Corporate social responsibility and financial performance: The case in Vietnam

Abstract This study aims to examine the impact of corporate social responsibility (CSR) on the financial performance of Vietnamese listed companies from 2012 to 2017. The study uses Fixed effects model and System Generalized Method of Moments to estimate models. The study contributes by analyzing the impact of social responsibility under three perspectives, namely economic, environmental and social responsibility in a developing country’s setting. The results suggest that overall CSR disclosure has a negative impact on firm performance, but the perspectives provide a more complete view: environmental responsibility shows a clear negative influence, while social responsibility demonstrates a preferable but weak impact on financial performance. The economic aspect does not show a significant effect on firm performance. The most burdensome category of CSR is the environment-related one, which calls for more careful employment of this investment and governmental support to ensure that it is more efficient.


PUBLIC INTEREST STATEMENT
Corporate social responsibility (CSR) refers to the activities firms conduct to contribute to improving the well-being of the society. Firms are engaged in CSR to gain societal approval and resources to be able to attain better performance. However, it is obvious that CSR is a costly realm, too. In this study, we use a sample of listed firms in Vietnam from 2012 to 2017 and several research strategies to ascertain the effect of CSR on firm performance. Since CSR could be conducted in a number of fields, we use three measures of CSR in order to more comprehensively examine its impact. The results show that social responsibility has a weak positive impact on performance, while environmental responsibility demonstrates a clear negative relationship with firm performance. Meanwhile, economic responsibility does not show a significant effect. As the most burdensome category of CSR is environment-related one, this calls for more efficient investment in this field and governmental support to relieve the burden for firms.

Introduction
The growing international concerns about environmental sustainability have incentivized firms to disclose corporate social responsibility activities (Y. C. Chen et al., 2018). There is a clear trend for corporations to recognize the importance of and take part in CSR activities (Mark-Herbert & Von Schantz, 2007), so managers are obliged to understand and manage these activities effectively. According to Orlitzky et al. (2011), firms of different sizes and types have to be socially and ecologically responsible and sustainable while remaining economically competitive. The critical importance of CSR has led to its receiving great attention from professionals and academics (Madueno et al., 2015).
There are three possible reasons why firms disclose CSR information: to fulfil "social contract", to adopt and enhance legitimacy and to improve economic performance (Mathews, 1997). CSR disclosure might enhance financial performance if firms could positively impress the investors by providing CSR information that meets or even exceeds expectations of stakeholders (Brooks & Oikonomou, 2018;Pham & Tran, 2020), and this activity may be beneficial to both stakeholders and firm owners, creating a win-win relationship (Wua & Shen, 2013). Nonetheless, the information on CSR can negatively affect firm performance if firms do not have sincere motives and only provide information with presentational rationales. In such instances, managers can be more prone to exaggerated performance in some areas of CSR while concealing their poor performance in others (Brooks & Oikonomou, 2018). The superficial substance of those activities may adversely affect the firm's reputation, rather than improve it, if investors and stakeholders could uncover the true story. As a result, theoretically, due to the potential complexed nature of CSR reporting, more studies on the link between CSR disclosure and firm performance are imperative.
There have been a large number of empirical works on the linkage between CSR (disclosure and performance) and firm performance, but the results are quite inconclusive (Al-Malkawi & Javaid, 2018). Margolis and Walsh (2001) conduct meta-analyses and find that for the link between CSR activities and firm performance, about half of the empirical studies documented a positive effect of CSR, while a quarter suggesting no significant relationship, 5% indicating a negative relationship and the rest documenting ambiguous results. More recently, the situation has not changed much. For example, Choi et al. (2010), Sun (2012), Van der Laan et al. (2008), H. Chen et al. (2011), andWu (2006) provide evidence on a positive link between CSR and financial performance. On the other hand, a negative association between the two has also been frequently documented, e.g., Lopez et al. (2007), Cavaco and Crifo (2014). Another meta-analysis conducted by Wang et al. (2016) shows support for a positive association between CSR and firm financial performance. However, Wang et al. (2016) note that the impact of CSR tends to be more beneficial in developed countries, i.e., the CSR-CFP link is moderated at least by some institutional factors.
