Corporate Social Responsibility Expenditures and Bank Performance: Role of Size Among Listed Banks in Ghana

Abstract

This study investigates the relationship between listed Ghanaian banks’ financial performance and corporate social responsibility (CSR), given the anticipated increase in businesses’ social duties. This study utilizes a panel autoregressive distributive lag model (Panel ARDL) to examine the impact of CSR on bank financial performance, as well as the moderating effect of bank size on CSR and financial performance, using return on assets as the measure of financial performance. All banks listed on the Ghana Stock Exchange (GSE) whose financial statements are readily accessible online, in print, or on their websites are chosen using convenience sampling. The sample spans 14 years, from 2010 to 2023. The results are shown for both the long and short run. Contrary to the expectations of many proponents of CSR, we find that firms incorporating CSR in their undertakings have negative financial performance. Additionally, the study finds that, relative to smaller banks, larger banks are able to alleviate this negative effect of CSR on performance by a certain magnitude. Therefore, not only should banks be strategic in their CSR implementation, but they should strive to grow their assets to the level where the negative effects of undertaking CSR could be reduced, if not entirely eliminated. To achieve this growth, the level of assets to keep is found to be above GHC 3922.52 million.

1. Introduction

The traditional pursuit of profit maximization by businesses has often exacerbated societal issues, such as environmental degradation, labor rights violations, and food safety concerns, creating challenges for sustainable growth. In response, businesses, particularly publicly listed ones, face growing pressure from stakeholders to integrate sustainable practices into their operations. One such practice is corporate social responsibility (CSR), which emphasizes a company’s obligation to its community. CSR has evolved significantly since its conceptual origins in the 1920s, gaining prominence as businesses try to balance financial objectives with broader stakeholder interests. Under the Triple Bottom Line (TBL) framework, CSR addresses economic, social, and environmental dimensions, aligning profitability with sustainability (Peloza & Shang, 2011). While financial management remains critical for operational success (Alexander & Buchholz, 1978), CSR has emerged as a complementary dimension that can enhance both reputation and financial outcomes (Kumar, 2015).

CSR’s benefits extend beyond compliance to include voluntary strategies that enhance corporate reputation and financial performance. As CSR becomes a global phenomenon, nations have introduced campaigns and regulatory measures to encourage its adoption. CSR disclosures—which inform stakeholders about business practices affecting society and the environment—play a crucial role in enhancing transparency and accountability (Gray et al., 1996; Salehi et al., 2020). Research highlights the numerous societal benefits of CSR, including improved living standards, environmental conservation, and technological advancements, while businesses gain customer loyalty, investor confidence, and competitive advantages (Bui & Huynh, 2020; Van Nguyen et al., 2022). The various benefits illustrate how CSR has grown to become an integral component of contemporary business strategy. By addressing social and environmental concerns alongside financial goals, CSR empowers businesses to contribute to the broader societal good while fostering long-term success.

The banking sector, a cornerstone of Ghana’s economy, has historically demonstrated robust profitability, albeit recently challenged by events like the Domestic Debt Exchange Program (DDEP). With their extensive branch networks and heightened public visibility, banks are under greater social pressure to engage in CSR compared to other industries (Cornett et al., 2016). Recognizing the strategic value of CSR, banks leverage such initiatives to mitigate risks, improve reputation, and secure long-term sustainability. CSR practices by Ghanaian banks often extend beyond compliance, reflecting a deliberate alignment with community development priorities. For example, banks frequently undertake CSR projects that involve building schools, hospitals, and clean water facilities, addressing critical societal needs (R. E. Hinson & Adjasi, 2009; R. Hinson et al., 2010). Such initiatives not only underscore the banks’ commitment to societal well-being but also enhance their corporate image and customer loyalty.

Notably, the financial sector in Ghana contributes significantly to the country’s GDP and employs a large portion of its population, placing banks in a unique position to drive sustainable development. Despite the growth and visibility of CSR in Ghanaian banks, empirical studies on its impact remain limited and inconclusive (Maama, 2021). For instance, most studies examining the relationship between CSR activity and the performance of banks in Ghana report a positive link (Anim et al., 2021; Boachie, 2020; Mensah et al., 2017). However, Gasti et al. (2019) find a negative relationship, and Gatsi et al. (2016) also report a negative relationship between CSR and financial performance in the long term when analyzing a broader sample of all firms listed in Ghana. Interestingly, both Boachie (2020) and Maama (2021) construct indexes of CSR activity—or ESG in the case of Maama (2021)—while Gasti et al. (2019) focus on CSR expenditures as their key measure. These divergent methodologies and findings underscore the need for further investigation into the complexity of the CSR–financial performance nexus.

We extend Gasti et al. (2019) by broadening the timeframe of analysis and employing different performance measures. While Gasti et al. (2019) assess financial performance using the ratio of market to book value, we utilize more conventional measures, such as return on assets (ROA) and return on equity (ROE), to provide a more comprehensive understanding. Additionally, whereas Gasti et al. (2019) control for growth using changes in interest income, we adopt a broader macroeconomic perspective by incorporating growth in Ghana’s GDP to capture industry-wide and economic effects. These methodological advancements enhance the robustness and relevance of our findings, contributing to a better understanding of CSR’s impact on financial performance.

Additionally, recent studies have emphasized the moderating effects of variables on the CSR–performance relationship (Li & Zhang, 2019; Ye et al., 2021). There is evidence, according to several researchers (Christmann & Taylor, 2001; Muller & Kolk, 2010; Gu, 2012), that businesses engaged in international trade can more effectively carry out their CSR. Similarly, according to Hernández et al. (2020), the correlation between CSR and economic performance is stronger for larger firms. This study suggests that it is possible that the size of the bank can influence the relationship between CSR and financial performance. Larger banks, by virtue of their resources and stakeholder influence, may derive greater financial benefits from CSR compared to smaller institutions. Conversely, smaller banks might achieve comparable results through targeted CSR investments, raising questions about the moderating role of bank size. Understanding these differences is essential for developing tailored CSR strategies that maximize impact across banks of varying sizes.

The primary objective of this study is to examine the impact of CSR on the financial performance of listed banks in Ghana, with a specific focus on the moderating effect of bank size. We aim to clarify the relationship between CSR and financial outcomes while uncovering how bank size influences this relation. The study not only evaluates the direct impact of CSR on financial performance but also sheds light on the conditions under which CSR initiatives are most effective, providing actionable insights for banks striving to enhance their societal and economic contributions. This research provides valuable insights for academics, practitioners, and policymakers. For academics, it offers empirical evidence to enrich theoretical frameworks and spark further investigation into the factors influencing CSR outcomes. Policymakers can leverage these findings to design incentives and regulatory structures that promote sustainable banking practices. Banks, as key stakeholders, can utilize the results to optimize their CSR strategies, ensuring alignment with financial and reputational objectives. Moreover, the study addresses the practical challenges that banks face in balancing profitability with social responsibility, offering a framework for integrating CSR into their core operations.

