The Impact of CSR Reporting on Russian and Dutch Companies Performance

To sum up, the analysis of previous studies demonstrates a shortage of quantitative researches about the impact of CSR on the performance of Dutch companies. Therefore, this paper will contribute to the existing literature by providing empirical evidence on CSR-CFP relations in The Netherlands.

Chapter 2. Setting up CSR-CFP model

CSR-CFP relations have been statistically examined in multiple researches. Even though most studies have determined positive relations between these variables, there exist several papers where the authors managed to find negative, neutral or U-shaped relations. For instance, Toit and Lekoloane (2011) did not find any correlation between CSR practices and financial performance of the companies listed on the Johannesburg Stock Exchange. Elouidani and Zoubir (2015) constructed their own CSR Index for the publicity traded companies on Casablanca stock exchange. The authors determined a strong negative impact of CSR on companies' financial indicators(Elouidani & Zoubir, 2015). A research of Barnett and Salomon (2012) found U-shaped relations between CSR and CFP(Barnett & Salomon, 2012). Quite interesting results were obtained in the work of Braga-Alves and Shastri (2011) where the authors found out that higher CSR scores lead to higher market value but not to higher profitability(Braga-Alves & Shastri, 2011). A study of Pernamasari (2015) determined a positive effect of CSR on companies' operating cash flow(Vena Purnamasari, 2015). Another research found a positive correlation between CSR and ROA of Indonesian Telecommunication companies(Firli & Akbar, 2016). Table 5 provides a summary of the results from the existing literature.

Table 6 - Types of CSR-CFR relations

Basing on the results of previous research papers and the fact that CSR-CFP relations have not been examined extensively for Russian and Dutch companies, we developed a two-tailed hypothesis:

H1: CSR reporting practices have an effect on CFP of Russian and Dutch companies.

A research of Black and Gorga (2012) demonstrated strong influence of country characteristics on CSR performance studying this concept on companies from emerging economies (Black & Gorga, 2012). Moreover, considering the fact that CSR concept exists in the Netherlands for a longer period of time than in Russia, we can expect that Dutch consumers and investors are more aware about the importance of social responsibility and, consequently, are more likely to distinguish between responsible and non-responsible practices. This leads to the following hypothesis:

H2: CSR effect on CFP differs for Dutch and Russian companies.

The above-mentioned hypotheses will be tested using the regression analysis. A full description of the research design, methodology and data is provided in the next Section.

2.2 Methodology

The research is based on the regression analysis that enables to determinea quantitative impact of CSR reporting practices on companies' financial performance. Previous researches have actively used statistical methods to measure the return from CSR activities. McWilliams (2000) has built an econometric model that defines financial performance as a function of CSR, firms' size, business risk and industry. The study of Vong and Wong (2013) performed statistical regression for the gaming industry. The authors examined the relations between various financial performance measurements including profit margin, revenue, income, earnings per share and CSR performance (Vong & Wong, 2013). Considering the methodology of recent research papers and the fact that econometric models are the most common way to measure correlation between two variables, a multiple regression analysis is performed to examine CSR-CFP relations.

2.2.2 Operationalization of variables

Return on equity (ROE) is determined as a measurement of business financial performance. ROE is a common measure used by most investors to access business efficiency and the performance of companies' top management(Horngren, 2015). That is why, this indicator has been chosen as a dependent variable for the study.

CSRhub scores for each organization in our sample are used as operationalization of our main explanatory variable. The method of calculation the scores and their validity have been discussed earlier in previous section. In order to isolate CSR impact on companies' financial performance, we included profit margin, assets turnover and financial leverage as control variables. The choice of additional explanatory variables is based on the DuPont analysis that assumes that ROE is composed of three ratios:

ROE = , where

Profit Margin =

Assets Turnover =

Financial Leverage = .

Profit margin will adjust for possible variation in ROE across the companies due to different profitability levels. Assets turnover will explain variation in ROE caused by differences in firms' size and efficiency of production. Financial leverage will capture the level of business indebtedness, the increase of which can artificially boost companies' ROE. Thus, the inclusion of the chosen control variables will enable us to avoid the problem of endogeneity in the model and, consequently, obtain valid measurements of CSR impact on business financial performance.

