Printer Friendly

Did 'work first' work? the role of employment assistance programs in reducing long-term unemployment in Australia (1990-2008).


Over the last two decades Australia's progress in reducing its long-term reliance on unemployment payments was disappointing; this was despite an improving labour market, tighter work requirements and reformed employment assistance. After the introduction of the Job Network in 1998, the focus of employment assistance for long-term unemployed people shifted from a human capital approach towards a 'work first' approach. We review evidence from microeconomic evaluations of employment programs. Generally, job search assistance central to work first--is relatively effective. Gaps in the research may be a reason for the apparent discrepancy between these findings and Australia's slow progress overall in reducing long-term reliance on unemployment payments. Short-term average measures mask the distribution of program outcomes and results over the longer term. As unemployment fell, a growing proportion of unemployment payment recipients were disadvantaged in the labour market, and the work first approach may be ineffective for this group. The paper concludes with a brief assessment of the Job Services Australia program.

1. Introduction

For the past 20 years, Australian governments have pursued activation policies the extension and intensification of activity requirements to reduce reliance on income support (Kinnear, Grant and Oliver 2003; DEEWR 2009a). Linked to these policies was a succession of employment assistance reforms including the introduction of a Job Guarantee in 1994 and its replacement by the Job Network three years later (Freeland 1998). In aggregate, unemployment fell substantially over this period. Even before the Global Financial Crisis, Australia had one of the lowest unemployment rates in the OECD: 4.3 per cent in 2008 compared with an OECD average of 6 per cent (OECD 2009a). High levels of Long-term reliance on unemployment payments persist. In May 2009, 310,000 people--more than half of all recipients of Newstart and Youth Allowances--received those payments for over a year; 220 thousand received them for more than two years; and 110 thousand people relied on them for over five years. The number of long-term recipients remains higher today than it was before the recession of 1991 (DEEWR 2008c; DEEWR 2009b; DEWR 2003a; Warburton, Okopu and Vuong 1999). Long-term unemployment is associated with higher levels of structural unemployment, declining health and social exclusion, and has large fiscal costs (Layard, Nickell and Jackman 1991; Chapman 1994; Dockery and Webster 2002; Social Inclusion Board 2010).

This paper charts the development of activation policies and employment assistance reforms since the early 1990s and assesses their impact on reliance on income support, especially Long-term unemployment, by reviewing relevant Australian and international evaluation studies. It begins by charting trends in reliance on unemployment payments and the profile of recipients to assess whether they have become more disadvantaged in the labour market.

The literature indicates that employment assistance programs work in two ways. First, in conjunction with activity requirements, the programs may improve the efficiency of job search and thereby offset the negative effects of social security payments on work incentives. Second, they may improve the work capacity of unemployed people by assisting them to overcome barriers to work such as low skills and lack of recent work experience (Frijters and Gregory 2006; Layard, Nickell and Jackman 1991 ). The introduction of the Job Network in 1998 signalled a shift in policy away from a capacity-building approach towards improving the intensity and efficiency of job search. Job Network providers were to a large extent paid according to the short-term employment outcomes they achieved. These price incentives led them to favour job-search assistance over more costly interventions such as paid work experience and training that may have had a delayed impact on the job prospects of their clients (Dockery and Stromback 2001; DEWRSB 2001a). This focus on job search assistance was consistent with the work first approach pursued in other countries then including the United States and the United Kingdom, which prioritised immediate and intensive engagement with the labour market over efforts to strengthen the human capital of unemployed people (Finn and Schulte 2007; Loedemel and Trickey 2001).

The paper outlines and critiques the major evaluations of employment assistance programs in Australia. Such evaluations typically use regression techniques to estimate the net impact of a program on the average short-term employment prospects of participants (DEETYA 1997b; Stromback and Dockery 2000; DEWR 2006a). I compare the average net impacts on employment of three types of program: job search assistance, vocational training and work experience. I compare the results with those of a recent meta-analysis of similar evaluations in Europe and the United States by Kluve (2006). On the face of it, this evidence suggests that the shift towards job search assistance in Australia should have improved the overall efficiency of employment assistance. The paper identifies three key gaps in Australian evaluations of employment assistance programs which may help to explain why Long-term reliance on unemployment payments remains high, despite favourable economic conditions and the switch to seemingly more efficient programs.