The societal implications of this research are equally significant. CSR practices have the potential to drive sustainable development, particularly in emerging markets, like Ghana, where banks play a pivotal role in economic growth. By aligning corporate objectives with societal needs, banks can foster stronger community relationships, enhance customer trust, and contribute to national development goals. This study underscores how CSR can serve as a catalyst for economic and social progress, emphasizing the importance of strategic CSR investments for fostering sustainable growth. As CSR practices continue to evolve, their transformative potential for businesses, communities, and economies becomes increasingly evident. By illuminating the link between CSR and financial performance, this research provides a roadmap for leveraging CSR as a tool for achieving holistic development outcomes.

In summary, this research advances our understanding of CSR’s role in the banking sector, highlighting the impact of bank size on its effectiveness. The findings have practical implications for promoting sustainable growth within Ghana’s banking industry while addressing broader societal challenges. By exploring these critical intersections, the study contributes to the growing body of knowledge on CSR and offers actionable insights for building a more inclusive and sustainable financial ecosystem.

The reminder of this paper is organized as follows. The theoretical framework, literature review, and hypothesis development will be covered in Section 2. Section 3 will also address the methodology, research design, sample, and data sources. The results of the study will be discussed in Section 4. Finally, conclusions and recommendations will all be summarized in Section 5.

2. Literature Review and Hypothesis Development

2.1. Theoretical Framework

Corporate social responsibility (CSR) is the notion that a business should expend resources to further social goals that are neither among the firm’s objectives nor are legally required (McWilliams & Siegel, 2001). Corporate social responsibility (CSR) is frequently conceptualized within the broader scope of stakeholder theory, legitimacy theory, and the resource-based view of the firm.

Stakeholder theory, popularized by Freeman (1984) and later expanded by Phillips et al. (2003), views that firms bear obligations not only to shareholders, but to all stakeholders, comprised of individuals, groups, or institutions that can influence the firm in reaching its objectives. (Phillips et al., 2003). Under CSR, business is expected to address the Triple Bottom line accounting concept of social, ecological, and economic sustainability (Peloza & Shang, 2011). Pursuing CSR may be incongruent with the firm’s core goals, because pursuing activities outside a firm’s goals may compromise financial performance. But in a longer-run perspective, firms that consider stakeholder welfare, and CSR as well, may be regarded as pursuing a forward-looking, strategic perspective. Such activities are effective investments against future uncertainty (Freeman, 2023). If successful, these activities should improve long-run firm performance. The business case for CSR then rests on an underlying societal expectation that firms should address goals beyond the narrow economic and legal obligations and extend objectives to include ethical and philanthropic considerations (Carroll, 1991; Carroll & Shabana, 2010). CSR can be viewed as an expansionary form of stakeholder strategy by widening the realm of stakeholders to include the disenfranchised. Banks, in particular, engage with multiple constituencies, such as depositors, borrowers, regulators, and local communities, whose welfare and perceptions can shape the bank’s reputation and long-term viability. Under this theory, CSR is not solely an ethical imperative but a strategic tool to mitigate risks by building trust and goodwill with diverse stakeholders.

Legitimacy theory complements stakeholder theory by emphasizing how organizations seek to align their activities with the values and norms of the larger society in order to maintain their “license to operate.” As posited by Suchman (1995), legitimacy is conferred when societal members believe that an organization’s actions are desirable, proper, or appropriate within a socially constructed system of norms and values. In the banking sector, where oversight by regulators and scrutiny by the public are persistent, CSR initiatives can function as visible demonstrations of adherence to broader societal expectations. By engaging in socially responsible lending, environmental sustainability programs, and community development projects, banks reinforce their legitimacy and reduce the likelihood of reputational damage or regulatory constraints that could compromise financial performance.

The resource-based view (RBV) (J. Barney, 1991; J. B. Barney & Clark, 2007) further enriches the theoretical foundation by proposing that firms attain a competitive advantage through the possession and effective use of valuable, rare, inimitable, and non-substitutable resources. CSR activities can generate and strengthen intangible resources, such as brand equity and reputational capital, which are particularly potent in differentiating banks from their competitors. By fostering community trust and a robust reputation for social and environmental responsibility, banks accumulate intangible assets that may lower funding costs, attract loyal customers, and enhance employee commitment. These benefits can, over time, translate into stronger financial performance.

Taken together, stakeholder theory, legitimacy theory, and RBV underscore that CSR, rather than merely representing an altruistic or compliance-driven endeavor, can serve as a vital strategic instrument that both mitigates short-term risks and contributes to a sustainable competitive advantage in the long run.

2.2. Literature Review

Since the work of Waddock and Graves (1997), there have been a large number of empirical studies attempting to find the link between CSR and firm performance. Soana (2011) highlight the heterogeneity of results for different measures of CSR, including reported activities, surveys of stakeholders, measures of reputation, and one-dimensional and multi-dimensional indicators of CSR engagement. Much of the literature has focused on the manufacturing and resource sectors, but there is growing literature on CSR and related issues, like ESG and sustainability, in banking. Banks in particular have substantial indirect effects through the activities of firms they support through their loans and are thereby seen as being keys to change (Weber & Feltmate, 2016).

Studies typically use standard measures to capture financial performance, like return on equity and return on assets. Other indicators of performance include measures of risk (Hojer & Mataigne, 2024), liquidity creation (W. D. Chen & Chen, 2024; Zheng et al., 2023), non-performing loans (Boussaada et al., 2023), bank stability (Gaies & Jahmane, 2022), and cost of bank loans to borrowers (Goss & Roberts, 2011). CSR activity is typically captured by spending on such activities but can also be captured by an index of CSR activities or multi-output measures (Wu et al., 2017; Wu & Shen, 2013). Belasri et al. (2020) use efficiency as their preferred measure of financial performance. In their study of 184 banks in 41 countries, they find that a discernible impact of CSR on efficiency is evident only in developed countries and then only where investor protection is high. Alam et al. (2022), also using an efficiency measure of performance, find a link between CSR and efficiency in a sample of four MENA countries: Kuwait, Qatar, Saudi Arabia, and the United Arab Emirates.

CSR activities may be of less interest to firms in the developing world. To the extent CSR is income-elastic, consumers in countries with lower per-capita incomes may be less willing to pay for such activities. Evidence is mixed. Hsu and Bui (2022) find consumer willingness to pay for CSR in Indonesia but not in Vietnam. Vahdati et al. (2015) find support for consumer willingness to pay in Iran, albeit their study is for one industry in one city. Fatma and Rahman (2016) in a study of Indian banks find that customers will pay for banks’ CSR activities if they are aware of the activities. More generally, Kim and Bae (2016) find that communication is key for consumers to accept firm’s CSR activities, and more importantly, communication must be in a culturally relevant form.

