RDP 1999-02: Reservation Wages and the Duration of Unemployment 2. The Australian Evidence

As there are relatively few cross-section or longitudinal data sets available, most Australian studies of unemployment have been based on macroeconomic data. In general, studies which have used data on individuals have estimated models which explain the probability of leaving unemployment, either after a given duration of unemployment or over a fixed period of time. The probability of leaving unemployment indirectly provides information about the characteristics which are likely to affect an individual's duration of unemployment. To date, only Miller and Volker (1987) have directly investigated those factors which affect unemployment duration using individual level data.

The most frequently used empirical model for investigating unemployment at the individual level explains the probability of leaving unemployment given the duration of unemployment already experienced. This framework, also known as the hazard model, has been used by Brooks (1986) and Miller and Volker (1987) to look at the youth labour market, by Aungles and Stewart (1986) to look at the behaviour of job seekers registered with the Commonwealth Employment Service (CES) in the Brisbane metropolitan area, and by Stromback, Dockery and Ying (1998) to look at a large cross-section of Australian job seekers surveyed in the SEUP.

In general, these studies find that among the general population, older people are less likely to leave unemployment than younger people, but when the analysis is restricted to the youth labour market, unemployed youth in their early twenties are more likely to exit unemployment than unemployed teenagers. Work experience and the level of education increase the probability of leaving unemployment, as does being married. To the extent that the duration of unemployment affects the probability of exiting unemployment, the effect appears to be restricted to the very short term.

Although this framework uses job-search theory to suggest the variables which should be included in an econometric model, the estimated parameters cannot be directly related back to the theory, and in this sense this analysis is essentially ‘reduced form’. In this respect it is noteworthy that none of the studies above simultaneously estimated the probability of leaving unemployment after a given duration of unemployment and reservation wages, as might be suggested by theory. It is also likely that the results are affected by the functional forms which are assumed for the different components of the hazard model.

The second indirect approach to understanding those factors which affect unemployment duration is to model transitions between labour market states between two points in time. This method has been used by Brooks and Volker (1985), who focus on the youth labour market, and Strombach, Dockery and Ying (1998). The effects of different characteristics on the probability of moving out of unemployment are consistent with the results obtained by studies noted earlier.

Brooks and Volker (1985) investigate the probability that an individual will either stay in unemployment, exit unemployment for full-time work or exit unemployment for part-time work. The reservation wage and duration of unemployment are included as explanatory variables. Across different specifications, one or other of these variables appeared to be significant, which suggests that they are highly correlated. Brooks and Volker also estimate an equation for reservation wages as a function of unemployment duration as well as other personal and background characteristics. Strombach, Dockery and Ying (1998) estimate similar models for the probability of remaining in unemployment or leaving unemployment for full-time or part-time work.^[1] They do not consider the reservation wage and find that longer spells of unemployment significantly reduce an individual's chances of getting a job.

Miller and Volker (1987) is the only Australian study of which we are aware that estimates an equation for unemployment duration directly. They exclude the reservation wage from their analysis, citing evidence that minimum wages, not reservation wages, are the real constraint facing job seekers in the youth labour market.

As already mentioned, the estimation methods outlined above use job-search theory to suggest which variables should be included to explain unemployment duration, but it is not possible to relate the estimated parameters back to the theory. The analysis which follows extends this literature by deriving an empirical model for unemployment duration which is directly related to job-search theory. The development of the empirical model in Section 4 ensures that the analysis is consistent with the economic model and that complications suggested by theory, such as the joint determination of the duration of unemployment and the reservation wage, are correctly dealt with.

Although these issues have not been addressed with Australian data, they have received some attention in the non-Australian literature.^[2] In part, both the Australian and overseas literatures have been constrained by the lack of appropriate data. The framework we have used to analyse the duration of unemployment is developed in Section 4, and follows the work of Lancaster (1985), Jones (1988) and Gorter and Gorter (1993).

Footnotes

In contrast to Brooks and Volker (1985), this study considers these transition probabilities individually and therefore cannot impose the constraint that the probabilities sum to one. [1]

An excellent survey of the empirical job-search literature, including structural models of unemployment duration and reservation wages, can be found in Devine and Kiefer (1991). [2]