The first gap that this research focuses on is the average employment outcomes of program participants. This tells us little about the effectiveness of the programs in assisting those most likely to remain unemployed for the long term. Second, because the evaluations focused on short-term employment outcomes, they do not inform us of the longer-term effects of the programs. Third, because they focus on the effects of a single program, they tell us little about the impact of the broader institutional environment on program effectiveness--in particular the interaction between activity requirements and employment assistance--and the effects of different combinations and sequences of employment programs. A number of suggestions are made to overcome these research gaps. I suggest that independent researchers have better access to the government's administrative and survey data. The paper concludes with a brief assessment of the Rudd government's reforms to employment assistance, including the Job Services Australia program that replaces the Job Network, and the new Productivity Places program.

2. Trends and Profiles of Long-term Recipients of Unemployment Payments

I begin by examining trends in recipients of unemployment payments (Newstart Allowance (NSA) and the Youth Allowance, Other (YAO)) since the recession of the early 1990s. (1) Figure 1 shows that both short and long-term reliance on these payments rose sharply for two years after the 1991 recession, but that Long-term reliance fell much more slowly throughout the recovery. After peaking in 1993, the number of Long-term recipients was almost static for five years before it gradually fell. It took another five years to begin to reduce the number of very long-term recipients substantially (those on payments for over two years). As noted, in May 2009 over half the recipients of these payments were long-term recipients and over a third were very long-term recipients.


These data suggest that substantial reductions in Long-term reliance on unemployment payments will only be achieved if policy measures are targeted to this group. Once people become unemployed long term, their probability of obtaining ongoing employment declines substantially due to scarring effects (Layard, Nickell and Jackman 1991 ). Further, as unemployment falls, a greater proportion of unemployed people are likely to come from groups that are more disadvantaged in the labour market (Chapman 1994).This is one possible explanation for our slow progress in reducing Long-term reliance on unemployment payments despite an improving labour market.

Long-term unemployment payment recipients have a low skill profile and face added barriers to employment than simply the duration of their unemployment. For example, among recipients of Intensive Support Customised Assistance within the Job Network (which was targeted to Long-term unemployment payment recipients) in 2006:

* 61 per cent lacked Year 12 qualifications or equivalent, compared to 33 per cent of people of working age (DEEWR 2008a; OECD 2007a);

* 20 per cent reported that their main barrier to work was a disability, and 15 per cent reported that they were 'too old' (DEEWR 2008a).

Further, 35 per cent of people on Long-term Newstart Allowances, 45 per cent of those on a Parenting Payment and 30 per cent of Disability Support Pensioners are estimated to have mental health disorders, mainly anxiety and depression (Butterworth et al. 2006). There is also evidence to suggest that unemployment payment recipients have become more disadvantaged since the recession of the early 1990s. Figure 2 shows that as the overall level of reliance on unemployment payments fell, the proportion of very Long-term recipients rose. Of particular concern, the proportion of people on unemployment payments for more than five years rose from one in 10 in 1999 to almost one in four in 2008 (O'Connor 2008b).


Unfortunately, there is limited, detailed time series data on the characteristics of recipients of Newstart and Youth Allowances. Given this, and the paper's focus on employment assistance, I examine changes since the early 1990s in the profile of participants in the major employment assistance programs for Long-term unemployed people: the Job Compact within the Working Nation strategy, and Intensive Assistance and Customised Assistance within the Job Network. Each year, over 200 thousand people participated in each of these programs. The targeting of the programs meant that the individuals' profiles were not identical to those of recipients of Long-term unemployment payments generally. The programs were all targeted mainly at Long-term unemployed people and those at risk of Long-term unemployment; participation was compulsory in most cases.