Throughout the literature review, there are still several significant gaps to fill. First, importantly, Al-Malkawi and Javaid (2018) and Oh and Park (2015) claim that there has been little research in the context of emerging markets. Meanwhile, Cui et al. (2015) and Wang et al. (2016) pinpoint that developing countries have different customers' concerns and less developed institutional systems and inefficient market mechanisms, which might hinder the preferable effects of CSR on firm performance, and the positive CSR-CFP link usually found in developed countries might not repeat in developing economies. Second, most studies only focus on one measure of the overall CSR and were not able to offer a more detailed and well-rounded analysis of the CSR impact on firm performance. Studies including Han et al. (2016) and those studied in Brooks and Oikonomou (2018) used ESG data (Environmental, Social and Governance Score), but the last component focuses more on the governance aspect of a firm and is available chiefly for firms in developed countries. Currently, firms in developing economies adopt reporting standards such as Global Reporting Initiatives (GRI) that also focus on environmental and social aspects, but the third component is economic responsibility, not the corporate governance. According to GRI disclosure guidance, economic responsibility covers a wide range of activities, from anti-competition behavior, support in the infrastructure investment, priority in local suppliers and staff sourcing. This economic performance is expected to affect the economic conditions of the firm's stakeholders as well as the economic systems at local, national and global levels (GRI, 2016). Extant studies have examined the social and environmental aspects of CSR disclosure, but little has been done with regard to economic performance. Understanding which aspect of CSR tends to have a positive effect on firm performance is important in directing resources for investment in such aspects, while ensuring that the less efficient aspects are monitored more closely. This suggests that it is important to also examine this aspect of CSR activity, which remains quite silent. Vietnam is a developing country that is currently adopting GRI standards, which is a suitable research setting.
This research seeks to extend the empirical literature in many aspects. First, it investigates the three aspects of CSR to analyze the impact of different categories of CSR. This is important in the sense that the results would help to uncover the impact of different aspects of CSR, rather than just an overall aspect of social responsibility. The economic responsibility aspect of CSR has seldom been researched before. Secondly, through this analysis, we can expect to comprehend more deeply the relevance of theories in a developing economy context. Thirdly, we conduct the research using data covering non-financial firms in Vietnam, an emerging market to provide more evidence on the relationship between CSR activities and firm performance, thus contributing to filling the gaps in developing countries' context.
The remaining of the research is structured as follows. Section 2 provides a discussion of relevant theories and empirical studies, which serve as the basis to form our testable hypotheses. Section 3 presents our empirical strategy, covering research models, variable construction and data sources, and estimation strategy. Section 4 provides estimation results and the discussion of the results. Finally, Section 5 summarizes the findings and provides implications based on the research findings.

Literature review and hypothesis development
Firms disclose their socially responsible activities through the disclosure in annual or separate CSR reports. The relationship between CSR and financial performance is a field of study that is based on a number of foundational theories. To name a few, there are agency theory (Eisenhardt, 1989), legitimacy theory (Brown & Deegan, 1998), stakeholder theory (Freeman, 1984;Garriga & Melé, 2004;Orlitzky et al., 2003) and slack resource theory (Waddock & Graves, 1997). A single theoretical perspective cannot fully explain CSR practices because CSR activities are complex. In this research, the legitimacy and stakeholder theories are employed to provide a framework for the analysis. Stakeholder theory focuses on all stakeholder groups, while legitimacy theory seems to reflect the expectations of society at large (Freeman, 1984).
According to stakeholder theory (Freeman, 1984), different types of stakeholders, including customers, investors, employees and suppliers, have different positions and negotiation powers depending on their control of resources that affect corporate decisions. Upon responding to stakeholder expectations, firms can alleviate the risk of stakeholders withdrawing resources, harming the survival and operations of the firms (Wang et al., 2016). Furthermore, appropriate CSR activities also build up and maintain strong and decent corporate images, satisfying different stakeholders (Franco et al., 2020;Rhou & Singal, 2020). This effect will help investors to have more favorable considerations for the firm stocks (Flammer, 2015;Madsen & Rodgers, 2015).
