Number 10: The duration of unemployment benefit spells: a comparison of Indigenous and non-Indigenous persons

This report was published by the former Department of Families, Community Services (FaCS).

Executive Summary

This paper  reports  the results  from an analysis of the duration of spells  of unemployment- related income  support  for all persons and for persons of Aboriginal and Torres Strait Islander origin.  The theoretical framework envisages individuals choosing whether to remain  on income support  depending on the value  of alternative opportunities. In this framework, the duration of an unemployment spell  depends mainly on how  job opportunities evolve over time. The empirical model  based  on this framework is a regression model  in which  the exit rate from unemployment is a function  of individual characteristics. This model  is estimated using  data  from the Department of Community and Family  Services Longitudinal Data Set.

The research has found  that the exit rate declines over the duration of an unemployment spell but that most individual characteristics have  a relatively small  effect. As there  are few indicators of long  expected duration apart  from having  already been  unemployed for a long time, policy intervention is best directed to the long-term unemployed. Using the data  on all Aborigines and Torres Strait Islanders, the research has found  that their exit rate is lower  than for other persons. However, in contrast  to most other indicators of labour  market  performance, the difference between Aborigines and Torres Strait Islanders and other persons in terms of exit rate must be regarded as quite  small.  Thus, the much  higher  unemployment rate of Aborigines and Torres Strait Islanders can be attributed to the much  larger  proportion of their number that becomes unemployed.

[ Return to Top   Return to Section ]

1  Introduction

This paper  reports  the results  from an analysis of the duration of spells  of receipt of unemployment-related income  support  in Australia.  This analysis has been  made  possible by the development of a longitudinal data  set of fortnightly  records  for income  support  clients  by the Department of Family  and Community Services (FaCS). The data  set is unique in that it is one of the few in which  there  are adequate numbers of Aboriginal and Torres Strait Islander persons to analyse whether their pattern  of spells  on income  support  differs from the patterns of the rest of the population. A central  theme  of the paper  is to investigate whether differences in the experiences of Aboriginal and Torres Strait Islander  peoples can help  us understand their relative exclusion from mainstream employment.

Although  the effect of the amount  of payment unemployed persons receive has been  the focus of most economic analysis, the data  used  here  do not permit  this effect to be identified. Rather, the focus is on the frequency and duration of spells  and the impact  of a limited  range  of personal characteristics such  as age,  sex,  marital  status,  number of children and country  of birth on the duration of spells  experienced.

The following two sections contain  an overview of the literature relating to the theory  and evidence of the duration of unemployment spells  and a discussion of the labour  market situation  facing  Indigenous Australians. The model  and data  used  in the analysis are then described. Descriptive analyses and results  from the estimation of hazard  or exit rate models for spells  of unemployment receipt  are presented in turn for the representative sample and for the full file of Aboriginal and Torres Strait Islander  client  records. The major findings  and avenues for future work  are summarised in the concluding section.

[ Return to Top   Return to Section ]

2  The duration  of unemployment benefit spells in theory

Whether  or not a person  claims  unemployment benefits  depends first on whether they  are eligible for benefits  and,  if they  are,  whether claiming benefits  is the best alternative for them. One of the principal eligibility criteria  is, of course, being  unemployed. There is a large  body of literature examining unemployment experience and other welfare payment experience independently of each  other,  as well  as the common  issue  of the utilisation of unemployment benefits. These  are largely based  on static models of participation, but attention  on entry  and exit rates is increasing with the availability of longitudinal data.

The incidence of unemployment among  different  subsections of the labour  force has been  well documented. Miller & Le (1999)  provide the most comprehensive recent  analysis of the characteristics of the unemployed in Australia.  Those with low  levels  of educational attainment, English language difficulties, disabilities, a previous history  of unemployment and those  from the younger and older  age  groups are all disproportionately represented in unemployment. Structural  changes, such  as in employment by skill  category, occupation, industry and spatial  region  can also contribute substantially to unemployment for those  from the ‘declining’ sectors.  Unemployment is extremely high  among  Aboriginal and Torres Strait Islander  peoples, as discussed further below. The dynamics of unemployment have  also been extensively studied, although not as thoroughly in Australia  as in other developed countries. Factors that increase the likelihood of being  in unemployment typically have  the effect of increasing spell  duration. There  is general consensus that unemployment displays ‘negative duration dependence’—a person’s chance of leaving unemployment declines the longer  she has been  unemployed(see, for example, Stromback  & Dockery  2000).

Studies  of welfare utilisation have  been  motivated by the so-called ‘welfare explosion’ witnessed in many  countries in recent  decades and by the associated issues  of welfare dependency and disincentive effects.  When  a person  becomes unemployed, their further time in unemployment is often modelled through  ‘search’  models. Critical parameters in these  models are the wages they might earn,  the rate at which  they receive offers, the cost to them of continuing to search  for employment, and their reservation wage (the lowest  acceptable wage). The specification is usually such that higher  unemployment benefits  reduce the cost of continued search  and/or increase the reservation wage, leading to longer  unemployment spells.

Other models follow  a standard discrete choice  model  where the individual maximises their benefit  in choosing between welfare payments and paid  work.  The disincentive effect of welfare payments has been  borne  out empirically in many  studies, most convincingly in countries in which  benefits  (or insurance) are not open-ended and exit rates from unemployment increase sharply as the expiration of benefit  eligibility approaches (see  Meyer 1990, Katz & Meyer  1990, Moffit 1992, Fortin & Lacroix 1997, Holmlund 1997).

The international findings  are not expected to translate directly to the Australian situation due  to differences in unemployment income  support. Most developed countries have  an unemployment insurance system  in addition to welfare benefits. Unemployment insurance usually provides earnings-related income  support  for a limited  period, while government-provided welfare supplements the unemployment insurance and provides for those  who  are not entitled  to insurance benefits. With a finite-duration unemployment insurance system, the exit rate from unemployment is strongly influenced by the time limit. Thus, the exit rate tends to rise sharply as this limit is approached.

Unemployment insurance can also affect the duration of welfare spells  by strengthening the incentive to return  to work  to re-establish eligibility for unemployment insurance. Among the Australian studies, Webster  (1999)  finds the replacement ratio to be one of the important  ‘shift variables’ for the Australian Beveridge curve,  with a 10 per cent increase in ratio of unemployment benefits  to minimum award wages increasing the level  of unemployment for a given  level  of vacancies by 5 per cent. 1 Stromback  & Dockery  (2000)  find that the receipt of income  support  markedly reduces the exit rate from unemployment (increases the expected duration of spells).

[ Return to Top   Return to Section ]

3  Labour market experience of Aborigines and Torres Strait Islanders

The degree of exclusion of Australia’s  Indigenous population from mainstream economic and labour  market  activity  is striking. This exclusion had force in law  until the 1967 national referendum granted full citizenship rights for Indigenous Australians. Only from the following 1971 Census  of Population and Housing  were  Indigenous people included in the data collection (Smith 1995, p. 29). Subsequently, a more  thorough statistical picture  of the contrast between the labour  market  fortunes  of Indigenous and non-Indigenous Australians has emerged. This evidence indicates that gains  in equality in legal  and administrative status for Aborigines have  not meant  equality in labour  market  outcomes.

Aboriginal and Torres Strait Islander  peoples comprise just 2 per cent of the Australian population. Hence,  none  of the mainstream labour  force or population surveys based  on representative sampling offer adequate sample sizes  to analyse the labour  market  experience of this group. Thus, most of the analysis of Indigenous population has relied  on Census  data.  Table 1 compares the labour  force status of Indigenous and non-Indigenous Australians in Census  data. It can be seen  that, according to the standard labour  market  constructs, Indigenous Australians have  faced  unemployment rates varying from two and a half to four times greater  than for other Australians over the past 30 years, despite a far lower  participation rate.

Table 1: Labour force status of Indigenous and non-Indigenous Australians, 1971, 1981, 1991 and 1996 Censuses
  Indigenous   Non-Indigenous
  1971 1981 1991 1996   1971 1981 1991 1996
Participation rate 46.1 47.3 53.5 52.7   58.7 61.3 63.2 62.0
Employment rate 42.0 35.7 37.0 40.7   57.7 57.6 55.8 56.4
Unemployment rate 4.1 11.6 30.8 22.7   1.0 3.7 11.4 9.0

Note: All calculations exclude those  who  did not indicate their labour  force state. Sources: Committee of Review  of Aboriginal Employment and Training  Programs  1985, Table  3.5 p. 46; Taylor  & Hunter 1998, Table  2, p. 13.

Even Census  data  contain  serious  technical and conceptual deficiencies for the purposes of describing Indigenous labour  market  experience. The true unemployment rate for Indigenous persons is likely to be higher  in absolute terms and relative to the wider  population due  to non-response bias.  The classification of persons as Aboriginal or Torres Strait Islander  relies  on self-identification. This has always presented limitations to the use of administrative data  from government departments for wider  statistical purposes, particularly as the rate of self- identification has been  sensitive to administrative rules.  The Census  counts  show  the Indigenous population growing 33 per cent from 265 371 in 1991 to 352 970 in 1996 (Taylor  & Hunter 1998, p. 10), clearly a result  of data  collection or identification effects.  The Indigenous population is actually estimated to be expanding at around 2 per cent a year,  twice  the rate of the wider  population.

