of 674 companies listed on the Indonesia Stock Exchange in 2020 distributed dividends during the research period and 11 were subsequently used based on their conformity with the criteria during the 10-year research period as attached in Table 1.

Table

Company Code on IDX	Company
AGRO	PT Bank Raya Indonesia Tbk.
BBKP	PT Bank KB Bukopin Tbk.
BDMN	PT Bank Danamon Indonesia Tbk.
BJBR	PT Bank Pembangunan Daerah Jawa Barat dan Banten Tbk.
BNBA	PT Bank Bumi Arta Tbk.
BNGA	PT Bank CIMB Niaga Tbk.
BNII	PT Bank Maybank Indonesia Tbk.
BNLI	PT Bank Permata Tbk.
MAYA	PT Bank Mayapada Internasional Tbk.
MEGA	PT Bank Mega Tbk.
SDRA	PT Bank Woori Saudara Indonesia 1906 Tbk.

Source: www.idx.co.id.

This research uses 8 variables with 9 required proxies as described in Table 2. The measurement of the crisis conditions resulting from the pandemic utilizes GDP growth and inflation rate. These two measurements are considered relevant because in 2020, Indonesia experienced a significant decline in GDP growth, reaching −2.07%, which represents a crisis situation. Restrictions on movement and trade resulted in limited economic activities and systemic negative impacts. Additionally, the inflation rate in 2020 decreased to 1.68% compared to previous years, indicating a hindered circulation of money during the crisis, which further contributed to the slowdown of economic activities and the occurrence of the crisis. Other predictors were measured based on the alignment with previous studies (de Leon, 2020; Romus et al., 2020; Silalahi et al., 2021; Tinungki et al., 2022).

Table 2. Variables and their proxies

Variable	Proxy	Formulation	Expected Sign
Dividend Policy	Dividend Payout Ratio	$DPS=\frac{TotalDividend}{OutstandingShares}$	—–
Crisis due to the COVID-19 Pandemic	Gross Domestic Product Growth	$GDPGrowth=\frac{GD{P}_t-GD{P}_{t-1}}{GD{P}_{t-1}}$	(+)
	Inflation Rate	The data obtained from the Central Statistics Agency of Indonesia	(+)
Profitability	Earnings Per Share	$EPS=\frac{NetIncome}{OutstandingShares}$	(+)
Financial Leverage	Debt to Equity Ratio	$DER=\frac{TotalLiability}{TotalEquity}$	(-)
Investment Opportunity	Market Price to Book Value ratio	$PBR=\frac{MarketPriceperShare}{BookValueperShare}$	(-)
Bank Size	Total Assets	$TA=naturallogarithmicofTotalAsset$s	(+)
Banking Liquidity	Loan to Deposit Ratio	$LDR=\frac{ThirdPartyCreditAmount}{ThirdPartyTotalFunds}$	(-)
Previous Year Dividend	Lagged-1 Dividend	$PYD=Lagged1ofDPS$	(+)

Sources: Alzomaia and Al-Khadhiri (2013); Basse et al. (2014); Hartono and Robiyanto (2023); Hartono et al. (2023); Imran et al. (2013); Kosasih et al. (2021); Kullab et al. (2022); Nadeem et al. (2018); Naveenan et al. (2021); Ong et al. (2018); Patra et al. (2012); Rahma and Syarif (2019); Ramadani and Jumono (2020); Silalahi et al. (2021); Wahjudi (2020); Yusuf and Muhammed (2015).

The statistical analysis instrument used was dynamic panel data regression with the first difference generalized method of moments (FD-GMM) parameter estimation method and the system generalized method of moments (SYS-GMM) with a two-step estimator. The general form of the regression is presented as follows (Biørn, 2017):

(1) $Y_{it}={\alpha }_{it}+\delta Y_{it-1}+{\beta }^{\prime }{X}_{it}+u_{it}$(1)

Where,

(2) $u_{it}={\mu }_i+v_{it}$(2)

Description:

$Y_{it}$: $i$- th exogenous variable for the $t$th period

$\delta$: Predictor influence coefficient of $Y_{it-1}$

$Y_{it-1}$: lagged 1 from $Y_{it}$ as the exogenous variable

${\beta }^{\prime }$: $\left({\beta }_1,{\beta }_2,\cdots ,{\beta }_n\right)$ the coefficient on the influence of exogenous variables from - $i$th individual for the $t$th time period

