4.1. Preliminary Analysis
Table 1 illustrates the descriptive statistics. The companies in our sample are characterized by good financial performance. Their market values represent, on average, 1.12 of their book values. Similar results were found during the period 2009–2011 by
Boubaker et al. (
2014) for the SBF 120 companies. The authors noted an average of Tobin’s Q equal to 1.09 and a standard deviation of 0.84 with a minimum and a maximum of 0.22 and 7.86, respectively.
The average score of corporate social performance was 56.09%, with a standard deviation of 14.98%. This finding indicates the dispersion of the level of social commitment of French companies, which can be explained by the fact that French companies are only required by law to disclose ESG information in their reference documents or annual reports, but they have no commitments to fulfil. Furthermore, the average social responsibility score was 64.68%, while the irresponsibility score was 52.41%.
Boards have an average of 13 members, which indicates an expanded size of boards. A large number of directors can create agency conflicts, as it can be a factor of the diversity of resources and knowledge. Independent directors represent 51.47% of board members, which indicates that SBF 120 non-financial companies are aligned with the recommendation of the good governance code AFEP-MEDEF (33%). Moreover, CEO duality is widespread in French companies as 58.4% of our sample companies have the same person occupying the positions of chairman of the board and general manager. This last result is comparable to that of
Ahmadi et al. (
2018), who found that 59% of CAC 40 companies between 2011 and 2013 have CEO duality. We found that about 52% of companies appoint auditors from the Big 4 audit firms. The average size of the companies was 9.2969 (logarithm transformation of book value of total assets). The debt-to-equity ratio was around 0.87 for all the companies in our sample. The average age of the companies in our sample was 65.5 years (3.87 for logarithmic age transformation), indicating that the SBF 120 index is composed of mature companies. Finally, we note that the asset growth average was 5.87% and is characterized by high volatility (standard deviation of 12.76%).
Table 2 depicts the number of observations and descriptive statistics of key variables per industry. We preferred to use the ICB classification (industry classification/ten industries) to the SIC classification (20 sector classification) to obtain more industry observations.
We noticed that most of the firms in our sample (27%) belong to the “Industrial” sector. Moreover, the companies belonging to the “Health Care” industry are the most financially efficient. In terms of social performance, these companies are characterized by a weak commitment to good practices (CSR), but also by a low level of irresponsibility.
Companies in the “Utilities” industry had the lowest financial performance. Despite the commitment of these companies to good social practices, they had the lowest social performance score. This last finding is explained by the high average of irresponsibility score. Indeed, good social practices are offset by bad ones. As a result, companies have, on the one hand, to be careful to avoid irresponsibility and, on the other hand, to implement good practices to have better social performance. Businesses belonging to the “Consumer Services” industry had the highest CSP scores and CSR scores averages. Additionally, companies in the “Technology” industry had the lowest irresponsibility score.
Table 3 presents the matrix of correlations between the different variables of our study. We found that the correlation between the explanatory variables of the same model was not high except for the correlation between the board size and the firm size. However, the variance inflation factor (VIF) values did not exceed two. These results indicate that there was no multicollinearity problem in our case.
Regarding the correlation matrix, we noted that CSR was positively correlated with social performance. However, CSI was negatively correlated with social performance. This result indicates that good social practices increase the company’s social performance, while bad practices reduce it. We found that the CSR and CSI measures were positively correlated. This observation reveals that the more engaged in good social practices the company is, the more likely it is to have bad practices, as companies involved in bad social practices try to invest in good ones to hide irresponsibility.
Measures of good and bad social practice were positively correlated with firm size. We also noted that the size of the board and its independence were positively correlated. This finding shows that the larger the board, the more independent it is.
4.2. Estimation Results
Table 4 illustrates the regression results. We noted that CSP had no significant impact on performance. Thus, our first hypothesis on the association between CSP and firm performance (H1) was not supported by our results. One explanation comes from the multidimensional nature of CSP scores and the overlap of the different components (responsibility/irresponsibility). Furthermore, we can explain this result by balancing the use of valuable resources by investing in socially responsible activities and the profits associated with this investment. Indeed,
McWilliams and Siegel (
2001) estimated that the costs of social responsibility consume the profits generated, which leads to a situation of equilibrium.
The results of the estimation of the effect of CSR and CSI on firm performance showed that CSR and CSI exert opposite effects on firm performance. In particular, a higher CSR score reflecting a higher social commitment improved the company’s performance, but in a non-significant way. However, CSI had a negative and significant effect on financial performance. Accordingly, Hypothesis 2 is validated. Similar results were found by
Price and Sun (
2017) and
Fatemi et al. (
2018).
Furthermore, the results showed that the impact of CSI was stronger than that of CSR on the financial performance of the company. Indeed, we observed a negative and significant impact at the 1% threshold for CSI compared with a non-significant coefficient for CSR. These results confirm Hypothesis (H3) and are consistent with the results of
Price and Sun (
2017).