The conceptual problems relate  to the appropriateness of the standard labour  force status constructs  for Aboriginal and Torres Strait Islander  persons. To the non-Indigenous population, favourable labour  market  status is largely synonymous with paid  employment in the market sector.  Many Aboriginal and Torres Strait Islanders live in remote  communities with very limited  integration with the mainstream economy, and thus opportunities for what  would normally be considered market  work  are largely non-existent. Further,  traditional Indigenous roles  relating to culture  or a hunting  and gathering existence, which  are clearly legitimate economic roles  in the context  of Indigenous societies, do not fit well  into our perceptions of a labour  market:

…non-Aboriginal ideas  of what  constitutes ‘real’  work,  ‘real  employment’ and the ‘real’ labour  force have  informed  official  assessments of Aboriginal labour  market  activities. The validity of associated labour  force indicators needs to be critically evaluated then, especially in respect to their conceptual limitations (Smith 1995, p. 35).

This has created a difficult  challenge for policy makers in addressing the labour  market outcomes of Indigenous Australians. Improving  outcomes involves an inevitable tension between the need  for integration into mainstream economic activities to achieve the goal  of equality in economic opportunity, and the goals  of self-determination and preservation of traditional society  and culture. This tension  is particularly relevant to the formulation of labour market  programs designed to improve  the labour  market  outcomes of Aborigines and Torres Strait Islanders. In 1985, the Committee of Review  of Aboriginal Employment and Training Programs  questioned whether ‘…the assumption implicit  in the programs that work  for wages and salaries is the most appropriate and practicable basis  for earning a livelihood’ was  valid  of all Aboriginal people. The Committee recommended that assistance measures respect the desire of Indigenous Australians for varying mixes  between their own  traditional lifestyle and appropriate components of the wider  market-based economy (pp.  5–6).

The Community Development Employment Project (CDEP) is a scheme that attempts  to chart such a delicate balance. It arose  originally in response to requests from communities for bulk payment of their unemployment benefit  entitlements. The CDEP allows remote  communities to administer the unemployment benefits, which  would otherwise have  been  payable to persons within  the community, as wages for designated activities. The community determines the work to be undertaken, though  guidelines provide that the work  should  be meaningful to the community and contribute toward  self-sufficiency. The program thus promotes self-determination as well  as partly  ameliorating the potential for the negative effects of welfare dependency.

The CDEP has grown  to be the largest  employment program  for Aboriginal and Torres Strait Islander  peoples, with an estimated 33 600 individuals ‘employed’ through  the 280 community organisations in 1999 (DEWRSB 1999).

Returning  to the problem of applying standard constructs  to measuring Indigenous labour  market status,  whether CDEP participants are treated  as employed, on labour  market  programs or on welfare thus has a significant impact  on any statistics  used  to describe the aggregate picture. According  to the 1994 National  Aboriginal and Torres Strait Islander  Survey  (NATSIS), the CDEP accounted for over one-quarter of Indigenous employment (Taylor  & Hunter 1997, p. 296).

Data from recent  censuses indicate that Indigenous employment has been  growing at a far higher  rate than overall  employment. Much of this growth  is clearly due  to expanded CDEP and other program  employment, and to positive employment policies within  the public  sector. According  to Altman & Taylor  (1995),  Indigenous employment in the private  sector  increased between 13 to 22 per cent between 1986 and 1991, depending on the measure used.  This compares to a 37 per cent increase in private  and public  sector  employment as measured from Census  data,  indicating a growing reliance on public  sector  employment among  Indigenous workers. However, the growth  in private  Indigenous employment has been  much  higher  than for the rest of the population (8.4 per cent),  suggesting some  real  gains  in relative employment status (Altman  & Taylor  1995, p. 78).

Notwithstanding the limited  statistics  and the problems of its interpretation, it is abundantly clear that Indigenous Australians who  wish  to participate in the mainstream labour  market  face severe  disadvantages. Beyond the static aggregate measures that estimate the extent  of this disadvantage, the data  problems outlined have  precluded further econometric analysis of the nature  of this disadvantage. The availability of the FaCS Longitudinal Data Set (LDS) offers a much-needed opportunity to undertake such  an analysis.

The capacity to distinguish between Aborigines and Torres Strait Islanders living  in metropolitan and rural areas  also permits  us to overcome two major hurdles in applying standard labour  market  constructs  analysis to data  on Indigenous peoples. First, the statistical impact  of the CDEP is removed. According  to the 1994 NATSIS, 80 per cent of all Indigenous employment in rural areas  was  in the public  sector,  particularly community organisations (Taylor  & Hunter 1997, p. 299).  Second, those  individuals living  in metropolitan areas  can be assumed to ‘have  chosen  or have  felt compelled to accept’  mainstream economic activity  as the basis  for their livelihood to a greater  extent  than those  who  remain  in isolated communities (Committee of Review  of Aboriginal Employment and Training  Programs  1985, p. 6). Thus, mainstream indicators of labour  market  status can be accepted as being  more relevant.

[ Return to Top   Return to Section ]

4  The model

In common  with most of the literature, it is assumed that the choice  to claim  or not to claim benefits  can be modelled as a discrete choice  problem in continuous time. Several approaches to the derivation of empirical models can be found  in the literature. What follows  should  be seen  as example of how  an empirical model  could  be rationalised using  a very  simple framework.

Consider  an individual who  is entitled  to a base  level  support  of b per unit of time if she has no other income. For the sake  of notational simplicity, the time subscript is suppressed. If she has an unearned income, a, the benefit  is reduced at a rate of q for every  dollar.  Thus, working zero hours she receives a benefit  g = b-θa <b. If she works  positive hours,  h>0, at an hourly  rate of w, the benefit  is subject  to the same  claw-back rate. Thus, the benefit  received is g = b-θ(a+wh). Ignoring  taxation, her net income, y, is given  by y = b+(1-θ)(a+wh)if bwh and y = (a+wh)  if θ(a+wh)³b, that is, if her income  exceeds the level  at which  the benefit  cuts out.

The individual’s kinked budget constraint is illustrated in Figure  1. The slope  of the segment AB reflects  the higher  effective  marginal rate of tax associated with the benefit  reducing with earned income. The point  where the benefit  cuts out is labelled B. This is at best a highly stylised description of the much  more  complex underlying reality. The budget constraint along AB may  contain  several kinks  and discontinuities that arise  from the precise rules  governing the determination of benefits.

In the case  of married  persons, the situation  is more complex. In general, each  person’s decision depends on the decision taken  by the spouse. The simplest approach to this problem takes  the labour  supply decision of a spouse as given  and assumes that the income  of the spouse is equivalent to own  unearned income. This approach is also broadly consistent with the Australian income-testing regime. Although  this regime makes  a distinction between own and spouse income, the two types  of income  have  a similar  effect on reducing benefit entitlements.

Figure 1:  The choice between  claiming or not claiming unemployment benefit

The choice between claiming or not claiming unemployment benefit

 

Conditional on new  information received the individual’s decision between claiming g >0 and not claiming depends on a comparison of the utilities  associated with the two alternatives.

The indirect  utility  from working can be represented by the function:

(1)   v = v(θ, a, h, x) + ε v

This function  has two parts.  In the deterministic part, the unobserved wage rate is replaced by its determinants x, a vector of individual characteristics. The random  term reflects  the arrival  of new information and it is this stochastic component of the utility  function  that drives  the model.

Conditional on remaining on income  support, the individual’s direct  utility  function  can be represented as:

(2)      u = u(b, θ, r, a, h, x) + εu

Let μd  = v(.)-u(.) denote  the difference between the deterministic components of the utility function  and ms  = ε u - εvthe difference between the stochastic components. The arrival  of new information is modelled as a Poisson  process with a constant  arrival  rate n. On the assumption that the error terms ε vand ε u are temporally independent, the instantaneous rate of exit from being  a benefit  recipient is given  by:

(3)      λ(z) = nProb[μs≤ μd(z)]= nF[μd(z)]

where z = (θ,b,a,x).

It is straightforward to show  that the level  of unearned income  (a) has a negative effect on the exit rate from income  support, providing that leisure is a normal  good.  The exit rate is also declining in the level  of benefit  (b) and in the claw-back rate.  The predicted effect of personal characteristics (x) depends on their effect on the wage rate.  Since  the wage rate is increasing in age,  at least  up to a point,  the expectation is that the exit rate would increase with increasing age.  Similarly, other characteristics that are associated with a higher  wage rate,  for example, being  married, having  children and being  native  rather  than foreign  born,  would also be expected to increase the exit rate.  However, these  predictions depend critically on the assumed constancy of the Poisson  parameter n. In general, n would depend on the same  personal characteristics as the wage rate.  Considering this largely precludes any  predictions about  the effect of personal characteristics. For example, while the wage rate is increasing in age,  the rate at which  new  information arrives  is most likely declining in age,  making the net effect ambiguous.

For the purpose of empirical estimation, the hazard  function  is assumed to be of the proportional form:

(4)    λ(z) = λ(t) exp(z’ β)

where λ0(t) is the baseline hazard  (that is, the hazard  when  z = 0) and β is a vector  of parameters to be estimated. In general, the elements of z might vary  over the duration of a spell.  However, in the empirical work  reported here  the initial  values of the z variables are used  unless  otherwise noted.

