3  The Data

The data used in this analysis are taken from the 1991/92 to 2000/01 Household Economic Surveys collected by Statistics New Zealand. These surveys were released on a yearly basis from 1991/92 to 1997/98, but are currently undertaken only once every three years. The survey collects information on the sources and amounts of income received by persons resident in private dwellings throughout New Zealand, along with data on a range of characteristics for all individuals within the household. The individuals in each household are linked by a household number and family number. The survey is held continuously over the year with around 2000 individuals interviewed every quarter during the financial year, except for 1992/93 when over 3000 individuals were surveyed per quarter. In the surveys from 1991/92 to 2000/01, information is available for 68,711 individuals.

The details of hours worked are required for the calculation of wage rates, which is obtained for each individual as the ratio of total earnings to hours worked. Hence the following analysis ignores the possibility that individuals may obtain overtime premiums, or may work in more than one job at different wage levels

The majority of the data used as explanatory variables were recoded as zero-one dummy variables. To keep all the variables to a similar scale, all of the non-wage income variables were divided by 1000 while age was divided by 10. Any individuals with inconsistent observations on income from wages or salaries and hours worked, that is positive earnings for zero hours or zero earnings for positive hours, are excluded from the wage equation (as sensible wage rates cannot be calculated for them). However, these observations do remain in the selection equation assuming that we correctly observe whether or not they are working.

In the survey, individuals could name up to three ethnic groups with which they associate themselves. These three ethnic variables are recoded into one variable with the following categories for this analysis. Anyone who names Maori or a Pacific Islander group as one of the ethnic groups to which they belong is categorized as such. Those who do not name other ethnic groups are classified separately from those who also name alternative groups. Those who do not name Maori or a Pacific Islander group, but who associate themselves with ethnic groups other than the Pakeha (or European) group, are grouped into the other non-European category. This group is also subdivided into two groups if sufficient observations are available. One group which only names other non-European ethnic groups to describe their ethnicity and one group which also names the European group.

As the emphasis of the analysis is on obtaining results to be used in labour supply analysis for people of working age, individuals over 65 years (the current age of eligibility for New Zealand Superannuation) are excluded from the sample. In the earlier survey years, individuals would be eligible for the New Zealand Superannuation[4] at 60 years of age, but this was gradually increased to 65 years of age. This change during the survey period provides an opportunity to look at the effect of this change on the labour force participation of individuals aged between 60 and 64. Other groups which have been excluded are those with a disability and those in full-time education, because they are unlikely to participate in the labour force and the factors determining their participation decision would be quite different from other people of working age. Finally, the self-employed are omitted from the sample, because their decision to work an additional hour has a less direct link to a wage rate for that additional hour than the link between wage rate and labour supply for the wage and salary earners.[5] In addition, the concepts of market wage and reservation wage underlying the selection equation in the wage model do not seem as relevant to the self employed as to wage and salary earners.

The eight surveys were pooled and the sample was divided into five demographic groups.[6] These are sole parents, single females without dependents, single males without dependents, married females and married males. Summary tables of sample characteristics are provided for each demographic group in Table A.2 in the Appendix. It was not possible to estimate separate equations for sole mothers and sole fathers, given the small number of sole fathers in the sample[7].

Examples of distributions of the logarithms of observed hourly wage rates for the five demographic groups are shown in Figures 1 to 5. These are based on 2001 wages for workers. The histograms suggest that the wage distributions are approximately lognormal, although the distributions of the logarithm of the wages are slightly more peaked than the corresponding normal distributions with the same mean and variance. Individuals reporting wage rates more than 50 per cent below the relevant minimum wage at the time[8] or greater than $100 an hour are considered outliers and were omitted from the wage equation. A total of 227 observations were excluded for this reason. As expected, the graphs show that the modal wage rate is higher for married men and women than for singles. The difference between men and women is less pronounced.

Besides the information in the HES data sets, we used information on quarterly unemployment rates by gender from the second quarter of 1996 and yearly information up to that time. The unemployment rates are used to create a time-dependent unemployment measure for each observation.

Figure 1 – Log hourly wage rates for sole parents, December 2001 wages

Figure 2 – Log hourly wage rates for single females without dependents, December 2001 wages

Figure 3 – Log hourly wage rates for single males without dependents, December 2001 wages

Figure 4 – Log hourly wage rates for married females, December 2001 wages

Figure 5 – Log hourly wage rates for married males, December 2001 wages