Chapter 5: Results
5.3 Results
5.3.1 Hypothesis 1: Direct effect
The first hypothesis states: Being located in global cities increases firm financial performance.
To test the first hypothesis, an independent sample t-test and a hierarchical linear regression model were chosen. An independent sample t-test helps to compare the means of two groups and identify whether there is a significant difference between the means. The hierarchical linear regression model allows predicting the outcome of the dependent variable based on the independent variable. Moreover, it helps to statistically control for multiple variables simultaneously. The hierarchical regression model is shown with and without the covariates to indicate the potential bias which is present without adding the covariates. The control variables in this study include the age of the firm, the liquidity ratio, the employees and the digital nature of the company.
First, AV_ROA was investigated by using the independent t-test which is shown in table three. To use the independent sample t-test, it is necessary to establish whether equal variances are assumed. For AV_ROA, equal variances are not assumed based on the Levene’s test of variance value; 0,000 < 0,1). The t-test shows that there is no equality of means
(p-35 value; 0,089 < 0,1) which means there is a significant difference between the means of
AV_ROA between being located in a global city vs a non-global city.
Second, ROS was investigated. Similar to ROA, an independent sample t-test was performed which is shown in table 10. Equal variances are not assumed based on Levene’s test of equal variances (p-value; 0,023 < 0,1). No statistically significant evidence was found that the means of the groups are different (p-value; 0,142 > 0,1).
Independent Sample T-test
Levene’s test T-test
F Sig t df Sig
(two-tailed) AV_ROA Equal
variances
20,182 ,000 1,807 343 ,072
No equal variances
1,707 236,330 ,089
AV_ROS Equal variances
5,220 ,023 1,664 330 ,097
No equal variances
1,476 177,285 ,142
Table 3 Independent Sample t-test
To predict a more robust estimation of the relationship between AV_ROA and AV_ROS and the global city factor, a hierarchical regression model was computed which is shown in table four and five. To compare both models, the adjusted R2 was used as it takes the number of predictors into account. The overall model of AV_ROA was found not to be significant (0,131>0,1), however shows a trend towards significance. The model explains 0,7% of the variance of the model based on the adjusted R2. However, this model does not include covariates which are proven in the literature to affect firm financial performance.
Therefore, the control variables were included, namely employees, years active, liquidity and digital nature presented in table six and seven. In this case, 4,1% of the variance was
explained by the model. The model is not significant as 0,434 > 0,1. In the second block, the dummy for global cities was included. The second model explains only 3,2% of the variance of the model. This number is higher compared to the direct effect indicating that the
covariates do have explanatory power. The overall model is however not significant (0,374 >
0,1). Two of the variables were found the be significant, namely employees (0,038<0,1) and liquidity (0,076<0,1). This indicates a negative effect on ROA for both employees (β =-0,003) and liquidity (β =-0,001).
Model summary
36
Model R R2 Adj
R2
Std.
error
R2 chang e
F chang e
Df1 Df2 Sig. F chang e
1 ,097 ,009 ,007 ,01229 ,009 3,264 1 343 ,072
Predictor: Dummy_gc
Table 4 Model summary AV_ROA (Gc only)
Coefficients
Unstandardized Coefficients
Standardized Coefficients
Model B Std.
Error
Beta t Sig
1 (Constant) ,001 ,001 1,515 ,131
Dummy_gc -,002 ,001 -0,97 -1,807 ,072
Dependent variable: AV_ROA Table 5 Coefficient AV_ROA (Gc only)
Model summary
Model R R2 Adj
R2
Std.
error
R2 chang e
F chang e
Df1 Df2 Sig. F chang e
1a ,291 ,085 ,041 ,01234 ,085 1,962 4 85 ,108
2b ,294 ,087 ,032 ,01240 ,002 ,187 1 84 ,667
a. Predictors: (Constant), DIGITAL_DUMMY, NEW_Employees, NEW_YEARSACTIVE, NEW_LIQ b. Predictors: (Constant), DIGITAL_DUMMY, NEW_Employees, NEW_YEARSACTIVE, NEW_LIQ, dummy_gc
Table 6 Model summary AV_ROA
Coefficients
Unstandardized Coefficients
Standardized Coefficients
Model B Std.
