Regression and Moderation Analyses

In this part, we analyze the hypothesized relations as structured in the

conceptual model by running multiple regression analyses with SPSS and PROCESS.

The multiple regression analyses present R² and the adjusted R², which indicate the variation in the dependent variable affected by the independent variables (Field, 2013). The analyses also provide the standardized Beta values (β-value), which describe the relations between the dependent variable and predictors. Besides, we consider the β-value significant when the P-value is lower than 0.1, which means the chance of a relationship between the variables is higher than 90% (Field, 2013).

Lastly, we include F-value to compare the model with and without predictors to check the fit of our models (Field, 2013).

5.3.1 Direct Relationship: CPA – ROA

Table 5 presents the regression analyses for the direct relationship between CPA and CFP, measured by ROA. First, we only include the dependent variable ROA and three control variables in Model 1. After that, we add the independent variables -- Aggregated CPA, Lobbying expenditure, and PAC contributions -- in Model 2,3 and 4, respectively.

The first model tells us that the control variables can explain 8.1% of the variance in the dependent variable ROA significantly with the R² of 0.081. There is a significant negative relationship between Firm Size and ROA (β = -0.008, P < 0.05), that is to say, an increase in firm size leads to a decrease of ROA. On the contrary, R&D intensity shows a significant positive effect on the ROA (β = 0.617, P < 0.01).

However, this model displays no indication of a significant relationship between industry regulation and ROA.

In Model 2, we include an independent variable, Aggregated CPA. The result suggests an increased adjusted R² of 0.106 and a significant positive CPA-CFP relationship (β = 0.019, P < 0.01), which means aggregated CPA improves the model fit. Besides, the more investment in aggregated CPA, the better performance in ROA representing CFP. We have hypothesized the positive impact of aggregated CPA on CFP in previous chapters. Hence, in the ROA perspective, we can accept our first hypothesis.

Model 3 and Model 4 illustrate significant positive relationships between ROA and both CPA strategies, namely Lobbying expenditure (β = 0.009, P < 0.01) and PAC contributions (β = 0.016, P < 0.01). The R² value of these two models is 0.095 and 0.114, which indicates the models consider 9.5% and 11.4% of the variance as significant, while adjusted R² in each model shows an improvement of fit comparing to Model 1. Therefore, in terms of ROA, we can accept the hypotheses H1b and H1c, which assert a positive relationship between informational CPA strategy (lobbying) and CFP and financial CPA strategy (PAC contributions) and CFP.

5.3.2 Direct Relationship: CPA – ROI

In Model 5-8, we keep the same process as last section but switch the dependent variable to ROI. Model 5 includes only control variables that R&D intensity has significant positive impact on CFP (β = 1.063, P < 0.01), whereas there

is no significant relation between firm size and CFP (β = -0.006, P > 0.1), neither between industry regulation and CFP (β = 0.004, P > 0.1).

In Model 6, we include aggregated CPA in the regression analysis as an independent variable. This model suggests 11.4% variance of the dependent variable at 0.01 significant level (R² = 0.114, P < 0.01). The adjusted R² of 0.104 presents the aggregated CPA as a new predictor that improves the model fit. The impact of aggregated CPA on ROI is a significant positive result (β = 0.043, P < 0.01), which indicates that firms spend more on CPA, achieve higher ROI. Combining the result of Model 1, We can conclude a significant positive relationship between aggregated CPA and CFP that our hypothesis 1a is accepted.

In Model 7 and Model 8, we replace the independent variable with an individual CPA strategy. Model 7 shows a significant positive effect of lobbying expenditure on ROI (β = 0.022, P < 0.01) and Model 8 reveals the similar result on the PAC contributions-ROI relation (β = 0.151, P < 0.01). It signifies that the

informational CPA strategy and financial CPA strategy positively correlate with CFP when measuring ROI. Thus, judging from the ROA regression analyses in the last chapter, we can accept our hypotheses H1b and H1c that each CPA strategy positively affects CFP.

5.3.3 Moderation Effect of CSR on CPA – ROA

Table 7 displays the moderation effect of both internal and external CSR on the relationship between each CPA strategy (aggregated CPA, lobbying expenditure, and PAC contributions) and CFP measured by ROA via PROCESS tool in SPSS.

In Model 9, we take aggregated CPA as the independent variable, and the R² indicates a significant model that the variables explain 16.3% of the variance in the dependent variable ROA (R² = 0.163, P < 0.01). However, the output shows that the interaction effect of both external CSR (β = 0.011, P > 0.1) and internal CSR (β = -0.029, P > 0.1) on ROA is not statistically significant. The results reject our

hypothesis H2a in the dependent variable measurement of ROA that neither internal CSR nor external CSR has a moderation effect on aggregated CPA and CFP, let alone a comparison between the moderators.

Model 10 focuses on CSR’s moderation effect on informational CPA strategy, namely lobbying expenditure, which is the independent variable in this model. The R² reflects 14.5% explanatory power at 0.01 significant level of the variance on ROA

this time. The internal CSR’s interaction effects with the direct relation between lobbying expenditure and ROA show a moderate significant impact (β = 0.212, P <

0.1), but there is still no significant effect of external CSR on ROA (β = 0.009, P >

0.1). In the ROA aspect, this model rejects hypothesis H2b, which describes the external CSR has more of a moderation effect on the relation between information CPA and CFP than internal CSR.