To test our second research hypothesis and capture potential differences in CSR effect between the countries, a dummy variable “Country” and cross variable “CSR#Country” are added in the model. Table 7 contains all the determined variables for our regression model.

Table 7 - Research variables

2.3 Sample

Given the fact that CSRhub provides the scores only for 45 Russian and 55 Dutch companies, the study is limited with a sample of 100 organizations. The research includes companies which are either listed on national stock exchanges (Euronext Amsterdam and MOEX) or officially register and have corporate headquarters in The Netherlands and Russia. Our sample also contains the data from the firms operating in different industries such as oil/gas, energy, mining, consumer goods, transport and financial services. Nevertheless, this study does not provide a cross-industry analysis due to quite a small number of companies related to a particular industry. Hence, the division of our sample per business industry would not enable us to obtain reliable statistical evidence.

The research variables were calculated based on the information from financial reports of sampled firms for the year 2017. All the data is presented in euro to enable comparison of our findings between Dutch and Russian organizations. However, since this study is conducted only for 2 particular countries, our results may not be generalized over the determined sample framework.

2.4 Building econometric model

In order to develop a valid econometric model, we will firstly conduct an analysis of our data set. Table 8 presents summary statistics for the research variables.

Table 8 - Summary statistics

Remarkably that there are quite a few differences in summary statistics for the Russian and Dutch companies. For instance, mean value for all the variables except CSR is almost equal in two subsamples. Given the above, we can conclude on small variation in ROEs by country and, consequently, expect insignificance of a single dummy variable “Country”. The difference in mean values of CSR scores for two subgroups is larger than for other variables. Therefore, we can assume the existence of various CSR effects for Russian and Dutch companies and, as a result, expect statistical significance of cross variable “CSR#Country”.

To determine a correct specification for our model, we will examine the type of relations between the dependent and explanatory variables.

Figure 3 - Scatter plots of ROE against Explanatory variables

Figure 3 shows positive correlation between ROE and CSR: companies with high CSR scores tend to have higher ROE. Thus, we assume the existence of liner relations between ROE and CSR. Moreover, we expect to obtain a positive Beta coefficient for CSR variable in our model. Other control variables demonstrate a weak positive correlation with our dependent variable. According to Higgins (2011), management of companies develops strategies to increase profit margin, assets turnover or financial leverage in order to improve corporate ROE. Considering this argument and graphical evidence, we can expect positive coefficients Beta for all the control variables reflecting the existence of the direct relations between profit margin, assets turnover, financial leverage and ROE. Finally, our scatter plots do not show any potential non-linear relations between the variables. Therefore, we are going to use a lin-lin specification of the model.

To ensure the absence of multicollinearity between our regressors that can lead to biased results due to low t-statistics and insignificant coefficients, we will examine a correlation matrix and variation inflation factor for our explanatory variables.

Figure 4 - Correlation Matrix and VIF for explanatory variables

According to the rule of thumb, VIF higher than 5 indicates the existence of multicollinearity in the model. Figures 4 shows that VIF for profit margin, assets turnover, financial leverage and CSR is below 5 which reflects the absence of significant correlation between explanatory variables. However, VIF for CSR#Country and Country has extreme values of 50.3 and 44.63 respectively. That is an evidence of strong relations between those regressors, which is also proved by their high correlation index of 0,97.Hence, the inclusion of both variables in the regression will cause a sever multicollinearity that can lead to a poor quality of our model. To avoid this problem, we will exclude variable “Country” from the regression keeping only a cross variable CSR#Country. The analysis of summary statistics has demonstrated similar results in two subsamples implying that on average our variables except CSR do not vary much by country. Therefore, we assume that the exclusion of dummy variable “Country” will not significantly influence our results.

In order to ensure that our model does not suffer from heteroskedasticity that causes low standard errors, high t-statistics and, consequently, declines the precision of estimations, we will examine the distribution of model error term. To be more specific, we are going to test the hypothesis that the variance of our residuals is homogeneous.

