Hypothesis 1, stating that a higher CFP predicts a higher CSP, was supported in the case of both net impact and all its four dimensions. Since no previous research exists on the CFP–CSP relationship in which CSP is measured using the net impact method, it is not possible to directly compare the models to previous research results.
However, the theory that posits that ROA is the most common predictor of CSP in CFP–CSP research is supported when other methods are used—including the use of net impact as a measurement method (Daniel et al., 2004; Busch & Friede; 2018;
Makni et al., 2008). Moreover, the theory that accounting-based CFP measures predict CSP best is also supported (Busch & Friede, 2018). In this research, ROA was the only accounting-based measure that qualified in the reduced models. Since ROA predicts CSP in all other dimensions of net impact except in societal impacts, then there must have been profitable growth earlier in the companies that allowed them to reach the respective financial positions measured with ROA.
Additionally, the theory that service companies have higher CSP than product companies is supported (Cheung et al., 2013; Huang & Yang, 2014). The fact that a company was a service company was the most common predictor variable of all the models. This supports also the result of Hypothesis 3, as noted later in this section.
The reason for the worst fits of the environmental impact models may be the fact that even if, from the kurtosis and skewness point of view, the environmental impact models showed normal distribution, the visual inspection of the histogram showed slight skewness to the left (Figure 9). Moreover, the P-P plot (Figure 10) shows some deviation from the straight line in the middle despite the fact that the ends follow the straight line.
The overall results do not indicate a difference in the CSP of growth companies when compared to the CSP of other companies in most of the models. Thus,
multiple regression results support the findings of Hypothesis 2, as noted later in this section. However, in the societal impact model, S_2016, the turnover growth is a significant predictor of societal impacts, as personnel growth is a significant predictor of societal impacts in model S_2017. Consequently, turnover growth is a predictor of future turnover, which is related to the future tax footprint—if the company is growing in a profitable manner. Furthermore, since absolute measures of the net impact were used, the higher tax footprint is related to the company turnover. This may explain turnover growth as a predictor of societal impacts.
However, in the environmental impact model, E_2016, personnel growth has a negative effect and turnover growth has a negative effect on environmental impacts in the E_2017 model. The same trend can be seen in the following knowledge models: K_2016, K_2017 and K_2018. The common fact that these results do not occur in the best fits of the environmental and knowledge impacts, makes explanation difficult. The phenomenon cannot be grounded on previous research.
It should be noted that company age is a predictor of net impact only—the younger a company is, the higher its net impact is in all net impact models. It makes sense that younger companies, which typically consist of younger entrepreneurs, would be more sensitive to responsible operation. Company age does not, however, seem to predict the four impact dimensions. On the other hand, Moore (2001) and Roberts (1992) have found that the older a company is, the higher its aggregate CSP. This is related to a company’s brand image, which is often stronger with established companies. This then elevates the stakeholder expectations regarding CSP. The contradiction may be explained by both the size and industry of the sample in this study—the average size of the sample is smaller than those in previous research and the companies are in the business-to-business market, whereas previous research examined the effect of age in larger consumer business.
To conclude the discussion on Hypothesis 1, the company being a service company and the ROA are the most common predictors of two to four independent variables that predict CSP measured using the net impact method in this study. Moreover, the different dimensions and net impact are not predicted by exactly the same independent variables. The statistically most significant models of the net impact and its four dimensions do not represent the CFP measurements of the same years. This is discussed in the next section when considering the findings for Hypothesis 4.
Hypothesis 2, claiming that high-growth companies have higher CSP than other companies, was not supported. First, in Section 2.2.1, it was concluded that in order for company growth and CSP to work together as a combined strategy, a company has to be proactive towards CSP, which means that the company must aim at or already be on the third or fourth level of CSP maturity (Baumgartner & Ebner, 2010).
This means that companies are either already working on their handprint or, in the best case, are already basing their strategy on a combined CSP and competitive growth strategy. It seems, however, that none of the companies in the dataset were at level 4 and that less than 20% of the companies were at level 3, with at least some proactive CSR activity. In addition, only a few of the companies in the dataset seemed to consider the sustainability issue to be a business opportunity and source of new products and services leading to growth (Halme & Korpela, 2014).