$X_{it}$: Exogenous variable from the $i$th individual for the $t$th time period

${\alpha }_{it}$: Intercept coefficient of each $i$th individual for the $t$th time period

$u_{it}$: Residual in the $i$th individual period and the $t$th time period

Dynamic panel data regression with FD-GMM and SYS-GMM involved three model specification tests (Biørn, 2017). The first is the instrument validity test conducted through the Sargan test to determine the existence of a correlation between the instrument variable ($Y_{it-1}$) and the error component ($u_{it}$). The second is the parameter consistency test through Arellano–Bond test to determine the existence of a serial correlation between $\mathrm{\Delta }{v}_{i,t}$ and $\mathrm{\Delta }{v}_{i,t-2}$ and the third is the unusual test which involved comparing $\delta$ from $Y_{it-1}$ on GMM with $\delta$ from $Y_{it-1}$ for parameter estimation of Ordinary Least Square Robust Standard Error (OLS-RSE) and Least Square Dummy Variable Robust Standard Error (LSDV-RSE).

The dynamic panel data regression with FD-GMM and SYS-GMM overcomes the problem of habitability due to endogeneity from $u_{it}$ towards $Y_{it-1}$ to ensure static panel data regression and this indicates it has the ability to produce robust parameter estimates (Arellano & Bond, 1991; Arellano & Bover, 1995; Blundell & Bond, 1998). According to Li (2016), GEN is the parameter estimation method with the highest corrective effect on endogeneity problems among exogenous variables in a tested empirical model, while Dang et al. (2018) stated that the GMM estimation method can solve the endogeneity problem in causality test.

The empirical panel data regression models formed were:

(3) $DP{R}_{it}={\beta }_0+{\beta }_{1}GD{P}_{it}+{\beta }_{2}EP{S}_{it}+{\beta }_{3}DE{R}_{it}+{\beta }_{4}PB{R}_{it}+{\beta }_{5}T{A}_{it}+{\beta }_{6}LD{R}_{it}+{\beta }_{7}DP{R}_{it-1}+{\epsilon }_{it}$(3)

(4) $DP{R}_{it}={\beta }_0+{\beta }_{1}IN{F}_{it}+{\beta }_{2}EP{S}_{it}+{\beta }_{3}DE{R}_{it}+{\beta }_{4}PB{R}_{it}+{\beta }_{5}T{A}_{it}+{\beta }_{6}LD{R}_{it}+{\beta }_{7}DP{R}_{it-1}+{\epsilon }_{it}$(4)

Where $DP{R}_{it}$ is the dividend payout ratio on $i$ individual in the $t$ time, $GD{P}_{it}$ is the growth of gross domestic product on $i$ individual in the $t$ time, $IN{F}_{it}$ is the inflation rate on $i$ individual in the $t$ time, $Dumm{y}_{it}$ is the binary dummy variable for the condition of a crisis in the COVID-19 pandemic and no crisis on $i$ individual in the $t$ time, $EP{S}_{i,t}$ is the earning per share on $i$ individual in the $t$ time, $DE{R}_{i,t}$ is the debt-to-equity ratio on $i$ individual in the $t$ time, $PB{R}_{i,t}$ is the market price to book value ratio on $i$ individual in the $t$ time, $T{A}_{i,t}$ is the natural logarithm transformation of the total asset on $i$ individual in the $t$ time, $LD{R}_{i,t}$ is the loan-to-deposit ratio on $i$ individual in the $t$ time, $DP{R}_{it-1}$ is the lagged 1-year period from dividend payout ratio on $i$ individual in the $t$ time,${\beta }_0$ is the constant parameters, ${\beta }_{1\cdots 7}$ is the coefficient on the effect of exogenous variables on endogenous variables, ${\epsilon }_{i,t}$ is the error on $i$ individual in the $t$ time.

4. Results and discussion

4.1. Results

Table 3 shows the descriptive statistics of the sample data and each proxy has overdispersion and equidispersion data distribution with the overdispersion observed to have occurred in DPS used as a proxy for endogenous variables and EPS as a proxy for profitability predictors, while GDP, INF, DER, PBR, TA, and LDR proxies are equidispersion for other exogenous variables. The minimum DPS value of 0 indicates that the object observed in the year did not distribute dividends and this led to the assumption that IDR 0 was distributed as dividends. Moreover, the GDP minimum value of −2.070% or −0.0207 implies that the growth was negative in 2020, while the negative value of EPS proxy shows that the companies in a certain period experienced negative profit. It is important to note that the extreme values were included as observation objects to determine the level of influence the extreme behavior has on dividend policy.

