Hypothesis 1

According to table 8 there is a correlation of 0.073 between the EnvironmentalPracticesL and FirmGrowthOT and the correlation between EnvironmentalPracticesO and FirmGrowthOT is –0.0929. The independent variables EnvironmentalPracticesL and EnvironmentalPracticesO were formed in a factor analysis, which was explained in section 3.4.2. The results of the first regressions can be seen in Appendix 6 (Tables 18 and 19).

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 6 FirmGrowthOT 0.073 -0.0929 -0.2116 -0.0729 -0.3039** 1 7 FirmGrowthET 0.073 0.0916 -0.0419 -0.0461 -0.1361 0.8042* 1 8 FirmGrowthNI -0.045 0.0485 0.0046 -0.1054 -0.0106 0.6166* 0.9150* 1 9 PLforperiodNetIncometh 0.1144 0.1062 -0.021 0.112 -0.027 0.4724* 0.7845* 0.8240* 1 10 Profitmargin2018_n -0.0436 0.0312 0.0991 -0.0392 0.0584 0.3728* 0.7834* 0.9094* 0.8010* 1 11 ROAusingNetincome2018 -0.1068 -0.0274 0.0329 -0.0807 0.0659 0.3332* 0.7123* 0.8907* 0.7667* 0.9320* 1 12 EBITDAMargin2018 0.0694 0.183 0.1316 0.0122 0.0638 0.4192* 0.8315* 0.8451* 0.7639* 0.9176* 0.8034* 1 13 EBITDAthEUR2018 0.3226** 0.1772 -0.0343 0.2128 -0.077 0.5312* 0.7260* 0.6217* 0.8648* 0.5942* 0.5029* 0.6582* 1 14 CA_Product -0.0183 0.209 0.3025** 0.0197 0.2931** -0.0559 -0.0176 0.052 -0.0625 0.0369 -0.0166 0.0679 -0.0517 1 15 CA_Quality -0.0076 0.3711* 0.3982* 0.1981 0.3689* -0.1855 -0.1429 -0.0887 -0.1604 -0.0673 -0.098 -0.0604 -0.1166 0.6793* 1 16 CA_Distribution 0.1915 0.4124* 0.4104* 0.3169** 0.2103 -0.1663 -0.0434 -0.0441 0.0336 -0.08 -0.1191 0.0448 0.0625 0.5838* 0.4600* 1 17 CA_Promotion -0.0478 0.2206 0.2880** 0.2188 0.2507 -0.143 -0.0115 0.036 0.0485 0.0432 0.0253 0.1168 0.0645 0.6210* 0.5796* 0.6865* 1 18 CA_Process 0.0882 0.1282 0.1682 0.0197 0.0073 0.215 0.106 -0.0289 -0.0207 -0.0886 -0.142 0.0398 0.1544 0.5225* 0.5051* 0.3168** 0.4110* 1 19 Intper_S -0.1069 0.0151 0.0336 0.0547 -0.0034 0.0113 0.0285 0.0272 0.1375 0.0562 -0.0321 0.0619 0.0672 0.2445 -0.0066 0.0102 0.0971 0.1323 1 20 FirmAge 0.3597* 0.1585 -0.1172 -0.0087 0.0298 0.0562 0.2101 0.2233 0.3800* 0.2488 0.1791 0.2743** 0.4125* -0.0638 -0.1122 0.18 0.0363 -0.1232 -0.1271 1 21 FirmSize 0.4280* 0.2085 0.0526 0.4013* 0.1624 -0.3107** -0.2673** -0.2723** 0.1587 -0.2181 -0.2141 -0.235 0.3616* -0.0546 0.0382 0.2461 0.101 -0.028 0.0259 0.3841* 1

49 Model 1a

According to Ramsay’s RESET test the model 1a has no omitted variables with the p-value 0.971 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[5] = 5.48) with the p-value 0.361 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.927) in the model, as the p-value (0.336) is greater than 0.05. The VIF value of 1.08 is acceptable. The tolerance values indicate that 92.5% of variation in firm age and 92.5% of variation in EnvironmentalPracticesL is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

