RDP 2002-07: An Exploration of Marginal Attachment to the Australian Labour Market 4. The Dynamics of Unemployment and Marginal Attachment

This section compares the labour market dynamics of the unemployed and marginally attached with the dynamics of the employed and other NILF. As discussed above, the behaviour of the employed and other NILF categories is only estimated using weighted labour market outcomes for individuals who were in the PRG sample. The behaviour of the unemployed and marginally attached workers is estimated using weighted information about individuals in the JS sample. Although the individuals in the PRG sample who were unemployed or marginally attached workers in June 1995 will also be representative of these groups, the sample sizes are small, and as it is not possible to combine the JS and PRG weights, these individuals have not been included.

4.1 Distribution of the Number of Episodes of Each Labour Force State

One way of describing the labour market dynamics is to examine the distribution of the number of episodes spent in each labour force state across respondents. This information is presented in Table 3. The analysis is restricted to respondents who were still in the sample at the time of the final (wave 3) interview, and had experienced at least one of the relevant labour market episodes. Note that it is conceptually difficult to identify the number of employment episodes because respondents may hold several jobs at the one time and consequently such estimates are omitted from Table 3.

Table 3: Distribution of Number of Episodes by Labour Force State
Per cent
Number of episodes Unemployed Marginally attached Other NILF
One 55.5 78.1 78.1
Two 28.5 17.5 13.2
Three 11.6 3.8 6.9
Four 2.7 0.5 1.0
Five 1.1 0.1 0.6
Six or more 0.6 0.0 0.2
Average number of episodes 1.67 1.27 1.33

Note: The sample includes all individuals that were present at the end of the survey period, and weights that account for sample attrition.

Amongst the respondents who experienced at least one episode of unemployment, the average number of episodes was 1.67. This is higher than the average of 1.33 episodes of other NILF and 1.27 episodes of marginal attachment. The distribution of unemployment episodes is relatively dispersed – 55.5 per cent of respondents experienced only one episode, 28.5 per cent experienced two episodes and 11.6 per cent had experienced three episodes of unemployment.

Overall the distribution of the number of episodes spent in other NILF is similar to that for marginal attachment. This similarity is not surprising given that neither the marginally attached nor other NILF are actively looking for employment and therefore are less likely to have an episode interrupted by employment. The tendency for there to be fewer episodes of marginal attachment compared to unemployment episodes indicates that there may be differences in the labour market dynamics of the unemployed and the marginally attached.

4.2 Labour Market Transitions

Transition probabilities are often used to describe labour market dynamics by presenting the extent to which the reported labour force status of individuals changes over time. A transition probability matrix identifies the extent and path of movement between various labour market states of individuals from one time period to another. Four labour market status categories are examined – employed, unemployed, marginally attached and other NILF.

Table 4 presents the transition probability matrix for the 3, 12 and 24-month periods. Each row in the table relates to the labour market status of individuals at June 1995 and each column is their labour force status 3, 12, or 24 months on. The interpretation of the cells in the table is best described with an example. Reading across the first row, the figure in the column titled ‘Employment’ indicates that, of all individuals who were employed in June 1995, 97.2 per cent were employed three months on. The figure in the next column suggests that 1.3 per cent of those employed in June 1995 were unemployed three months on. The column titled ‘Marginally attached’ indicates that 0.1 per cent of the employed in June 1995 were marginally attached three months on. The final column indicates that 1.4 per cent of those employed on 1 June 1995 were other NILF three months on.

