**3.1 Sample and data collection**

The sample was obtained from the STOXX Europe 600 Index and the information was accessed during February 2018. Financial information was taken from Bureau Van Dijk's Amadeus database, version 14.07. Additional data was taken from FTSE Group and World Bank.

Our measure of CSR is based on The STOXX Europe Sustainability Index, a subset from the STOXX Europe 600 Index. It aggregates the selected companies according to a sector-business analysis together with sustainability assessments based on environmental, social, and economic criteria. The assessment is considered positive if the combination of company valuation and sector valuation results in a shaded matrix field in the Sarasin Sustainability Matrix [42].

From the initial sample of 600 companies, all those belonging to the public and financial sectors were excluded due to their specific rules and legislation, thus avoiding possible bias in the results. Companies for which it was not possible to calculate all the variables under study were also removed.

Besides, in order to avoid bias due to the extreme values found, outliers were also removed. Outliers are defined as the values of the variables below percentile 5% and above percentile 95%. Therefore, the final sample used for the study consists of 266 companies, with a total of 2660 observations.

Through the analysis of the sample composition by country (**Table 1**), we can see that most of the companies are from the United Kingdom with 84 companies (32%), France with 50 companies (19%), and Germany with 36 companies (14%).


**Table 1.**

*Sample by country.*


**Table 2.** *Sample by sector.*

According to the North American Industry Classification System (NAICS), the 266 companies were divided into 14 sectors. The most represented sectors are the professional, scientific, and technical services sectors with 85 companies (32%), followed by the industrial and other services sectors with, respectively, 53 companies (20%) and 34 companies (13%) (**Table 2**).

#### **3.2 Empirical model**

Prior research on CSR has measured financial performance using accountingbased measures or market-based measures (e.g., see [14, 43, 44]). As accountingbased measures, those authors have used return on equity (ROE) and return on assets (ROA) and, as market-based measures, they use the Tobin's Q . These two types of measures can capture the two dimensions of financial performance: the short-term through ROE and ROA [15] and the long-term and future evaluation through Tobin's Q [14]. Indeed, several authors choose to use *Tobin's Q* in order to study the relation between CSR and financial performance in a long-term perspective [39, 45].

Following previous studies, financial performance is measured by both ROE and ROA as accounting measures of short-term financial performance. ROE provides information on how efficient the company is in using its shareholder's invested capital [39], while ROA measures the efficiency that comes from using all company's assets during a fiscal year, that is, the ability to generate earnings [15]. Both profitability ratios are based on the company's net income over a given fiscal period because it is what effectively "remains" after all expenses are deducted from the income obtained, thus presenting the impact of financial policies and also the tax burden incurred by companies in different countries. We also use Tobin's Q as a market-based measure of long-term value which has proven to be an important variable to assess the future financial performance [46].

*Social Responsibility and Financial Performance: The Case of STOXX Europe Index DOI: http://dx.doi.org/10.5772/intechopen.93573*

Therefore, based on [43], three estimation models were developed to test the hypothesis:

$$\begin{array}{l} \text{ROE}\_{i,t} = \beta\_{\alpha i,t} + \beta\_i \text{CSR}\_{i,t} + \beta\_z \text{Size}\_{i,t} + \beta\_y \text{Leverage}\_{i,t} + \beta\_4 \text{Industry}\_{i,t} \\ \quad + \beta\_\circ \text{Country}\_{i,t} + \beta\_\circ \text{Financial Slack}\_{i,t} + \varepsilon\_{i,t} \end{array} \tag{1}$$

$$\begin{array}{l} \text{ROA}\_{i,t} = \beta\_{\text{ol},t} + \beta\_{\text{i}} \text{CSR}\_{i,t} + \beta\_{\text{z}} \text{Size}\_{i,t} + \beta\_{\text{y}} \text{Leverage}\_{i,t} + \beta\_{\text{i}} \text{Industry}\_{i,t} \\ \quad + \beta\_{\text{y}} \text{Country}\_{i,t} + \beta\_{\text{6}} \text{Financial Slack}\_{i,t} + \varepsilon\_{i,t} \end{array} \tag{2}$$

$$\begin{array}{c} \text{Tobbin's } Q\_{i,t} = \beta\_{\alpha i,t} + \beta\_i \text{CSR}\_{i,t} + \beta\_z \text{Size}\_{i,t} + \beta\_y \text{Leverage}\_{i,t} + \beta\_4 \text{Industry}\_{i,t} \\ \quad + \beta\_\circ \text{Country}\_{i,t} + \beta\_6 \text{Financial Slack}\_{i,t} + \varepsilon\_{i,t} \end{array} \tag{3}$$

*CSR* is a dummy variable that assumes the value 1 if the company belongs to the STOXX Europe Sustainability Index and value 0 otherwise. The STOXX Europe Sustainability Index aggregates companies based on their sustainability ratings. The index i represents each of the companies in the sample, and the index t refers to the year. The estimation method used was the pooled Ordinary Least Squares. We controlled for unobserved country and year heterogeneity using country and year fixed effects. The standard errors were grouped by company in order to correct the presence of autocorrelation.