Banks may be motivated to undertake CSR activities, but unless they can credibly signal that the activities are legitimate, and not greenwashing, they may not be able to accrue the benefit which will reduce their incentive to pursue CSR activities (Houston & Shan, 2022). To address this problem, banks may employ third-parties to verify reporting, they may signal by lending to firms already noted for CSR activities (K. J. Huang et al., 2024), and they may support government adoption of reporting standards (Khan et al., 2021). While viewed as a problem by the general public, it may not occur as frequently as assumed. Pope and Wæraas (2016) find that greenwashing depends on the link from firms to consumers. But they also note that competition among firms may also help rein in false CSR reporting (Pope & Wæraas, 2016, pp. 183–184). The Ghanaian banking sector may be somewhat monopolistically competitive (Azumah et al., 2023; Biekpe, 2011), while there is evidence suggesting that competition in the sector has been increasing (Alhassan & Ohene-Asare, 2016). This suggests an environment in which greenwashing may be mitigated.

Several studies of CSR in banking explore the relationship between CSR and financial performance within developing economies, including Africa (Bugandwa et al., 2021), Asia Pacific (Alamsyah & Muljo, 2023; Nathania & Ekawati, 2024), Sub-Saharan Africa (Deigh & Farquhar, 2021; Siueia et al., 2019), Middle East and North Africa (Buallay et al., 2020; El Khoury et al., 2023; Jamali et al., 2020; Maside-Sanfiz et al., 2024), and a sample of emerging market economies (Azmi et al., 2021; Shakil et al., 2019). These studies find a relationship between CSR and their measure of financial performance. Laskar and Maji (2017) looking at a mix of developed and developing Asian countries, find evidence of the link between CSR and financial performance for the developed countries and for one of the two developing countries. Their measure of CSR is an index constructed from firm CSR disclosure under the Global Reporting Initiatives.

Because banks face different regulatory regimes, studies of the link between CSR and firm performance explore firms within a single country. Recent studies include China (J. Huang et al., 2017), Egypt (El Sayad & Diab, 2022; Shahwan & Habib, 2023), Ethiopia (Eyasu & Arefayne, 2020) India (Ampoumet et al., 2024; Chowdhury et al., 2024; Maqbool & Zameer, 2018), Kenya (Barako & Brown, 2008; Muchiri et al., 2022), Lebanon (Hassanein & Yeşiltaş, 2021), Malaysia (Abueid, 2022), Mauritius (Ramlugun & Raboute, 2015), Morocco (Abbass et al., 2023), Mozambique (Siueia et al., 2019), Nigeria (D. Chen et al., 2024; Madugba & Okafor, 2016), Poland (Matuszak et al., 2019), Saudi Arabia (Shaddady & Alnori, 2024), South Africa (Naurikay & Obalade, 2023), Turkey (Akben-Selcuk, 2019), and the European Union (Bătae et al., 2021; Boussaada et al., 2023; Chiaramonte et al., 2022; Di Tommaso & Thornton, 2020).

Not all CSR activities are as important so the evidence of the link between CSR and financial performance may depend on which measure is chosen. López-Penabad et al. (2023) find no evidence for an impact of bank investment in environmental initiatives on financial performance. Similarly, Laskar and Maji (2016) find that CSR activities on employee and human rights initiatives are most important, while more general initiatives with an external focus are least important. Shabbir et al. (2020) for a large panel of firms in Pakistan find a link between social and governmental activity reporting with firm performance but not for environmental activities, and they find that the relationships are non-linear. Shakil et al. (2019) in a sample of banks in developing countries find an impact of environmental and social activities but not of governance. However, studies breaking down governance more finely find some aspects important. Rojas Molina et al. (2023) and Shakil et al. (2021) find gender diversity important and that board independence is not significant. Pham et al. (2024) in a study of Southeast and East Asian banks find that governance is important, while social and environmental initiatives are detrimental to performance. Alam et al. (2022) in their sample of the four Gulf states find that environmental CSR has a positive effect on efficiency, but social CSR activities only have an effect on non-Islamic banks in their sample.

Ghana too has been the focus of several studies of the impact of CSR. There are several qualitative studies characterizing CSR in Ghanian banks, investigating the type of CSR information the banks report. Nyarku and Hinson (2017) characterize the types of CSR disclosures of banks, noting that banking disclosures on CSR activities tend to focus on activities related to population health, education, and sports development. R. Hinson et al. (2010) link quantity and type of CSR information banks report on their website to the bank’s characteristics. Boateng (2016) finds that foreign banks provide more CSR reporting than domestic banks; and among domestic banks, state-owned banks focus more on their philanthropic activities. Anaman et al. (2023) assess bank employees’ views on bank Environmental, Social, and Governance (ESG) performance, with positive findings on Social and Governance measures.

Most studies of the relationship between CSR activity and financial performance of banks in Ghana find a positive link (Anim et al., 2021; Boachie, 2020; Mensah et al., 2017), although Gasti et al. (2019) find a negative relationship. Gatsi et al. (2016) also find a negative relationship between CSR and financial performance for a wider sample of all firms listed in Ghana. Both Boachie (2020) and Maama (2021) construct indexes of CSR activity (or ESG in the case of Maama 2021), while Gasti et al. (2019) use CSR expenditures.

2.3. Hypothesis Development

The relationship between corporate social responsibility (CSR) and financial performance has been widely studied, with mixed findings. While some research suggests that CSR enhances financial performance by improving reputation, customer loyalty, and stakeholder trust, others argue that CSR can be a financial burden, diverting resources from profit-maximizing activities. Given these conflicting perspectives, understanding the factors that influence this relationship is crucial.

One key factor that may shape the CSR–financial performance link is bank size. Larger banks generally have greater financial capacity, broader stakeholder reach, and stronger risk management structures, allowing them to absorb CSR costs more effectively and leverage CSR initiatives for competitive advantage. In contrast, smaller banks with limited resources may struggle to balance CSR commitments with profitability.

Although studies on CSR in Ghanaian banks have examined its direct impact on profitability, few have explored how bank size influences this relationship. Musah et al. (2018) provides rare empirical evidence, showing that Ghanaian banks engaging in CSR tend to report higher profitability, particularly in terms of return on equity (ROE). The study also found that larger banks are more likely to participate in CSR, suggesting that firm size plays a role in CSR engagement. However, whether bank size alters the effectiveness of CSR in enhancing financial performance remains unclear.

To address this gap, this study examines both the direct effect of CSR on financial performance and whether bank size plays a moderating role in this relationship. Based on this, we propose the following hypotheses:

H1. 

CSR is associated with the financial performance of banks.

H2. 

Bank size moderates the relationship between CSR and financial performance.

3. Research Methodology

3.1. Sample and Data

This study focuses on Ghana’s banking sector, specifically institutions listed on the Ghana Stock Exchange (GSE). The selection of this population addresses gaps in prior research and is particularly relevant given the limited use of banks as case studies in CSR research, which often focuses on manufacturing and other sectors. The choice of the banking industry is significant due to the critical role banks play in economic development. To ensure the inclusion of reliable and accurate data, we employed convenience sampling. All GSE-listed banks operating from 2010 to 2023 were included in the study sample. During the study period, six commercial banks listed on GSE had published the necessary data. The selected timeframe reflects data availability on CSR expenditure and related publications for all sampled banks, making the analysis comprehensive and relevant.

