1  Introduction

This paper reports estimates of wage functions for a number of demographic groups in New Zealand, using pooled information from the 1991/92, 1992/93, 1993/94, 1994/95, 1995/96, 1996/97, 1997/98 and 2000/01 Household Economic Survey (HES). Similar to a number of other papers in this field, the main aim of estimating wage equations is to impute wage rates for those who are not currently working, so that they can be used in labour supply models.[1] This imputation of wage rates is complicated by the fact that wage equations should ideally contain variables, such as industry and occupation, which are not observed for non-workers (for the same reason that wage rates are not available). These variables are important determining factors of wage rates. This paper therefore follows the same approach as Creedy et al. (2001) to allow the use of these variables.

The availability of eight data sets covering a period of ten years, in which several changes occurred in New Zealand’s economic situation and social policies (see for example Knutson, 1998), allow us to explore several interesting issues. For example, this long time span allows insights into changes over time, response to unemployment levels, the effect of changed work requirements for benefit recipients in 1998 and the increase in the age of eligibility for the New Zealand state pension from 60 to 65 years of age (gradually introduced from 1991 to 2000).

As in other articles, the estimation procedure corrects for the sample selection bias that would otherwise arise from the fact that only the wage rates of those currently working are observed. The approach of Creedy et al. (2001) is extended by estimating wage and employment (or selection) equations simultaneously, allowing for correlation between the unobserved components of the two equations, instead of using the standard Heckman procedure (Heckman, 1979), which is a two-step approach. The results using the two approaches are compared. There have been few studies estimating New Zealand wage functions, exceptions are Chiao and Walker (1992), who did not have good data, and Maloney (1997), whose equation had only a few explanatory variables. Earlier Australian wage functions were discussed for example by Miller and Rummery (1991), Murray (1996), Kalb (2000), Creedy et al. (2001) and Kalb and Scutella (2002). All these studies used the standard Heckman approach.

An advantage of the New Zealand data over the Australian data used in Creedy et al. (2001) and Kalb and Scutella (2002) is that there is no censoring in the recorded hours of labour supply, which simplifies the estimation procedure. Furthermore, a longer period, including more recent years, is covered by the available data in New Zealand. That is, a period from 1991 to 2001 compared to a period from 1994 to 1998 in Australia.

The standard selection model is described briefly in section 2. The data are described in section 3. Estimates of the wage equations are reported in section 4. The approach of assigning wage rates to non-workers and the prediction of wage rates for some hypothetical individuals is discussed in section 5. Brief conclusions are in section 6.