Error
Beta t Sig
1 (Constant) ,004 ,006 ,787 ,434
37
New_employees -,004 ,002 -,238 -2,238 ,028
New_yearsactive ,003 ,004 ,088 ,786 ,434
New_liq -,001 ,000 -,187 -1,774 ,080
Digital_dummy -,003 ,004 -,071 -,658 ,512
2 (Constant) ,005 ,006 ,893 ,374
New_employees -,003 ,002 -,230 -2,112 ,038
New_yearsactive ,003 ,004 ,071 ,602 ,549
New_liq -,001 ,000 -,191 -1,797 ,076
Digital_dummy -,002 ,004 -,060 -,541 ,590
Dummy_gc -,001 ,003 -,052 -,445 ,657
Dependent variable: AV_ROA
Table 7 Coefficient AV_ROA
The same econometric exercise is performed using ROS as the performance measure.
Results are reported in tables eight, nine, ten and eleven. The overall model explains 0,5% of the variance of the model and is significant (0,021<0,1). Furthermore, the global city variable was significant (0,097<0,1). This means that being located in a global city decreases the ROS by 0,079 when no attention was given to control variables. Hierarchical regression was performed to show the effect of these covariates. The control variables were entered first and explained 0,40% of the variance of the model. The model was not significant as 0,490>0,1.
The independent dummy variable of being located in a global city was added in the second block. Adding the dummy variable made the model to explain 0,28% of the variance which means that the variable did not add value to the model. Moreover, the final model was found not to be significant as 0,517>0,1. One variable was found to be significant, namely the years a company is active (0,073<0,1), with a coefficient of β =0,166. Moreover, the variable for employees was found to be close to the threshold (0,184>0,1), with a coefficient β= 0,51.
Model summary
Model R R2 Adj
R2
Std.
error
R2 chang e
F chang e
Df1 Df2 Sig. F chang e
1 ,091 ,008 ,005 ,41653 ,008 2,769 1 330 ,097
Predictors: dummy_gc
Table 8 Model summary AV_ROS (Gc only)
38 Coefficients
Unstandardized Coefficients
Standardized Coefficients
Model B Std.
Error
Beta t Sig
1 (Constant) ,067 ,029 2,325 ,021
Dummy_gc -,079 ,047 -,091 -1,664 ,097
Dependent variable: AV_ROS
Table 9 Coefficients AV_ROS (Gc only)
Model summary
Model R R2 Adj
R2
Std.
error
R2 chang e
F chang e
Df1 Df2 Sig. F chang e
1a ,291 ,085 ,040 ,25318 ,085 1,897 4 82 ,119
2b ,291 ,085 ,028 ,25474 ,000 ,001 1 81 ,982
a. Predictors: (Constant), DIGITAL_DUMMY, NEW_Employees, NEW_YEARSACTIVE, NEW_LIQ
b. Predictors: (Constant), DIGITAL_DUMMY, NEW_Employees, NEW_YEARSACTIVE, NEW_LIQ, dummy_gc
Table 10 Model Summary AV_ROS
Coefficients
Unstandardized Coefficients
Standardized Coefficients
Model B Std.
Error
Beta t Sig
1 (Constant) -,091 ,131 -,693 ,490
New_employees -,051 ,038 -,157 -1,359 ,178
New_yearsactive ,166 ,086 ,218 1,929 ,057
New_liq ,071 ,056 ,147 1,270 ,208
Digital_dummy ,003 ,081 ,004 ,032 ,975
2 (Constant) -,092 ,141 -,651 ,517
New_employees -,051 ,038 -,157 -1,340 ,184
39
New_yearsactive ,166 ,092 ,219 1,814 ,073
New_liq ,071 ,057 ,147 1,259 ,212
Digital_dummy ,002 ,083 ,003 ,026 ,979
Dummy_gc ,001 ,062 ,003 ,023 ,982
Dependent variable: AV_ROS
Table 11 Coefficients AV_ROS
To conclude, for ROA no statistically significant relationship was found between the global city location and the dependent variable. For ROS, the relationship was found to be negative only when the control variables were not added to the model. The control variables of this study have been chosen based on previous research and are proven to affect the performance of a company, thus the focus will be on the model which includes them, however, it is important to consider that the size of the sample strongly decreases due to missing values for the control variables. It is worth nothing that it is interesting to find a negative relationship between global city locations and ROS, that needs to be further
investigated. Based on the above considerations, this research does not find a robust support of hypothesis 1.