The research data was sourced from the financial reports of the sampled banks. Excel was used to organize the data, and the dependent variables were Return on Equity (ROE) and Return on Asset (ROA), while the independent variable was the amount spent on CSR. Use of CSR expenditures, available from Ghanaian banks, is a direct measure, and its use means that there is no need to construct an index with arbitrarily assigned weights or to rely on ratings agencies. Additional variables, like capital adequacy, economic growth, and bank size, were also added to the variables. The financial report’s data are quantitative, and it covers the years 2010 to 2023. For analysis, E-Views was used to evaluate the quantitative data. The analysis displays a descriptive summary of the variables, correlation, and regression.

3.2. Model Specification

The aim of this study is to investigate the possible effect of CSR spending on banks’ financial performance, while also analyzing the moderating effect of bank size. In order to determine the association, this study uses the panel data technique. Comparing the panel data technique to the cross-sectional and time-series approaches has shown it to be superior over time, more convincing, and accurate in terms of the research’s outcome and findings. This is largely because the panel data technique is able to combine the benefits of both cross-sectional and time-series approaches while also mitigating their disadvantages. According to Stock and Watson (2001), the panel data technique helps adjust for variables that have been left out and addresses the variables unique to each bank, allowing for both long- and short-term effects. Thus, the application of the panel data technique mitigates the drawbacks of using the cross-sectional or time-series approaches separately. For this investigation, the panel data methodology is appropriate due to a few benefits. The methodology makes the assumption that many companies are heterogeneous, with widely differing elements, a greater degree of freedom, and fluctuation in data. Consequently, the general form of the model is:

$${LFP}_{it}={{\alpha }_{1}LCSR}_{it}+{{\alpha }_{2}LBS}_{it}+{\alpha }_3 \times ({LCSR}_{it} \times {LBS}_{it}) + \xi Z_{it}+{\mu }_i+{\nu }_t+{ϵ}_{\mathrm{it}}$$

(1)

where

L = Log

${FP}_{it}$ = Financial performance for bank i at time t;

${CSR}_{it}$ = Corporate social responsibility expenditure for bank i at time t;

${BS}_{it}$ = Bank size for bank i at time t;

${(LCSR}_{it}\times {LBS}_{it})$ = Interaction term of between CSR and bank size (log-transformed);

$Z_{it}$ = Vector of control variables firm i at time t;

${\mu }_i$, = Bank specific effects;

${\nu }_t$ = Time specific effects;

$ϵ$_it = Error terms at time t.

Since the variables were log-transformed, some adjustments were made to variables with zero and negative observations. For zero observations, they were changed to values of one. For negative observations, they were dropped from the data. Hence, some missing values are present.

Figure 1 integrates three core theoretical perspectives—stakeholder theory, legitimacy theory, and the resource-based view—to illustrate how CSR expenditure (LCSR) and bank size (LBS) jointly shape financial performance (LFP). On the left, CSR expenditure is linked to the development of intangible resources, such as stakeholder trust, reputation, and brand equity, reflecting how socially responsible initiatives can align with stakeholder expectations and strengthen a bank’s legitimacy. On the right, bank size is associated with stakeholder pressures, visibility, and resource availability, highlighting that larger institutions may have greater capacity to implement and leverage CSR activities, whereas smaller banks could face more constraints.

The moderating effect of size on the CSR–performance relationship is captured by the interaction term (LCSR × LBS), signifying that the benefits of CSR may vary based on the scale and reach of the bank. These intertwined factors ultimately drive outcomes captured under the Triple Bottom Line—encompassing social, ecological, and economic dimensions—which includes financial performance (LFP). At the bottom of the framework, control variables, such as capital adequacy (CAP) and GDP growth (GDPG), account for bank heterogeneity and wider market conditions, ensuring that the analysis addresses contextual factors that can influence how CSR and bank size impact overall performance.

3.3. Variables

Appendix A summarizes the definition of our main variables. Financial performance is the dependent variable for this study. Two dependent variables were assessed for this study: ROE and ROA. ROE is obtained by dividing a company’s net income by shareholders’ equity. Put another way, it measures the profits made for each dollar that investors invest. ROA is a measure of how well bank managers manage their assets to produce profits (Davydenko, 2010). This is calculated by dividing the bank’s net profit after taxes by the entire amount of assets it has. A higher return on assets (ROA) suggests that management is using funds more wisely and efficiently. It has been a trusted and very helpful measure of banks’ financial health for a very long time. For this study, ROA was chosen since it was found that some banks reported positive ROE in some periods despite reporting negative values for both total equity and net profit. The interpretation of ROE is different in these cases.

Following Gasti et al. (2019), we focus on CSR expenditures as our key independent variable. The size of a bank reveals its magnitude. Dang et al. (2018) conducted a review of 100 prominent empirical articles published in finance, accounting, and economic journals. The results indicated that the three most commonly used indicators of business size are total assets, total sales, and market value of equity. Nonetheless, total assets are the most often used and favored metric of business size by authorities in the banking sector. It summarizes the amount of a bank’s operations, assesses the firm’s resources, and has a clear, comparable definition (Schildbach, 2017). It is biased in terms of valuation, particularly with regard to certain asset classes. The internet superhighway’s advancements in technology and the evolution of one branch have made it unnecessary for banks to expand their branches, which has limited the amount of assets they may own. Large banks often tend to have access to funding, which comes at a lower cost. This variable is therefore selected since the ability of the bank to undertake diversification will largely depend on the availability of resources or its asset base. The log of total assets value is used to measure bank size.

3.4. Estimation

3.4.1. Unit Root Tests

Unit root testing is a useful tool for determining the order of integration in econometric data analysis. Stationarity is essential to this investigation because non-stationary time-series variables are known to yield statistical results that are inaccurate for inferences and hence have reduced predictive ability. This is important because the integration order tells you which statistical estimator to use to prevent spurious regression. According to Granger and Newbold (1974) and Engle and Granger (1987), spurious regression estimates frequently result in inflated performance statistics, which leads to a high prevalence of Type 1 errors among researchers. Consequently, two types of panel unit root tests were run: Levin, Lin, and Chu’s (Levin et al., 2002) (LLC) test, which assumes a common unit root process, and Im, Pesaran, and Shin’s (Im et al., 2003) test (IPS) and Maddala and Wu’s (1999) Fisher-type ADF and PP tests, which assume individual unit root process. The null hypothesis for these tests is the presence of a unit root. A p-value less than 0.05 provides more evidence to reject the null hypothesis and conclude stationarity.