With regard to the legitimacy theory proposed by Suchman (1995), an entity is expected to be able to meet societal expectations because it is a member of the society. If it does not meet the social standards, it can risk losing the required legitimacy to survive and operate in the society. In other words, firms that perform poorly in terms of CSR practices can be seen as illegitimate. On the contrary, firms that show strong CSR commitments can earn legitimacy more efficiently, so that their economic performance can be improved.
As mentioned earlier, the empirical results tend to be mixed. It is interesting to note that Margolis and Walsh (2001) find that the majority of the dated literature points to a positive association between CSR and financial performance, agreeing with the value of CSR in improving corporate images and garnering support from stakeholders to enhance economic outcomes. Similarly, Orlitzky et al. (2003) reviewed previous papers and concluded that CSR tends to have a positive effect on firm performance, even in a wide range of contexts and industries. Qiu et al. (2020) and Al-Malkawi and Javaid (2018) also find a positive impact of CSR on firm performance. Specifically, CSR (Zakat contribution) is found to have a positive impact on the financial performance of 107 firms from 2004 to 2013 in Saudi Arabia (Al-Malkawi & Javaid, 2018). The authors believe that CSR in the form of Zakat contribution is a win-win strategy, serving for the benefit of the society while improving profitability and value of firms. Qiu et al. (2020) examine the role of CSR in terms of protection of firm value during the spread of the novel Coronavirus. The research provides evidence that CSR engagement is conducive to improving stock returns as well as stakeholder attention during the pandemic. One more notable result is that in the pandemic context, community-related CSR tends to be more preferable and shows a more immediate and positive effect on stock returns, compared to activities focusing on customers and employees.
In summary, as suggested by legitimacy and stakeholder theories, CSR activities are those that firms could use to garner the attention and approval of stakeholders who possess resources that are prerequisite to firm survival and growth, and to obtain social legitimacy. Therefore, more CSR activities could be positively related to firm performance. Our first hypothesis is as follows:

H1: CSR activities are positively associated with firm performance
On the other hand, various studies have documented no significant linkage between CSR and firm performance, e.g., Aras et al. (2010), Lee et al. (2013), andAras et al. (2010) find no significant relationship between financial performance and CSR in Turkey from 2005 to 2007. Lee et al. (2013) also find that CSR activities not related to operations, such as those that focus on community, can even decrease firm value, and only operations-related CSR activities, including those for employees and environment, could help increase firm value. Franco et al. (2020) also find that CSR activities do not necessarily generate adequate financial benefits for firms.
Rather than focusing solely on firm profitability, the impact of CSR activities on social externalities is also examined. Using a dataset of firms in China, Y. C. Chen et al. (2018) find that the mandate that requires firms to provide information on CSR activities actually forces the firms to conduct real responsible activities, creating pressure on firms to spend resources and reducing firm profitability. Those expenses really aid in the reduction of wastewater and SO2 emissions, especially in cities that are regulated by the CSR disclosure mandate. The evidence suggests that even though the stakeholders benefit from the behavior of firms or the mandate helps create positive externalities, firms still suffer from loss of profitability. Qiu et al. (2020) examine whether firms should invest more in CSR activities in difficult times. Even though Qiu et al. (2020) agree that CSR tends to improve long-term financial performance, in line with Feng et al. (2018) and Flammer (2015), the authors argue that CSR can also comprise investments to improve social well-being without benefiting corporate well-being. Furthermore, according to slack resource theory, CSR often involves huge costs that negatively affect financial well-being, especially in the difficult times, such as industrial crises or disasters.
Finally, Wang et al. (2016) provided a meta-analysis showing support for a positive association between CSR and firm financial performance, but noted that the impact of CSR could be dependent on some institutional factors. Campbell (2007), Wang et al. (2008), and Wang et al. (2016) have pointed to several different institutional factors that encourage firms to be socially responsible in developed countries, e.g., tax breaks. Firms in developing economies are not able to reap such a preferred treatment, so CSR could be more devastating to their performance. Pablo et al. (2019) indicate the importance of simultaneous combination of three dimensions of CSR. Qiu et al. (2020) find that in the pandemic context, community-related CSR tends to be more preferable and shows a more immediate and positive effect on stock returns, compared to activities focusing on customers and employees. Therefore, it should be of managerial interest to examine the impact of different aspects of CSR on firm performance. In this research, we also aim to examine three dimensional CSR in the context of an emerging country.