Table 3. Descriptive statistic

Proxies	N	Min	Max	Mean	St. Dev.
DPS	110	0.000	301.561	35.169	50.780
GDP	110	−2.070	6.170	4.587	2.270
INF	110	1.680	8.380	4.234	2.180
EPS	110	−99.718	433.957	98.893	114.468
DER	110	3.209	14.748	7.207	2.721
PBR	110	0.293	5.695	1.532	1.108
TA	110	14.902	19.454	17.879	1.263
LDR	110	0.537	1.623	0.886	0.157

Source: www.idx.co.id, processed

The dynamic panel data regression analysis was initiated with three model specification tests. The first was the consistency conducted through the Arellano–Bond test as presented in Table 4. It was discovered that a p-value of 0.374 for FD-GMM and 0.066 for SYS-GMM were recorded in the 2nd-order Arellano–bond test on the estimated parameter with GDP proxy while the values were 0.573 and 0.197, respectively, for INF proxy. These results showed that the overall parameter estimates are consistent with p-value > α (5%).

Table 4. Arellano–Bond test

Order	GDP				INF
	FD-GMM		SYS-GMM		FD-GMM		SYS-GMM
	z	Prob. > z	z	Prob. > z	Z	Prob. > z	Z	Prob. > z
(1)	−0.825	0.409	−2.705	0.007	−0.518	0.604	−2.096	0.036
(2)	−0.890	0.374	−1.840	0.066	−0.563	0.573	−1.290	0.197

Source: www.idx.co.id, processed.

The second is the instrument validity determined through the Sargan test as presented in Table 5 and the results for the parameter estimation with GDP proxy showed that the p-value for FD-GMM and SYS-GMM was 1.000 each and the same was also recorded for the INF proxy. This signifies the overall parameter estimation has a p-value > α (5%), thereby, indicating the instrument variable in the model is valid because it is not correlated with the error component.

Table 5. Sargan Test

Indicator	GDP		INF
	FD-GMM	SYS-GMM	FD-GMM	SYS-GMM
${\chi }^2$	4.249	4.102	2.672	5.273
$Prob.>{\chi }^2$	1.000	1.000	1.000	1.000

Source: www.idx.co.id, processed.

The third is the unbiased test which is presented in Tables 6 and 7, and the results showed that the parameter estimation with GDP proxy for FD-GMM was δ LSDV-RSE < δ OLS-RSE < δ FD-GMM and this means it is upward biased while SYS-GMM had δ LSDV-RSE < δ SYS-GMM < δ OLS-RSE which indicates it is unbiased. Meanwhile, the INF proxy for FD-GMM was δ LSDV-RSE < δ FD-GMM < δ OLS-RSE which means it is unbiased while the SYS-GMM produced δ LSDV-RSE < δ SYS-GMM < δ OLS-RS and this means it is also unbiased. The three model specification tests conducted showed that the parameter estimation with GDP proxy for FD-GMM is unfit for use but GDP proxy for SYS-GMM and INF proxies for FD-GMM and SYS-GMM were fit and feasible to be applied in the research.

Table 6. Dynamic panel data regression analysis and unbiased test for GDP proxy

Proxy of Exogen Variable	LSDV-RSE	FD-GMM	SYS-GMM	OLS-RSE
Constant	236.671*** (98.845)	499.737 (545.247)	260.104 (422.821)	111.695*** (45.040)
DPS L1.	0.113** (0.074)	0.564** (0.370)	0.180*** (0.097)	0.230** (0.156)
GDP	−3.051** (1.944)	−1.905*** (0.895)	−1.901* (1.775)	−2.612** (2.044)
EPS	0.436*** (0.115)	0.497*** (0.089)	0.374*** (0.064)	0.348*** (0.064)
DER	−3.253*** (1.244)	4.018 (4.570)	−4.496*** (1.350)	−0.481 (0.896)
PBR	2.211 (2.290)	−3.188 (3.951)	−8.637 (10.514)	0.136 (1.795)
TA	−11.708*** (5.069)	−29.046 (36.406)	−7.278 (23.244)	−3.899*** (1.904)
LDR	−9.084 (24.026)	−31.297 (163.392)	−97.261 (84.996)	−40.704*** (19.316)
No. of Instrument	—–	43	51	—–
No. of obs.	99	88	99	99
No. of group	11	11	11	—–
Wald	—–	6547.92***	1663.57***	—–
F-statistics	76.09***	—–	—–	29.19***
R2	0.728	—–	—–	0.798

Source: www.idx.co.id, processed.