Aside from duration of unemployment, a number of personal characteristics for which data were available were associated with poor employment prospects in the short term, including being of mature age, Indigenous backgrounds, having a disability and low education levels (DEEWR 2008c; Stromback and Dockery 2000). On the other hand, sole parents had better job prospects on average than other Long-term unemployment payment recipients did, although a minority is clearly severely disadvantaged in the labour market (FACS 2001).

Table 1 shows trends in the abovementioned characteristics of those who commenced these programs in 1995, 1999, 2003 and 2008. For completeness, data are also included for commencements in specialist programs for people disadvantaged in the labour market: the Personal Support Program (PSP) for people with social barriers to work such as homelessness, addictions or a mental health disorder; specialist disability programs established after 2006 for people with a partial work capacity and who are on unemployment payments; and the programs comprising the Indigenous Employment Strategy (IES). (ii) The last column compares the average employment outcomes attained by each group three months after participation in Job Network Customised Assistance in 2007-08.

The table shows that since the mid-1990s there were substantial increases in the proportions of program participants who were of Indigenous backgrounds, of mature age, sole parents, people with disabilities, Personal Support Program participants and those unemployed for at least two years. There were significant reductions in the proportion lacking Year 10 or equivalent educational qualifications, but it does not follow that they became less educationally disadvantaged compared with the workforce as a whole, since the proportion of people of working age with low educational qualifications also declined. (iii)

Similar changes in the profile of unemployed people were also noted in qualitative research citing the views of employment service providers:

As these changes [to the Welfare to Work policy introduced in 2006] were taking place Job Network members observed that the numbers of job seekers with hard barriers to employment, such as ill-health, were increasing in their caseloads relative to those with soft barriers, such as limited vocational skills (DEEWR 2008a, p. 122).

To summarise, as overall unemployment levels fell over the last two decades, the proportion of unemployment payment recipients who received these payments long term--especially for over five years--increased substantially. Long-term unemployment payment recipients are a diverse group. Characteristics associated with poor employment outcomes include being of Indigenous backgrounds, having social barriers to work such as homelessness, addictions or a mental illness, being of mature age, having a disability and low education levels. With the exception of low skills, their incidence among Long-term unemployed people assisted by employment services has increased since the mid-1990s.

3. Activation and Employment Assistance in Australia since the Early 1990s

3.1 Activation

Following the recession of the early 1990s, many OECD countries pursued activation policies. This refers to a redesign of social security and employment assistance for jobless people to give priority to active engagement with the labour market. Common features of activation policies include the extension of activity requirements to a wider range of jobless people, and their intensification through stricter requirements, closer monitoring and sanctions for non compliance (Carcillo and Grubb 2006).

In Australia, activity requirements were progressively extended to mature age women by closing off access to dependency-based payments from the mid 1990s; to sole parents with school age children and people with disabilities able to work part time by limiting access to parenting and disability payments from 2006; and to Indigenous peoples in rural and remote areas by removing activity test exemptions from 2007 (Bond 2001; DEEWR 2009a; Altman and Jordan 2008). Requirements for Long-term recipients of unemployment payments were intensified though the introduction of Newstart activity agreements in 1991 and then through a requirement to undertake six months paid employment under the Working Nation policy in 1995. This was replaced by a requirement to undertake six months of training, or unpaid work experience through the Work for the Dole program which applied to young unemployed people from 1997 and extended progressively to recipients up to 50 years of age. From 2003, Long-term unemployed people participating in the first three months of the Customised Assistance phase of the Job Network (the most intensive phase of Job Network assistance) were interviewed fortnightly by their Job Network provider and required to undertake work preparation and job-search activity for at least 25 hours a week. Monitoring and compliance with job-search requirements was intensified in 1997 through employer contact certificates and job seeker diaries (Kalisch 1991; Freeland 1998; DEWRSB 2000a, p. 19; DEEWR 2008a).