In Vietnam, firms could face more constraints in terms of resources, and the slack resources theory predicts that activities that are not related to stakeholders that are close to the firms, e.g., employees, customers, could bring more negative effect on firm performance (Qiu et al., 2020). The slack resource theory should be highly relevant in Vietnam, a developing country, together with the fact that firms that are responsible may not receive financial incentives as in developed economies (Campbell, 2007;Wang et al., 2008Wang et al., , 2016. Compliance with environmental requirements disclosure could be a typical example because the disclosure would mean the firms have to conduct the costly activities in reality (Y. C. Chen et al., 2018). Other costly activities are those to fulfil economic responsibilities, e.g., anti-collusion and anti-competition requirements, or to be responsible to the local people, e.g., only seek local suppliers. The economic responsibility is not investigated earlier, and in this research, with the disclosure requirement of such responsibility, we are able to test its impact. Finally, social responsibility, which is in fact more directly related to firms, through requirements to address the concern of employees and customers, could be expected to be more visible and to bring more benefits compared to the previous two categories of responsibility.
In summary, in addition to positive effects, CSR could impose negative effects on financial performance. Therefore, our second hypothesis is as follows:

H2: CSR activities are negatively associated with firm performance.
Our final hypotheses for the components in the CSR department are: H3: Economic responsibility and environmental responsibility are negatively related to firm performance H4: Social responsibility is positively related to firm performance.

Research methodology
To fulfil the research objectives, the dynamic model (1) is used to examine the impact of CSR in year t-1 on financial performance in year t, while model (2) is to investigate the impact of three aspects of CSR in year t-1 (Pham & Tran, 2020).
CFP it = α 0 + α 1 CFP it-1 + α 2 CSR it-1 + α 3 Size it + α 4 Lev it + α 5 IND i + ε it CFP it = β 0 + β 1 CFP it-1 + β 2 ECO it-1 + β 3 ENV it-1 + β 4 SOC it-1 + β 5 Size it + β 6 Lev it + β 7 IND i þ μ it Where: i,t represent firm i in year t. The dependent variable (CFP) is proxied by accounting-based measure, namely return on total assets (Abu Farha & Alkhalaileh, 2016;Al-Malkawi & Javaid, 2018;Choi et al., 2010;T. L. Nguyen et al., 2019;Scholtens, 2008). ROA adequately captures the historical aspect of financial performance and is widely used for this purpose (Cavaco & Crifo, 2014). We refrain from the use of market-based performance indicator for a developing market like Vietnam because the market is still small and could be prone to manipulation and problems of herding as documented in Vo and Phan (2017). All of these factors may affect the market assessment of firm value.

CSR variables
The independent variables of interest include economics-related CSR (ECO), environment-related CSR (ENV) and society-related CSR (SOC). We also generate an overall index representing the overall CSR activities (CSR). CSR has been measured using different approaches. However, studies that use only one overall index of CSR could provide results that are vague, because CSR activities expand quite a number of categories. Based on the combination of disclosure requirements of the Ministry of Finance in Circular 155 and GRI-GSSB series of standards, the three dimensions of CSR disclosure cover the economic, environmental and societal effects of a firm' operations. The analysis is based on a list of indicators classified into three groups of aspects in accordance with the GRI standards: 1) economic responsibility (6 standards) including the economic impacts of the business on the locality and government; 2) environmental responsibility (8 standards) including corporate impacts on the environment, and 3) social responsibility (19 standards) includes business support activities to employees, citizens and local government, government, customers and suppliers.