Figure 5 - Tests for heteroskedasticity

We performed White and Breusch-Pagan tests to reveal the problem of heteroskedasticity in our model. The results of White test allow to reject the null hypothesis about homoskedasticity and, consequently, conclude on the existence of heteroskedasticity in the model. On the contrary, Breusch-Pagan test enables to accept the null about homogeneous residuals at 1% and 5% levels of significance. The inconsistence of our results may be explained by high sensitivity of those tests to the model assumption of normality(IDRE, 2019). Hence, we are going to combine our findings with graphical representation of model residuals.

Figure 6 - Residuals vs Fitted values plot

Figure 6 demonstrates small concentration of the residuals in the middle of the scatter plot. That can be an indication of heteroscedastic residuals' variance. In order to draw a final conclusion on heteroskedasticity in our model, we will test our error term on normality.

Figure 7 - Testing residuals for normality

Graphical representation of error term distribution shows a bell-shaped form that is close to normal distribution. Moreover, summary statistics demonstrates a low skew ness of 0,35 indicating about quite a symmetric errors distribution. The kurtosis of residuals is higher than 3 which is a threshold for normality. However, the deviation is not critical to conclude on non-normal distribution of model errors. Finally, the results of Shapiro-Wilk test enable us to accept the null hypothesis about normal distribution of residuals at 1% level of significance. Even though White test and a scatter plot of model residuals against fitted values have given some evidence of heteroskedasticity, further investigation demonstrates that our error term is distributed closely to normality. Hence, we can conclude that the severity of heteroskedasticity in the model is not crucial and should not substantially affect the validity of our results.

Considering the character of the relations between our variables and the data analysis performed above, we will use a lin-lin econometric model to test our research hypotheses. Table 9 presents the expected results of our estimations that are discussed in the next chapter.

Table 9 - Expected results for econometric model: ROE = Я₁ + Я₂ Profit Margin + Я₃ Assets Turnover + Я₄ Financial Leverage + Я₅ CSR + Я₆ CSR#Country + e_i

Chapter 3. The effect of CSR on ROE

Table 10 - The results of estimated CSR-CFP model: ROE = -0,459+ 0,143 Profit Margin + 0,159 Assets Turnover + 0,183 Financial Leverage + 0,00022CSR - 0,00004CSR#Country + e_i

The estimations of our regression do not fully coincide with our initial expectations. Nevertheless, our control variables demonstrate positive impact on companies' ROE. Thus, an increase in profit margin by 1 monetary unit causes an increase in ROE by 0,14 units. A unit growth of assets turnover results in improvement of ROE by 0,16 monetary units. An increase in financial leverage by 1 unit will boost ROE by 0,18 monetary units. P-value for these repressors is quite small which enables to reject the null hypothesis about the insignificance of control variables. Hence, we can conclude that an improvement in one of those three indicators may indeed increase companies' return on equity. Therefore, management could use our results to predict ROE at different levels of companies' profit margin, assets turnover and financial leverage.

The coefficient for CSR and CSR#Country variables turned out to be significant at 5% level. Since CSR scores have been developed basing on the information disclosed by companies, our research hypothesis that CSR reporting have an effect on Russian and Dutch companies' financial performance can be accepted. Moreover, the hypothesis that CSR effect on CFP differs between the countries cannot be rejected as well. Thus, a unit increase in Russian companies CSR score improves their ROE by 0,00022 monetary units. On the contrary, a unit increase in Dutch companies CSR scores improves their ROE only by 0,00018 monetary units. Thus, we conclude that CSR positively affects CFP both for Russian and Dutch companies. However, the effect from sustainable reporting for the Dutch businesses is 0,00004 units less than for the Russian organizations.