The results of the multiple regression analysis of model 1a are presented in Appendix 6. The total amount of observations is 73. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The model with 73 observations with three independent variables is acceptable and thus results can be generalizable. Model 1a is not statistically significant according to the Prob > F = 0.305 at the significance level of 0.05. The R2 of 0.033 indicates that the independent variables, EnvironmentalPracticesL and FirmAge can be used to explain about 3.3 % of the variation in the dependent variable, FirmGrowthOT. The coefficients indicate that one unit increase in EnvironmentalPracticesL would results in an increase of 4.223 in the FirmGrowthOT. A one unit increase in Firm age would result in a decrease of 1.862 in the FirmGrowthOT.

According to table 8 there is a correlation of -0.045 between the EnvironmentalPracticesL and FirmGrowthNI and the correlation between EnvironmentalPracticesO and FirmGrowthNI is -0.0485. The results of the regressions can be seen in Appendix 6 (Tables 18 and 19).

50 Model 1a2

According to Ramsay’s RESET test the model 1a2 has no omitted variables with the p-value 0.621 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[5] = 3.8) with the p-value 0.578 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.053) in the model, as the p-value (0.817) is greater than 0.05. The VIF value of 1.07 is acceptable. The tolerance values indicate that 93.2% of variation in firm age and 93.2% of variation in EnvironmentalPracticesL is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

The results of the multiple regression analysis of model 1a2 are presented in Appendix 6. The total amount of observations is 75. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The model with 75 observations with three independent variables is acceptable and thus results can be generalizable. Model 1a2 is not statistically significant according to the Prob > F = 0.716 at the significance level of 0.05. The R2 of 0.009 indicates that the independent variables, EnvironmentalPracticesL and FirmAge can be used to explain about 0.9 % of the variation in the dependent variable, FirmGrowthNI. The coefficients indicate that one unit increase in EnvironmentalPracticesL would results in an increase of 1.055 in the FirmGrowthNI. A one unit increase in Firm age would result in an increase of 0.061 in the FirmGrowthNI.

Model 1a3

According to Ramsay’s RESET test the model 1a3 has no omitted variables with the p-value 0.700 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows

51

that the model is homoscedastic (chi2[5] = 2.16) with the p-value 0.827 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.001) in the model, as the p-value (0.973) is greater than 0.05. The VIF value of 1.06 is acceptable. The tolerance values indicate that 94.0% of variation in firm age and 94.0% of variation in EnvironmentalPracticesL is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

The results of the multiple regression analysis of model 1a3 are presented in Appendix 6. The total amount of observations is 73. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The model with 73 observations with three independent variables is acceptable and thus results can be generalizable. Model 1a3 is not statistically significant according to the Prob > F = 0.554 at the significance level of 0.05. The R2 of 0.017 indicates that the independent variables, EnvironmentalPracticesL and FirmAge can be used to explain about 1.7 % of the variation in the dependent variable, FirmGrowthET. The coefficients indicate that one unit increase in EnvironmentalPracticesL would results in an increase of 2.603 in the FirmGrowthET. A one unit increase in Firm age would result in a decrease of 0.019 in the FirmGrowthET.

Model 1b

According to Ramsay’s RESET test the model has no omitted variables with the p-value 0.836 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[5] = 6.72) with the p-value 0.242 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 1.408) in the model, as the p-value (0.235) is greater than 0.05. The

52

VIF value of 1.02 is acceptable. The tolerance values indicate that 97.9% of variation in firm age and 97.9% of variation in EnvironmentalPracticesO is dependent from the other variables.

It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

The results of the multiple regression analysis of model 1b are presented in Appendix 6. The total amount of observations is 73. The model with 73 observations with three independent variables is acceptable and thus results can be generalizable. Model 1b is statistically not significant according to the Prob > F = 0.249 at the significance level of 0.05. The R2 of 0.039 indicates that the independent variables, EnvironmentalPracticesO and FirmAge can be used to explain about 3.9 % of the variation in the dependent variable, FirmGrowthOT. The coefficients indicate that one unit increase in EnvironmentalPracticesO would results in a decrease of 15.863 in the FirmGrowthOT. A one unit increase in Firm age would result in a decrease of 1.642 in the FirmGrowthOT.