The legislated penalty regime for breaches of activity requirements was altered a number of times from the mid-1990s to 2006, with the emphasis shifting from a complete loss of a fortnight's payments for a first or second breach (such as failure to attend an interview), towards partial loss of payments, warnings or payment suspensions. Third and subsequent breaches attracted the highest penalty--complete loss of payments for eight weeks (reduced from 12 weeks in 1997). While on the face of it the legislated penalties were reduced, they were more frequently applied, including the maximum penalty. During periods of strong employment growth around 2000 and 2007, monitoring and enforcement of activity requirements was intensified. By 2001, the number of breaches applied was twice that of 1998, and 18 per cent of unemployment payment recipients (26 per cent of young unemployed people) were breached resulting in losses of 16 to 24 per cent of their payments for up to 26 weeks. This prompted concerns from welfare advocates about the resulting financial hardship, and a subsequent easing of administrative efforts to enforce compliance. From 2006 to 2007, a fresh intensification of compliance monitoring led to a doubling of the number of eight-week non-payment penalties applied from 16,400 to 32,000 (Bond 2001; Pearce, Disney and Ridout 2002; DEEWR 2010).

Two broad approaches to activation have been pursued across the OECD. The first, designated as the human capital approach, prioritised investment in the work capacity of unemployed people to improve their prospects of sustained employment. The second, referred to as work first, emphasised job search to improve their immediate employment prospects (Loedemel and Trickey 2001; Bruttel and Sol 2006).

3. 2 Employment Assistance Programs

The distinction between the work first and human capital approaches is reflected in national expenditures on employment assistance programs. These programs work in two ways: by improving the intensity and efficiency of job search, and by helping people overcome barriers to work such as poor skills or lack of recent work experience (Frijters and Gregory 2006; Kinnear et al. 2003). The programs take three main forms: job search assistance, vocational training and further education, and work experience. Job search assistance is a common foundation stone of employment assistance, but countries vary in their levels of investment in capacity-building programs such as training and work experience (OECD 2005).

In the 1980s and early 1990s, training and work experience programs dominated public expenditure on labour market assistance in Australia (Figure 3).


Investment in vocational training rose at the peak of the economic boom of the late 1980s, when concerns about skill shortages were elevated (Freeland 1998). Following the 1991 recession, when policy concern shifted to inadequate demand for labour, the balance shifted towards work experience programs. From 1994 to 1996, the Working Nation--a strategy adopted by the former Labor government to reduce unemployment--guaranteed to give unemployment payment recipients of 18 months' duration six months paid employment in either the private, public or community sectors under a Job Compact. This required their participation in the program in return for a training wage. The relatively high cost of these programs is reflected in a peak in overall expenditure on employment assistance at that time.

As Figure 3 shows, after 1997 the focus shifted towards job search assistance. (iv) The watershed was the replacement of the Job Compact and other employment programs with the Job Network, a system for the regular purchase of a bundle of employment services by the government from non-government providers. At this time, policy makers in Australia and other OECD countries were increasingly concerned about the apparently high deadweight costs of traditional employment programs such as subsidised public sector employment and classroom-based vocational training (that is, the extent to which participants would have secured outcomes without assistance) (Vanstone 1996; Wiseman 2001; Calmfors, Forslund and Hemstrom 2002). The OECD advocated careful targeting of more costly programs towards those individuals most likely to benefit in lieu of a mass rollout of programs to entire categories of income-support recipients such as long-term unemployed people (Martin 2000). Official evaluations also suggested that loose targeting undermined the cost effectiveness of the Working Nation programs (DEETYA 1997a).

The Howard government's response to these concerns went well beyond a redesign of individual programs (Vanstone 1996; Freeland 1998). With the introduction of the Job Network, most national employment assistance programs were abolished and the Commonwealth Employment Service (the former public employment service) was replaced by a network of services contracted by government. They were paid a fiat service fee for each job seeker, together with an outcome fee based on short-term employment outcomes (three to six months after assistance). Outcome fees were scaled according to the assessed level of disadvantage faced by each unemployed person. The highest level of support, Intensive Assistance, was reserved for long-term unemployed people and those assessed as being most at risk of long-term unemployment based on an assessment tool called the Job Seeker Classification Instrument. Most of the government's investment in the Job Network was devoted to Intensive Assistance. The form of assistance offered was not specified by the purchaser. Instead, it was hoped that fees tied to employment outcomes would encourage providers to invest efficiently in assistance tailored to individual needs.