3.4.2. Panel Autoregressive Distributed Lag (ARDL) Model

Pesaran et al. (2001) proposed the panel autoregressive distributed lag (ARDL) model. It is possible to generate a mixture of integrated variables of order 0, 1, 2, or more after the stationarity tests. This means that testing for cointegration under a framework is necessary. Cointegration tests permit us to ascertain long run relationships that exist among the variables. The ARDL model offers several advantages relative to other models. The ARDL does not require the same order of integration of variables as other modes do in order to test for cointegration. Whether the variables are integrated to order 0, order 1, or partially to orders 0 and 1, the ARDL model enables researchers to evaluate long-run correlations (Sanusi et al., 2019). The ARDL model may provide both short-run and long-run coefficients for conclusions while also being suited for small sample numbers. With Akaike Information Criteria automatically choosing the maximum lag length, the general ARDL equation is given as:

$${∆LROA}_{i,t}={\Phi }_i({LROA}_{i,t-1}-{{\beta }_{i}X}_{i,t})+{\sum }_{j=1}^{p-1}{\gamma }_{ij}{∆(LROA}_{i,t-j})+{\sum }_{j=0}^{q-1}{\lambda }_{ij}{∆(X}_{i,t-j})+{\mathsf{\mu }}_i+{ϵ}_{\mathrm{it}}$$

(2)

where LROA is the log of return on assets, X denotes all the independent variables, while γ and $\lambda$ represent the short-run coefficients of dependent and independent variables, respectively. The subscripts i and t stand for cross-section and time, respectively, β stands for long-run coefficients, $\mathsf{\mu }$ stands for fixed effect, and $ϵ$ is the error term. The estimation of this ARDL model is based on the Panel Pooled Mean Group (PMG) ARDL estimation.

3.4.3. Cointegration Test

Long-term relationships between the variables in the ARDL model can only be determined by a cointegration test. To examine the cointegration (interdependence) of the variables in this regard, the Kao (1999) cointegration approach was employed. The strategy was selected because Kao’s test had a higher power than other panel cointegration tests, according to Gutierrez’s (2003) comparison of the power of several panel cointegration test data. The null hypothesis for this test is ‘no cointegration’. The null hypothesis is rejected if the p-value is less than 0.05. In contrast, we fail to reject the null hypothesis if the p-value is higher than 0.05. In the event that the null hypothesis is rejected, the error correction model is estimated. This is given by the error correction term (ECT) in Equation (2) as: ${\Phi }_i$ (${LROA}_{i,t-1}$ − ${{\beta }_{i}X}_{i,t}$), with Φ representing the group specific speed of adjustment.

4. Results and Discussion

4.1. Descriptive Statistics

Table 1. Descriptive statistics.

	CSR	BS	ROA	CAP	GDPG
Mean	1,414,107	6,390,000,000	2.946	15.229	5.721
Maximum	10,430,000	33,700,000,000	7.000	44.149	14.047
Minimum	0.000	198,000,000	−8.900	−1.098	0.514
Std. Dev.	1,790,717	6,330,000,000	2.725	5.182	3.417
Observations	84	84	84	84	84

Source: Authors’ computation from data collected (2024).

gives the descriptive and statistical summary of each variable used to measure the impact of CSR on the financial performance of commercial banks listed on the GSE from 2010 to 2023. The mean for the observed sample in this study is 2.946% for the dependent variable, ROA, which measures the financial performance of banks. In addition, it has a minimum value of −8.9% and a maximum value of 7%, with a standard deviation of 2.725%. CSR, the primary independent variable, has a standard deviation of GHC 1,790,717 and an average of GHC 1,414,107. GHC 0 and GHC 10,430,000 are the minimum and maximum values of CSR, respectively. The mean and standard deviation of a bank’s size are GHC 6,390,000,000 and GHC 6,330,000,000, respectively. The bank size has a minimum of GHC 198,000,000 and a maximum of GHC 33,700,000,000. Another variable being examined is capital adequacy (CAP), which has a mean of 15.229% and a standard deviation of 5.182%. The capital adequacy variable has minimum and maximum values of −1.098% and 44.149%, respectively. With a mean of 5.721% and a standard error of 3.417%, the GDP growth (GDPG) is the final but equally important metric to take into account. Additionally, its values range from 0.514% to 14.047%, at the lowest and maximum points.

4.2. Correlation Matrix

When the explanatory variables are almost linearly dependent on one another, this is known as multicollinearity. Multicollinearity is the state in which there is an approximately linear dependence between the explanatory variables. Correlation coefficients with magnitudes between 0.7 and 0.9 suggest that variables might be considered highly correlated. Correlation coefficients with magnitudes between 0.5 and 0.7 indicate that variables are moderately related. Multicollinearity only becomes an issue, according to Gujarati (2004), when any two repressors of pairwise correlation have a correlation coefficient greater than 0.9.

Table 2. Correlation matrix.

	CSR	BS	ROA	CAP	GDPG
CSR	1
BS	0.4813	1
ROA	−0.0825	−0.1336	1
CAP	−0.3293	−0.5055	0.3556	1
GDPG	−0.3239	−0.5222	0.1176	0.1193	1

Source: Authors’ computation from data collected (2024).

displays the correlation matrix for variables to be used in our regression model. With the highest absolute correlation coefficient value of 0.5375, it is evident that the majority of the measurement variables exhibit little or no substantial correlations with each other. Thus, multicollinearity is not a problem.

4.3. ROA, Untransformed, and Log-Transformed Results

Subsequent result discussions will focus on the natural log (represented as L) of the above variables. Besides the easy interpretation under log-transformed variables, it assures issues of heteroskedastic errors and outlier issues are addressed. It also helps in identifying cointegrating relationships which is important in Panel ARDL models. Under the regression result discussion, results are discussed in two sections (Section 4.6 and Section 4.7). The first section (Section 4.6) presents results considering all variables log-transformed, including ROA. In contrast, the second section (Section 4.7) presents the results considering all variables log-transformed, excluding ROA. The reasoning for excluding log-transformed ROA is to ascertain the robustness of the results under log-transformed ROA. Due to the eight negative values dropped before logging the ROA, it is possible that the regression results may not be robust enough. Therefore, the second section (Section 4.7) presents the results without taking the natural log of ROA. Interestingly, the results did not differ much.

4.4. Panel Unit Root Test Results

Table 3. Panel unit root test results (with constant).

Variables	Levin, Lin, & Chu	Im, Pesaran, & Shin	ADF-Fisher	PP-Fisher	Order of Integration
At Levels
LCSR	0.0680	0.0368	0.0375	0.0001
LROA	0.0002	0.0064	0.0020	0.0000	I(0)
LCAP	0.0097	0.0495	0.0550	0.1434	I(1)
LGDPG	0.0009	0.0028	0.0065	0.2326	I(0)
LBS	0.0194	0.8558	0.6459	0.5596
At First Difference
LCSR	0.0017	0.0001	0.0003	0.0000	I(1)
LROA	0.0005	0.0015	0.0015	0.0000
LCAP	0.0088	0.0913	0.1132	0.0000	I(1)
LGDPG	0.9936	0.0711	0.1302	0.0000
LBS	0.0000	0.0000	0.0000	0.0000	I(1)

Source: Authors’ computation from data collected (2024); All values reported in the table are probability values.

displays the test results for the stationarity of the variables, as indicated in the previous sections. The null hypothesis is the presence of unit root. The results suggest that at the 5% significance level, the variables, bank performance (LROA) and economic growth (LGDPG), are integrated at levels (I(0)), whilst corporate social responsibility (LCSR), bank size (LBS), and capital adequacy are integrated at first difference (I(1)). The mixture of I(0) and I(1) without I(2) permits us to use the PMG/Panel ARDL model and test for both the cointegration relationship and its effects.