The empirical work uses a flexible (non-parametric) hazard function; the Cox proportional hazard model as implemented in the software used (SAS). Specific parametric forms are subject to a number of problems. First, the theoretical basis for a particular form is rarely clear. Second, the empirical hazard is often irregular and may not be well approximated by a simple parametric form. Thirdly, misspecification of the baseline hazard can lead to inconsistent estimates of β. Fourthly, the estimates are sensitive to unobserved heterogeneity if this is ignored or incorrectly specified. While the Cox model overcomes these problems, the proportionally assumption may be an invalid restriction. Our use of SAS for estimation entails further limitations. First, it was not possible to allow for unobserved heterogeneity, that persons with the same values of z may have a different hazard. Second, because SAS does not recover the underlying base hazard from the estimated parameters and the data, a useful piece of information is missing from the results.

[ Return to Top   Return to Section ]

5  The data

5.1   Measurement of spell duration

The LDS is a 1 per cent representative sample of the fortnightly  records  of income  support clients  of FaCS. The sample was  selected from those  in receipt of income  support  at a point in time, and then expanded by moving  both forwards and backwards in time through  the fortnightly  records  of FaCS clients.  For each  new  fortnight  added to the panel  (moving both forwards and backwards in time),  all newly appearing income  support  clients  had an equal (1 per cent)  chance of being  selected. Length bias (oversampling those  with long  spells) is thus present for those  selected at the initial  time point,  but this effect is diminished as the length  of the panel  increases. Once selected into the sample, clients’  records  over the full panel  are known, including ‘null’ records  for fortnights  when  they  were  not recipients. Persons’  activity  and circumstances while not in receipt of income  security are,  however, not known. This means, for example, that the destinations of persons who  exit spells  of income support  receipt are not known.

At the time this analysis was  undertaken, the panel  spanned the 105 fortnights  from 23 June 1995 to 18 June 1999, almost  four years. The data  cover  persons in receipt  of most types  of income  support  payments, but the focus of this paper  is on spells  of unemployment- related benefits. The two major types  of unemployment benefit  are known as Newstart Allowance and its counterpart for 18–21 year  olds,  Job Seeker Allowance. Also included are several variants  or predecessors of these  payment types:  Youth Training  Allowance, Newstart Mature Age Allowance and Mature Age Allowance. In what  follows, all the above  types  of benefits  are simply  referred  to as unemployment benefits  (UB).2

5.1   Measurement of spell duration

Spell  data  are derived from the series  of cross-section records. There  are two potential means of defining continuous spells  of benefit  receipt  for individuals. One is simply  as a series  of consecutive fortnightly  payment dates  for the one individual. The other is through  the ‘duration on current  payment’ variable in the data  set. In either  case,  a spell  is deemed to end in the fortnight  after which  there  is no record  for an individual for the following two fortnights.

Relying  on the duration of current  payment variable, the start of the spell  would be taken  to be the date  corresponding to the end  date  minus  the duration of current  payment. This method has the advantage that the start date  of all spells  observed to take  place  during  the observation period, and hence  the duration, would be measured. The previous discussions between FaCS and other researchers using  this data  set concluded that the duration of current payment should  not be used  in defining and measuring spells  for two main  reasons. First, the duration of current  payment is not always reset  to zero when  persons re-apply for UB after a short absence (for example, a few weeks of work). In such  cases, their case  might be treated as a re-application rather  than as a new  application, for administrative convenience. Second, in many  instances the duration on current  benefit  variable indicated a spell  of shorter  length  than the number of continuous payments. This should  not occur  and FaCS has confirmed problems with the integrity of the duration on current  payment variable.

The alternative method  is to take  as the start date  the first fortnight  there  is a record  for the individual in question. By using  this method,  the start date  of spells  that began before  the observation period  cannot  be determined. The break  of up to two fortnights  treats short interruptions as part of a current  spell  in line  with the administrative practice. This practice also suggests that a variation of this method  should  be used  to define  the end  of the spell.  This variation would take  a spell  to end  if either  there  are no records  for the individual for at least two fortnights  or if there  are two consecutive fortnights  for which  the relevant benefit  variables are zero.  However, this variation could  not be implemented, because correctly measured benefit  variables were  not available at the time this part of the analysis was  undertaken.

[ Return to Top   Return to Section ]

6  Patterns of utilisation of unemployment benefits

The data  set derived by the above  method  resulted in 45 147 spells  of unemployment benefit receipt by 26 747 individuals. Table  2 shows  that around  57 per cent of these  individuals experienced one spell  only  in the period, but that these  accounted for 34 per cent of all spells recorded. Thus, about  two-thirds of spells  in the sample are experienced by repeat  recipients, suggesting a significant degree of churning into and out of UB receipt.3 The maximum number of spells  experienced by any  one individual was  nine  (six  people). In the analysis that follows, no distinction is made  between first and repeat  spells.

Table 2: Number of spells by individuals
No. of spells No. of individuals experiencing that number of spells Total no. of spells
No. % No. %
1 15 385 57.5 15 385 34.1
2 6 794 25.4 25.4 25.4
3 2 880 10.8 8 640 19.1
4 1 128 4.2 4 512 10.0
5 404 1.5 2 020 4.5
6 113 0.4 678 1.5
7 26 0.1 182 0.4
8 11 0.0 88 0.2
9 6 0.0 54 0.1
Total 26 747 100.0 45 147 100.0

Around 19 per cent of the spells  (8 378) were  in existence at the beginning of the observation period. Of all spells  that began before  the beginning of the observation period, only  those  that are sufficiently long  will  be observed. These  ‘old’ spells  were  omitted  from the analysis. Effectively,  this means  that the sample comprises all spells  that began during  the observation period  June 1995 to July 1999. In another  paper  using  this data  set, Barret (2000)  also excludes ‘old’ spells.4A further 16 per cent of all spells  (7 364) were  right-censored—that is, ongoing at the end  of the observation period. For spells  that began and ended within  the observation period, the mean  duration was  14.2 fortnights,  just over half a year.  Clearly, this represents an underestimate of the actual  mean  duration because the excluded spells  (the continuing spells) are more  likely to be relatively long  ones.  For continuing spells, the mean  elapsed duration was 41.6 fortnights  (just over one and a half years). Including only  newly commencing spells, the mean  and quartiles from non-parametric estimation of the survival  function  are shown  in Table  3. This takes  into account the right-censoring of the continuing spells. The estimated mean  duration is around 23 fortnights.  The estimated survival  functions  are similar  for UB spells for males  and females, with males  on average having  slightly  longer  average spells.Likewise, those  with a partner  have  longer  spells  than singles, as have  older  persons and those born in a non-English-speaking country. These  estimates were  obtained using  the lifetable or actuarial method  for estimating the hazard  and survival  function.

Table 3: Non-parametric estimates for the survival function for UB spells – fortnights
  Mean 25% quantile 50% quantile 75% quantile
Males 23.5 5 12 29
Females 21.3 5 11 26
Single 21.8 5 11 26
With partner 25.2 5 13 32
Age 20–24 19.9 5 11 24
Age 25–44 23.6 5 12 29
Age 45–54 28.7 6 16 40
Country  of birth: Australia 22.2 5 11 27
Main Eng-spkg 21.1 5 11 26
Non-Eng-spkg 26.3 6 14 36
Persons 22.8 5 12 28

As discussed earlier, a deficiency with the data  is the lack  of information on the destination of exits.  It is possible, however, to identify  those  who  moved  onto other payment types.  To allow some  time for transition, we  investigated whether individuals were  receiving another  payment type  two fortnights  after the end  of a UB spell.  Spells  with end  dates  in the last or second  last fortnight  of the observation period  are censored in the sense  that we  cannot  determine whether or not the individual moved  on to a different  income  support  payment. For the other completed spells, 12.4 per cent terminated because the person  moved  onto another  form of income  support  payment. The remainder, presumably, gained work,  left the labour  market  or ceased to claim  unemployment benefit  for other reasons. The most common  categories of income  support  to move  into were  Disability  Support  Pension  (3.2 per cent),  Youth Allowance (2.4 per cent),  Age Pension  (1.4 per cent),  Sole Parent Pension  (1.1 per cent)  and various categories of Parenting Allowance (1.1 per cent).  The duration of spells  that did end  in a transition  to another  form of income  support  are considerably longer, with mean  duration of 23.6 fortnights,  compared to 14.8 for other completed spells.5

Figure  2 shows  the non-parametric estimates of the hazard  function  up to 52 fortnights  (two years). Comparing males  and females reveals that the shape of the hazard  function  is the same but that females have  a higher  hazard  than males  at most duration times.  In general, however, the differences are quite  small.  The hazard  initially rises  from just under  5 per cent to a peak in fortnight  4 or 5. Around 7.5 per cent of those  still receiving unemployment benefits  at this point  are no longer  on UB in the following fortnight.  From then on, the hazard  declines monotonically with duration. The graphs  of the hazard  function  by age,  country  and birth, and marital  status all display the same  unimodal hazard  declining after a peak  at four or five fortnights.

Figure 2:  Non-parametric estimates of the hazard function for UB spells

(a) by gender

Non-parametric estimates of the hazard function for UB spells - by gender

(b) by age group

Non-parametric estimates of the hazard function for UB spells - by age groups

(c) by marital status

Non-parametric estimates of the hazard function for UB spells - by marital status

(d) by country of birth

Non-parametric estimates of the hazard function for UB spells - by country of birth

[ Return to Top   Return to Section ]

7  Hazard function estimation

The effects of covariates on the hazard  function  are estimated using  the Cox proportional hazards model.  This method  allows for an arbitrary  baseline hazard, but the disadvantage is that the baseline hazard  is not estimated. The data  permit  only  a relatively small  number of covariates to be included in the estimation. Hence,  there  are omitted  variables that we  know from previous research to be important  in determining the rate of exit from spells  of unemployment. Of particular importance is educational attainment. Though  this is available as a variable in the data  set, it is not known or not specified in over 90 per cent of cases.