Model 1b2

According to Ramsay’s RESET test the model 1b2 has no omitted variables with the p-value 0.306 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[5] = 3.61) with the p-value 0.607 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.082) in the model, as the p-value (0.775) is greater than 0.05. The VIF value of 1.02 is acceptable. The tolerance values indicate that 98.3% of variation in firm age and 98.3% of variation in EnvironmentalPracticesO is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

53

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

The results of the multiple regression analysis of model 1b2 are presented in Appendix 6. The total amount of observations is 75. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The model with 75 observations with three independent variables is acceptable and thus results can be generalizable. Model 1b2 is not statistically significant according to the Prob > F = 0.823 at the significance level of 0.05. The R2 of 0.005 indicates that the independent variables, EnvironmentalPracticesO and FirmAge can be used to explain about 0.5 % of the variation in the dependent variable, FirmGrowthNI. The coefficients indicate that one unit increase in EnvironmentalPracticesO would results in an increase of 0.754 in the FirmGrowthNI. A one unit increase in Firm age would result in an increase of 0.080 in the FirmGrowthNI.

Model 1b3

According to Ramsay’s RESET test the model 1b3 has no omitted variables with the p-value 0.477 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[5] = 1.91) with the p-value 0.862 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.027) in the model, as the p-value (0.870) is greater than 0.05. The VIF value of 1.01 is acceptable. The tolerance values indicate that 98.7% of variation in firm age and 98.7% of variation in EnvironmentalPracticesO is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram

54

and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

The results of the multiple regression analysis of model 1b3 are presented in Appendix 6. The total amount of observations is 73. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The model with 73 observations with three independent variables is acceptable and thus results can be generalizable. Model 1b3 is not statistically significant according to the Prob > F = 0.726 at the significance level of 0.05. The R2 of 0.009 indicates that the independent variables, EnvironmentalPracticesO and FirmAge can be used to explain about 0.9% of the variation in the dependent variable, FirmGrowthET. The coefficients indicate that one unit increase in EnvironmentalPracticesO would results in an increase of 3.038 in the FirmGrowthET. A one unit increase in Firm age would result in a decrease of 0.018 in the FirmGrowthET.

Model 1c

According to Ramsay’s RESET test the model has no omitted variables with the p-value 0.200 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[9] = 4.1) with the p-value 0.905 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.051) in the model, as the p-value (0.821) is greater than 0.05. The VIF value of 1.07 is acceptable. The tolerance values indicate that 91.0% of variation in firm age, 91.8% of variation in firm size, and 96.8% of variation in EnvironmentalPracticesL is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

55

The results of the multiple regression analysis of model 1c are presented in Appendix 6. The total amount of observations is 86. The model with 86 observations with three independent variables is acceptable and thus results can be generalizable. Model 1c is not statistically significant according to the Prob > F = 0.841 at the significance level of 0.05. The R2 of 0.010 indicates that the independent variables, EnvironmentalPracticesL, FirmAge and FirmSize can be used to explain about 1.0 % of the variation in the dependent variable, Intper_S. The coefficients indicate that one unit increase in EnvironmentalPracticesL would results in an increase of 0.003 in the Intper_S. A one unit increase in Firm age would result in a decrease of 0.001 in the Intper_S and a one unit increase in firm size would results in an increase of 0.001 in the Intper_S. None of the coefficients are statistically significant at the significance level of 0.05, as the p-values are all above 0.05.

Model 1d

According to Ramsay’s RESET test the model has no omitted variables with the p-value 0.134 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[9] = 8.04) with the p-value 0.531 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.047) in the model, as the p-value (0.829) is greater than 0.05. The VIF value of 1.06 is acceptable. The tolerance values indicate that 91.9% of variation in firm age, 91.9% of variation in firm size, and 98.2% of variation in EnvironmentalPracticesO is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution.