4.5. Cointegration Test Results

Next,

Table 4. Cointegration test results.

Kao Residual Cointegration Test
Series: LCSR LBS LCAP LROA LGDPG
Sample: 2010 2023
Null Hypothesis: No cointegration
Automatic lag length selection based on AIC with a max lag of 1
			t-Statistic	Prob.
ADF			−1.8570	0.0317
Residual variance			5.2447
HAC variance			0.8296
Augmented Dickey-Fuller Test Equation
Dependent Variable: D(RESID)
Method: Least Squares
Sample (adjusted): 2011 2022
Variable	Coefficient	Std. Error	t-Statistic	Prob.
RESID(−1)	−1.1078	0.1229	−9.0162	0.0000
R-squared	0.5592	Mean dependent var		0.0623
Adjusted R-squared	0.5592	S.D. dependent var		2.4137
S.E. of regression	1.6025	Akaike info criterion		3.7962
Sum squared resid	164.3467	Schwarz criterion		3.8297
Log likelihood	−122.3777	Hannan–Quinn criter.		3.8094
Durbin-Watson stat	2.1635

shows the cointegration test results computed using Kao (1999). To find out if these variables move together over time, the cointegration between the dependent variable (LROA) and the independent variables (LCSR, LCAP, LBS, and LGDPG) was examined. At the 5% significance threshold, the null hypothesis states that there is no cointegration. With a maximum lag of 1, the automatic lag duration was chosen using the Akaike Information Criterion (AIC). The probability value, 0.0317, is less than the 5% significance level when considering the null hypothesis, which states that there is no cointegration among the variables. The null hypothesis that there is no cointegration is thereby rejected. Thus, this study concludes that a long term cointegration relationship exists among the variables.

4.6. Regression Results from PMG/Panel ARDL Model (with Log-Transformed ROA)

Table 5. Regression results from the PMG/Panel ARDL model (with log-transformed ROA).

Panel A. Long Run Results (Without Interaction Term)
Dependent Variable: LROA
Included observations: 65
Model selection method: Akaike info criterion (AIC)
Model: PMG/ARDL(1,0,1,1,1)
Variable	Coefficient	Std. Error	t-Statistic	p-Value
LCSR	−0.0288	0.0058	−4.9235	0.0000 ***
LCAP	0.9001	0.1296	6.9448	0.0000 ***
LBS	−0.2516	0.0329	−7.6458	0.0000 ***
LGDPG	−0.2522	0.035	−7.1963	0.0000 ***
CONSTANT	4.717	0.5923	7.9638	0.0000 ***
Panel B. Short Run Results (Without Interaction Term)
Dependent Variable: D(LROA)
Included observations: 65
Model selection method: Akaike info criterion (AIC)
Model: PMG/ARDL(1,0,1,1,1)
Variable	Coefficient	Std. Error	t-Statistic	p-Value
ECT	−0.6413	0.1126	−5.6965	0.0000 ***
D(CSR)	−0.1185	0.1028	−1.1527	0.2535
D(LCAP)	2.4372	0.4895	4.9795	0.0000 ***
D(LBS)	3.4446	1.0614	3.2453	0.0019 **
D(LGDPG)	0.0834	0.0478	1.746	0.0858
CONSTANT	−0.5468	0.2971	−1.8402	0.0706
Panel C. Long Run Results (With Interaction Term)
Dependent Variable: LROA
Included observations: 65
Model selection method: Akaike info criterion (AIC)
Model: PMG/ARDL(1,0,1,1,0,1)
Variable	Coefficient	Std. Error	t-Statistic	p-Value
LCSR	−1.0272	0.3485	−2.9471	0.0046 ***
LCAP	0.9941	0.1071	9.2847	0.0000 ***
LBS	−0.8152	0.2174	−3.7488	0.0004 ***
LCSR*LBS	0.0465	0.016	2.9131	0.0050 ***
LGDPG	−0.0516	0.0082	−6.3299	0.0000 ***
CONSTANT	16.409	4.7802	3.4327	0.0011 ***
Panel D. Short Run Results (With Interaction Term)
Dependent Variable: D(LROA)
Included observations: 65
Model selection method: Akaike info criterion (AIC)
Model: PMG/ARDL(1,0,1,1,0,1)
Variable	Coefficient	Std. Error	t-Statistic	p-Value
ECT	−0.6392	0.0996	−6.4182	0.0000 ***
D(LCSR)	−0.0585	0.1035	−1.2405	0.2196
D(LCAP)	1.0737	0.533	2.0145	0.0482 **
D(LBS)	1.2083	0.6179	1.9554	0.0550 *
D(LCSR*LBS)	0.0025	0.0041	0.6117	0.543
D(LGDPG)	0.0177	0.0187	0.9472	0.3472
CONSTANT	−4.7955	0.6477	−7.4037	0.0000 ***

Source: Author’s computation from data collected (2024); Significance level: 1% (***), 5% (**), 10% (*).

reports the results from the PMG/Panel ARDL model considering all log-transformed variables. Besides the easy interpretation under log-transformed variables, it assures that issues of heteroskedastic errors and outlier issues are addressed. Panel A presents the long run results without the interaction term of CSR and bank size, while Panel B presents the short run results without the interaction term of CSR and bank size. Panel C presents the long run results with the interaction term of CSR and bank size, and Panel D presents the short run results with the interaction term of CSR and bank size.

From Table 5 Panel A, corporate social responsibility (LCSR), bank size (LBS), and economic growth (LGDPG) have negative effects on return on assets in the long run. These relationships are statistically significant at the 1% level of significance. Panel B reports the short run results. The error correction term (ECT) coefficient, which is negative, lies between −1 and 0 and is significant at the 5% alpha level. This confirms the presence of cointegration and a return to equilibrium should there be any deviation. The ECT suggests that 74.49% of disequilibrium in bank performance (return on assets) for the past years is corrected within a year. This implies a high rate of convergence toward the long-run equilibrium. While capital adequacy and economic growth were insignificant in the short run, bank size shows a positive significant (at the 5% level of significance) effect on bank performance in the short run. Large banks may be able to lower their risk exposure and increase profitability by diversifying their portfolios more. This finding corroborates the findings of Adam (2014), Musah et al. (2018), and Awunyo-Vitor and Badu (2012), who found a significant positive relationship between bank size and performance.

Panel C reports long run results with inclusion of the interaction term between CSR bank size. The key result is the significance of the interaction term for LCSR and LBS. While the signs on the coefficients of LCSR and LBS are negative, the coefficient on the interaction term is positive and statistically significant. An increase in CSR spending for any given bank size will reduce the negative impact of CSR spending and will turn positive within the range of bank sizes in our sample.