These results suggested that, for the French market, addressing bad social practices is more valued and has a greater impact on the financial performance of companies than the development of good social practices. This observation can be attributed to the fact that when faced with bad social practices, stakeholders react strongly and no longer appreciate the company’s good social practices, hence the need to take a closer look at the duration and strength of CSR and CSI effects.
To examine the duration of CSR and CSI effects on financial performance, we used a vector autoregressive (VAR) model designed for panel data. This method allowed for the exploration of the dynamic relationship between the variables of the panel data (
Abrigo and Love 2016). These models include systems of equations that assume that the variables within each system influence each other in the time series. The results of their long-term effects are reflected by the impulse response functions (IRF). The IRFs show the magnitude of the change that one variable will have as a function of the change (one standard deviation) of another variable.
This elasticity can be visualized to illustrate the direction and strength of the relationship. To this end, we first determined the number of lags to be performed based on the three model selection criteria by
Andrews and Lu (
2001). Indeed, the VAR analysis for panel data is based on the choice of the optimal lag order in the model specification.
Andrews and Lu (
2001) proposed selection criteria analogous to the various criteria for selecting models based on the commonly used maximum likelihood, namely the Akaike information criteria (AIC) (
Akaike 1969,
1981), the Bayesian Information Criteria (BIC) (
Schwarz 1978;
Akaike 1977), and the Hannan-Quinn Information Criteria (HQIC) (
Hannan and Quinn 1979). For these different criteria, the rule is to select the model for which this criterion is minimal. For our study, the first-order VAR was the selected model because it has the smallest MBIC, MAIC, and MQIC.
Then, based on the selection criteria, we used a GMM method for VAR model to estimate the equations using the lagged terms as instruments. This method uses a Helmert transformation on variables by default.
Table 5 presents the results of the estimation by the VAR model.
Taking into account the dynamic nature of the effects of CSR and CSI, the impact of CSR became significant at the 10% level. CSI had a negative and significant impact at the 5% level.
Before evaluating the impulse response functions (IRF) and forecast error variance decomposition (FEVD), we first checked the stability condition of the estimated panel VAR. The resulting table and the graph of eigenvalues confirmed that the estimate was stable.
The long-term effect was then demonstrated by generating IRFs from the system. We calculated the significance intervals of the elasticity estimators using a 1000-fold Monte-Carlo simulation. This analysis aimed to determine the significance of the effects of CSR and CSI over a 10-year forecast period.
Table 6 shows the results of the forecast error variance decomposition.
Following the FEVD estimates, we can see that the variation in financial performance measurement was explained by CSI up to three times more than CSR over the 2nd forecast period. This gap narrowed and stabilized as of the 3rd period, from which CSI exerted a double effect on the financial performance of companies compared with CSR. Nonetheless, we noticed that the effect of a CSI shock lasted longer than a CSR shock. Indeed, the impact of CSR was established from the 4th period, while the impact of CSI only stabilized from the 5th period; thus, our Hypothesis 4 was confirmed. More germane to this study is the reaction of stakeholders to CSI reflected in the financial performance of firms. We found that financial performance decreased more after CSI shocks than increased after CSR investment, in absolute value. This is consistent with the prediction of the stakeholder’s theory and psychology literature.
As can be seen in
Figure 1, the effects of CSR and CSI were best reflected when we assumed that the impacts of the two components of CSP were not instantaneous and lasted over time. Based on these results, we can conclude that CSI not only produces a much stronger effect than CSR, but also has a more persistent effect. This finding adds to the stakeholder theory that it is crucial to avoid CSI practices to satisfy stakeholders. The stakeholders’ disappointment following a negative event has a stronger impact that lasts longer than their satisfaction following a positive event on financial performance. Indeed, irresponsible behavior persists more than responsible behavior in the memory of stakeholders.
Our findings may be explained from the following three aspects:
First, the media tend to pay more attention to negative news rather than to positive ones. In particular, firms are supposed to behave ethically, and thus CSI practices attract media agencies (
Sun and Ding 2021). Hence, CSI practices will be widespread among stakeholders and will affect the analyst recommendations and investors and consumers’ perceptions, and therefore firm performance and value (
Luo et al. 2010).
Second, board characteristics, as a governance mechanism, may help to understand the opposite effects of CSR and CSI on CFP. In fact, many board characteristics impact the quality of CSR engagement. For example, a better CSR use associated with better performance is detected in larger and more experienced boards, as well as those with more independent directors (
Pekovic and Vogt 2021;
Rossi et al. 2021). Likewise, board gender diversity was shown to exert a greater impact on avoiding CSI than on investing in CSR (
Boulouta 2013;
Boukattaya and Omri 2021).
Third, CSR and CSI practices vary significantly depending on the company’s industry. In fact, firms belonging to high competition intensity industries are particularly focused on maximizing competitive advantages and minimizing disadvantages because the latter will deteriorate firm reputation and decrease the brand equity and the firm value (
Sun and Ding 2021).