6 Conclusion
This paper has reported estimates of wage equations for New Zealand workers, using pooled data from the Household Economic Surveys between 1991/92 and 2000/01. Selection bias was taken into account by using the alternative methods of estimating wage and selection equations simultaneously and in two steps using the Heckman approach, in both cases allowing for correlation between the two equations. The effects of all the variables are similar in both approaches, but the estimated correlation differs for singles and married individuals, resulting in somewhat different predicted wages depending on the approach.
The results for the wage equations are as expected for the usual characteristics, with education and age (up to the early forties) increasing the expected wage. People over 43 to 45 experience decreasing wages with increasing age. Except for married women, people at higher education levels have more steeply increasing wages and the maximum wage rate occurs at an older age. From the models, it is also clear that occupation and industry affect the level of the wage. People working in managerial, professional or associate professional positions are clearly paid more than people in other occupations and people in elementary occupations and agriculture or fishery workers are paid the least. Individuals living in the two main cities of Auckland and Wellington are paid more than people living elsewhere. The ethnicity variable shows that people from different ethnic groups are affected in different ways, individuals from European descent are paid higher wages than individuals from Maori or Pacific Islander descent, who are again paid more than individuals from the remaining ethnic groups.
The long time period for which data were available, allows a time trend to be estimated. The effect of business cycle changes was taken into account to some extent by including national unemployment rates in the model. It is found that changes took place for some groups over time and that unemployment rates do not seem to affect wage levels. Wage rates for married and single women have increased over time to some extent.
Employment rates are affected by the usual characteristics as well, that is, women with younger or more children are less likely to be employed; individuals living in cities; more highly educated individuals; European New Zealanders, except for married women; and men married to a partner with a medium-level education are more likely to be employed. Individuals with more income from other sources, with a partner on a higher wage income, with a partner with a postgraduate degree, and individuals from other non-European backgrounds (and Maori and Pacific Islander descent to a lesser extent) are less likely to be employed. The highest probability of employment occurs around the age of 35 years.
In addition to the usual characteristics include in wage models, some other variables were included. First, the gradual change in the age of eligibility for New Zealand Superannuation from 60 to 65 years during the survey allowed us to examine the influence of eligibility for the Superannuation while controlling for several other characteristics. An individual’s own eligibility plays a significant role in the decision to participate, whereas for women the partner’s eligibility is also important although to a lesser extent than their own eligibility. This effect is insignificant for sole parents, which may be explained by the smaller proportion of sole parents in the age group for which this change took place. No evidence was found for the effect of the changed work requirements for Domestic Purposes Benefit recipients or for the effect of teenage births on employment rates. Secondly, time trends were included and show that employment rates for sole parents increased over time and decreased with the unemployment rate for married men, and single women and men. Finally, including a dummy variable for singles and sole parents that indicates whether they are living with their parent(s) shows that singles are less likely to be employed when living with their parents, whereas sole parents are more likely to be employed when living with their parents. This may indicate the existence of childcare opportunities provided by the grandparents.
The process of assigning a wage rate to non-workers, needed in the context of labour supply analysis, was examined with special attention given to dealing with the situation where the wage equation includes variables that are not available at an individual level for the unemployed (such as occupation and industry). For these two characteristics, annual distributions over the categories were used for the unemployed instead of individual observations. These annual aggregate observations on industry and occupation, based on the value for these characteristics in their previous job (excluding the unemployed who have never had a job), were used in the imputation. On average, non-workers have lower wages than the workers, even for sole parents, where there is no evidence of selection bias. In this latter case, it is the difference in characteristics between the two groups that is causing the difference in average wages.
