The Relationship Between Corporate Social Responsibility and Corporate Financial Performance: A Moderating Effect of Economic Policy Uncertainty

Within recent decades, researches on corporate social responsibility (CSR) has been receiving more attention over the world. The existing literature on CSR is very diverse, both in evaluating the performance of CSR activities as well as and the relationship between CSR disclosure and firms‟ outcome. This paper extends the literature of the latter case, that is, not only it aims to purely examine the relationship between CSR disclosure activities and corporate financial performance (CFP), but also consider this nexus under economic policy uncertainty (EPU) context. Our primary data is collected from more than 500 listed companies in the Vietnamese stock market from 2013 through 2017, while secondary data (CSR and EPU) are self-calculated under serial criteria. Our results support the hypothesis that the more companies intensively disclose CSR, the higher financial performance (both ROA and Tobin‟s Q) they could obtain. More interestingly, we find that while EPU seems to weakly moderate the relationship between CSR disclosure and “internal financial performance” (ROA), it will significantly diminish the effect of CSR toward “external financial performance” (Tobin‟s Q). The research shed light on an approach to measure CSR disclosure indexes for the emerging market as in Vietnam. Our findings encourage the firm‟s managers to pay more attention to CSR disclosure activities due to the positive benefit that their firm could obtain and suggest policymakers to maintain a stable economic background for a sustainable market.

adopting a more rigorous classification method, more evidence of a positive correlation between the two was found. A commonly identified reason for these diverse and contradictory findings is the measurement issue pertaining to both concepts of interest. It is, therefore, suggested to investigate the measurement approaches for the concepts of CSR and CFP (Galant and Cadez, 2017).
There are only a handful of outstanding papers that include researches related to CSR in Vietnam. As a developing country, there is a lack of context-sensitive researches on CSR in Vietnam, because it is inappropriate to utilize the universal CSR theories, given various cultural, political, and economic specificities (Nguyen et al., 2018). However, along with the diffusion of global CSR certificates, the awareness toward environmental or human issues of domestic firms has improved. Wrana and Revilla Diez (2018) advises policymakers to consider CSR support systems as a means to motivate firms to improve Vietnamese companies" involvement with social activities such as the environment and human rights. Subsequently, there has been a number of managers who express a high degree of positive attitude towards CSR behaviors is increasing. For instance, Thang & Fassin (2017) observes a positive and significant correlation between internal CSR and organizational commitments related to the employee's work environment. These cultural and historical characteristics also influence the way that a firm"s board makes decisions on CSR activities. Hoang et al. (2018) studied the relationship between board diversity on corporate social disclosure (CSD) of listed companies in Vietnam from 2008-2010 and points out that the factor that would influence CSD is not "off-board" but "in-board" diversification. This finding reflects the shareholder"s expectations and how they exercise CSR behaviors, which can assist in navigating for regulators and corporate executives.
Recently, Yekini and Ho (2014) studied the relationship between CSR and CFP of the top 20 listed companies in Hanoi andHo Chi Minh City Stock Exchange (2010-2012). The results show that there is a weak relationship between CSR and CFP of these companies in Vietnam. Vice versa, by exploring the relationship between social responsibility, market orientation (MO), and company efficiency, Long (2015) discloses that CSR, in both aspects, has a positive impact on business performance. This result recommends senior managers, CEOs, and shareholders of Vietnamese companies should raise their awareness of the importance of CSR so that they could improve their competitiveness in the global market.
However, most of the previous empirical studies have a drawback. That is, they consider CSR as a factor that directly impacts the performance of a firm"s operation. However, as to the perspective of financial managers, CSR is nothing but information disclosure that should not be an estimator of financial performance. To overcome this problem, we put the relationship between CSR and CFP under a moderating effect of economic factors to reflect both direct and indirect influences. The literature of those economic factors is presented below.

