Despite other authors, Barnett and Salomon (2012) took a different approach to solve this long-lasting discussion. In contrast to ideas of instrumental stakeholder approach or classical view, Barnett, and Salomon (2012) concluded that the association between these two phenomena could be both (negative or positive) and support their ideas that the key factor here is how well corporations are able to manage to capitalize their social responsibility efforts. Barnett and Salomon (2012) emphasized the importance of stakeholder influence capacity (SIC) in their investigation and pointed out that corporations with sufficient SIC should invest in social issues to make a return from a social investment. Furthermore, based on the logic behind SIC, they underline that the link between social and financial performance is neither negative nor positive, but U-shaped.The 3 model multiple regression results of Barett and Salomon revealed interesting aspects. While the first two model failed to support a U-shaped relationship between CSP and CFP, after using quadratic net KLD scores in the third model, the results failed to reject the main hypothesis. The effect of social responsibility on financial performance became first negative and then turned to positive. The results indicate that corporations with the highest and lowest social performance score (net KLD score) have the highest return on asset. On the other hand, regressing the second dependent variable (net income ) into a model, the first two model again did not fit with a hypothesis. However, in the third model, U shaped relationship is achieved. Yet, while the results for the third depended variable are somehow similar, there is some difference between them which we can see also from the figures 1 and 2. Authors find evidence that to have low net KLD score is not profitable as in case of ROA rather than net income. This is illustrated in Figure 1 and 2 for the visual representations. In the end, Barnett and Salomon (2012) concluded that to have the higher social performance score is much more profitable rather than having lower social performance score.
Barnett and Salomon (2012) also conclude the cost that is derived from the social performance can compensate by improving relations with stakeholders. As mentioned earlier, according to empirics the key factor in CSP and CFP relations is SIC, and authors support their idea that the return from the social investment is strongly associated with SIC. They explained that in short-term corporations should not expect the profit from the social investment, and those investments in social issues would produce only costs for corporations. However, in long-term these investments in social issues would improve corporation’s stakeholder relations and sensibly would increase SIC score which finally would produce a return from social responsibility. Finally, even though the relationship between CSP and CFP is U-shaped for Barnett and Salomon, they find evidence that to have high CSP would lead higher financial performance for corporations than lower CSP.
In contrast to others, McWilliams and Siegel (2000) directly indicated the problem of omitted variable bias on their paper and did their research to find a relationship and prove a correlation between R&D and CSP, by developing Waddock and Grave’s (1997) model. Note to mention, the detailed explanation of their research will be explained in Section III. McWilliams and Siegel’s (2000) main hypothesis was that there is a positive correlation between CSR and research and development (R&D) cost of the corporation. All things considered, after controlling the R&D and advertising expenses they concluded that the relationship between social responsibility and financial performance is neutral.
By contrast, when it comes to Waddock and Grave (1997), the particular attention paid to analyse and to find an answer to two main questions. Paper has presented solutions to explain the relationship between CSP and CFP and to define reverse causality between these two variables. The data obtained by 500 Standard and Poor firms for regression analysis. The answer to the second question will be explained in section IV. When it comes to the association between CSP and CSR, Waddock and Grave (1997) concluded that the relationship between financial and social performance is bidirectional and the direction of the relationship is positive.
Two main factors differentiated McGuire, Sundgren and Schneeweis’s (1988) investigation from other empirics. First is related to the measurement of financial performance, which authors went beyond to traditional measures and added risk factors in their regression. Second, they analyse the reverse causality between social performance and financial performance which will be looking at more detailed in section IV. The regression results of McGuire, Sundgren and Schneeweis’s (1988) support that there is no significant relationship between market-based indicators and corporation social performance, however, the results are significant in case of account-based measurement. These findings obtained are compatible with their view that account-based indicators have positive association with CSR activities. Likewise, it has been demonstrated that firm with better social scores generally has a lower financial risk.
In order to verify the relationship between CSP and CFP, Callas and Thomas (2008) carried out a research, which took the financial performance of the firm as a depended variable and control variables which likely misspecify the model. Notably, one of the main advantages of their research is related to control variables. Particular attention paid to control variables which highly criticised by early researches and authors try to adjust their explanatory variables accordingly. Callas and Thomas (2008) also find that there is a positive relationship between CSP and CFP. Moreover, they also concluded that this positive relationship changes throughout industries. Another interesting point is that they add the relationship between CSP and CFP could also be quadratic and further researches should be done to validate this relationship. Interestingly, Barnett and Salomon (2012), which mentioned above, support their and idea tested the quadratic relationship and found significant results.
Misspecification of model
Although a large number of researches, there is not still a clear evidence that shows the relationship between CSR and CFP. Different empirical results have found various; a direct, negative and neutral link between these 2 phenomena. (McWilliams and Siegel, 2000) This section is aimed to analyse the omitted variable bias problem and to find an answer which variable essentially generate the misspecification problem.