Model 11 changes the independent variable to PAC contributions, the financial strategy, with a higher variance percentage of 16% at a significant level to explain the dependent variable ROA (R² = 0.160, P < 0.01). The results of the moderating

interaction effect are still not significant regarding the moderator, external and

internal CSR (β = -0.012, P > 0.1; β = -0.020, P > 0.1). Our hypothesis H2c states that external CSR moderates the nexus between financial CPA strategy and CFP more strongly than internal CSR. Based on Model 11, we have to reject this hypothesis.

5.3.4 Moderation Effect of CSR on CPA – ROI

Model 12, 13, and 14 in Table 8 align with the same structure as Table 7 in the last section, but we alternate with ROI as the dependent variable. Although the R² value is relatively lower than in the previous moderation effect regression analyses, these models are all significant, with a 99% confidence interval.

Model 12 tests the internal and external CSR moderating the relationship between aggregated CPA and ROI. The result suggests no significant effect by interacting with two moderators, external CSR (β = -0.036, P > 0.1) and internal CSR (β = -0.043, P > 0.1). Considering the dependent variable ROA result, we reject our hypothesis H1a and summarize that the internal and external CSR has no moderation effect on the relationship between aggregated CPA and CFP. At the same time, it is not necessary to compare the moderation effect.

In Model 13 and 14, we run the regression analyses with each individual CPA strategy separately. The outcomes are similar with what we have obtained in Model 10 and 11. The moderating interaction effects of external and internal CSR on

lobbying expenditure (β = 0.014, P > 0.1; β = -0.048, P > 0.1) and PAC contributions (β = -0.015, P > 0.1; β = 0.011, P > 0.1) are insignificant. Therefore, we have to reject our hypotheses H2b and H2c.

5.3.5 Moderation Effect of CSR in different Industry Type on CPA – ROA Table 9 displays the results from Model 15, 16, and 17. We test whether the aggregated CSR in different industry type (B2B vs. B2C) moderates the relationship between CPA and CFP, in specific ROA measurement, by running a regression with PROCESS template model 3. In Model 15, we examine the aggregated CSR's moderation effect in different industry types on the aggregated CPA and CFP. The outcome shows that the model is significant at 20.4.1% of variance to explain the dependent variable. Besides, the aggregated CPA has a positive effect on ROA but insignificant (β = 0.008, P > 0.1), whereas the interactive effect of aggregated CSR and industry type is significant but negative (β = -0.140, P < 0.01). It means that the dummy variable with a high value, B2C, has a significant negative effect when interacting with aggregated CSR on the relation between aggregated CPA and ROA.

Since we hypothesize the CSR in the B2C sector will impact CPA and CFP more positively, we have to reject hypothesis H2d in the CFP measurement of ROA.

Moreover, Model 16 shows the moderation of the aggregated CSR in different industry types on the relation between CFP and individual CPA information strategy by measuring ROA and lobbying expenditure. The model gives an insignificant positive relationship between lobbying expenditure and ROA (β = 0.003, P > 0.1), but significant negative interaction effect of aggregated CSR and industry type on this direct relationship (β = -0.044, P < 0.1). We have similar results in the other individual CPA strategy, financial incentives, in Model 17. The direct relationship between PAC contributions and ROA is insignificant and positive (β = 0.009, P >

0.1), while the aggregated CSR and industry type’s interaction effect is significant negative (β = -0.087, P < 0.05). The result shows that the CSR in the B2C industry has a significant negative effect on CPA and CFP relationship, but in the B2B industry, there is no significant moderating influence. Therefore, we reject the hypotheses H2e and H2f in the ROA perspective since we hypothesized that firms operating the CSR strategy in the B2C industry have a more positive moderation effect on the relationship between each CPA strategy than the firms in the B2B industry.

5.3.6 Moderation Effect of CSR in different Industry Type on CPA – ROI

In this section, we test the moderation effect of the aggregated CSR in different industry types on another dependent variable ROI with three CPA independent

variables in Model 18, 19, and 20, summarized in Table 10. Model 18 indicates a significant model that the control variables and independent variable aggregated CPA explain 19.8% of the model (R²= 0.198, P < 0.01). There is a mild significant positive effect between aggregated CPA and ROI (β = 0.026, P < 0.1). The interaction

moderation effect of aggregated CSR and the dichotomous moderator variable, industry type, on the CPA-CFP relationship is also significant (β = -0.304, P < 0.01).

In Model 15, we have a similar ROA outcome that CSR in B2C reveals a significant negative moderating impact on aggregated CPA and CFP, but CSR in B2C gives no moderation. Considering that, we reject our hypothesis H2d.

Subsequently, we test the aggregated CSR's moderation effect in different industry types on each CPA strategy in Model 19 and 20. These two models are significant: 14.1% and 16.5% of the variance can explain the dependent variable separately. Information strategy (Lobbying expenditure) has a significant positive effect on ROI (β = 0.011, P < 0.05), but PAC contributions, the financial strategy, presents an insignificant positive direct relation with ROI (β = 0.018, P > 0.1).

However, the interaction effect of aggregated CSR and the dummy variable industry type shows a significant negative result, both on lobbying expenditure and ROI (β = -0.098, P < 0.1) and on PAC contributions and ROI (β = -0.232, P < 0.01). The results are consistent with what we have obtained on ROA that CSR in B2C offers significant negative moderation, but CSR in B2B does not. As the results for both ROA and ROI contradict our theoretical assumptions, we reject Hypotheses H2e and H2f.