CHAPTER 6: THE SURVEY: DATA ANALYSIS
6.9 BINARY LOGISTIC REGRESSION
This study aimed to determine whether, or not, any of the set of push factors namely governance framework, infrastructure, legislation, economic uncertainty, future limiting and narcissism society, are statistical significant predictors of the likelihood that a person will emigrate to New Zealand rather than Australia by using binary logistic regression.
In assessing the model adequacy and fit, table 6.18 sets out the information regarding the predictors included in the model of push factors, while table 6.19 provides information regarding overall model fit. The number of respondents included in this model was 216 as a case-wise removal process was used for missing data on any of the variables.
The results are tabled on page 167.
Table 6.18: Variables in the equation of predictors included in the model of push factor
B S.E. Wald df Sig. Exp(B)
Step 1 Gov_framework .153 .236 .424 1 .515 1.166
Infrastructure -.018 .166 .012 1 .914 .982
Legislation -.288 .145 3.957 1 .047 .750
Economic_Unc .240 .220 1.185 1 .276 1.271
Fut_limiting -.069 .209 .109 1 .741 .933
Narc_Soc -.027 .195 .019 1 .889 .973
Constant -.142 .840 .029 1 .866 .868
The overall correct prediction classification of the model is 56.5%.
The results indicate that legislation is the only statistically significant predictor, at the 5%
level of significance. The odds ratios indicate further that:
For each one point increase in the importance level of legislation as a push factor, the odds of a person emigrating to New Zealand decrease to 0.75.
Table 6.19: Variables in the equation of the overall model fit for push factor
Model summary Hosmer and Lemeshow Test
-2-Log-likelihood Cox & Snell R Square
Nagelkerte R Square
Chi-square df Sig
Step 1 291.754 .021 .028 10.334 8 .242
The Hosmer and Lemeshow test shows non-significance indicating that the data fits the model adequately (Hosmer & Lemeshow, 2000). The R2 measures (Nagelkerke R2 , Cox and Snell R2), however, are low (0.028 and 0.021 respectively) and indicate that follow-up studies are needed to validate and improve the current model.
In assessing the model adequacy and fit, table 6.20 sets out the information regarding the predictors included in the model of pull factors, while table 6.21 provides information regarding overall model fit. The number of respondents included in this model was 75 as a case-wise removal process was used for missing data on any of the variables.
Table 6.20: Variables in the equation of predictors included in the model of pull factor
B S.E. Wald df Sig. Exp(B)
Step 1 Eff_gov_services .037 .382 .010 1 .922 1.038
Econ_certainty -.645 .368 3.068 1 .080 .525
Lower_costliving .684 .286 5.698 1 .017 1.982
Fam_circumstances -.409 .260 2.467 1 .116 .664
Better_Future .328 .295 1.238 1 .266 1.388
Constant .617 1.592 .150 1 .698 1.854
The Overall correct prediction classification of the model is 61.3%.
The results indicate that lower cost of living is the only statistically significant predictor, at the 5% level of significance. The odds ratios indicate further that:
For each one point increase in the importance level of a lower cost of living as a pull factor, the odds of a person emigrating to New Zealand decrease to 1.982.
Table 6.21: Variables in the equation of the overall model fit for pull factor
Model summary Hosmer and Lemeshow Test
-2-Log-likelihood Cox & Snell R Square
Nagelkerte R Square
Chi-square df Sig
Step 1 89.399 .133 .180 10.187 7 .178
The Hosmer and Lemeshow test shows non-significance, indicating that the data fits the model adequately (Hosmer & Lemeshow, 2000). The R2 measures (Nagelkerke R2 , Cox and Snell R2), however, are low (0.133 and 0.180 respectively) and do indicate that follow-up studies are needed to validate and improve the current model.
In assessing the model adequacy and fit, table 6.22 sets out the information regarding the predictors included in the model of personal needs factors, while table 6.23 provides information regarding overall model fit. The number of respondents included in this model was 162 as a case-wise removal process was used for missing data on any of the variables.
The results are tabled on page 169.
Table 6.22: Variables in the equation of predictors included in the model personal needs
B S.E. Wald df Sig. Exp(B)
Step 1 PhysNeeds .310 .209 2.192 1 .139 1.363
SafetyNeeds .000 .229 .000 1 .998 1.000
Belonging -.096 .200 .232 1 .630 .908
Esteem_Needs .137 .240 .325 1 .568 1.147
Constant -1.519 1.290 1.386 1 .239 .219
The Overall correct prediction classification of the model is 50.6%.
The results indicate that safety needs is the only statistically significant predictor, at the 5%
level of significance. The odds ratios indicate further that:
For each one point increase in the importance level of safety needs as a push factor, the odds of a person emigrating to New Zealand decrease to 1.000.
Table 6.23: Variables in the equation of the overall model fit for personal needs
Model summary Hosmer and Lemeshow Test
-2-Log-likelihood Cox & Snell R Square
Nagelkerte R Square
Chi-square df Sig
Step 1 220.937 .022 .029 7.964 8 .437
The Hosmer and Lemeshow test shows non-significance indicating that the data fits the model adequately (Hosmer & Lemeshow, 2000). The R2 measures (Nagelkerke R2 , Cox and Snell R2), however, are low (0.029 and 0.022 respectively) and do indicate that follow-up studies are needed to validate and improve the current model.
6.10 CONCLUSION
The purpose of this chapter was to determine what the main push factors and pull factors (both internal and external) are that contribute to the emigration or moving abroad of economically active South Africans. It also set out to determine whether there is a correlation between the push and pull factors for South African emigrants in Australia and New Zealand.
The chapter reviewed the demographics profile of the respondents. Secondly, the push, pull
and personal needs items were analysed through exploratory factor analysis to determine whether meaningful factors emerged. Inferential statistics explored the relationship between these factors by focussing on differences in importance of the push, pull and personal needs factors between emigrants residing in Australia and New Zealand. Binary Logistic regression showed whether there were factors that are statistically significant predictors of the odds that a South African emigrant will choose to locate to New Zealand. The results will be discussed in Chapter 7.