Research on EPU
The negative impact of uncertainty on economic activities has been well documented in the literature. Baker et al. (2016) indicated that the degree of uncertainty of economic policies leads to lower output, investment, and employment rate in both the US economy and in the international environment. In fact, there is no regional or technological border for the impact of an uncertain shock to the outcome of the economy, such as productivity, unemployment, or investment (Gulen and Ion, 2015;Kang et al., 2014;Leduc and Liu, 2016;Rodrik, 1991). The stock market, an important institution of every economy, is also being forced under a high level of EPU. For instance, Pá stor and Veronesi (2012), by using a general equilibrium model, features a stock price fall, in general, at the time of declaration of policy changes. Not only does the whole stock market"s value decline, the risk premium of the market, in sequence, also leads the individual stock return to be more volatile and interrelated (Pá stor and Veronesi, 2013).
In broad term, not only do researches about EPU spread out the term, but also dig into a specific component. One important element should be taken into consideration regarding EPU is political uncertainty. Though it is still a matter of debate whether political uncertainty is part of EPU, researchers are unified in accrediting the negative impact of political risk and uncertainty toward economic factors such as investment and capital diminishing (Jens, 2017;Lensink et al., 2000). Among varieties of sources, fiscal policy, especially tax policy, is judged as the most important factor contributes to political and economic uncertainty (Baker et al., 2016), for example, weaken economy recovery after crises (Kydland and Zarazaga, 2016). Meanwhile, microeconomic factors appear to be more affected by monetary policy uncertainty. Husted et al. (2019); Sinha (2016) and Aghion et al. (2009) focus on the interest rate and currency uncertainty and productivity and explore a depression of those variables on production outputs.
Standing from the perspective of industrial companies, economists have been trying to figure out the optimal solution to face the challenge of uncertainty. During an economic uncertainty period, they suggest firms to raise investment in their research and development activities (Vo and Le, 2017). This can be further explained that competitive firms tend to implement preemptive strategies when they face high uncertainty and being supported by the theory of strategic growth options. Other investments, such as hiring, high return futures/options, or capital expenditure should be deduced to maintain firms" financial ability (Caballero, 1991;Gulen and Ion, 2015) until the uncertainty is resolved.
The above argument leads us to a question of whether EPU plays a role in the relationship between CSR and CPF. Therefore, different from the previous studies in this relationship, we treat EPU as a moderating variable and hypothesize that EPU will have both direct and indirect regulatory effects on the firm"s financial performance.

Research Model
To examine the impact of CSR on CFP, similar to several previous studies (Harjoto et al., 2015) we estimate the following equation: CFP , = η , + β * CSR , + ′ * X , + ε i,t ; (1) where CFP i,t is the measure of the corporate financial performance of firm i in year t; η i,t is the firm-year fixed effect that controls all factors that are firm-specific (level of development, geography, institutions, etc.) and time-specific (other factors rather of uncertain economic policies such as changes in regulations, market risks…); X i,t is a matrix of control variables to capture the effect of firm-specific factors such as financial leverage, size, risk, or ownership.
In order to clarify the relationship between social responsibility and CFP, similar to previous empirical studies in developing countries, this study also examines the impact of different aspects of social responsibility on society and the needs of different stakeholders for CFP. Therefore, we also estimate the effect of CSR"s components to CFP through 4 equations, respectively.
The moderating effect of economic policy uncertainty (EPU) factors is then estimated by the following regression model: CFP , = η , + (β + γ * EPU ) * CSR , + α * EPU + ′ * X , + ε i,t ; Here, EPU t represents for the economic policy uncertainty factors in year t. Because EPU is a macroeconomic variable, it has the same value for all observations in a specific year. Not only is the purpose of this regression to examine the effects of EPU to CFP, but also to explore how these factors moderating the relation between CSR and CFP (through the coefficient of cross-product).

Variable Definition
-Corporate social responsibility (CSR): There are several ways to measure the CSR index. The first is the reputation index, such as the CEP (The Council of Economic Priorities), the Milton Moskowitz, and the Fortune reputation index. The second is the corporate rating, such as Kinder, Lydenberg and Domini Index (KLD); Dow Jones Sustainability Index (DJSI); Global Initiative Reporting Indicator (GRI); and Domini 400 Social Index (DSI). The third one, related to assessing social responsibility, is carried out through a survey. The final way is to analyze the content of secondary data. Many studies have applied this method to analyze the extent of CSR activities in public companies, especially in their reporting units.
Each measuring method has its pros and cons, and only reflects a certain aspect of social responsibility, that is, considering it to be a proxy variable for social responsibility. This study, therefore, proposes a different approach, incorporating the "GRI-4 Guideline" and the actual situation of Vietnam, -an emerging market, through content analysis to develop social responsibility criteria involving four types of responsibilities. The reason that we use the "GRI-4 Guideline" is that it provides systematic guidance and support for companies willing to publish their CSR activities and sustainable development measures (Brown et al., 2009). These responsibilities include the environment, working, community, and product. Subsequently, the social responsibility criteria are benchmarked and the social responsibility index is developed for developing countries. Arriving from existing literature, the study adopts accounting and market-based measures to indicate CFP. In detail, the two indicators are ROA and TBQ (Choi et al., 2010;Peters and Mullen, 2009). These two factors are chosen since they are the most common measures to test the relationship between social responsibility and CFP in previous studies that imply profitability, financial structure, and stock price.
-Economic policy uncertainty (EPU): There are many ways to determine the EPU index. In the context of this study, we adopted the approach of Baker et al. (2016) to develop measurements of EPU in Vietnam. The spirit of our process is to construct a scaled frequency count in articles about policy-related economic uncertainty in the Vietnamnews (VNN), one of the Vietnamese leading English-language newspapers. We first collected all VNN"s articles related to at least one term from each of the Vietnamese economic uncertainty information, such as Vietnam, Vietnamese, economy, economic, uncertain, and