Description: the numbers in parentheses are the standard error and robust standard error values. In one-tailed statistics, *** means a significance level of α = 5%, ** means a significance level of α = 10%, and * means a significance level of α = 15%.

Table 7. Dynamic panel data regression analysis and unbiased test for INF proxy

Proxy of Exogen Variable	LSDV-RSE	FD-GMM	SYS-GMM	OLS-RSE
Constant	261.873*** (79.442)	5874.540*** (3141.251)	415.581** (302.442)	126.686*** (39.811)
DPS L1.	0.156*** (0.057)	1.633*** (0.734)	0.252*** (0.057)	0.268*** (0.143)
INF	−3.140*** (1.192)	−14.200*** (6.484)	−2.641*** (1.028)	−3.449*** (1.315)
EPS	0.412*** (0.109)	0.560*** (0.088)	0.373*** (0.037)	0.337*** (0.058)
DER	−2.455*** (0.769)	24.199** (16.343)	−5.421*** (1.795)	0.021 (0.844)
PBR	2.072 (2.339)	−94.957*** (52.470)	−1.041 (9.526)	0.071 (1.750)
TA	−13.683*** (4.537)	−343.994*** (184.163)	−20.970** (15.566)	−4.722*** (1.723)
LDR	−2.657 (25.825)	270.426*** (137.920)	−5.947 (41.640)	−40.113*** (19.382)
No. of Instrument	—–	43	51	—–
No. of obs.	99	88	99	99
No. of group	11	11	11	—–
Wald χ2	—–	11249.00***	3754.55***	—–
F-statistics	78.21***	—–	—–	26.72***
R2	0.724	—–	—–	0.806

Source: www.idx.co.id, processed

Description: the numbers in parentheses are the standard error and robust standard error values. In One-tailed statistics, *** means a significance level of α = 5%, ** means a significance level of α = 10%, and * means a significance level of α = 15%.

The parameter significance was also determined with a simultaneous test through the application of Wald and partial test combined with the Z-test. The Wald test showed that the p-value of the estimated parameter with GDP proxy on SYS-GMM was 0.000 < α (5%) and the value for INF proxy was 0.000 < α (5%) on FD-GMM and 0.000 < α (5%) on SYS-GMM. This implies that at least one exogenous variable affects the exogenous variable. The partial with Z-test was also conducted with the hypothesis test.

Statistically, partial parameter significance tests on estimated parameters indicate that several predictors have no significant impact. These findings were observed in the PBR proxy of the SYS-GMM model for the crisis proxy that utilizes GDP and INF, the TA proxy of the SYS-GMM model for the crisis proxy that employs GDP, and the LDR proxy of the SYS-GMM model for the GDP and INF proxies. These results can be examined through descriptive statistical analysis, revealing an imbalance in the dispersion conditions of the proxies measuring the independent variables with respect to the proxy of the independent variable. The dispersion conditions of PBR, TA, and LDR are equidispersion, while the DPS proxy exhibits overdispersion. This imbalance in dispersion conditions further strengthens the assertion that the results of the significance tests on the parameters are insignificant (Tinungki, 2019).

4.2. Discussion

The crisis caused by the COVID-19 pandemic which was proxied by GDP growth and the annual inflation rate was observed not to have a positive effect on dividend policy and this denotes H1 and H2 were rejected. This was observed to be contrary to the findings of Abdulkadir et al. (2015), Krieger et al. (2021), Ong et al. (2018), and Romus et al. (2020) that crisis conditions usually reduce dividend rates and that GDP growth has a positive effect on dividend policy but in line with the results of Tinungki et al. (2022) that the crisis caused by the pandemic proxied by GDP growth has a negative effect on dividend policy. This shows that the banking companies in Indonesia tend to pay higher dividends during the economic crisis caused by the pandemic. It was also discovered that the robust test for the results is consistent with the annual inflation proxy. Moreover, the dividend policy behavior was perceived to have provided a positive signal to the capital market where trading conditions were observed to be sluggish due to the crisis (Renitia et al., 2020). This assumption was supported by the findings of Tinungki et al. (2022) that there was a positive and rapid reaction to the announcement of dividend distribution.