Table 4: Transaction Probabilities between Labour Force States
  Later labour force status  
From 1 June 1995 Employment Unemployment Marginally attached Other NILF Total number
3-month horizon probabilities
(1 September 1995)
Employment 97.2 (0.6) 1.3 (0.4) 0.1 (0.1) 1.4 (0.4) 1,401
Unemployment 22.4 (0.9) 70.0 (1.0) 3.6 (0.4) 4.1 (0.5) 2,455
Marginally attached 11.3 (2.1) 18.5 (2.7) 68.3 (3.2) 1.9 (1.1) 296
Other NILF 10.6 (2.9) 4.6 (1.9) 1.1 (0.8) 83.7 (3.4) 283
12-month horizon probabilities
(1 June 1996)
Employment 94.8 (0.8) 1.5 (0.3) 0.7 (0.2) 3.1 (0.6) 1,401
Unemployment 43.2 (1.1) 41.2 (1.1) 5.8 (0.5) 9.7 (0.7) 2,455
Marginally attached 25.0 (2.9) 17.2 (2.5) 52.8 (3.3) 5.1 (1.5) 296
Other NILF 22.1 (3.5) 4.0 (1.1) 1.6 (0.7) 72.3 (3.5) 283
24-month horizon probabilities
(1 June 1997)
Employment 91.2 (0.9) 2.3 (0.5) 1.3 (0.4) 5.2 (0.7) 1,401
Unemployment 49.4 (1.1) 30.4 (1.0) 7.5 (0.6) 12.7 (0.7) 2,455
Marginally attached 33.3 (3.0) 20.7 (2.7) 39.3 (3.2) 6.9 (1.8) 296
Other NILF 30.3 (3.7) 6.3 (2.1) 4.3 (1.6) 59.1 (3.9) 283

Notes: The sample includes all individuals that were present at the end of the survey period, and weights that account for sample attrition. The estimates for those who are initially employed or other NILF are from the PRG sample, and for those who are initially unemployed and marginally attached are from the JS sample. Estimates for those who are initially unemployed and marginally attached for the PRG sample are presented in Appendix B. The standard error for each of the transition probabilities is presented in parentheses and is estimated using standard variance estimators. Further details can be found in StataCorp (2001b, pp 69–71).

We can test whether the probability of remaining in the initial labour force state differs between the unemployed and marginally attached by asking whether the point estimate of one transition lies within the 95 per cent confidence interval of the other. In other words, the estimate of 70 per cent for the probability of remaining unemployed after three months is not significantly different, statistically speaking, from the estimate of 68.3 per cent for the marginally attached, because it lies within the 95 per cent confidence interval 61.9 to 74.7 per cent. Therefore, we can conclude that in the short-run (three months), marginal attachment and unemployment are equally stable labour force states.

4.2.1 Short-run labour force transitions

While the probability of remaining in the same labour force state in the short-run is statistically similar for marginally attached and the unemployed (Table 4), it is much lower than the probability of remaining in employment or other NILF. However, the unemployed are significantly more likely to leave the labour force to move to other NILF (4.1 per cent), than are the marginally attached or the employed (1.9 and 1.4 per cent respectively).

Turning to movements into the labour force as it is conventionally defined (employment plus unemployment), a much higher proportion of the marginally attached enter the labour force (29.8 per cent) than do those who were other NILF (15.2 per cent). This is primarily due to differences in the probability of moving into unemployment. This suggests that over the short-run, the marginally attached have a much stronger attachment to the labour force than people classified as other NILF – a finding that confirms that the definition of marginally attached used in this paper is reasonable and accords with the underlying concept it attempts to capture.

In the short-run, the probability of moving from marginal attachment to employment (11.3 per cent) is similar to the probability of entering employment from other NILF (10.6 per cent), but is only half the probability of entering employment from unemployment (22.4 per cent). It is not really surprising that the unemployed are more likely to move into employment in the short-run given they are actively searching for work. The similarities between the transitions into employment of the marginally attached and other NILF are likely to be driven by different factors. The other NILF who move into employment in the short-run are likely to be people who were always highly employable, but changed personal circumstances mean that they have decided that they now want to work. On the other hand, the marginally attached are more likely to be responding to changes in personal circumstances that facilitate job search activity and improvements in labour market conditions, which increase chances of finding employment.