GRI4 Guidelines CSR reporting in Vietnam
Building the CSR Assessment Content Analysis CSR Criteria uncertainty. We identified whether the sample articles contain any of the following terms: policy, spending, budget, interest rates, reform, tax, government, Hanoi, authorities, central bank, State Bank of Vietnam, regulation, deficit, and WTO. Finally, we divided the numbers of selected articles in a particular month using the total numbers of all VNN articles in the same month and normalize the resulting series to a mean value of 100 from January 2013 to the most recent by applying a multiplicative factor. The yearly EPU indexes are then calculated by the average of monthly EPU indexes in that year.
-Control variables: The debates between social responsibility disclosure and financial efficiency are controversial. This nexus is controlled by many factors, such as company size, and age. Kumar et al., (2014) believes that the practice of social responsibility disclosure is related to the size of the firm, as stakeholders expect large companies to participate more effectively in social activities than in smaller ones. Because company size is related to environmental standards, the bigger one is more concerned about their social image, therefore, only small errors in management behaviors can lead to unpredictable consequences, such as product quality, emissions, employment, and so on. Following the spirit of literature, this study utilizes several control variables that could increase the estimated probability of the relationship between social responsibility and CFP, such as (1) Firm size, (2) Financial leverage, (3) Firm Risk, (4) Age and (5) SOE.

Descriptive Statistics
This study collects a sample based on data of listed companies in Vietnam (using financial reports, annual reports, sustainability reports, etc.). The total number of observations is 592, corresponding to the total number of enterprises with CSR disclosures for the five years 2013 to 2017. The descriptive statistics of the variables in our sample are as follows: The average value of CSR_EMP is the largest among the CSR criteria of 0.764. It is reasonable because prior to CSR disclosure under Decision 155 of the Ministry of Finance, enterprises also presented information on employees in accordance with previous regulations. Hence, the information about the employees that the companies publish will be easier and more comprehensive than other information. Only this indicator has the smallest value other than 0 (Min of CSR_EMP = 0.166). Meanwhile, CSR_ENV has the smallest average value (CSR_ENV = 0.439) in the four components of the total CSR. In fact, the information items from 6.1 to 6.3 in the annual report can be exempt from declarations for enterprises in the finance, banking, insurance and securities industries. Therefore, usually in the annual reports of businesses in these industries, there is no information on environmental responsibility.
The two indicators show the CFP of enterprises is ROA and TBQ with an average value of 0.071 and 1.168, respectively. Each indicator has different meanings, with a ROA of 0.071 showing that firms in the sample, the average across the research period, have a return on assets of more than 7%. This figure is a positive sign that businesses are profitable. On the other hand, the fact that the average value of TBQ is 1.168534 implies that the value created by the business is greater than the cost of the enterprise. Regarding the variation, data on the maximum and minimum value of ROA are 0.784 and -0.5172, respectively with the standard deviation of 0.09, which shows a very large difference in the profitability index of assets. A similar trend is also discovered with TBQ.
Other control variables, including SIZE, LEV, RISK, AGE, and SOE reveal some interesting insights. The size of the enterprise in the sample varies widely from -1.729 to 9.394, corresponding to the firm"s largest and smallest total asset of 78 billion and 9,964 billion VND. In addition, the minimum and maximum of LEV values are 0 and 0.984, with an average value of 0.392. This means that a business does not use a loan while a business uses a loan greater than its total asset value. The average value of SOE is 0.206, which means that most businesses in the sample are not under State ownership. With the selection of market-oriented mechanisms, Vietnam is gradually transforming former state-owned enterprises into the form of common joint-stock companies. The largest value of AGE is 18 years while the smallest one is 1 year with an average value of 8.212, indicates that the Vietnamese stock market is still emerging and companies have just started their listing.