Profitability was found to have a positive effect on dividend policy based on all feasible parameter estimates and this means H3 was accepted. The results are in line with Kullab et al. (2022), Marfo-Yiadom and Agyei (2011), and Silalahi et al. (2021) that dividend policy is positively affected by profitability. It is also important to note that the use of profitability as the main predictor of dividend policy is relevant considering the fact that the dividends distributed are based on the profit earned by the company (Dewasiri et al., 2019; Lestari, 2018). The same trend was reported to have occurred during the economic crisis caused by the COVID-19 pandemic (Tinungki et al., 2022). The findings of this study indicate a positive impact, demonstrating that even in times of crisis, the banking sector in Indonesia maintains positive profitability, aligning with a positively established dividend policy. Consequently, the positive profitability exhibited by these companies enables Indonesian banks to distribute dividends in a positive manner (Hartono & Raya, 2022).

The results also showed that financial leverage has a negative effect on dividend policy as indicated by the parameter estimation with GDP growth proxy on SYS-GMM as well as annual inflation proxy on SYS-GMM, and this indicates H4 was accepted. Meanwhile, it was observed from the parameter estimation on the annual inflation proxy for FD-GMM that financial leverage has a positive effect on dividend policy but it was corrected by the estimation of the SYS-GMM parameter which had a stronger robust nature. The findings were found to be in line with Naveenan et al. (2021), Sharma and Bakshi (2019), and Wahjudi (2020) that financial leverage has a negative effect on dividend policy, and this simply infers that a higher level of corporate debt is expected to provide burden on interest expenses on debt expected by a company with subsequent effect on its profitability. This further leads to the reduction of dividends to be distributed to shareholders (Sharma & Bakshi, 2019; Wahjudi, 2020). Meanwhile, during the pandemic, there was lower debt and this allowed companies to distribute higher dividends than the previous year. This is contrary to the findings of Tinungki et al. (2022) that there is a positive effect of financial leverage on dividend policy. This signifies that there are differences in the influence of financial leverage on the banking sector companies examined. On the other hand, the high liabilities of banks are also attributed to the increased level of funds accumulated. This is due to the decrease in the circulation of money during times of crisis. During such crises, the systemic disruption of business activities hampers companies’ ability to channel funds to debitors, resulting in a tendency for liabilities to increase. Therefore, the tendency for increased liabilities can have a positive impact on dividend policies during the pandemic.

Investment opportunity was also found to have a negative effect on dividend policy as evident in the parameter estimation with the annual inflation proxy for the FD-GMM parameter estimation method and this implies H5 was accepted. This is similar to the findings of Nadeem et al. (2018), Patra et al. (2012), and Rizqia et al. (2013), and this indicates a higher investment opportunity is expected to cause a reduction in dividend rate because the profits obtained by the company tend to be earmarked for further investment (Hartono et al., 2021; Patra et al., 2012). This was observed to be relevant to stock prices during the economic crisis caused by the relative decline in the pandemic through the use of the market price to book value ratio as the proxy. This simply shows that banking companies in Indonesia have low investment opportunities during the pandemic and this tends to increase dividend rates.

Bank size was reported not to have any positive effect on dividend policy and this led to the rejection of H6. The finding is that bank size affected dividend policy negatively. This was observed from the parameter estimation with the annual inflation proxy for both the FD-GMM and SYS-GMM and discovered to be in line with the findings of Hartono and Matusin (2020) and Lestari (2018). It shows that large banks have greater expenses but the expenses were reduced during the pandemic, thereby, leading to an increase in the dividend rates (Kaźmierska-Jóźwiak, 2015; Lestari, 2018; Tinungki et al., 2022).

Banking liquidity was also found to have both positive effects on dividend policy and this means H7 was rejected. This is in line with the findings of Badu (2019), Bostanci et al. (2018), and Karauan et al. (2017) which indicate that higher loans provided to creditors normally lead to a higher profit due to interest to be collected. It is important to reiterate that an increase in the profit made by the bank usually leads to an increment in the dividends distributed (Karauan et al., 2017). The observation from the pandemic period reveals that a higher dividend rate to shareholders is an indication of the higher profits recorded by the bank from the interest on creditors’ debts. In the context of a crisis, bank liquidity tends to increase due to the accumulation of a substantial amount of funds, which cannot be effectively disbursed to borrowers as a result of hindered business cycles. Consequently, this argument reinforces the notion that banking liquidity does not exert any influence on dividend policies.