We show this by calculating the marginal effect of CSR expenditure on ROA from the regression results in Panel C. Mathematically, the total effect of a percentage increase in CSR on ROA is −1.0272 + 0.0465LBS. This means an increase in CSR by 1% initially reduces ROA by −1.0272%. But depending on the bank size (LBS), the negative effect could be reduced by 0.0465% or more if the bank increases in size beyond a turning point. The turning point can be found as:

−1.0272 + 0.0465LBS = 0

LBS = 22.0903 ≈ 22.09

⇒ BS = GHC 3922.52 million

The turning point falls within the bank size range used for this study. This means that CSR expenditures by banks whose size exceeds the critical threshold of LBS = 22.09, or a bank size of GHC 3922.52 million, have a positive effect on ROA. This addresses the second objective of the study by indicating that bank size positively moderates the negative relationship between CSR and ROA. Thus, for larger banks, CSR tends to have a positive effect on their financial performance. This could be due to economies of scale or reputation as the bank expands and CSR becomes part of its cost.

The signs on capital adequacy (LCAP) and economic growth (LGDPG) coefficients remain unchanged, positive for capital adequacy and negative for economic growth. The negative effect of economic growth (LGDPG) in the long run can be attributed to inflation and currency depreciation that may reduce purchasing power of money and exchange rate losses. Capital adequacy and the interaction term (LCSR*LBS) both have a positive effect on the bank performance. These relationships are statistically significant at the 1% level of significance. A bank with a higher capital adequacy ratio is typically better insulated against risk. A higher ratio indicates that a larger portion of the bank’s assets is financed by capital rather than liabilities, enhancing financial stability. This will lower the amount paid to the firm’s creditors or depositors, improving the net return of the business and raising its return on equity and assets. As a result, as the ratio rises, banks are able to enhance the relative returns that are generally attributed to the company while also reducing its liability. The results obtained here corroborate those obtained by Bateni et al. (2014).

The negative effect of CSR on ROA is consistent with the assumption that higher corporate social performance must come at a cost, resulting in lower bank performance. Therefore, this long-run result was in line with the shareholder theory and the views of Friedman (1970, 2007), who hold that social activities should not be promoted if they do not add to the value of the businesses. But we find that the result is contingent on bank size. Larger banks are able to benefit from CSR expenditures, offsetting the increased cost.

Panel D reports the short run results from our interaction model. The coefficients in Panel D are virtually identical with those from Panel B, the non-interacted model. The error correction term (ECT) coefficient is negative and lies between −1 and 0 and is significant at the 5% level. This confirms the presence of cointegration and a return to equilibrium should there be any deviation. The ECT suggests that 83.9% of disequilibrium in bank performance (return on assets) for the past years is corrected within a year. This implies a high rate of convergence toward the long-run equilibrium. While economic growth was insignificant in the short run, capital adequacy and bank size show are positive and significant (at 5% and 10% level of significance, respectively). Unlike the results without the interaction term, capital adequacy is significant in both long run and short run when the interaction term is introduced.

4.7. Regression Results from PMG/Panel ARDL Model (Without Log-Transformed ROA)

Table 6. Regression results from the PMG/Panel ARDL model (without log-transformed ROA).

Panel A. Long Run Results (Without the Interaction Term)
Dependent Variable: ROA
Included observations: 73
Model selection method: Akaike info criterion
Model: PMG/ARDL(1,0,1,1,0)
Variable	Coefficient	Std. Error	t-Statistic	Prob.
LCSR	−0.1235	0.056	−2.205	0.0314 **
LCAP	1.8646	0.5508	3.3853	0.0012 ***
LBS	−0.6543	0.2042	−3.206	0.0021 ***
LGDPG	−0.0988	0.0395	−2.497	0.0145 **
CONSTANT	7.6543	1.9367	3.951	0.0002 ***
Panel B. Short Run Results (Without Interaction Term)
Dependent Variable: D(ROA)
Included observations: 73
Model selection method: Akaike info criterion
Model: PMG/ARDL(1,0,1,1,0)
Variable	Coefficient	Std. Error	t-Statistic	p-Value
ECT	−0.7443	0.2696	−2.7609	0.0073 ***
D(LCSR)	−0.0117	0.0469	−0.2495	0.0803 *
D(LBS)	7.1458	3.0899	2.3126	0.0236 **
D(LCAP)	7.7209	2.5013	3.0868	0.0029 ***
D(LGDPG)	0.0158	0.1478	0.1067	0.9153
CONSTANT	3.4333	2.0503	1.6745	0.0984 *
Panel C. Long Run Results (With Interaction Term)
Dependent Variable: ROA
Included observations: 73
Model selection method: Akaike info criterion (AIC)
Model: PMG/ARDL(1,0,1,1,0,1)
Variable	Coefficient	Std. Error	t-Statistic	p-Value
LCSR	−1.534	0.632	−2.429	0.0163 **
LCAP	1.904	0.577	3.299	0.0015 ***
LBS	0.768	1.161	0.662	0.5104
LCSR*LBS	0.070	0.033	2.121	0.0362 **
LGDPG	−0.176	0.159	−1.108	0.2716
CONSTANT	−20.167	25.475	−0.792	0.4312
Panel D. Short Run Results (With Interaction Term)
Dependent Variable: D(ROA)
Included observations: 73
Model selection method: Akaike info criterion (AIC)
Model: PMG/ARDL(1,0,1,1,0,1)
Variable	Coefficient	Std. Error	t-Statistic	p-Value
ECT	−0.7029	0.2694	−2.6089	0.0110 **
D(LCSR)	−0.1780	0.3437	−2.5853	0.0103 **
D(LCAP)	8.0781	2.5322	3.1901	0.0021 ***
D(LBS)	7.2985	3.1486	2.318	0.0233 **
D(LCSR*LBS)	0.0077	0.0156	0.4926	0.6238
D(LGDPG)	0.0798	0.1673	0.4766	0.6351
CONSTANT	5.4557	2.3157	2.356	0.0212 **

Source: Author’s computation from data collected (2024); Significance level: 1% (***), 5% (**), 10% (*).

reports the regression results from PMG/Panel ARDL model without log-transformed ROA. Panel A presents the long run results without the interaction term of CSR and bank size, while Panel B presents the short run results without the interaction term of CSR and bank size. Panel C presents the long run results without the interaction term of CSR and bank size, and Panel D presents the short run results without the interaction term of CSR and bank size.

The results obtained in Table 6, where ROA was not log-transformed, are similar to earlier results in Table 5, where ROA was log-transformed. From Table 6 Panel A, without the interaction term, all variables are statistically significant at the 5% level of significance in the long run. All variables except capital adequacy have a negative effect on ROA. The effect is the same for log-transformed ROA. The difference between the ROA results, non-transformed and log-transformed, other than the coefficient magnitudes, are the signs on the LBS and LCAP terms. From Table 6, an increase in CSR by 1% reduces ROA (2.096/100 = 0.021 percentage points) in the long run.

From Panel B, capital adequacy and bank size are statistically significant in the short run. The ECT, which is negative and significant, confirms the long-run relationship among variables. The ECT coefficient suggests that 70.29% of disequilibrium in bank performance (return on assets) for the past years is corrected within a year. This implies a high rate of convergence toward the long-run equilibrium.