The personal characteristics that determine the wage rate include age,  sex,  country  of birth, race,  marital  status,  children, housing circumstances and unearned income. Also included are additional variables that may  proxy  the effect of unobservable characteristics on the wage rate. This set of variables includes whether and from whom  a person  rents the place  where they live, partner’s  activity  (receiving unemployment benefit  or not) and previous labour  market history  (previous receipt  of income  support  and repeat  spell). Some of these  variables could also be expected to affect the arrival  of new  information and thus the chance of finding  a job.

The identification of the benefit  effect depends on exogenous variations in the benefit  level over time or between individuals. The benefit  level  did not change over the period  in question apart  from cost of living  adjustments. The variation between individuals is entirely related to differences in individual circumstances, of which  marital  status,  number and age  of children and housing arrangements exert  the largest  influence. For that reason  it is uncertain whether the effect of the benefit  level  can be identified—separated from the influence of these  other characteristics. It should  be noted  that most studies  of duration of welfare spells  have nevertheless included a benefit  level  variable. We follow  this approach by including the potential benefit  rate measured by the sum of maximum basic  rate,  maximum rate assistance rate,  maximum pharmaceutical allowance rate,  remote  area  allowance and the add-on-amount. However, this variable is closely related to the determinants of the benefit  level—age, marital status,  number and age  of children and the amount  of rent paid.

Unearned income  is directly measured in the data  set. In addition to the direct  measurement, the amount  of unearned income  is also represented by a set of home  ownership variables. Home ownership constitutes ownership of an asset,  which  in most case  yields an imputed income  that is not included in the measured unearned income.

As discussed, partner’s  income  could  affect the choice  between claiming and not claiming, in the same  way  as unearned income  could  affect it. For each  individual with a partner,  there  is a matching record  of that person’s partner.  That record  includes the partners’  earned and unearned income. If the partner  is also in receipt  of income  support, the amount  of that support is also recorded.

The variables included in the data  set but not used  in the analysis include activity  (job search or other),  type  of breach  of the conditions for income  support  and the penalty imposed, if any. The principal reasons for omitting  these  are that it would be unreasonable to assume that these  variables are constant  during  a spell  and the proportional hazard  framework is not well suited  to analysis of their effect.

Several of the covariates used  in the analysis vary  over the duration of the spell.  In this paper, the estimation does  not take  this into account. Instead,  the initial  value  of a variable is taken  to represent the values of the variable for the duration of a spell.  This limitation  of the analysis is a consequence of the software used  (SAS). Consequently, several activity  status variables available in the data  set, such  as participation in training  or employment assistance programs, were  not included. These  variables are highly likely to be related to the duration of the spell, and hence  any  identified effect would be spurious unless  time-varying variables are constructed.

The proportional hazard  model  was  estimated from the 36 769 spells  that began during  the observation period. The results  are given  in Table  4 separately for persons with and without partners. Positive  coefficients indicate that the hazard  function  or exit rate is increasing with that variable, and the expected duration of spells  is decreasing. A negative coefficient implies a falling  hazard  and longer  expected duration of the UB spells. The magnitude of the coefficients gives  the approximate per cent effect of a unit change in the variable. Taking  age as an example, if the hazard  for a person  in the omitted  age  group  (25–44)  is 5 per cent per fortnight,  the hazard  for a person  aged  15–19 with a coefficient of 0.2463 is exp(ln 0.05+0.2463)*100=6.4 per cent – that is, 28 per cent higher.

Table 4a: Estimates of the effect of individual characteristics on the hazard function: single persons
  (1) No benefit variables (2) With benefit variables
Age:
15 to 19 0.2463 *** 0.2134 ***
20 to 24 0.1325 *** 0.1297 ***
25 to 44
45 to 54 -0.3187 *** -0.3191 ***
55 to 59 -0.4813 *** -0.4804 ***
60 to 64 -0.6132 *** -0.6388 ***
Born:
Australia
Main Eng-spkg country 0.1154 *** 0.1184 ***
Non-Eng-spkg country -0.0256 -0.0245
Female 0.0423 *** 0.0418 ***
Indigenous -0.1824 *** -0.1750 ***
Home Owner -0.1223 *** -0.0835 **
Purchaser 0.0600 0.0836
Lives elsewhere 0.1073 0.1326
Other owner -0.4659 ** -0.4143 **
Non-owner    
Rents:
Private -0.0009 0.0647 **
Government -0.4567 *** -0.4030 ***
Board/lodging -0.0296 0.0143
Site/other fees -0.1856 ** -0.1284
No rent paid -0.0952 *** -0.0503 **
Free board/lodging -0.1379 -0.2430
Non-renter
Sharer -0.4399 *** -0.4399 ***
Has child aged:
0 to 6 0.1254 0.1040
7 to 12 0.1331 0.1032
13 to 15 -0.2923 -0.2794
over 15 -0.4769 *** -0.5101 ***
no depdt. children
Partner receives UB
Benefit ($ eligible) -0.0006 ***
Benefit squared 2.9x10-7 **
Unearned income 0.0013 *** 0.0012 ***
(Unearn inc.)2 -8.3x10-9 3.4x10-8
Partner income
(Partner inc.)2
Rent paid -0.0012 *** -0.0012 ***
(Rent paid)2 2.0x10-6 *** 1.8x10-6 ***
Previous IS receipt -0.1084 *** -0.1388 ***
Repeat spell -0.1034 *** -0.1340 ***
N (spells) 27 027 27 027
Censored spells 4 525 4 525
Model χ2 (-2 Log L) 1 650 *** 1 650 ***
Degrees of freedom 30 32

Notes:
***, ** and * denote significance at the 1, 5 and 10 per cent levels, respectively.
—: denotes ‘omitted’

                        

Table 4b: Estimates of the effect of individual characteristics on the hazard function: persons with partners
  (1) No benefit  or partner  income  variables 2) With benefit variables (3) With benefit  and partner  income  variables
Age:
15 to 19 0.2823 *** 0.2980 *** 0.3018 ***
20 to 24 0.1346 *** 0.1360 *** 0.1306 ***
25 to 44
45 ot 54 -0.1961 *** -0.1992 *** -0.2002 ***
55 to 59 -0.6044 *** -0.6076 *** -0.6022 ***
60 to 64 -0.6693 *** -0.6827 *** -0.6739 ***
Born:
Australia
Main Eng-spkg country 0.1105 *** 0.1096 *** 0.1157 ***
Non-Eng-spkg country -0.1785 *** -0.1773 *** -0.1667 ***
Female -0.0208 -0.0249 -0.0286
Indigenous -0.2692 *** -0.2705 *** -0.2578 ***
Home owner 0.0532 0.0765 0.0715
Purchaser 0.2209 *** 0.2386 *** 0.2316 ***
Lives elsewhere 0.3972 *** 0.4047 *** 0.4101 ***
Other owner -0.0412 -0.0198 -0.0260
Non-owner
Rents:
private -0.0315 -0.0052 -0.0134
Government -0.2555 *** -0.2283 *** -0.2210 ***
Board/lodging -0.0216 0.0035 0.0094
Site/other fees -0.0376 -0.0130 -0.0351
No rent paid -0.0816 -0.0460 -0.0435
Free board/lodging 1.2585 1.1983 1.2246
Non-renter
Sharer -0.9168 *** -0.9184 *** -0.9184 ***
Has child aged:
0 to 6 -0.0657 ** -0.0673 ** -0.0539 *
7 to 12 -0.0354 -0.0374 -0.0310
13 to 15 -0.0638 * -0.0664 * -0.0681 *
over 15 -0.0292 -0.0453 -0.0652
no depdt. children
Partner receives UB -0.0661 ** -0.0634 * -0.0473
Benefit  ($ eligible) -0.0005  *** -0.0004  **
(Benefit)2 4.2x10-7 4.0x10-7
Unearned income 3.6x10-7 8.0x10-7 1.1x10-6
(Unearn inc.)2 2.8x10-8 2.9x10-8 2.9x10-8
Partner  income 0.0003 ***
(Partner  inc.)2 -1.2x10-7  ***
Rent paid -0.0004  *** -0.0004  *** -0.0004  ***
(Rent paid)2 9.3x10-7 *** 9.1x10-7 *** 9.4x10-7 ***
Previous  IS receipt -0.0058 -0.0173 -0.0196
Repeat  spell -0.1099  *** -0.1247  *** -0.1185  ***
N (spells) 9 742 9 742 9 741
Censored spells 2 071 2 071 2 071
Model χ2 (-2 Log L) 481 *** 496 *** 539 ***
Degrees of freedom 31 33 35

Notes:
***, ** and * denote  significance at the 1, 5 and 10 per cent levels, respectively.
—: denotes ‘omitted’

Considering first the results  for single  persons in Table  4a, the hazard  for young people is higher  than the omitted  prime  age  group  (25–44).  Persons  in the 45–54 and 55–59 and 60–64 groups  are found  to have  a lower  hazard. In fact, the exit rate decreases linearly with age across  these  ranges, rather  than the ‘U’ shaped relationship often observed. Persons  born overseas in a main  English-speaking country  have  a higher  exit rate than their Australian-born counterparts. However, there  is no difference between Australian born and those  born in a non-English  speaking country. The male-female difference in the hazard  is quite  small,  with females having  a slightly  higher  exit rate than males. A larger  difference is found  between the Indigenous and non-Indigenous persons. This is taken  up in detail  in a following section. Several of the home  ownership and rental  status indicators have  a large  and significant effect on the hazard. Sharer  status in particular is associated with a much  lower  hazard  than persons in other living  arrangements.

