This paper presents results for four separately estimated sets of discrete choice labour supply models using the Household Economic Surveys from 1991/92 up to 2000/01. The New Zealand working-age population is divided into sole parents, single men, single women, and couples. The labour supply models use imputed wages for the non-workers. Some of the preference parameters for work and income are made dependent on personal and household characteristics to allow for heterogeneity in preferences among households. In addition, allowance is made for unobserved heterogeneity in preferences. The estimated parameters for the different groups are used to calculate confidence intervals for expected labour supply and the probability of working at the different discrete hours points. The effect of particular characteristics on labour supply is illustrated by computing marginal effects across the samples. The wage elasticities fall within the range of values found in other studies. Expected labour supply, predicted by using the estimated models, results in values close to the observed averages and confidence intervals around the expected values are reasonably narrow in most groups. The results are as anticipated and similar to results in other countries, with preferences for work being higher for people with higher education, who are in their thirties. Furthermore, for women the presence of young children decreases the preference for work. In addition to these variables, which are usually included in labour supply models, the "eligibility for New Zealand Superannuation" indicator and a "living with parents" indicator are included. For all groups, the delayed eligibility for the state provided superannuation scheme is found to increase labour supply. The indicator for living with one's parents is found to increase labour supply for sole parents (indicating that living with one's parents may be a childcare strategy), although the effect was not significant.
New Zealand labour supply discrete choice labour supply model simulated maximum likelihood simulated confidence intervals