Panel C presents the long run results with the interaction term when the ROA is not log-transformed. Results here are very similar to those in Table 5 for the log-transformed dependent variable results. While all variables are significant under the log-transformed ROA, capital adequacy alone is not significant. GDP growth and bank size are only significant at the 10% level of significance while CSR is significant at 5% level.

The main result is the positive sign and statistical significance (5%) of the interaction term. As with the model run on log-transformed ROA, the signs on LCSR and LBS are negative, meaning increases in either alone yield lower financial performance. But with the positive interaction term, an increase in LCSR coupled with larger bank sizes, leads to increases in financial performance. The value of bank assets (LBS) at which this turning point occurs can be found as follows:

−1.534 + 0.070LBS

At turning point, −1.534 + 0.070LBS = 0

LBS = 21.9142 ≈ 21.91

⇒ BS = GHC 3276.36 million

Again, this turning point falls within the bank size range used for this study. This means an increase in CSR by 1% initially reduces ROA (1.534/100 = 0.01534 percentage point). But depending on the bank size (LogBS or BS), the negative effect could be reduced by 0.070/100 = 0.0007 percentage points or more if the bank increases in size beyond the turning point. Banks larger than the threshold size of LogBS = 21.91, GHC 3276.36 million, see an increase in ROA with CSR expenditure. As the bank increases in size toward the turning point, the negative effect of 0.01534 percentage points in ROA diminishes and becomes positive. This addresses the second objective of the study by indicating that bank size positively moderates the negative relationship between CSR and ROA. Thus, for larger banks, CSR tends to have a positive effect on their financial performance. This could be due to economies of scale or reputation as the bank expands and CSR becomes part of its cost.

5. Conclusions

5.1. Summary of Findings

The research focuses on the impact of CSR on the financial performance of banks listed on the Ghana Stock Exchange and how this relationship is moderated by bank size. The study explores the relationship between firm performance and CSR activity. CSR expenditure is used to capture bank CSR activities, and control variables used are bank size, capital adequacy, and GDP growth. The performance variable was proxied with return on assets. The study used annual bank-level data spanning 2010 to 2023 analyzed with a PMG/Panel ARDL model. Bank-level data were collected from annual reports while macroeconomic data were collected from the World Development Indicators (WDI).

In the simple model, the CSR variable, the variable of interest in this study, was found to be significant and negatively related to ROA in the long run, consistent with findings in Giannopoulos et al. (2024). This supports hypothesis H1 that CSR is associated with banks’ financial performance. Capital adequacy was significant and positively related to the return on assets in the short and long runs. GDP growth was also significant and negatively related to the return on assets in the long run, although insignificant in the short run. That might suggest the Friedman view that spending on CSR is a cost without benefit.

Our main finding is the mediating effect of bank size on CSR expenditure. In our simple model without interaction, we find that both CSR and bank size negatively related to financial performance in the long run, However, when we interact bank size with CSR expenditure, we find the interaction term to be positive. This supports hypotheses H2 that bank size moderates the impact of CSR expenditures on financial performance. That means that as bank size increases, the negative impact of CSR expenditure on firm performance is reduced, and in fact, ultimately becomes positive for the banks in our sample period. This moderating role indicates that larger banks see a reduction in the negative impact of CSR on financial performance. For our sample, the range of bank size for which CSR may have a negative effect on financial performance lies somewhere below the range from GHC 3276.36 million to GHC 3922.52 million. Banks larger than the upper end of the range display a positive relationship between CSR expenditures and financial performance.

We find that simple result that increases in CSR expenditure appear to reduce financial performance (ROA). However, once we condition on bank size, this negative effect is diminished with increases in bank size. The larger banks display economies of scale. These scale economies may be realized along several dimensions, including allowing the banks to be able to gain standing with governments, investors, and a community that may bring some economic benefit to the firms. We also find a positive relationship between capital adequacy and ROA. This may be attributed to the improved risk absorption capacity associated with a higher capital adequacy ratio. A stronger capital position reduces the likelihood of financial distress, enhances investor confidence, and may lower the bank’s risk-adjusted funding costs. Additionally, well-capitalized banks may have greater flexibility in asset allocation, allowing them to pursue more profitable lending opportunities, which in turn supports higher returns on assets.

5.2. Policy Recommendations

The findings suggest that certain variables should be closely monitored by banks listed on the Ghana Stock Exchange, as they have the potential to have a positive or negative effect on their financial performance. The study’s conclusions suggest that banks should realize that CSR performance should be carefully planned and implemented to improve bank performance rather than being seen as a pointless addition. To be more effective, banks should evaluate how well CSR programs are strategically aligned with bank objectives. Furthermore, researchers should focus on whether the short-term relationships last over time, rather than just focusing on the short-term position of CSR as it is currently practiced. Diseconomies of scale can cause problems for banks in terms of financial performance if not curtailed. As the banks expand, it is necessary to make efforts to improve managerial and organizational effectiveness. This can help reduce the diseconomies of scale. As much as there is a rising concern about sustainability among banks, they should generally scale up all aspects, especially in size, so as to gain the benefits that could accrue from CSR. It would be best for banks to keep assets above GHC 3922.52 million while engaging in strategic CSR.

As a policy goal, recommending that banks increase in size may assist in terms of CSR, as long as growth in bank size does not reduce the sector’s competitiveness, itself an important goal. This is of some concern as Ghana’s banking sector suffers from scale limits given the size of Ghana’s economy and population. If growth in bank size comes at the cost of reduced numbers of banks, it will lead to a more monopolistic sector. This will lead to improved bank performance but at the expense of customer welfare. Therefore, growth in bank size should be generated from increased international operations. Policy makers should therefore be concerned with growth in individual banks through mergers and acquisitions. These activities alone would likely be detrimental to customers. Future research should also address banking competitiveness and how competition mediates the CSR-financial performance link.

Also, it is recommended that banks should be intentional in keeping their capital adequacy ratio high. In addition, the negative effect of economic growth on bank performance can be reduced if the macroeconomic indicators, like inflation and exchange rate, are stabilized, but those variables are mostly exogenous to the banks, unless the banking sector itself has influence with the government over its setting of macroeconomic policy.

5.3. Limitations and Future Research

Our study has limitations. First, and most obviously, we explore banking in one country, Ghana. To the extent Ghana is unique will limit the applicability of our results more generally. But we think the results are useful and extend beyond the one country. Data availability limits the timeframe of study. There is no evidence to suggest our study period is exceptional in any way, but regardless, inferences out of sample should be made cautiously. We rely on CSR spending as our measure of CSR. Our data do not disaggregate CSR spending by category. Therefore, we are not able to draw inferences on or make recommendations about the form in which CSR activity is most effective for the banks’ financial performance. It is particularly important to further investigate this channel because while we find bank size is an important determinant, we do not know how the larger banks in Ghana are allocating resources for CSR activities. Studying how banks in Ghana manage their CSR expenditures is a recommended next research goal. It is possible and likely that results from the larger body of work on CSR is relevant here. An approach that prioritizes CSR on employee and governance issues may be most important, but that is merely conjecture.