Another explanation for the rejection of Hypothesis 2, besides the low CSP maturity, may be related to the fact that the net impact method is new. The first time the net impact of the sample had been measured was in the year 2018 and companies do not even know what their net impact is nor its four dimensions. Hence, the feedback loop, which is needed to start optimising the net impact, is missing (Witek-Hajduk
& Saborek, 2016).
The third reason for the results may be that the sample consisted of small- and medium-sized companies, which are less under the scrutiny of stakeholders (Clarkson, 1995; Baron 2009; Chiu & Sharfman, 2011). Only 4 of the 37 test group companies were public companies that have been reporting their ESG activities prior to measuring the net impact.
Fourth, as Chen et al. (2015) found in their CFP–CSP study, only the categories of diversity and equal opportunity in the GRI labour practices dimension (content analysis) had a positive and significant correlation with sales growth and cash flow/sales. When regressing all other dimensions and categories, the relationship was insignificant. This research is, however, using the net impact for the first time and the results cannot thus be grounded on previous results. On the other hand, Busch and Friede (2018) concluded in their meta-study that the CFP–CSP relationship does not depend on the CSP measurement method used.
Fifth, according to previous research, growth-based CFP measures do not seem to predict CSP as well as accounting-based CFP measures (Busch & Friede, 2018).
Unfortunately, sales growth as a growth-based CFP measure in CFP–CSP research
has seldom been investigated. It seems natural, however, that company growth alone, as a CFP measure, does not explain CSP—hence, ROA as well as other CFP measures are needed. Moreover, the CFP–CSP relationship is complex and not a direct result (Xie & Wang, 2017), which supports that company growth alone does not necessarily explain CSP.
Hypothesis 3, stating that service companies have higher CSP than product companies, was supported in the case of net impact and three out of its four categories. On the one hand, this supports the theory about service companies having higher CSP than product companies (Cheung et al., 2013; Huang & Yang, 2014). On the other hand, previous research uses aggregate CSP measures when comparing the CSP of product and service companies. Additionally, the hypothesis was rejected for environmental impacts. Since there is no previous research in which CSP was divided into categories and the categories investigated separately between service and product companies, it is not possible to ground the results of the environmental dimension on the results of previous research. Furthermore, servitisation is good for business because it improves both profitability and customer intimacy (Kowalkowski et al., 2017).
The criterium for dividing product companies from service companies in this study is based on how the turnover of a company is divided between its service and product components—if either one exceeds 50% of the turnover, then the company was categorised as belonging to the respective group (Finnish Technology Industries, 2018). This criterium seems somewhat vague, especially if both components represent close to 50% of the turnover. When looking at the industries in the dataset, the extreme of a service company is, for example, an engineering consulting company that designs machinery. A corresponding extreme of a product company could be a manufacturing company that produces physical machinery. Very often, however, the same product company offers after sales services or alternative/additional support to another company, which is concentrating in providing services. In the case of software companies, the division between
“product” and “service” companies may be more difficult to make. Should a software company that produces software as a service, like other software companies, be counted as a product company or a service company? In this research study, such companies were counted as service companies, whereas companies producing licenced software were counted as product companies.
Hypothesis 4 relates to temporal lags. Based on previous research, these temporal lags can be interpreted as the lag between the time period between the year of the independent variables producing the most statistically significant multi-regression model and the year 2018 when the dependent variables were measured (Busch &
Friede, 2018). The multi-regression models of the four different dimensions, as well as the net impact, differ from one another in terms of the independent variables explaining them. Moreover, as discussed below, it seems logical that the temporal lags of the four dimensions and their different categories also vary. Nevertheless, the most common temporal lag found in the CSP literature is one year (McGuire et al., 2003; Waddock & Graves, 1997; Fischer & Sawczyn, 2013). However, according to Moore (2001) the reason for a longer time lag than one year could be to avoid distortions caused by rogue figures. Additionally, there is no previous research on temporal lags of different dimensions of net impact nor its categories.