The previous year’s dividend has a positive effect on dividend policy as signified by the overall feasible estimated parameters and this implies H8 was accepted. This is similar to the previous research by Alzomaia and Al-Khadhiri (2013), Maladjian and El Khoury (2014), and Tinungki et al. (2022), thereby, denoting that a higher dividend rate in a certain period influences another period. It was discovered that banking companies in Indonesia tend to increase their dividend levels during the economic crisis caused by the pandemic to provide a positive signal to the capital market experiencing sluggish trading activity (Hartono & Matusin, 2020; Tinungki et al., 2022). This exemplifies the relevance of the dividend signaling theory, which posits that dividends should be set positively compared to the previous year, even amidst a crisis.

5. Conclusion

This research found that the economic crisis caused by the COVID-19 pandemic had a negative effect on dividend policy with most of the banking companies in Indonesia discovered to have the tendency to increase their dividend in order to provide a positive signal to the sluggish trading in the capital market. Profitability, banking liquidity, and previous year dividends all had a positive effect while financial leverage, investment opportunity, and bank size had a negative effect on dividend policy. These predictors further strengthened the argument related to the negative influence of the pandemic on dividend policy.

Statistical testing reveals that the estimations of suitable parameters are capable of representing dividend policy behavior during crises for conventional banks in Indonesia. This is attributed to the statistical analysis tool being a highly powerful method for estimating panel data structures (Biørn, 2017; Hartono & Robiyanto, 2023). Moreover, this analytical tool is effective in addressing endogeneity issues and generating coefficients that align most appropriately with the direction of influence compared to other estimation methods (Li, 2016). This assertion is supported by congruent research conducted by Tinungki et al. (2022) examining dividend policies during the pandemic crisis for non-financial companies in Indonesia, Tinungki et al. (2022) investigating dividend policies during the pandemic crisis for green index companies, as well as Hartono and Raya (2022) studying dividend policies during the pandemic crisis for manufacturing companies.

The findings serve as a reference for investors to make informed investment decisions towards achieving optimal returns and also for banking companies to implement a dividend policy governance that can increase firm value (Salvatori et al., 2020). This is expected to be based on the consideration of the determinants with positive signals to maintain the stability of trading activities of each issuer in the capital market. Moreover, for banks, a positive dividend policy during a crisis is expected to send a positive signal to the sluggish market affected by the crisis. Furthermore, during a crisis, banking institutions have limited investment opportunities, making a positive dividend policy more relevant in preventing agency conflict.

It is, however, important to note that this research only examines the dividend policy during the COVID-19 pandemic. Therefore, it is suggested that further research focuses on the dividend policy in pre, during, and post-crisis conditions with special attention on the COVID-19 pandemic (Tinungki et al., 2022). There is also the need to analyze the market reaction to dividend distribution during this period, specifically for the banking sector in Indonesia (Hartono & Raya, 2022). Moreover, mediation-moderation test was not conducted on the predictors used, and this signifies that there is a need to focus on testing the effect of predictors on dividend policy moderated by investment opportunities, which is one of the main predictors to determine the dividends associated with growth and subsequent investment opportunities (Hartono et al., 2021).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. The datasets obtained from some financial platforms which can be accessed freely, i.e. www.idx.co.id.

Competing interests

The authors declare that they have no competing interests.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes on contributors

Puguh Budi Santosa

Puguh Budi Santosa is a Ph.D. student from Doctoral Program in Economics, Faculty of Economics and Business, Universitas Diponegoro, Semarang, Indonesia. His research interests are banking, corporate governance and financial management.

Irene Rini Demi Pangestuti

Irene Rini Demi Pangestuti is an Associate Professor of the Department of Management, Faculty of Economics and Business, Universitas Diponegoro, Semarang, Indonesia. Her research interests are banking, financial management, capital market and corporate governance.

Sugeng Wahyudi

Sugeng Wahyudi is a Professor of the Department of Management, Faculty of Economics and Business, Universitas Diponegoro, Semarang, Indonesia. His research interests are banking, financial management, capital market, behavioral finance and corporate governance.

Harjum Muharam

Harjum Muharam is a Professor of the Department of Management, Faculty of Economics and Business, Universitas Diponegoro, Semarang, Indonesia. His research interests are Islamic finance, banking, financial management, capital market and corporate governance.