Regression Results
Our first concern is to explore the impact of CSR on a firm"s performance. Using the Stata analysis software, the result of the first regression is summarized in Table 2.   Note: t statistics of regression coefficients are shown in parentheses. All variables are winsorized at the 5% level; *, **and *** indicate significance at the 10%, 5% and 1% level, respectively.
We also divide our sample into some subsets, as by stock market and by firm"s operational characteristic to test the robustness of CSR effects. The column 2,3,4,5 in Panel A and Panel B describe the results of the regression of ROA, TBQ on the values of CSR for different sub-samples. According to the market-based samples, there are 555 firms listed on the Ho Chi Minh stock exchange (HOSE) and 37 on the Ha Noi stock exchange (HNX). It is evident that regardless of which stock exchange firms listed, their CSR information publishing will have a statistically significant positive impact on financial performance. However, HOSE is the bigger and older stock exchange with stricter requirements and disclosure regulations in comparison with HNX. Therefore, it is expected that, on average, firms listed in HOSE will enjoy less benefit corresponding to the same CSR disclosure level of ones listed in HNX. Indeed, our results depict that this impact is lower in HOSE than those in HNX (0.099 vs 0.298 for ROA; and 1.182 vs 7.607 for TBQ).
Regarding the subsets of manufacturing and non-manufacturing firms, we find some statistical evidence that encourages manufacturing firms to enhance their CSR disclosing more than service-firms (the impact of CSR on ROA is 0.122 vs 0.104 and TBQ is 1.548 vs 1.365 for manufacturing and non-manufacturing firms, respectively). As in Vietnam, policymakers believe that manufacturing firms should have more environmental influence, and therefore, require these companies to reveal more details about their activities. Subsequently, the more these companies publish their CSR information, the more easily they attract investment flows and other benefits from the government and investors.
We further assess the impact of CSR components on CFP. The results of the regressions are presented in Table 3. As of expectation, it can be generally seen that all of these components have the same influences toward CFP in the overall CSR index.   1019 Note: t statistics of regression coefficients are shown in parentheses. All variables are winsorized at the 5% level; *, **and *** indicate significance at the 10%, 5% and 1% level, respectively.
In detail, we first discuss CSR_ENV, which is indicated for more environmental activities. The positive and significant coefficients of this variable (0.057 at 1% significant level) imply that businesses, who have more environmental activities, will increase their financial performance. This advocacy for environmental activities is similar to what was found in Elsayed and Paton (2005); Endrikat et al. (2014); McWilliams and Siegel (2001).
The other dimensions of CSR, namely CSR_EMP, CSR_COM, and CSR_PRO, also have very similar impacts to CFP compared to what of CSR_ENV (both sign and magnitude). This assists us to assert that firms who better aware of matters of employee"s issues, community factors, as well as production processes will enjoy more financial benefits (Branco and Rodrigues, 2006;Joshi and Gao, 2009;Kang et al., 2010). In other words, these results once again emphasize the importance of CSR, in every aspect of the activities.

Moderating Effects of EPU
So far, we just concentrated on analyzing the impact of CSR on a firm"s financial performance without taking into account the moderating effects of the EPU factor. In this section, we add this factor into our regression to see whether it changes the CSR -CFP nexus. Table 4 shows some interesting results.   Firstly, it can be seen that CSR still has a sustainable effect on ROA and TBQ, that is all of the coefficients of CSR variables are positive and significant at least at 5% of significant level. However, when we consider this relationship in the context of uncertainty, the level of impact is relatively low. This implies that in an emerging market like Vietnam, when the economy becomes more uncertain, in order to ensure better service, firms" managers have to pay more attention to other activities rather than CSR, which is, in turn, dragging down the importance of these responsibilities. Indeed, even though our results do not illustrate the direct impact of EPU on CFP (coefficients of EPU are diverse across CFP variables), economic shocks, to a certain extent, tend to diminish the advantages of CSR activities. More specifically, the betas of cross-product between CSR and EPU are uniquely negative across columns 6 to 10 and ambiguously in columns 1 to 5. Only "external financial performance" is driven down by economic uncertainty and "internal financial ability", in the short term, should not be affected. In most developing countries, EPU factors quantified by information counting are more sensitive to the market price of the company"s stocks rather than to its real operation outcomes.
When dividing the research sample by listing location, we see that EPU has negative effects on the relationship between CSR and TBQ (-0.009 for HOSE and -0.008 for HNX). This shows that EPU has the same impact on all types of stocks, regardless of where they are listed. This result is consistent with the system risk theory, that is, systemic hazards will affect all factors in the market (economic policy risk is a general risk or systemic risk). A similar phenomenon is found when we divide our sample by a firm"s operational features, that is, for economic shocks, the imagination of manufacturing or non-manufacturing firms both being underestimate by investors.