Hence, the temporals lags (Hypothesis 4) for each dimension and examples of the interrelation of the different dimensions and net impact could perhaps be considered as follows:
Environmental impacts. The financial numbers of 2018 provided the best fit for the multi-regression models. This might mean that an investment deployed for improving environmental impacts can be seen in environmental impacts within a year of the change made. For example, if a company made and implemented an investment for reducing GHG gases, non-GHG gases, waste or water usage, then the impact of this change can be seen quite quickly after measuring the environmental impacts again after a year. It can be argued that if the improved impact is visible in one year, then this impact will increase in the subsequent years because the optimisation process of emission reduction will continue beyond the first year. Hence, Hypothesis 4 can be accepted for environmental impacts. As an example of the interrelation between the dimensions, it can also be argued that reducing emissions will have a positive impact on reducing lung defects, which represents the “diseases” category in the health impacts dimension.
Health impacts. The financial numbers of 2015 provided the best fit for the multi-regression models. We do not know, however, anything about what kind of a fit 2014 or earlier financial numbers would give. Using the same logic as above, one could argue that it takes at least four years to see the impact of health-related investments of a company in the respective health impact measurement change. It
can also be argued that the health impact would improve after four years because not all the people get the disease or are cured from the disease at the same time.
Moreover, a health-related innovation will help reduce future diseases of the same sort. Hence, Hypothesis 4 can be accepted for health impacts. Additionally, as an example of the interrelation between the dimensions, a health-related innovation would help reduce sick-leave costs paid to the employees by the employers as well as the healthcare costs of governments. These reduced costs contribute positively to societal impacts in terms of receiving more tax income and reducing government costs. Hence, governments could use the additional resources for other citizen benefits.
Societal impacts. The model of 2016 represents the best fit and the model of 2017 was very close to the best fit. Again, one could argue that it takes one to two years to see the impacts of the changes made in terms of societal investments of a company. The fiscal year of all the companies in the sample is the calendar year. This means that the impact of an investment that, for example, increases the revenue and taxes or the number of personnel can be seen one to two years after the change is implemented. It can be argued that the societal impact will increase after 1 to 2 years because of the similar inertia demonstrated earlier. Hence, Hypothesis 4 can be accepted for societal impacts. As an example of the interrelation between the dimensions, the improved tax income of governments will hopefully increase government investments in either the operational environment of companies or in whatever might make the lives of citizens better.
Knowledge impacts. The model of 2015 provided the best fit by far for the multiple regression models. Consequently, using the logic above, it would take at least four years to see the impact of knowledge creation investments in a company. This makes sense because the R&D activities that produce new knowledge for a company are most often long-term investments. Hence Hypothesis 4 can be accepted for knowledge impacts. Moreover, as an example of the interrelation between the dimensions, it can be argued that if an innovation is made in basic research, it will often generate new innovations in applied research at different time delays—the commercial success of which will also come at different times. This will have a positive impact in the societal dimension.
Net impact. The net impact represents the sum of all the impacts and the best fit in this study was given by the 2015 multiple regression model. The models of the two
out of four dimensions—namely, health and knowledge impacts—also represented the best fits with the 2015 models. Again, we do not know anything about what kinds of models would be generated using numbers from 2014 and earlier. Hence, the temporal lag of the NS_2015 model seems to fit into the picture painted by the models of its four dimensions. Net impact, besides being the sum of the four dimensions, is a complicated combination and interaction of the 4 impact dimensions and the 19 impact categories therein. Consequently, Hypothesis 4—
stating that the longer the temporal lag is, the higher the CSP will be—can be supported. In conclusion, it can be said that when a company wants to optimise its net impact, the improvements it makes in the environmental and societal dimensions produce the most rapid positive changes in the net impact. However, the interrelations between the dimensions—and even between their categories—need to be considered.
To sum up the discussion on key findings of this study, growth alone does not increase corporate social responsibility, but it is needed to increase the profitability.
Profitability, in turn, is a requirement for responsibility investments. Servitization may, however, increase both growth and CSP. To increase the CSP even more, one has to take into consideration the fact that environmental and societal dimensions may react faster to the responsibility investments than health- and knowledge-dimensions. The cross impacts of different dimensions need also to be considered.