AN OCCASIONAL PAPER BY ANNIE CARINO-ABELLO, DAVID PEDERSON AND ANTHONY KING OF THE NATIONAL CENTRE FOR SOCIAL AND ECONOMIC MODELLING, UNIVERSITY OF CANBERRA
This Occasional Paper is intended to make the results of current research available to other interested parties.
Views expressed in this paper are those of the author and do not necessarily represent those of the Australian Bureau of Statistics. Where quoted or used, they should be attributed clearly to the author.
SUMMARY
A study of the dynamics of earned income is important for two basic reasons. First, it elaborates our picture of the labour market, providing the framework for looking at the distribution of earnings in greater depth, and providing inputs to policy development. Second, it is an input into modelling earnings dynamics, an area of growing interest in Australia.
This paper provides new evidence about how people’s earnings in Australia changed from year to year over the mid-1990s. The study was based on data from the Survey of Employment and Unemployment Patterns (SEUP), a longitudinal survey covering the period September 1994 to September 1997. Data for this survey were collected in three waves, in 1995, 1996 and 1997. This unique data set enabled us to track the pattern of people’s earnings on an annual basis and to relate the pattern of changes in earnings to personal characteristics such as sex, marital status, educational attainment and current and historical information on their work and earnings.
The first part of the report describes the mobility of earnings of the general population based on usual weekly earnings. Overall we find a general stability in earnings across all earnings deciles. Despite the fact that most people changed their earnings decile, 70-90% of people in each decile experienced an earnings shift of one decile or less (stayed in their current decile or moved to an adjacent decile). While there appears to be more variability in earnings among those on middle incomes relative to those at the upper and lower range of the distribution, this is due to the fact that the decile classes are narrow at middle incomes and wide at the income extremes.
Focusing on individuals at different points of the earnings distribution, we find a strong relationship between transitions to and from lower earnings and movements in and out of work. While this analysis of earnings mobility was restricted to respondents who were wage and salary employees at each date the data were collected, this does not preclude their being out of work at points in between. This is particularly the case for those in the lowest three earnings deciles. To focus on moves in and out of this broad sector of the earnings distribution, we defined the respondents falling in the bottom 3 deciles as the 'low earnings group' and the remainder as the 'higher earnings group' for every wave. Based on this definition, we looked at the pattern of changes in earnings over the three years for which data are available. Of the total sample of the general population, 9% had persistently low earnings while 72% had persistently higher earnings. Correspondingly, 28% of the general population had low earnings at least once over the three-year period.
Looking at the characteristics of the different earnings groups, we find that relative to the overall sample: (a) those with persistently low earnings have greater proportions of persons who are female, are part-time workers, have not completed secondary school, are unmarried, have less than excellent English speaking ability, have an employment handicap, and are working at middle and lower level clerical or transport and production occupations; (b) for those making a transition from low to higher earnings, there were changes associated with movement from part-time to full-time work, improvement in educational attainment, and changes in industry and occupation; and (c) for those making a transition from higher to low earnings, the most apparent changes were those from full-time to part-time work, and movement to lower level occupations.
In the second part of the report, earnings equations were estimated. Separate models were run for those who were employed in the previous period (transition groups 1 to 4) and those who were not (transition groups 5 to 8). Regressions were run for both the Population Reference Group and Jobseekers. We find that, generally, the coefficient estimates reflect what might be expected of a human capital model, and the equations have a good fit, explaining between 54% and 80% of the variance of log weekly earnings for the Population Reference Group, and between 49% and 67% for Jobseekers.
A key finding is that the effect of previous year’s earnings on current earnings varies on one’s earnings level as well as transition type. Generally, for those remaining in full-time work and with earnings below the middle of the range, we find no systematic relationship between current and previous year’s earnings. In contrast, for individuals remaining in full-time work and with earnings in the middle and upper ranges, we find a positive relationship. Finally, for those remaining in part-time work, the relationship between current and previous year’s earnings is positive at all earnings levels. The foregoing results hold true for both males and females, regardless of whether one is from the general population or a Jobseeker, and has implications concerning earnings stability for workers in various categories of employment and earnings.
The availability of actual data on labour market experience enabled us to estimate returns to experience for the pooled group of males and females. (If we had used imputed instead of actual years of experience, there would have been a need to allow for differences by sex, to take account of possible discontinuity in years in paid work particularly for females.) The coefficients on years in paid work are positive and show increasing returns with additional years of experience. Correspondingly, returns to years spent looking for work and not in the labour market are negative - the longer one stays in either of these two labour market states, the lower one’s earnings.
The results with respect to occupation, multiple jobs, educational attainment, English-speaking proficiency, employment handicap and marital status are as expected. Relative to the benchmark group and taking all other factors constant, higher earnings accrue to individuals at higher occupation levels, those with multiple jobs, those with post-school qualifications, those with high levels of English-speaking proficiency, those with no employment handicap, and to unmarried (relative to married) females. Those working in the public sector receive higher returns than those in the private sector, and the same holds true for those working in the manufacturing industry relative to those in agriculture and services, all other factors constant.
The foregoing findings enhance our knowledge of the dynamics of earned income and factors that affect the distribution of earnings. The results of the study can also serve as an input to modelling earnings dynamics in general, and for validating current models on the dynamics of earned income in Australia. Finally, the findings from these and other SEUP studies highlight the importance of the availability of longitudinal data over a reasonable length of time to enable analysis that would otherwise not be possible.