4.2.2 Medium-run and long-run labour force transitions

As the time horizon increases to the medium-run (12 months) and the long-run (24 months), the probability of movement across labour market states increases and the probability of remaining in the same labour market state declines. Unemployment appears to become less stable relative to marginal attachment, as the probability of remaining in the same labour force state only decreases to 52.8 per cent for the marginally attached but to 41.2 per cent for the unemployed over a 12-month horizon. This is consistent with the earlier observation that on average respondents experience fewer episodes of marginal attachment than of unemployment. In contrast, although the probability of remaining employed falls as the horizon increases, it remains over 90 per cent after 24 months.

The unemployed remain significantly more likely to move to other NILF than either the marginally attached or the employed at both the 12-month and the 24-month horizons. This suggests that the marginally attached have a similar attachment to the broadly-defined labour force as the employed, and a stronger attachment than the unemployed.

The probability of transition from outside the conventionally-defined labour force (employment plus unemployment) to employment increases for marginally attached and other NILF from around 10 per cent in the short-run to 25 per cent in the medium-run to over 30 per cent in the long-run. Despite this, both the marginally attached and other NILF remain significantly less likely to move to employment than the unemployed. This is probably a reflection of both differences in employability and in job search effort. In contrast, the marginally attached remain much more likely to move into unemployment than the other NILF at both the 12-month and 24-month horizons, although these transition probabilities do not increase significantly as the time horizon increases.

This suggests that there is a constant flow of marginally attached workers into unemployment where they actively look for work, and it is only after commencing an active search for work that the marginally attached enhance their employment prospects. In contrast, many of the respondents who indicated they moved from other NILF into employment appear to have done so without becoming unemployed, reinforcing our contention that such people are inherently more employable than the marginally attached.

4.2.3 Labour force transitions by broad age group and gender

Given the a priori expectations of differences in labour market participation for different age groups, we separately calculate the 12-month transition matrices for those under 25 years of age (younger), those aged over 25 to 44 years (prime-aged) and those aged 45 years or over (older) respondents (Table 5). Labour market behaviour is also likely to be different across gender, and these probabilities are presented in Table 6.

Table 5: 12-month Transition Probabilities by Age Group
  Later labour force status  
  Employmenl Unemployment Marginally attached Other NILF Total number
Aged less than 25 years
Employment 88.8 (2.6) 3.3 (1.4) 1.5 (1.0) 6.4 (2.1) 207
Unemployment 51.6 (2.1) 35.7 (2.0) 4.7 (0.9) 7.9 (1.2) 683
Marginally attached 30.9 (5.8) 21.0 (4.9) 42.2 (6.1) 6.0 (3.0) 80
Other NILF 27.6 (9.8) 10.6 (4.1) 3.6 (2.4) 58.2 (9.7) 50
Aged between 25 and 44 years 815
Employment 96.3 (0.9) 1.0 (0.4) 0.4 (0.2) 2.3 (0.8) 1,178
Unemployment 45.5 (1.6) 40.1 (1.6) 6.1 (0.7) 8.4 (0.9) 130
Marginally attached 28.4 (4.5) 16.1 (3.6) 50.0 (4.9) 5.6 (2.7) 124
Other NILF 29.1 (4.9) 3.1 (1.2) 1.6 (1.2) 66.2 (5.0) 379
Aged 45 years and over 595
Employment 95.8 (1.1) 1.1 (0.5) 0.7 (0.5) 2.4 (0.9) 86
Unemployment 27.0 (2.0) 51.1 (2.3) 6.9 (1.1) 15.0 (1.7) 109
Marginally attached 14.3 (4.4) 14.9 (4.8) 67.3 (5.9) 3.4 (1.9) 207
Other NILF 11.4 (3.8) 0.5 (0.4) 0.3 (0.3) 87.7 (3.8) 683
Notes: The sample includes all individuals that were present at the end of the survey period, and weights that account for sample attrition. The estimates for those who are initially employed or other NILF are from the PRG sample, and for those who are initially unemployed and marginally attached are from the JS sample. The standard error for each of the transition probabilities is presented in parentheses and is estimated using standard variance estimators. Further details can be found in StataCorp (2001b, pp 69–71).
Table 6: 12-month Transition Probabilities by Gender
  Later labour force status  
  Employment Unemployment Marginally attached Other NILF Total number
Notes: The sample includes all individuals that were present at the end of the survey period, and weights that account for sample attrition. The estimates for those who are initially employed or other NILF are from the PRG sample, and for those who are initially unemployed and marginally attached are from the JS sample. The standard error for each of the transition probabilities is presented in parentheses and is estimated using standard variance estimators. Further details can be found in StataCorp (2001b, pp 69–71).
Females
Employment 92.3 (1.4) 1.5 (0.6) 1.2 (0.5) 5.0 (1.2) 654
Unemployment 41.7 (1.7) 38.6 (1.7) 8.1 (11.5) 11.5 (1.1) 1,075
Marginally attached 22.3 (3.4) 15.8 (3.0) 4.0 (3.6) 3.6 (1.4) 201
Other NILF 16.2 (2.7) 4.1 (1.3) 0.8 (78.1) 78.1 (3.0) 231
Males
Employment 96.7 (0.8) 1.4 (0.4) 0.2 (1.6) 1.6 (0.6) 747
Unemployment 44.3 (1.5) 43.1 (1.5) 4.2 (0.6) 8.4 (0.8) 1,381
Marginally attached 29.9 (5.2) 19.8 (4.4) 42.5 (5.5) 7.8 (3.4) 95
Other NILF 40.3 (9.8) 3.8 (2.1) 1.6 (1.6) 54.3 (9.6) 52