We peruse the reports and find information regarding the above standards. If a firm mentions the content in accordance with a standard, it receives 1 point for that corresponding standard, and 0 otherwise. After determining the score for each standard of each firm in each year, the CSR dimension j for firm i in year t are calculated as follows: Where: CSR ij is the sum of all scores firm i receives for aspect j divided by total number of standards for aspect j (0 ≤ CSR ij ≤ 1). n equals 6, 8, 19, representing the number of criteria for each aspect j. The average weighting scheme has been adopted widely in the same field (Abu Farha & Alkhalaileh, 2016) The overall CSR is calculated by the following formula: Where: CSR it : overall CSR index of firm i in year t (0 ≤ CSR it ≤ 1). We apply the average weighting scheme in line with Cavaco and Crifo (2014).  (Cavaco & Crifo, 2014) and year effects. For Size, Bayoud and Kavanagh (2012) argue that large firms have more potential to generate profits compared to smaller peers. This variable is calculated as the natural logarithm of total assets. For Age (natural logarithm of firms' number of years listed on stock exchanges), Raymond and St-Pierre (2010) claim that firm age may affect performance. Lev is used to control for the effect of firm leverage and is calculated as the ratio of total debt to total assets. IND represents the business sector. According to Dierkes and Preston (1977) and Deegan et al. (1996), some industries may have a stronger linkage between CSR disclosure and financial performance. For example, some mining and oil and gas industries will have more incentives to disclose environmental and social impacts. This is because manufacturing firms tend to have a more direct impact on the environment and other stakeholders compared to non-manufacturing firms. Therefore, firms in the two sectors tend to have different CSR activities, and the linkage between CSR and firm performance can also be different for these two sectors. In line with Hossain et al. (2006) and Bayoud and Kavanagh (2012), we include business sectors to further control for sectoral effects. Manufacturing firms are those that belong to materials, industrial, transportation and food industries, while non-manufacturing firms are those in the real estate, construction, technology and services industries. If a firm is in the manufacturing sector, IND receives the value of 1, and 0 otherwise. The data for the control variables are obtained from Thomson Reuters Eikon.
The study applies the two-step System Generalized Method of Moments (GMM) estimator, in line with studies that use dynamic models (Al-Malkawi & Javaid, 2018;Cavaco & Crifo, 2014;Madorran & Garcia, 2016). The study employs GMM method to estimate dynamic panel data models thanks to its ability to solve endogenous problems due to the presence of a lagged dependent variable as an independent variable in the model. Roodman (2009) used the lagged differences of the explanatory variable as instrumental variables. In summary, the GMM regression must satisfy two tests (Hansen test and autocorrelation of order 2 test) to ensure that the estimates are valid for statistical inferences (Roodman, 2009). In addition to GMM estimation, we also conduct fixed effects regression to estimate the static model and compare the results.

Research results and discussion
We collect data covering 480 enterprises from 2012 to 2017, totalling 2,410 firm-year observations. We started with a full list of listed firms in Vietnam that do not belong to the financial sector. We then collect CSR information, but the information is available only for roughly 75% of the population of non-financial listed firms (the total number of listed firms in 2017 is 728 firms, and about six hundred firms are nonfinancial ones). We only collect information from 2012 because Vietnamese listed firms first started to include social responsibility content in their annual reports as required by the State Securities Commission. Firms disclosed more CSR activities when the Ministry of Finance issued Circular 155/TT-BTC on the guidelines on information disclosure on the disclosure of CSR information.