The results of the multiple regression analysis of model 1d are presented in Appendix 6. The total amount of observations is 86. The model with 86 observations with three independent variables is acceptable and thus results can be generalizable. Model 1d is not statistically

56

significant according to the Prob > F = 0.840 at the significance level of 0.05. The R2 of 0.010 indicates that the independent variables, EnvironmentalPracticesO, FirmAge and FirmSize can be used to explain about 1.0 % of the variation in the dependent variable, Intper_S. The coefficients indicate that one unit increase in EnvironmentalPracticesO would results in a decrease of 0.009 in the Intper_S. A one unit increase in Firm age would result in a decrease of 0.001 in the Intper_S and a one unit increase in firm size would results in an increase of 0.001 in the Intper_S. None of the coefficients are statistically significant at the significance level of 0.05, as the p-values are all above 0.05.

H1b. Responsible business practices related to employees are positively associated with financial performance of internationalizing SMEs.

According to table 8 there is a correlation of -0.212 between the FirmGrowthOT and EmployeeEngagement and the correlation between Intper_S and EmployeeEngagement is -0.034. The independent variable EmployeeEngagement was formed in a factor analysis, which was explained in section 3.4.2. The results of the regressions can be seen in Appendix 6 (Tables 18 and 19).

Model 1e

According to Ramsay’s RESET test the model 1e has no omitted variables with the p-value 0.650 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[5] = 5.05) with the p-value 0.409 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.353) in the model, as the p-value (0.552) is greater than 0.05. The VIF value of 1.03 is acceptable. The tolerance values indicate that 96.8% of variation in firm age and 96.8% of variation in EmployeeEngagement is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram

57

and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

The results of the multiple regression analysis of model 1e are presented in Appendix 6. The total amount of observations is 73. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The model with 73 observations with three independent variables is acceptable and thus results can be generalizable. Model 1e is statistically significant according to the Prob > F = 0.050 at the significance level of 0.05. The R2 of 0.082 indicates that the independent variables, EmployeeEngagement and FirmAge can be used to explain about 8.2 % of the variation in the dependent variable, FirmGrowthOT. The coefficients indicate that one unit increase in EmployeeEngagement would results in a decrease of 52.767 in the FirmGrowthOT. A one unit increase in Firm age would result in a decrease of 2.210 in the FirmGrowthOT. Although the model is statistically significant neither of the coefficients is statistically significant.

Model 1e2

According to Ramsay’s RESET test the model 1e2 has no omitted variables with the p-value 0.765 (>0.05). To check the assumption of homoscedasticity, White’s test was run. It shows that the model is homoscedastic (chi2[5] = 6.58) with the p-value 0.254 (>0.05). It can be concluded that the variance of error terms is constant in this model. To check the assumption of having no autocorrelation, Durbin’s alternative test for serial correlation was used. It shows that there is no autocorrelation (chi2 = 0.013) in the model, as the p-value (0.910) is greater than 0.05. The VIF value of 1.04 is acceptable. The tolerance values indicate that 96.4% of variation in firm age and 96.4% of variation in EmployeeEngagement is dependent from the other variables. It can be concluded that there is no significant multicollinearity in the model.

A histogram and normal probability plot were created to check the normality of the residuals (Appendix 5). In addition, it was checked by running the Shapiro-Wilk test. The p-value (0.000) was below 0.05, which means that the variance is not normally distributed. Also, the histogram and normal probability plot do not show normal distribution. But as the sample is quite large, this is not seen as problem.

58

The results of the multiple regression analysis of model 1e2 are presented in Appendix 6. The total amount of observations is 75. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The

The results of the multiple regression analysis of model 1e2 are presented in Appendix 6. The total amount of observations is 75. A generally accepted rule is that the ratio of observations to independent variables should not be below 5:1 and ideally being from 15 to 20 to one. The