3.2 Discussion of CSR-CFP relations

Even though the results of the regression analysis enabled to accept our research hypotheses and conclude that companies' CSR scores explain some variation in ROE, the strength of CSR-CFP relations is quite weak. This can be justified by the following reasons. Firstly, ROE is an accounting-based measurement of corporate performance which might not completely reflect expectations of companies' shareholders that are frequently affected by the level of business sustainability. For instance, companies with low CSR scores may not be perceived as sustainable, facing a higher risk of being penalized for unethical behavior or violation of environmental legislation. Since few investors would like to have a stake in such businesses, the demand for their shares would fall. Hence, total market value of unsustainable companies may be adversely affected by negative shareholders' expectations. Therefore, usage of market-based measures of business performance such as EPS (earnings per share) or market to book ratio could provide an evidence of stronger CSR-CFP relations than our model did. Secondly, weak CSR impaction ROE can also be explained by low perception of corporate social practices as an essential component for business success. Being an emerging concept, the level of corporate sustainability may still not be considered by companies' management and other stakeholders as an important business indicator. As a result, companies do not aim to develop appropriate sustainable strategies that could increase net income and, consequently, improve ROE. Thirdly, this study is based on cross-sectional analysis. Hence, our model does not capture the effect of time that is required to obtain the return from CSR strategies. Therefore, a panel data analysis should be performed for more precise investigation of CSR-CFP relations. Finally, operationalization of CSR concept might have affected our results as well. Thus, the usage of a content analysis of companies' CSR reports, in order to quantify sustainability concept, could provide different evidence of CSR-CFP relations.

Our study also demonstrates lower return from CSR reporting practices for Dutch companies in comparison with the Russian. This fact can be explained by the different levels of business risk in these countries. Dutch companies, especially those listed on national stock exchange, are obligatory to disclose non-financial information about their core business operations. At the same time, Russian organizations report on their non-financials on voluntary basis. This results in higher transparency of Dutch businesses in comparison with the Russian. Moreover, Dutch institutions which regulate the behavior of corporations are stronger than the Russian. Hence, Dutch government imposes more control on local businesses comparing to the Russian state. Given the above, investors may perceive Dutch businesses as less risky to invest rather than the Russian. This decreases the importance of CSR reporting for Dutch companies since local investors will be sure in high sustainability of national organizations in advance and, consequently, pay less attention to CSR disclosure. On the contrary, Russian businesses have been historically characterized as non-transparent. Therefore, CSR reports of Russian companies may be highly valued among the investors, becoming a source of additional information about businesses that is necessary for making solid investment decisions. Hence, different perception of business risk by investors as well as various levels of companies' transparency can justify lower CSR effect on the performance of Dutch companies comparing to the Russian.

3.3 Robustness of CSR-CFP relations

In order to derive a final conclusion on the relations between CSR practices and business performance, we will test our model for robustness. To ensure that our regression indeed reflects the dependence of ROE on CSR both for the Russian and Dutch companies, we will build two separate models for these countries.

Table 11 - Estimation of separate models

The results of estimations for two separate models are similar to our initial findings. All the coefficients stay significant meaning that our main regression is robust to changes in the sample. Hence, we can conclude that the impact of CSR on corporate ROE exists both for Russian and Dutch companies. However, new estimations demonstrate a variation in CSR effect for both subgroups in comparison with the results of our first model. Thus, CSR coefficient for the Russian companies slightly increased from 0,00022 to 0,00025 while for the Dutch organizations CSR effect fell from 0,00018 to 0,00013. Such changes may appear due to inadequate operationalization of CSR or omitted regressors. Therefore, in order to make sure that our main model is sustainable to changes in regressors and, consequently, measure a true CSR effect, we will re-estimate initial regression with new control variables. To account for business profitability and efficient usage of resources, we will replace profit margin and assets turnover by return on assets (ROA). ROA is a ratio of net income to total assets that shows how much of profit is generated by a company per each monetary unit invested into its assets. Given the above, we expect that ROA will have a direct positive effect on companies' ROE. The level of business indebtedness is another factor that influences return on equity (Higgins, 2011).Companies with big amounts of financial leverage are more likely to default on their obligations. Perceiving such businesses as high risky, investors will require higher return per each unit of their investment to compensate for the risk they undertake while providing their funds to these organizations. That makes companies to boost their ROEs up in order to attract new capital inflows. Hence, we will replace financial leverage by debt to equity ratio that demonstrates to what extent a business is financed by creditors rather than by equity holders.

Table 12 - New explanatory variables

1. Barnett, M. L., & Salomon, R. M. (2012). Does it pay to be really good? Addressing the shape of the relationship between social and financial performance. Strategic Management Journal, 33, 1304-1320.