Regression analysis

In this paragraph the regression analysis will be performed and the two hypothesis will be tested. This thesis considers two hypothesis whose regressions look as follows:

HI: EFF_FIN = β0 + β1D_MANUFACTURING + β2D_FSERVICES +

β3D_NFSERVICES+ β4D_PUBLICSEC + β5FIRM_SIZE + β6DEG_DIG + β7ANN_GROW + β8DIG_FIN

HII: EFF_FIN = β0 + Β1D_MANUFACTURING + β2D_FSERVICES +

β3D_NFSERVICES + β4D_PUBLICSEC + β5FIRM_SIZE + β6DEG_DIG + β7ANN_GROW + β9INV_SKILL

HIII: EFF_FIN = β0 + Β1D_MANUFACTURING + β2D_FSERVICES +

β3D_NFSERVICES + β4D_PUBLICSEC + β5FIRM_SIZE + β6DEG_DIG + β7ANN_GROW + β8DIG_FIN + β9INV_SKILL + β10*DIG_FIN*INV_SKILL

Multicollinearity was not an issue in the regressions. Multicollinearity means that there is a strong linear connection between the explanatory variables. The result is that the explanatory variables are associated with each other, leaving the regression model with no extra variance to explain. This could lead to poorer estimates of the regression coefficients in the regression model. High multicollinearity exists when the VIF-value of the collinearity statistics is higher than 10, but it could already lead to problems with a value of 5 (Hair et al., 2009). The results of the VIF values, which can be found in Appendix E, show that multicollinearity is not an issue.

Furthermore, one dummy variable will be left out for the control variable sector, otherwise, multicollinearity would emerge. The dummy variable that is left out serves as a reference group. In this thesis, D_NFSERVICES will be left out in the regressions.

4.3.1 Hypothesis I

The results from the regression analysis for the first hypothesis are shown in table 8. The first hypothesis states that digitalization in Finance is positively related to the effectiveness of the Finance function. There is a R² of 0,44. This means that 44% of the variance of the effectiveness of the Finance function is explained by digitalization in Finance and the control variables.

Regression Hypothesis I

EFF_FIN Unstand. B Std. Err. Coeff. Beta t-value p-value Sig

CONSTANT 2,873 0,238 11,905 0,000 ^**

DIG_FIN 0,431 0,060 0,510 7,161 0,000 ^**

D_MANUFACTURING 0,037 0,101 0,024 0,366 0,715

D_FSERVICES 0,121 0,142 0,054 0,849 0,397

D_PUBLICSEC -0,344 0,133 -0,169 -2,594 0,010 ^**

FIRM_SIZE 0,001 0,018 0,002 0,034 0,973

DEG_DIG 0,079 0,039 0,145 2,041 0,043 ^*

ANN_GROW -0,018 0,040 -0,026 -0,451 0,652

** p<0,01, * p<0,05

Adjusted R^{2 0,42} N 182

R² 0,44 F-test 19,54 ^**

Table 8 Linear regression hypothesis I. The dependent variable is the effectiveness of the Finance function (EFF_FIN), and the independent variable is digitalization in Finance (DIG_FIN). The control variables are also included.

The results in table 8 show that there is a positive association between digitalization in Finance and the effectiveness of the Finance function. The adoption of digital technologies in Finance has a positive effect on the effectiveness of the Finance function with a coefficient of 0,431 (p-value <0,01). The control variable public sector shows a significant negative effect with a coefficient of -0,344 (p-value <0,01). This means that the effectiveness of the Finance function is lower in companies in the public sector which is consistent with the findings in table 7. In this thesis the public sector is limited to companies from the following industries: public government; defense organizations and social security; education, human health, and social work; and, environment, culture, recreation, and other service activities. Another control variable, the degree of digitalization of the organization, shows a positive and significant effect with a coefficient of 0,079 (p-value <0,05). This means that the degree of digitalization of organizations has a positive influence on the effectiveness of the Finance function. The results show that hypothesis I is confirmed: digitalization in Finance is positively related to the effectiveness of the Finance function.

4.3.2 Hypothesis II

Table 9 shows the results for the second hypothesis. The second hypothesis states that investments in the technical skills of Finance professionals lead to a more effective Finance function. The R² of 0,67 means that 67% of the variance of the effectiveness of the Finance function is explained by the investment in technical skills of the Finance professionals and the control variables.