The continuous variables are included as second  order  polynomials to allow  for any  non- linearities. The higher  the amount  of rent paid  rent the lower  is the hazard, but with a positive quadratic term this effect becomes smaller at higher  levels  of rent. In this and most other cases, however, the coefficient on the quadratic term is small  so the effect is approximately linear  in the relevant range.

In theory,  the exit rate should  be increasing in unearned income  on the grounds that leisure is a normal  good.  The higher  the unearned income  the more  leisure a person  can afford and the more  likely it is that they  will  choose  to work  few hours  and claim  UB. The estimated coefficient on unearned income, however, is positive, implying that the exit rate is increasing in unearned income. To reconcile the theoretical predictions with the estimated effect, a possible explanation is that the unearned income  variable also captures the effect of other omitted  variables that have  a positive effect on the exit rate and,  when  combined, they dominate the predicted negative effect. Including a measure of the potential level  of benefits as in column  2 of the table  reveals that the benefit  level  reduces the hazard  but leaves the effect of other individual characteristics unchanged. The point  estimates of the effect of the benefit  amount  imply  that every  $100 reduces the hazard  by about  6 per cent.

Comparison of the estimates pertaining to persons with partners  to single  persons reveals both similarities and differences in the impact  of individual characteristics. The age  effect is similar for both groups  but the effect of birthplace and gender differ. There  are more  marked differences in the estimated coefficients on the variables representing type  of home  ownership and rental  status.  For both groups, however, living  in government rented  accommodation is associated with a much  lower  hazard. The effect of children differs even  more  between the two groups. For single  persons, having  a child  over 15 is associated with a significantly lower hazard. In the case  of persons with a partner,  having  a child  under  the age  of six is associated with a significantly lower  hazard.

As previously stated,  home  ownership represents the ownership of an asset  that in most cases yields only  imputed rather  than actual  income. Thus, home  ownership would be predicted to have  the same  effect as unearned income  – those  who  own  their home  can afford more leisure and are therefore more  likely to choose  few hours  and claim  UB. In the case  of single  persons, some  of the estimated coefficients on the home  ownership variables correspond to this prediction. It is difficult  to explain, however, why  the ‘Other owner’  category should  have  the largest  negative effect on the exit rate.  This category includes several types  on joint ownership that one would expect to be associated with lower  asset  values than the basic  ‘Home owner’ category. In the case  of persons with a partner,  home  ownership is associated with higher  exit rate.  The same  result  was  obtained for Indigenous persons with a partner,  as will  be shown  in the next section.  As before,  the most likely explanation is omitted  variables. Home ownership is associated with omitted  and possibly unobservable variables that tend to increase the exit rate.  Evidently,  these  effects dominate and result  in a positive coefficient on the home ownership variables. That single  persons are different  would then have  to be attributed to home ownership being  much  less common  among  single  persons. Single  persons are much younger and many  of those  who  do not yet own  their own  home  will  do so later in life. Thus, many  of those  not owning a home  at an early  age  do not possess the unfavorable characteristics that older  non-home owners do.

Living in government rental  accommodation is strongly associated with lower  individual and neighbourhood socioeconomic status.  The effect of this is highly significant and is to prolong the duration of spell  on benefit.  Living in shared  accommodation has an even  stronger  effect of the same  sign,  and is particularly strong  for persons with partners. Across all renters,  the duration of spells  increases with the amount  of rent paid.

If a person’s partner  is also in receipt of unemployment benefits  in their own  right,  their rate of exit from unemployment is lower, suggesting a concentration of unemployment within family  units.  For single  persons only,  the duration of spells  falls with the average fortnightly amount  of unearned income  received and increases if they  were  in receipt  of another  form of income  support  prior to moving  onto unemployment benefits.

Repeat  spells  are identified by the individual having  experienced a previous spell  during  the four-year observation period. This implies that many  spells  that began early  during  the period are not identified as repeat  spells  even  though  this should  really have  been  the case.  This censoring of repeat  spells  is not taken  into account in the estimation. The coefficient implies a reduction of the hazard  but, by definition, the average duration of spells  must be relatively short for a person  who  experiences multiple spells  within  the four-year observation period. Therefore, the coefficient is likely to understate the effect that would be observed given  a longer  panel. Further, repeat  spells  may occur for contrasting reasons. Some persons may experience repeat spells  because of a tendency to move in and out of employment rapidly. We would expect spells for these  persons to be short. Repeat  spells  may also be expected for the ‘hardcore unemployed’, but in this case  we would expect the spells  to be of long relative duration.

Column  2 of Table  4b adds  the amount  of benefit  to the model.  As was  the case  for single persons, the amount  of benefit  is estimated to have  a small  but significant negative effect on the hazard. As was  also the case  for singles, the inclusion of the benefit  variables does  not have  much  of an effect on the estimates of the other variables.

The third column  of Table  4b reports  the estimates obtained when  partner’s  income  was included as well  as the amount  of benefit.  Partner  income  is defined as the sum of partner’s earned and unearned income. The estimated coefficient is highly significant and the positive coefficient implies that the larger  is partner’s  income  the higher  is the hazard. In theory, partner’s  income  should  be equivalent to unearned income. Both are expected to increase the demand for leisure and thus to result  in a lower  hazard. However, the estimates contradict this prediction.

In terms of the magnitudes, the most important  determinants of the duration of UB spells appear to be living  in shared or governmental rental  accommodation and age.  As is evident from the previous discussion, however, the interpretation of some  of the results  is highly speculative and relies  on bold and untested assumptions. In the main,  one can only  note the effect of a particular variable and speculate about  the likely reason  for the direction and magnitude of the effect.

These  limitations are mainly a consequence of three  related factors.  The first is the omission of many  variables found  to be important  determinants in previous research. Secondly, instead  of these  variables, many  idiosyncratic variables that are specific  to this data  set are included to account for the variation between individuals. As stated,  this makes  the interpretation of the findings  somewhat speculative. Furthermore, and this is the third factor, it raises  doubts  that the included variables account for all the variation between individuals. Regrettably, it was  not possible to allow  for any  remaining and unobservable heterogeneity in the estimation.

7.1   Comparison with other estimates

The results  can be compared to a previous analysis of the hazard  rate for spells  of unemployment using  the Australian Bureaus of Statistics’  1994–97 Survey  of Employment and Unemployment Patterns  (SEUP) (Stromback and Dockery  2000).  The SEUP data  measure the length  of job search  spells  but not all job search  spells  result  in a spell  of unemployment benefit.6 Some persons are not eligible for unemployment benefit  and,  of those  who  are,  not all apply. Thus, spells  of UB can be regarded as a subset  of spells  of job search. Another difference between the data  sets is that the SEUP data  are a stock  sample dominated by persons classified as jobseekers when  the sample was  drawn. Since  most jobseekers were unemployed when  the sample was  drawn, the SEUP sample is subject  to length  bias—those with long  duration of unemployment are over-sampled.

The magnitude of the hazard  rate estimated from SEUP data  is slightly  lower  than in the LDS data,  with persons leaving spells  of job search  at a rate of around  6 per cent per fortnight  in the first month,  falling  to below 3 per cent per fortnight  after a year.  However, even  a small difference in the hazard  rate translates into quite  a large  difference in the survival  function. Thus, job search  spells  are very  much  longer  than spells  of unemployment benefit.  The SEUP median spell  is estimated to be 17 fortnights  compared to 12 fortnights  in the LDS data (see  Table  3). The different  sampling schemes provide one reason  for this difference. The SEUP job search  spells  are length-biased. They reflect  the length  of spells  of currently unemployed persons while the LDS data  refer to persons who  begin  a spell  of unemployment. However, apart  from this reason, non-UB job search  spells  may  well  be longer  than UB job search  spells. As previously discussed, many  UB spells  end  by transfer  to another  type  of income  support  rather  than by finding  a job. In addition, UB spells  can end  by the loss of eligibility due  to an increase in partner’s  income. Job search  spells, however, end  in either  of two ways: by finding  a job or giving  up the search.

The shape of the hazard  is also different  in the two data sets. The SEUP data showed that the hazard  is monotonically declining from the first month.  In contrast,  the LDS data indicate an increasing hazard  up to 4–6 fortnights  and declining monotonically thereafter. The difference could  arise  because the short spells  in LDS are atypical of all job search  spells. Persons  only claim unemployment benefit  if they expect to be out of work  for some  time, meaning that there will  be a considerable number of very short spells  of job that do not appear as spells  of unemployment benefit.  Thus, in the LDS data few spells  are observed to end after a short period.

