5.2  Predicted wages

In this subsection, we present an overview of the predicted wages, using the approach from the previous subsection assigning the average occupation and industry characteristics for the unemployed (see Table A.3). Table 4 shows the average predicted wage by employment status and education level. For those who are employed the average observed wages are presented as well. Average predicted wages and average observed wages are close to each other in the different education groups for all demographic groups. For married women and single women, the difference between the observed and predicted wage is somewhat larger, because for this group a larger proportion of the observed wages were too low to be included in the wage model.[18] When these observations are excluded, the predicted average wage is close to the observed average wage.

Comparing average predicted wages over the whole sample we find clear differences between participants and non-participants. As expected the predicted wages are lower for those who are currently unemployed. This is even true for sole parents (although to a much smaller extent) who did not suffer from selectivity problems, which indicates that the characteristics of participants and non-participants are different. An explanation of this smaller selectivity problem for sole parents is that the reason for non-employment of sole parents is different from the reasons in the other groups. Responsibility for a young child and the difficulty of combining care for a young child with employment may be the main reason of non employment for sole parents. This should not affect a sole parent’s expected wage rate, whereas in the other groups other (unobserved) factors may impede employment and wage levels at the same time. Comparing the average wages across the demographic groups confirms expectations that married men have the highest wage levels. Married women and sole parents have higher wage levels than singles, where single women obtain the lowest wage levels.

Note a: Wages are only observed for employed individuals. The first number represents the average predicted wage and the second number represents the average observed wage.

In addition to the average predicted wages presented above, as a further illustration, we provide some examples of predicted wages obtained when unemployed hypothetical individuals are assigned the average occupation and industry characteristics for the unemployed (see Table A.3). The two different approaches used in estimation also result in somewhat different predictions for singles and married men, although the relative wage rate levels remain similar. In the next stage of the project, we will use the imputed values obtained from the jointly estimated model.

In the following examples, we look at hypothetical individuals living in Auckland in 2001, who are not eligible for the New Zealand Superannuation. Consider first a hypothetical female unemployed sole parent with the following characteristics: aged 23 years; with a certificate; partly from Maori/Pacific Islander descent; with no other income unit income; not living with her parents; with one dependent child at the age of three; and a current female unemployment rate of 5.1 per cent. The predicted or imputed wage obtained using the distribution over industry and occupation groups for unemployed individuals (see Table A.3) is found to be $12.48 per hour using the joint estimation approach and $12.41 using the two-step estimation approach.

Second, consider a hypothetical unemployed single female without children; aged 19 years; from European descent; with a certificate; with $400 of other income; not living with her parents; and a current female unemployment rate of 5.1 per cent. The imputed hourly wage is found to be $8.98 for the joint model ($6.70 in the two-step model).

Third, consider a hypothetical unemployed single male without children; aged 28 years; with a certificate; from Maori/Pacific Islander descent; living with his parents; no other income; and a current male unemployment rate of 5.4 per cent. The imputed wage is $11.31 for the joint model ($9.65 in the two-step model).

Fourth, consider a hypothetical unemployed married female: aged 35 years; with six dependent children, where the youngest child is 4 years old; from European descent; without educational qualifications; partner has postgraduate degree and is employed at a wage of $1000 per week; no other income; and a current female unemployment rate of 5.1 per cent. The basic imputed wage is $12.40 per hour in the joint model ($12.28 in the two-step model).

Finally, consider a hypothetical unemployed married male: aged 33 years; with two dependent children where the youngest is two years of age; from European descent; without qualifications; partner has a certificate and is currently employed at $800 dollars per week; no other income; and a current male unemployment rate of 5.4 per cent. The basic hourly rate is $12.40 per hour in the joint model ($10.63 in the two-step model).

From the above results, it is clear that the groups where the estimated correlation is quite different between the two alternative specifications also display the largest difference between the imputed wage rates using the alternative models. As a result of the very strong correlation between wage rates and employment probabilities estimated in the two-step model for singles and married men, the two-step model predicts wage rates at a rather low level for non-working singles and married men.