Descriptive statistics & correlation matrix
CSR had increased its average value each year from 2012 (0.256) to 2017 (0.356), indicating the increased interest in fulfilling CSR activities over time. In the same period, ECO increased from 0.4 to 0.45, SOC from 0.27 to 0.34, and ENV from 0.1 to 0.27. This suggests environmental responsibility tends to be receiving more concern over time; however, since it started the period with the lowest value so it received the lowest average value for the whole period. In Table 1, the mean values of the independent variables representing aspects of social responsibility are arranged in order from high to low values: ECO (0.4268), SOC (0.3057) and ENV (0.1826). This indicates that firms listed on the Vietnamese stock exchange were engaged in and disclosed social responsibility activities. Economic responsibility (ECO) activities tend to be conducted the most, and firms in the research sample fulfilled approximately 40% of the economic standards. In the meantime, firms could only satisfy a third of social activities. We find that sampled firms could only complete 20% of environmentally responsible activities prescribed in GRI. This is partly because not all the firms in the sample are manufacturing ones. In fact, about two thirds of the firms are non-manufacturing, which might explain why ENV has low value. On average, about half of the assets are funded with debt, while ROA is approximately 6%. With regard to the age variable, on average, the logarithm of the number of listing years is 1.7022, and this is also consistent with the majority of the firms' listing years in the range from 6 to 8 years. This information indicates that the firms should have little experience in preparing CSR reports, compared to those in developed countries. Table 2 presents the pairwise correlation coefficients between the variables in the research model. CSR i,t-1 is positively and significantly correlated with ROA. This suggests that, the more companies disclose information about their social responsibility activities, the better their financial performance. In terms of aspects of social responsibility, we find that ECO i,t-1 , ENV i,t-1 , SOC i,t-1 are also positively and significantly correlated with ROA. Nonetheless, the correlation analysis just provides the correlation between each pair of variables; as a consequence, it does not consider the interaction of variables introduced in the same model. Therefore, we proceed with the estimation of models (1) and (2) using the System GMM estimator. The issue of multicollinearity should not be concernible, given the low correlation coefficients among the independent variables. We also conducted the Variance Inflation Factor test and all the values are lower than 2. Table 3 shows the results of the estimation of model (1). The p-values of autocorrelation of order two and overidentification tests are not statistically significant at 10%, indicating that the estimation results are valid for the statistical inferences (Roodman, 2009). Furthermore, the coefficients of lagged dependent variables are significant at 1% level, indicating that the use of dynamic models is necessary and appropriate in this paper.

Regression results
The estimation results point to a negative association between CSR and accounting-based financial performance measure, in line with Cavaco and Crifo (2014), Cui et al. (2015), and Kao et al. (2018) and hypothesis H2. This is consistent with the argument that firms have to spend resources on projects sustaining environmental conditions and improving the well-being of employees and the community to demonstrate their social responsibility. These activities could be expensive without direct benefit for firm financial performance and could render some activities unpaid (Feng et al., 2018;Flammer, 2015). Beldad et al. (2020) argue that inappropriately managed and executed CSR initiatives could even inhibit the firms from enjoying the CSR's benefits. Interestingly, Cui et al. (2015) suggest that in developing countries, customers are less concerned about CSR-related characteristics of products and prefer cheaper products. Ting (2021) also shows that large firms superficially disclose their CSR activities, so these activities do not have a positive impact on firm performance. This evidence is not consistent with the view that CSR activities enable firms to gain the trust of customers, employees and investors and social legitimacy, which enable firms to maintain value in developed countries (Madorran & Garcia, 2016).
We further perform fixed-effects regression with robust standard errors on model (1) and find that CSR is negatively associated with ROA. This result confirms the robustness of the findings that CSR has a more negative impact on firm performance in developing countries where there are less developed institutional factors and CSR activities have lower chance of visibility (Cui et al., 2015;Wang et al., 2016). Since the use of an overall index of CSR does not give much insight, we decompose the CSR activities into three components: economic, social and environmental responsibility, and investigate the impact of these components on firm performance. Since the use of an overall index of CSR does not give much insight, we decompose the CSR activities into three components: economic, social and environmental responsibility, and investigate the impact of these components on firm performance. The p-values of autocorrelation of order two and overidentification tests are not statistically significant at 10%, indicating that the estimation results are valid for the statistical inferences (Roodman, 2009). In terms of economic responsibility, estimation results in Table 4 show that there is no significant relationship between economic responsibility and ROA. On the one hand, the disclosure of information about economic responsible commitments, such as anti-fraud, corruption, anti-monopoly, the choice of local suppliers, and so on, is expected to enable firms to gain trust of stakeholders and gain societal legitimacy. From the descriptive statistics, it is clear that the firms in the sample seem to recognize the importance of this type of CSR, so they tend to be engaged in this activity, as evident by the highest mean value of this CSR category. This widespread report could make firms' activities in this category go unnoticed if firms do not have any noticeable achievements. On the other hand, this activity is costly. According to slack resources theory, CSR often involves huge costs that negatively affect financial well-being (Qiu et al., 2020). Therefore, the two different effects could lead to an insignificant impact of economic responsibility on firm performance.