Regression Hypothesis II

EFF_FIN Unstand. B Std. Err. Coeff. Beta t-value p-value Sig

CONSTANT 2,915 0,239 12,217 0,000 ^**

INV_SKILL 0,457 0,061 0,533 7,524 0,000 ^**

D_MANUFACTURING 0,010 0,100 0,007 0,919 0,601

D_FSERVICES 0,073 0,140 0,033 0,524 0,601

D_PUBLICSEC -0,377 0,131 -0,185 -2,877 0,005 ^**

FIRM_SIZE 0,010 0,018 0,033 0,579 0,564

DEG_DIG 0,063 0,039 0,115 1,607 0,110

ANN_GROW -0,025 0,040 -0,036 -0,628 0,531

** p<0,01, * p<0,05

Adjusted R^{2 0,43} N 182

R² 0,67 F-test 20,59 ^**

Table 9 Linear regression hypothesis II. The dependent variable is the effectiveness of the Finance function (EFF_FIN), and the independent variable is the investment in technical skills of Finance professionals (INV_SKILL). The control variables are also included.

The results show that there is a positive association between the investment in technical skills of Finance professionals and the effectiveness of the Finance function. This means that investing in the skills of employees will lead to a more effective Finance function (coef: 0,457;

p-value <0,01). Furthermore, the control variable public sector shows a significant negative effect again (coef; -0.377; p-value<0,01), which means that also in this regression the effectiveness of the Finance function is lower in the public sector. The results show that hypothesis II is confirmed.

4.3.3 Hypothesis III

The results for the third hypothesis are shown in table 10. This hypothesis states that digitalization in Finance, together with investing in the technical skills of Finance professionals will lead to higher effectiveness of the Finance function. The explanatory variables, DIG_FIN and INV_SKILL, have been standardized before estimating the moderation effect to avoid strong correlations of the moderation with one of the original variables. The results show an adjusted R² of 0,48, which means that 48% of the variance of the effectiveness of the Finance function is explained by the regression model. The adjusted R² of 48% is higher relative to table 8, but lower than table 9 which means that the regression model of hypothesis III explains more of the variance of the effectiveness of the Finance function compared to the regression model of hypothesis I but less compared to the regression model of hypothesis II.

Table 10 Linear regression hypothesis III. The dependent variable is the effectiveness of the Finance function (EFF_FIN), and the independent variables are digitalization in Finance (DIG_FIN) and the investment in technical skills of Finance professionals (INV_SKILL). The control variables are also included.

There is a positive and significant relationship between digitalization in Finance and the effectiveness of the Finance function in the first regression in table 10 (coef: 0,279; p-value

<0,01). Furthermore, investing in the technical skills of Finance professionals has a positive effect, with a coefficient of 0,318 (p <0,01), on the effectiveness of the Finance function as well. The two variables show an individual direct positive and significant effect on the dependent variable which is in line with hypotheses I and II.

The interaction of the two variables does lead to a positive effect on the effectiveness of the Finance function in the second regression in table 10 (coef: 0,026); however, this interaction is not significant (p-value of 0,645). This means that the expectation of hypothesis III is not confirmed. Digitalization in Finance, together with investing in the technical skills of Finance professionals will not lead to higher effectiveness of the Finance function. However, the high correlation of investing in digital technologies, as well as the high correlation of investing in the technical skills of the employees independently, does show that the variables are associated. Besides, due to the positive and significant effects of the variables independently, it seems that companies invest in both, technical skills and digital technologies, but separately and they contribute separately to a more effective Finance function.

Furthermore, the control variable public sector does show a significant negative effect (coef: -Regression hypothesis III

(1) (2)

CONSTANT 3,191 ^** 3,170 ^**

DIG_FIN 0,279 ^** 0,268 ^**

INV_SKILL 0,318 ^** 0,324 ^**

INT_DIGFIN_INVSKILL 0,026

D_MANUFACTURING 0,020 0,020

D_FSERVICES 0,112 0,109

D_PUBLICSEC -0,353 ^** -0,359 ^**

FIRM_SIZE -0,002 -0,002

DEG_DIG 0,015 0,018

ANN_GROW -0,027 -0,029

** p<0,01, * p<0,05

R^{2 0,51 0,51}

Adjusted R² 0,48 0,48

F-test 22,12 ^** 19,60 ^**

N 182 182

EFF_FIN

0,359; p-value <0,01). This means that also in this regression model the effectiveness of the Finance function is lower in companies from the public sector.