The effects of personal characteristics are estimated to be very  similar  in both data  sets. The exit rate is decreasing in age,  the difference between males  and females is small,  the effect of country  of birth is the same,  the labour  market  status of a spouse has a large  effect (in SEUP a working spouse increases the exit rate—in  LDS a spouse on unemployment benefit  decreases the exit rate)  and previous unemployment/income support  is associated with a lower  exit rate. In many  cases, the magnitude of the estimated effect of these  variables is also similar.  The SEUP data  also show  that the exit rate is lower  for those  receiving income  support  and unemployment-related income  support  in particular.

Another pertinent comparison is with the analysis of the duration on single  parent  pension (SPP) undertaken by Barrett (2000)  using  the LDS. In this case,  the comparison of greatest interest  is between single  persons on unemployment benefit  and single  parent  pension beneficiaries. This comparison reveals that SPP recipients have  (i) a lower  hazard, which  is (ii)  monotonically declining at a slow  rate.  That the hazard  is lower, and the duration longer, is not surprising, since  these  persons are under  no obligation to engage in job search. Likewise, that the hazard  does  not decline much  over the duration of a spell  could  be attributed to differences in exit destinations of UB and SPP recipients. If it were  the chance of finding  a job that decreases over the duration of a spell  (due  to a scarring effect),  this effect would be less pronounced for SPP recipients since  their spells  are more likely to end  for other reasons (such as re-partnering).

More difficult  to rationalise are the significant differences in the effects of individual characteristics on the exit rates of the two types  of beneficiaries. Although  some  of the estimated effects are of the same  sign and magnitude (for example, age,  unearned income, home ownership variables) in most cases  the estimated signs  differ. Thus, sex,  age  of youngest child, Aboriginal status,  repeat  spell  and rental  status have  a different  impact.  In the case  of Aborigines and Torres Strait Islanders, those on UB have  a 16 per cent lower  exit rate and those on SPP an 18 per cent higher  exit rate than other persons. As regards the effect of the benefit  level,  Barrett’s estimates range  from a 3 to 13 per cent lower  hazard  for each  $100 of benefit.  Our estimates all fall at the lower  end of this range  and to that extent  the findings  are similar.  However, in the case of SPP, the estimates of the benefit  effect are at best weakly significant.

[ Return to Top   Return to Section ]

8  Aborigines and Torres Strait Islanders

Hitherto,  persons of Aboriginal and Torres Strait Islander  origin  have  been  represented by a dummy  variable implying an exit rate proportional to that of non-Aborigines. Based  on this representation it was  found,  as expected, that the exit rate for Aborigines and Torres Strait Islanders was  significantly lower. To investigate this result  in more detail,  this section  reports results  of a separate analysis of Aborigines and Torres Strait Islanders. This is based  on a census file of all fortnightly  records  for Indigenous clients  receiving unemployment-related benefits. The file was  generated by the FaCS for the same  observation period  as in the preceding analysis. Although  the range  of explanatory variables in the data  is very  restrictive, the number of observations for Aboriginal and Torres Strait Islanders provides a unique opportunity for the analysis of labour  market  dynamics for this group. The Indigenous census file generated a sample of 116 147 spells  of UB receipt  experienced by nearly 65 000 individuals.7

Table 5: Number of spells by individuals, Aborigines and Torres Strait Islanders
No. of spells

No. of individuals experiencing Total no. of spells that number of spells

Total no. of spells
1 32 124 50.1 32 124 27.7
2 18 725 29.2 37 450 32.2
3 8 492 13.2 25 476 21.9
4 3 416 5.3 13 664 11.8
5 1 030 1.6 5 150 4.4
6 272 0.4 1 632 1.4
7 66 0.1 462 0.4
8 20 0.0 160 0.1
9 1 0.0 9 0.0
10 2 0.0 20 0.0
Total 64 148 100.0 116 147 100.0

Comparison of Tables  5 and 2 shows  that the usage of the UB system  is remarkably similar  for Indigenous and non-Indigenous persons, although just over 7 per cent more Aboriginal and Torres Strait Islander  people experienced repeat  spells. Eleven per cent of the Indigenous spells were  left-censored and 22 per cent right-censored.

As shown  in Table  6, however, Indigenous persons are much  more  likely to experience  UB spells  than non-Indigenous persons are.  Over the four-year period  covered by the data, 28 per cent of Indigenous persons of working age  experienced an UB spell  compared to 18 per cent for all persons.

Table 6: Incidence of unemployment benefit spells among the non-Indigenous and Indigenous population: June 1995–June 1999
  Indigenous persons All Persons
Working-age population (15 and over) 231 660a 14 600 000 b
Number of persons with a UB spell 64 148 2 674 700
Proportion of persons with a UB spell 0.28 0.18

Notes:
a: figure for 1996, taken from Taylor & Hunter 1998, Table 1
b: average figure for 1995 to 1998 ABS time series data, 3201.0 Table 9.

Due to the problems with the application of standard labour  force status concepts to Aborigines and Torres Strait Islanders living  in remote  areas  and of the impact  of CDEP participation on the statistics,  results  are reported separately for Indigenous persons living  in the major metropolitan areas. This is determined via the postcode variable at the commencement of the spell.  This approach is by no means  a definitive solution. However, it does  allow  us to single  out a group for which  the data are free of the impact  of the CDEP and who  face the same  external labour market  conditions as most other Australians. Thus, a more legitimate comparison can be made between the results  for this group  and those for the 1 per cent sample of the general population. For just under  one-quarter of the Indigenous sample (27 779 spells), the individual’s initial address was  identified as being  in a metropolitan area.  Figure  3 compares the non-parametrically estimated hazards for Indigenous persons living  in metropolitan and non-metropolitan regions with those for non-Indigenous UB recipients.

Figure 3:  Non-parametric estimation of the hazard function for UB spells; Indigenous persons in metropolitan and non-metropolitan area (FaCS one per cent sample and Indigenous census file)

Non-parametric estimation of the hazard function for UB spells; Indigenous persons in metropolitan and non-metropolitan area (FaCS one per cent sample and Indigenous census file)

For all Aboriginal and Torres Strait Islanders, the mean  duration of spells  is 26.2 fortnights,  3.4 fortnights  longer  than for the one per cent sample. It is lower  for Aboriginal and Torres Strait Islanders in metropolitan areas, but still higher  than for UB recipients as a whole. The Indigenous file included only  records  for unemployment-related benefits, so it is not possible to investigate the frequency or patterns  of transition  from UB spells  into other payment types. However, the means  for the variable ‘Previous IS receipt’  indicate that Aborigines and Torres Strait Islanders are twice  as likely to have  entered a spell  of UB receipt  straight  from an episode of income  support  receipt. It would seem  likely that they  are also more  likely to move onto other income  support  categories from spells  of UB receipt.

Table 7: Non-parametric estimates of the survival function for UB spells—fortnights; Aboriginal and Torres Strait Islander clients
All Indigenous Mean 25% quantile 50% quantile 75% quantile
Males 26.3 6 14 34
Females 25.6 5 13 33
Persons 26.2 6 14 34
Metro
Males 24.3 5 13 31
Females 23.0 5 12 30
Persons 23.9 5 13 30

Table  8 reports  the results  for estimation of the proportional hazard  model  for UB spells  by Aboriginal and Torres Strait Islander  clients.  Separate models were  estimated for those  living  in metropolitan and non-metropolitan areas  and for single  persons and persons with partners. Table  8a shows  that the vast majority  of Indigenous persons living  in metropolitan areas  are single  persons (87 per cent).  Thus, it seems  appropriate to focus mainly on the single  group. The signs  and order  of magnitude of most of the coefficients are similar  to those  reported for all persons. The main  differences are that among  Indigenous single  persons, older  age,  sex, type of home  ownership and rental  status have  no significant effect on the hazard, with the exception of sharer  status.  For those  with partners, however, type  of home  ownership matters, but the presence of children has no effect.

Although  non-metropolitan Indigenous persons face a very  different  labour  market,  the estimates of the effect of individual characteristics are very  similar  to the estimates for those  in metropolitan areas. One of the few differences that may  be noted  is the significant effect of home  ownership variables for single  persons. Among all Indigenous persons, owning or purchasing a home  is associated with a larger  hazard. This may  reflect  that home  ownership among  Indigenous persons is limited  to a far smaller proportion of the population and hence is more strongly correlated with other favourable labour  market  characteristics. The effect of remote  communities may  have  some  influence here,  but the result  is particularly strong  for married  Indigenous persons within  the major metropolitan regions. In addition, in contrast  to the results  for the wider  population, we  find no negative effect of living  in government rental accommodation for Aborigines and Torres Strait Islanders, but living  in shared accommodation is associated with a longer duration of spells.

Most of the other parameters have  the same  sign  and are of the same  order  of magnitude as for the population as a whole. This suggests that the lower  hazard  for Aborigines and Torres Strait Islanders is due  to their having  characteristics which  are associated with a longer  benefit duration. However, inspection of the average values of the covariates reveals that the differences between the two groups are usually quite  small.  Thus, the observed difference in the hazard  must be due  to a number of small  differences in average values of the covariates and their estimated effect.