Endogeneity Check
To explicitly examine the endogeneity issue, we use an endogeneity test in the OLS estimation. That is, we employ the Durbin-Wu-Hausman (DWH) test to verify the consistency of the OLS. Davidson and MacKinnon (1993) suggest an augmented regression test (DWH test), which can easily be formed by including the residuals of each endogenous right-hand-side variable, as a function of all exogenous variables, in a regression of the original model. Because CSR is suspected as an endogenous variable in the equation, we use any exogenous variables in the original regression and add as a new exogenous variable, namely Award, to get residual CSR_res, then perform an augmented regression. To check if our results are robust, we use two-stage least squares regressions. It is possible that the observed relationship between CSR and TBQ is not due to an actual relationship between these variables, but rather to a third factor that drives both CSR and TBQ. As instruments for CSR, we use the list of Vietnam listed company Awards as a dummy variable (Award = 1 if the company is in the list of Awards otherwise = 0) (Note 3). The rationale behind this instrument is that a firm"s CSR is systematically influenced by the Award. Companies that receive annual awards are having good social responsibility reports. The result shows a significant and positive impact of CSR on TBQ. This result is a line with these found earlier from OLS estimations. We also check the strong/weak level of IV variable on the IV regression. The minimum eigenvalue statistic (19.3552 > 10) assists us to conclude that our IV variable is strong, while estimates and tests base on the resulting IV estimator are reliable.

Conclusion
The main objective of this study is to examine the relationship between CSR disclosure activities and CFP under independent and EPU-moderating-effect contexts. By employing content analysis, we develop the CSR information publishing indexes, for the basis of examining the level of CSR disclosure practiced by Vietnamese listed companies and to explore the motivation of these activities. There have been several studies describing the development of socially responsible framework for developing countries. Nonetheless, this study employs four aspects of CSR (environment, employees, communities, product/customers with 23 operational criteria) in conjunction with GRI4 guidelines, to propose a framework that takes into account the context of Vietnam. Socially responsible practices that vary between countries and cultures (Kao et al., 2018;Sardana et al., 2020) suggest researchers develop CSR frameworks consistently with their national documents and customs.
The study utilizes a variety of panel data analysis techniques to test for our research questions. In detail, to analyze the relationship between CSR and CFP, multivariate regressions were performed for a sample of 592 corporate observations obtained from Vietnamese stock markets (HOSE and HNX) from 2013 to 2017. The results provide empirical evidence that in Vietnam CSR and CFP have a positive correlation. Additionally, all components of CSR also have a significant positive impact on CFP. Last but not the least, this study reveals that as a moderating variable, EPU could drive down the value of TBQ, but stand out of the EPU-ROA nexus. Our findings are consistent with the literature that has been identified from different countries, especially developing markets.
Because of the potential concerns regarding omitted variables and individual heterogeneity, we perform a panel regression analysis. We find that fixed-effects and random-effects results are consistent with those of pooled OLS regressions. Furthermore, the endogenous phenomenon is tested and eliminated to confirm the stability of our results. Future studies may develop a more diverse set of criteria for CSR assessment, control variables that can be strengthened, and CFP indicators that can address other aspects such as level of growth or is the net market value instead of ratios (ROA, TBQ) as in the study. Although there are some limitations, the research has done a great amount of manual work due to data service issues in Vietnam. Through this research, the authors hope to introduce readers to have more concerns and awareness of CSR activities and their positive impact on the performance of businesses. We believe this study will contribute to encouraging policy makers, managers to strengthen CSR activities for a globally sustainable economy.