In general, the rates of transition between labour force states decrease with age, indicating that the older group have more stable labour market behaviour. For the younger respondents, only 42.2 per cent of those marginally attached in the initial period were still marginally attached 12 months later. This compares to 50 per cent of prime-aged marginally attached and 67.3 per cent of older marginally attached respondents. The only labour force state that was not always more stable for older age groups, in the sense that the probability of remaining in that state increased, was employment where about 95 per cent of prime-age and older groups remained in a job over a 12-month period.

The higher propensity for the unemployed to move outside the broadly-defined labour force (compared to the marginally attached and the employed) is especially evident among prime-aged and older groups. That is, the ongoing desire to work of the older unemployed appears to be eroded by unsuccessful job search. In contrast, the older marginally attached retain the desire to work although their job search activity may be constrained.

The flip-side of this observation is that the older other-NILF group are significantly less likely to move into the conventionally-defined labour force (11.9 per cent) than the prime-aged other-NILF group (32.2 per cent), who are in turn less likely to make this transition than younger members of the other-NILF group (38.2 per cent). In general, the primary difference between a marginally attached person and another NILF person entering the conventionally-defined labour force, across all age groups, lies in the relatively high probability that the marginally attached person will enter unemployment. The high probability of the unemployed moving into employment relative to the probabilities for the marginally attached and the other-NILF groups can also be observed across all age groups, although the absolute probabilities fall with age.

In general, females have much higher probabilities of remaining marginally attached or other NILF than do males, and somewhat lower rates of remaining either employed or unemployed (Table 6). The relative stability of marginal attachment and other NILF is consistent with the evidence presented in Table 3. Again, the unemployed are more likely to leave the broadly-defined labour force, although this phenomenon is almost entirely concentrated among females. Given that this observation is also more prominent for the older age group, it appears that older females are the driving force behind the disproportionate number of unemployed who leave the broadly-defined labour force.

In terms of the conventional definition of labour supply, only 20.3 per cent of females in the other NILF category were either employed or unemployed after one year compared to 44.1 per cent of the analogous males. The sex differential was less stark for marginally attached females and males with 38.1 and 49.7 per cent respectively entering the labour force after one year. In contrast to the results for other groups, the probability of males in the other NILF category entering employment after one year is 40.3 per cent. This is comparable to the probability that an unemployed male gains employment and is better than the employment prospects of the marginally attached males, although the difference is not statistically significant. For both males and females, the general result holds that marginally attached people have a higher probability of moving into unemployment than those who are classified as other NILF.