Table 8a: Estimates of the effect of personal characteristics on the hazard function: Aboriginal and Torres Strait Islanders—metropolitan clients
  With partner Single
  (1) No benefit variables (2) With benefit variables (3) No benefit variables (4) With benefit variables
Age:
15 to 19 0.4336 *** 0.4600 *** 0.4025 *** 0.3851 ***
20 to 24 0.1488 *** 0.1647 *** 0.2011 *** 0.1977 ***
25 to 44
45 to 54 -0.1830 ** -0.1749 ** -0.1881 *** -0.1864 ***
55 to 59 -0.5264 *** -0.5216 *** -0.0541 -0.0392
60 to 64 0.1890 0.0840 0.4414 ** 0.4119 *
Born:
Australia
Main Eng-spkg country -0.5430 -0.5941 0.1816 0.1818
Non-Eng-spkg country 0.0077 -4.1x10-5 0.0377 0.0661
Female 0.0075 -0.0022 -0.0058 -0.0098
Home owner 0.3234 *** 0.3149 *** 0.0965 0.1078
Purchaser 0.5503 *** 0.5457 *** 0.2313 0.2602
Lives elsewhere 1.5219 *** 1.5434 *** 0.4286 0.4552
Other owner -0.0883 -0.1387 0.0445 0.0730
Non-owner
Rents:
Private 0.1166 0.1238 -0.0215 0.0310
Government 0.0577 0.0571 -0.0429 -0.0239
Board/lodging 0.1117 0.1101 -0.0721 *** -0.0322
Site/other fees -0.0181 0.0489 -0.0640 -0.0160
No rent paid 0.1526 * 0.1598 * -0.0497 * -0.0060
Free board/lodging 0.0811 -0.0481
Non-renter
Sharer -0.6744 *** -0.6627 *** -0.3387 *** -0.3597 ***
Has child aged:
0 to 6 -0.0117 -0.0126 0.3428 *** 0.3292 ***
7 to 12 -0.1049 -0.0990 0.1760 0.1568
13 to 15 0.0042 0.0057 -0.1389 -0.1224
over 15 0.2033 0.1787 -0.3531 -0.3902
no depdt. children
Partner receives UB -0.0814 -0.0749
Benefit ($ eligible) 0.0001 -0.0009 ***
(Benefit)2 -1.4x10-6 ** 6.2x10-7 ***
Unearned income 0.0012 *** 0.0012 *** 0.0006 * 0.0006 *
(Unearn inc.)2 1.5x10-9 1.4x10-9 8.3x10-7 6.9x10-7
Rent paid 0.0006 *** 0.0005 ** -0.0007 *** -0.0007 ***
(Rent paid)2 1.0x10-6* 1.1x10-6* 1.6x10-6*** 1.5x10-6***
Previous IS receipt -0.1721 *** -0.2187 *** -0.1897 *** -0.1883 ***
Repeat spell -0.1894 *** -0.2270 *** -0.0931 *** -0.1320 ***
N (spells) 3 261 3 261 21 791 21 791
Censored spells 772 772 4 443 4 443
Model χ2 (-2 Log L) 155 *** 175 *** 1 074 *** 1 219 ***
Degrees of freedom 29 31 29 31

Notes:
***, ** and * denote significance at the 1, 5 and 10 per cent levels, respectively.
—: denotes ‘omitted’

Table 8b: Estimates of the effect of personal characteristics on the hazard function: Aborigines and Torres Strait Islanders—non-metropolitan clients
  With partner Single
  (1) No benefit variables (2) With benefit variables (3) No benefit variables (4) With benefit variables
Age:
15 to 19 0.3108 *** 0.3268 *** 0.4268 *** 0.3869 ***
20 to 24 0.0928 *** 0.0990 *** 0.1387 *** 0.1360 ***
25 to 44
45 to 54 -0.1453 *** -0.1435 *** -0.2288 *** -0.2308 ***
55 to 59 -0.2119 *** -0.2115 *** -0.0970 ** -0.1032 **
60 to 64 0.0815 0.0812 0.3127 *** 0.2775 ***
Born:
Australia
Main Eng-spkg country 0.4666 0.4410 -0.0439 -0.0599
Non-Eng-spkg country -0.2245 ** -0.1902 * -0.0620 -0.0380
Female 0.0301 0.0307 -0.0429 *** -0.0416 ***
Home owner 0.2371 *** 0.2446 *** 0.1232 ** 0.1549 ***
Purchaser 0.3977 *** 0.3839 *** 0.4407 *** 0.4385 ***
Lives elsewhere 0.0321 0.0577 0.4140 *** 0.4381 ***
Other owner 0.4868 0.4749 -0.3645 -0.2835
Non-owner
Rents:
Private -0.0198 0.0104 -0.0507 *** 0.0084
Government 0.0546 * 0.0781 ** -0.0183 0.0030
Board/lodging 0.0404 0.0675 * -0.0747 *** -0.0263 *
Site/other fees -0.0143 0.0335 0.0162 0.0818
No rent paid -0.1101 *** -0.0641 ** -0.0863 *** -0.0267 *
Free board/lodging -14.2013 -14.2839 0.3697 0.2440
Non-renter
Sharer -0.4869 *** -0.4758 *** -0.3654 *** -0.3837 ***
Has child aged:
0 to 6 -0.0644 *** -0.0563 ** 0.1938 *** 0.2205 ***
7 to 12 -0.1001 *** -0.0955 *** 0.0752 0.0868
13 to 15 -0.0515 * -0.0501 * -0.1387 * -0.1395 *
over 15 0.0092 -0.0295 -0.1584 -0.2234
no depdt. children
Partner receives UB -0.2408 *** -0.2393 ***
Benefit ($ eligible) -0.0002 -0.0008 ***
(Benefit)2 -5.4x10-7 ** 3.1x10-7 **
Unearned income 0.0014 ** 0.0013 ** 0.0004 0.0003
(Unearn inc.)2 1.3x10-6 ** 1.2x10-6 ** 4.6x10-11 4.3x10-1
Rent paid 0.0000 -0.0001 -0.0011 *** -0.0011 ***
(Rent paid)2 1.6x10-6*** 1.2x10-6*** 3.2x10-6*** 3.2x10-6***
Previous IS receipt -0.2019 *** -0.2305 *** -0.1888 *** -0.2010 ***
Repeat spell -0.1497 *** -0.1815 *** -0.0976 *** -0.1380 ***
N (spells) 18 754 18 754 59 959 59 959
Censored spells 5264 5264 13 491 13 491
Model χ2 (-2 Log L) 327 *** 691 *** 3 071 *** 3 491 ***
Degrees of freedom 30 32 29 31

Notes: ***, ** and * denote significance at the 1, 5 and 10 per cent levels, respectively.
—: denotes ‘omitted’

[ Return to Top   Return to Section ]

9  Conclusions

The LDS has provided a unique opportunity for researchers to examine the temporal nature  of the utilisation of unemployment benefits  in Australia.  The analysis here  provides a summary of the structure  of spells  of unemployment benefit.  In the four-year period  for which  the 1 per cent sample is available, we  find that the average duration of a spell  on unemployment benefits  is almost  11 months.  Two-thirds  of the spells  observed are accounted for by clients  who experienced repeat  spells. Thus, the average stay on unemployment benefits  in Australia  is long and there  is a high  degree of churning – exit and re-entry to unemployment benefits. Given that the spells  of repeat  clients  are of longer  average duration, it can be surmised that a large  proportion of time on unemployment benefits  is experienced by a small  proportion of the population.

The hazard  or exit rate from UB spells  initially rises  to peak  in the fourth or fifth fortnight,  and then falls monotonically with duration. This finding  does  not necessarily reflect  duration dependence, but can arise  because of heterogeneity among  persons. That is, persons with a high  but constant  probability of exiting UB leave  after a short time, while those  with a low probability leave  after a long  time. In contrast,  estimates of the hazard  from spells  of job search  over a similar  period  show  a uniformly declining hazard. We believe this is likely to reflect  that many  persons do not claim  benefit  if they  expect to be unemployed for a short period. This may  arise  because of administrative rules,  the cost of applying for benefits  or the desire  to avoid  any  possible stigma  attached with receiving welfare payments, factors that may outweigh the gains  from benefit  receipt  for short spells.

The factors estimated to have  the greatest impact  on the duration of UB spells  are accommodation status and age.  However, the results  for individual covariates are difficult  to interpret  since  there  are a number of important  omitted  variables that cannot  be observed. Hence,  many  of the estimated effects may  not reflect  a direct  effect, but a proxy  for the effect of unobservable characteristics on the wage rate or arrival  rate of new  information. Unearned income  and partner’s  income  are two obvious examples. The expected direct  effect would be to increase the demand for leisure and reduce the exit rate.  Alternatively, and this effect seems to dominate, unearned (partner’s) income  may  be associated with positive human  capital characteristics that increase the exit rate.

The main  policy implication of this paper  follows  from the result  that the expected duration of a benefit  spell  is not strongly related to observable individual characteristic. This implies that there  is limited  scope  for targeted policies that seek  to reduce the duration of benefit  spells  for those  most likely to experience long  spells. Instead,  in view  of the declining exit rate,  those most likely to experience long  spells  are those  whose spell  has already lasted  a long  time. In other words,  disadvantage cannot  be identified in advance of unemployment. Instead,  long elapsed duration is the principal predictor of remaining long  duration. Thus, policies that seek to reduce the duration of unemployment should  in the first instance be directed towards the long-term unemployed. Note that this implication is not dependent on whether the declining exit rate is a result  of true duration dependence or heterogeneity.