Based on prior literature, the following control variables were chosen: *Size*, *Leverage*, *Industry*, *Country*, *Financial Slack*, and *Crisis.* Size is a relevant control variable since larger companies are assumed to have more visibility, and to generate a greater impact with their operations [47], they are more likely to adopt CSR policies compared to small companies [12, 23]. Financial leverage was also taken into account since high debt levels lead to high levels of financial leverage causing a negative impact on financial performance [48]. In line with this conclusion, [12] also showed that this negative impact continued to persist when financial leverage was introduced in a CSR regression.

In addition, the type of business activities [49] as well as the level of economic development of a country [50] may be related to a higher or lower CSR. Indeed, companies developing activities with high social and environmental impacts tend to adopt more CSR policies compared to others. Besides, companies with high liquidity are more likely to adopt CSR policies compared to others with less liquidity that can only focus on their own business activities [51]. Appendix 1 provides more detailed information about variables' measurement.

#### **4. Result analysis**

#### **4.1 Descriptive statistics**

**Table 3** presents the descriptive statistics of the variables for the total sample. *ROE*, *ROA*, and *Tobin's Q* present averages of 14.0, 5.4, and 97.6% and medians of 13.5, 5.1, and 0.786%, respectively. Regarding the standard deviation statistical measure, the values are small regarding the averages of each of the variables and do not show huge discrepancies, suggesting a certain normality in the sample distribution.

Furthermore, it is possible to observe that companies have, on average, a level of indebtedness of approximately 59%, suggesting that they rely more on external


#### **Table 3.**

*Descriptive statistics for the whole sample.*

capital than on equity to meet the asset needs. In terms of the current liquidity, that is, the ability to meet short-term liabilities, the result is higher than 1 (1.51), which means that companies have a favorable short-term financial situation.

In a next step, we divided the sample into two subsets, companies that pursue social responsibility-based policies (SRSE) and those that do not (NRSE). **Tables 4** and **5** present the values for the SRSE and NRSE, respectively.

It is possible to observe that the SRSE shows, on average, higher values than the NRSE for all financial performance measures, which means that, on average, SRSE has a higher financial performance compared to the NRSE. Moreover, the average of *Tobin's Q* in SRSE is higher than 1, while in NSRSE it is lower than 1, suggesting that companies pursuing social responsibility-based polices are more valued by the market.

For the remaining variables, on average, SRSE is larger than NRSE and the debt ratio is higher for SRSE compared to NRSE by approximately 4 percentage points


#### **Table 4.**

*Descriptive statistics for SRSE.*


#### **Table 5.**

*Descriptive statistics for NSRSE.*

*Social Responsibility and Financial Performance: The Case of STOXX Europe Index DOI: http://dx.doi.org/10.5772/intechopen.93573*


**Table 6.**

*Mean t-test results.*

(61.6% for SRSE and 57.3% for NRSE). On the contrary, *Financial Slack* presents higher value for NRSE (current liquidity of 1.545) on comparing to SRSE (current liquidity of 1.419).

**Table 6** shows the results of the mean equality test of the dependent variables *ROE*, *ROA*, and *Tobin's Q*. Results suggest that there is statistical evidence to assert that the means are different between SRSE and NSRSE, since the p-value is 0.000 in all dependent variables.

The correlation between the different variables is presented in Appendix 2. Most of the variables do not show strong correlations with each other and are statistically significant at 1%, except for the correlation of the *Leverage* and *Low Impact* variables, which are statistically significant at 5%.

The dependent variables *ROE*, *ROA*, and *Tobin's Q* are positively correlated with the independent variable *CSR*, suggesting that firms that pursue CSR activities have higher financial performance. Regarding the control variables, *Size*, *Low Impact*, *Country*, and *Financial Slack* are positively related to the dependent variables, suggesting that companies with higher financial performance values are larger, have low environmental impacts, belong to countries with high economic development, and have higher liquidity values. On the other hand, the dependent variables are negatively correlated with *Leverage*, *Medium Impact*, and *High Impact*, meaning that companies with high debt values and higher environmental impact have lower financial performance values.

A multicollinearity test was performed by calculating the variance inflation factors (VIFs). The values are less than 10, suggesting that there are no multicollinearity problems.