For environmental responsibility, the results of the model show an inverse relationship between the variable ENV and financial performance. This result is consistent with Y. C. Chen et al. (2018), which suggests that environmental disclosure is a burden for firms because they would have to spend considerable amount of resources to limit the emissions/wastewater discharge so that they could report some environmental performance indicators. The result is also consistent with the findings of Lioui and Sharma (2012), which suggest that implementing environmental responsibility is expensive and would obviously reduce financial performance. Cui et al. (2015) argue that consumers in developing countries tend to pay less attention to environment-related features of the products and be more interested in their prices, so investments to improve environmental performance could lead to lower financial performance in these economies. This factor remains significant when we use fixed-effects model to test for the robustness of the findings.
In Vietnam, beside the perennial problem of insufficient resources, firms also have to face the fact that environmental regulations are scattered in many legal documents, such as the Law on Environmental Protection in 2014, Law on Water Resources 2012, Law on Environmental Protection Tax and most recently Circular 155 that mandates the disclosure of environmental activities. Inconsistent guidance as well as the expensive compliance with the environmental protection regulations could pose a significantly negative impact on firm performance.
For social responsibility, there is a positive and statistically significant association between SOC and ROA. Information related to employees, communities, products and legal compliance is always the topics  (Franco et al., 2020;Rhou & Singal, 2020), thus improving firm performance. However, this effect is missing when estimating using a fixed-effects model, which suggests weaker evidence for the positive effect of social responsibility activities. This is in line with hypothesis H3, and consolidates the evidence that the overall impact of CSR is negative.
In summary, CSR disclosure has an impact on financial performance but with mixed effects. In particular, the overall CSR and environmental responsibility reduce ROA, which shows support for slack resources theory. Meanwhile, there is weak evidence in support of consistency with the stakeholder  theory and the legitimate theory that explains that CSR helps firms to build a good image with stakeholders and get recognized for their socially responsible activities, thereby improving ROA.

Conclusion and implications
CSR has been increasingly researched, but its impact on corporate financial performance is far from being conclusive. Furthermore, there is a dearth of research on this linkage in the context of emerging countries, especially when CSR is analyzed under several aspects rather than an overall index. Instead of considering CSR components using the framework of ESG systems that are more readily available in the context of developed markets, we use GRI guidelines that focus on social, economic and environmental responsibility in Vietnam, a developing country. This framework has received less attention; thus, our study is expected to bring new insight to the literature.
In this study, CSR disclosure in general has been shown to exert a negative impact on financial performance. In particular, the overall CSR reduces ROA for both System GMM and fixed-effects estimators. According to slack resources theory, CSR activities require resources and this may lead to a decrease in ROA, at least in the short term. The negative impact of CSR is most evident in the category of the environmental aspect, and this could be due to the increased environmental protection costs. This could prove to be a marked burden for firms in developing countries. On the other hand, society-related CSR tends to increase ROA, and this could be due to its wide-ranging effect on relevant stakeholders and increased legitimacy. However, this positive effect is not robust as shown in fixed-effects estimation.
The implication of this research is that CSR does not always improve firm performance, especially in the setting of developing countries. Firms should weigh up the pros and cons of CSR investments. Research findings show that environment-related CSR has a negative effect on firm performance, so the CSR activities in this category should be conducted with more efficient monitoring and planning to ensure better performance. Meanwhile, the government could support to ease the capital expenditure in this category, and release the environmental requirements that are harsh with the current technology and conditions. The limitation of this paper is that it only investigates the relationship between CSR and firm performance in an emerging market and only uses accounting-based performance indicator. Future studies could compare the effects of CSR on firm financial well-being in two countries or two groups of developed and developing countries, or among countries that have different levels of institutional quality. This would enrich the literature and provide important implications for improving the benefits of CSR investment.