The results  for Aborigines and Torres Strait Islanders reinforce this conclusion. Given what  is known of the extent  of the labour  market  disadvantage facing  Aborigines and Torres Strait Islanders, the surprising finding  from this paper  is the similarity in the pattern  of receipt  of unemployment benefits  between Indigenous and other Australians. By looking separately at Aborigines and Torres Strait Islanders living  in the major metropolitan capitals, we  know  that this finding  has not arisen  because of the impact  of peoples living  in regions away from ‘mainstream’ economic activity  or of participation in the Community Development Employment Program.  The important  message for policy-makers is that the greater utilisation of unemployment benefits  among  Aborigines and Torres Strait Islander  peoples is primarily due  to the higher  proportion that enters  spells  of unemployment benefits  rather  than differential behaviour or incentive effects once  receiving benefits.

While  this paper  provides a valuable analysis of the pattern  of the utilisation of unemployment benefits  over time, some  limitations with the quality of the data  set remain  to be resolved. There are also some  methodological issues  relating the construction of variables and identification of their effects,  notably with respect  to payment levels. We have  drawn  attention to these  limitations throughout the paper. Continued improvement in the data  quality, extension of the panel  duration and further development in methodology offers considerable potential to enhance the current  analysis, particularly with respect to changes in the dynamics of unemployment utilisation over the economic cycle, between regional labour  markets and in response to changes in policy parameters.

[ Return to Top   Return to Section ]

Appendix: Variable means for spells data used in hazard estimation

  One Percent Sample   Indigenous – metropolitan   Indigenous – non-metro
  Single Partner   Single Partner   Single Partner
Age:
15 to 19 0.2007 0.0224   0.2493 0.0791   0.2504 0.0708
20 to 24 0.2973 0.0898   0.2636 0.2082   0.2434 0.1979
45 to 54 0.0830 0.1932   0.0406 0.0678   0.0521 0.0865
55 to 59 0.0185 0.0748   0.0067 0.0156   0.0105 0.0202
60 to 64 0.0058 0.0491   0.0011 0.0012   0.0021 0.0060
Main Eng-spkg  country 0.0757 0.0953   0.0018 0.0009   0.0007 0.0005
Non-Eng-spkg  country 0.1150 0.2347   0.0107 0.0175   0.0036 0.0074
Female 0.3671 0.2258   0.3186 0.2352   0.3209 0.2578
Indigenous 0.0339 0.0241   1.0000 1.0000   1.0000 1.0000
Home owner 0.0627 0.3829   0.0077 0.0580   0.0072 0.0355
Purchaser 0.0096 0.0643   0.0023 0.0248   0.0013 0.0068
Lives elsewhere 0.0051 0.0072   0.0006 0.0009   0.0009 0.0007
Other owner 0.0013 0.0119   0.0002 0.0009   0.0002 0.0007
Rents:
Private 0.3638 0.3397   0.2773 0.3661   0.2033 0.3068
Government 0.0202 0.0624   0.0612 0.2349   0.0369 0.1505
Board/lodging 0.2084 0.0250   0.3402 0.1055   0.3199 0.0831
Site/other fees 0.0085 0.0128   0.0030 0.0067   0.0061 0.0059
No rent paid 0.2020 0.0527   0.1464 0.0754   0.2358 0.2199
Free board/lodging 0.0005 0.0002   0.0004 0.0000   0.0004 0.0002
Sharer 0.0985 0.0046   0.0782 0.0104   0.0414 0.0043
Has child  aged:
0 to 6 0.0020 0.3263   0.0072 0.4818   0.0091 0.4602
7 to 12 0.0013 0.1153   0.0035 0.0938   0.0052 0.1032
13 to 15 0.0017 0.1204   0.0033 0.1153   0.0045 0.1295
over 15 0.0036 0.0238   0.0016 0.0163   0.0017 0.0099
Partner  receives UB 0.1966   0.2186   0.2621
Benefit  ($ eligible) $284.76 $265.51   $236.29 $218.89   $234.68 $227.49
(Benefit)2 $127 006 $108 654   $87 900 $75 406   $86 156 $79 647
Unearned income $4.11 $16.61   $0.65 $2.32   $0.50 $0.96
- if have  unearn inc. $34.76 $47.89   $45.52 $307.85   $191.39 $62.74
(Unearn inc.)2 $879 $7 328   $191 $277 924   $136 509 $546
- if have  unearn inc. $7 039 $26 247   $12 417 $6.97m   $11.06m $21 026
Partner  Income $127.81    
- if have  ptnr income   $347.97    
(Partner  inc.)2 $936 228    
- if have  ptnr income   $2548854    
Rent paid $103.01 $114.29   $102.97 $134.74   $79.09 $87.70
- if renting $155.92 $251.70   $148.07 $210.26   $132.13 $165.46
(Rent paid) 2 $17 610 $31 214   $17 336 $34 719   $11 633 $18 206
- if renting $29 439 $73 186   $25 937 $53 254   $20 772 $35 174
Previous  IS receipt 0.2400 0.2338   0.4349 0.4195   0.4346 0.4660
Repeat  spell 0.5208 0.4439   0.5185 0.4851   0.5030 0.4778
N (spells) 27 027 9 742   21 791 3 261   59 959 18 754

Notes:
—: denotes ‘omitted’

[ Return to Top   Return to Section ]

Endnotes

  1. That is, by 0.4 percentage points  given  an 8 per cent unemployment rate.
     
  2. The payment of unemployment benefits  is normally conditional on meeting an activity  test. However, as some categories of recipients are excluded from the activity  test there  is a case  for excluding such  persons from the analysis as well  on the ground that their behaviour is different.  Since  the status of the relevant indicator variable was  uncertain in the preliminary versions of the data  set, this issue  had to be left for future research.
     
  3. Dawkins, Harris & Loundes  (2000)  provide an extensive analysis of repeat  spells  using  the same  data  set.
     
  4. In the statistical literature, the presence of ‘old’ spells  is referred  to as left-truncation. Rather than deleting these spells, unbiased estimates of the distribution of completed spells  can be obtained if the contribution of these spells  to the likelihood function  is properly accounted for.
     
  5. ‘Old’ spells  are included in this calculation.
     
  6. Spells  of unemployment defined by the standard criteria  cannot  be identified in SEUP. The spells  used  in the analysis comprised periods for which  a person  was  reported to be looking for work  and was  not working. This is a close  approximation, but does  not meet  all the technical criteria  of the standard definition of unemployment.
     
  7. In this data  set, a person  is an Aborigine or Torres Strait Islander  if they  identify  themselves as such.

[ Return to Top   Return to Section ]

References

Altman, J. & Taylor, J. 1995, ‘Calculating Indigenous Australian employment in the private sector’, Labour Economics and Productivity, vol. 7.

Barrett, G. 2000, The dynamics of participation in the Sole Parent Pension, paper presented to Department of Family and Community Service Workshop, 21st March, Canberra. Committee of Review of Aboriginal Employment and Training Programs 1985, Report, AGPS, Canberra. Commonwealth of Australia 1994, Review of Aboriginal Employment Development Policy, Aboriginal and Torres Strait Islander Commission, Canberra.

Dawkins, P., Harris, M. & Loundes, J 2000, Repeated spells on benefits: an analysis of ‘Churning’ using the FaCS longitudinal administrative data set, paper presented to Department of Family and Community Service Workshop, 21st March, Canberra.

Department of Employment, Workplace Relations and Small Business (DEWRSB) 1999, Regional Labour Market Information for the Employment Services Request for Tender 1999, DEWRSB, Canberra.

Fortin, B. & Lacroix, G. 1997, ‘Welfare benefits, minimum wage rate and the duration of welfare spells: evidence from a natural experiment in Canada’, CIRANO Scientific Series 97s–25, Centre Interuniversitaire de Recherche en Analyse des Organisations, Montreal.

Holmlund, B. 1997, ‘Unemployment insurance in theory and practice’, Discussion Paper No. 380, Centre for Economic Policy Research, Australian National University.

Katz, L. F. & Meyer, B. D. 1990, ‘The impact of the potential duration of unemployment benefits on the duration of unemployment’, Journal of Public Economics, 41, pp. 45–72.

Meyer, B. D. 1990, ‘On unemployment insurance and unemployment spells’, Econometrica, 58, pp. 757–782.

Miller, P. W. & Le, A. T. 1999, A risk index approach to unemployment: an application using the Survey of Employment and Unemployment Patterns, Australian Bureau of Statistics Occasional Paper, ABS Catalogue No. 6293.0.00.001, February.

Moffit, R. 1992, ‘Incentive effects of the US welfare system: a review’, Journal of Economic Literature, 30, pp. 1–61.

Smith, D. 1995, ‘Measuring the contemporary labour force status of Indigenous Australians: issues, implications and options’, Labour Economics and Productivity, vol. 7, no. 1, pp. 28–48.

Stromback, T. & Dockery, A. M. 2000, Labour market programs, unemployment and employment hazards: an application using the 1994–1997 Survey of Employment and Unemployment Patterns, Australian Bureau of Statistics Occasional Paper, ABS Catalogue No. 6293.0.00.002, February.

Taylor, J. & Hunter, B. 1998, The job still ahead: economic costs of continuing Indigenous employment disparity, Aboriginal and Torres Strait Islander Commission, Australia.

—— 1997, ‘Promoting growth in Indigenous employment: the role of the private sector’, Australian Bulletin of Labour, 23, 4, pp. 295–313.

Webster, E. 1999, ‘Labour market programs and the Australian Beveridge Curve’, The EconomicRecord, 75, 231, pp. 405–416.

[ Return to Top   Return to Section ]

Content Updated: 5 June 2013