McWilliams and Siegel (2000) stated that the main reason for the inconsistency in the results is the omission of the significant variables in the empirical research. In statistics, the exclusion of a variable which has a relationship with an explanatory variable or is the main determinant of a dependent variable is called omitted variable bias problem which leads to misspecification of a model. (Stock, Watson; 2008). One of the most important problems in CSP-CFP relationship is misspecification of the empirical model, which a very important variable of profitability is omitted from regression. (McWilliams, Siegel; 2012).
Interestingly, Cochran and Wood (1984) also stated that authors should explore and add new variables on their research which may play a significant role in CSP and CFP relationship. However, this variable also changes from different authors to authors. According to Cochran and Wood (1984), the omitted variable that cause misspecification is “asset age”. However, McWilliams and Siegel (2000) found the high correlation between research and development cost and social performance. Barnett and Salomon (2012) had also emphasized the importance of research and development cost, and therefore controlled that variable in their empirical research analysis.
Like Barnett and Salomon (2012) Cochran and Wood (1984) have also used more than one dependent variable in their multiple regression models and measured their financial performance as operation earnings to sale and operation earnings to assets ratios and excess market valuation separately. Interestingly, Cochran and Wood (1984) had developed two different regression models in their research and in the first regression model, they try to analyse the effect of the social performance on the financial performance of a firm while controlling industry effect. Reasonably, in the first model dependent variable was considered a financial performance of the corporation, while independent variables were the social performance and industry dummy variables. Cochran and Wood (1984) empirical analysis had very interesting results. While regressing the operating earnings to sales and excess value as a depended variable, firms with the best CSR ratings had also better financial performance.
However, this pattern was not consistent with operating earnings to asset variable and had reverse results. Therefore, to investigate this inconsistency authors added new variables to multiple regression model and find evidence that the asset age had played a significant role while measuring the link between CSP and CSR. When logit analyses also resulted in the same indicators as the first model did, Cochran and Wood (1984) proved that without adding asset age in a model, the regression results would be inconsistent, and the results of previous research also became suspicious because of misspecification problem. After inconsistency with regression results, authors added two additional variables; asset age and asset turnover in their model and find a strong correlation between asset age and a social performance rating of corporations.
McWilliams and Siegel (2000) developed Waddock and Graves (1997) econometric model and stated that that model was misspecified because omitting R&D and advertising expenses variables from the model. Like Cochran and Wood (1984) they also prove that R&D costs have a direct relationship with CSP ratings, therefore without controlling these variables the empirical model would give us biased result in terms of CSP. McWilliams and Siegel’s (2000) model data consisted of 524 corporations and time frame changed between 1991 to 1996. The regression results of them also proved that Waddock and Grave (1997) model was suffered from misspecification problem. Beside Waddock and Grave (1997), they regressed the second model where they adjusted R&D. So that, when the authors control variables, like R&D and advertising, the social performance variable of the firm became statistically insignificant and its effect on CFP became neutral.
After McWilliams and Siegel’s (2000) Elsayed and Paton (2005) also supported their view and stated that R&D and advertising expenses cause misspecification in the model. They further continue and argue advertising is a powerful tool which let their customers know how companies’ outputs meet social demands. Even though the authors asserted the consequence of controlling advertising expenses, they only adjust R&D costs on their regression. These results also consistent with Andersen and Dejoy (2011), which in their regression social performance of the firm become insignificant once R&D costs excluded model. Besides R&D, Andersen and Dejoy’s (2011) developed a model which indicated that financial risks and advertising expenses could be also a significant variable that causes the misspecification of a model. McGuire, Sundgren, and Schneeweis (1988) also mentioned the significance of controlling risk. They criticized the early analysis and asserted that the reason for researches who produce various results is that they did not control risk factors when they use market-based indicators to measure financial performance.
Hillman and Claim (2001) pointed out the importance of controlling corporate size while regressing CSP and CFP and expressed how it plays a significant role in the relationship with shareholders. Corporate size, research and development cost were also controlled by King and Lenox (2011) in their investigation. However, the authors also brought a new control variable which they believe it is important to adjust. They add the regulatory variable, which measures the how strict is environmental issues management is. Authors assert that these regulatory issues charge huge amount of cost for corporations. Because of it changes and not the same across locations it should be controlled in empirical research. Porter and Van de Linde (1995) regression results also provide confirmatory evidence that strict environmental regulation increases the competitiveness and the profitability of the firm in the long term. However, there is also inconclusive debate about whether environmental investment improves financial performance. Palmey and Porney (1995) research on this environmental issue does not support abovementioned authors support (especially Porter and Van de Linde (1995)) and develop the claim that strict regulation on environment decrease the financial gain of corporations.