Outcomes-based funding shifted the risk of investment in employment assistance from the government to non-government providers. The pressure to achieve quick employment outcomes was reinforced by the new performance management system based on a 'star rating' system of providers that determined each provider's share of referrals of unemployed people. The star ratings were based on estimates of the average improvement in short-term employment outcomes achieved by different providers derived from a regression model that controlled for differences in the estimated job prospects of their clients. Further, the budget allocation for the Job Network was roughly half that of the previous programs. This limited the level of investment that providers could undertake on behalf of disadvantaged jobseekers (ACOSS 2004).

Providers responded to these incentives and constraints by prioritising job-search assistance. They judged this as being more likely to achieve outcomes quickly at a lower cost than investment in overcoming barriers to work such as training or paid work experience (Dockery and Stromback 2001, p. 447). For example, Webster and Harding (2001) demonstrated that investment in wage subsidies was not financially worthwhile for the providers, whether or not they significantly improved employment outcomes. This was consistent with a broader policy shift towards work first approaches. Thus, the name given to the Job Network's third iteration from 2003--09 was the Active Participation Model, and its accompanying slogan was activity is the key. Unemployed people moved through a fixed continuum of activity requirements and assistance from job-search training courses at three months' unemployment to mutual obligation (six months of employment, training or Work for the Dole) at the six-month point and then a six-month period of intensive job preparation and job search called Customised Assistance at the 12-month point. Flexibility in employment assistance was sacrificed to maximise compliance effects, whereby jobseekers able to find work left income support to avoid anticipated participation requirements (DEEWR 2008a).

An early concern about the Job Network model was that providers were encouraged to 'cream' by assisting those closest to employment while 'parking' harder to place unemployed people (ACOSS 2004; Productivity Commission 2003). The first official evaluation of the Job Network reported that: 'In qualitative research, providers confirmed that they often "give up" on jobseekers who are too hard to assist. They indicated that they were extremely unlikely to obtain employment for these job seekers and that their time would be better spent helping others. Such "hard-to-help" jobseekers receive the minimum assistance required to meet contractual obligations'. Jobseekers who were more likely to fit into this category included:

* people with physical barriers ranging from mild physical disabilities through to those who have serious drug dependencies;

* people with emotional barriers ranging from lack of motivation, through to mild mental disabilities;

* older jobseekers, particularly those who have few skills and (or) a low propensity to learn new skills; and

* people who have been unemployed for a long time.

In addition, the bias in the fee structure toward employment outcomes may, in some cases, favour the pursuit of short-term employment outcomes over training and educational outcomes which may, in the longer term, be of greater benefit to jobseekers' (DEWRSB 2001a, p. 96).

Concerns about creaming and parking led to the introduction from 2003 of tighter administrative controls over the assistance offered by providers, and the introduction of a Job Seeker Account for providers to draw upon to invest in help for the most disadvantaged to overcome their barriers to work such as low skills and a lack of work experience. The Account funded very modest investments in employment assistance: an average amount of approximately $1,000 for a typical long-term unemployed person (DEEWR 2008a).

When data became publicly available on average levels of provider investment in work experience and training through the Account, they confirmed the retreat from those forms of assistance. The official evaluation of the Job Seeker Account indicated that in 2006 25 per cent of the mainly long-term unemployed jobseekers receiving Customised Assistance (the highest level of Job Network assistance) received training funded through the Account at an average cost of $300 each. This amount included job-search training as well as vocational training, although no breakdown between the two was provided. Only 10 per cent obtained subsidised employment financed from the Account, although the department identified a trend towards increased investment in wage subsidies (DEWR 2006b).

With the introduction of the Job Network in 1997 there was a shift in employment assistance away from wage subsidies and training to overcome barriers to work, and towards a work first strategy that relied mainly on less-costly job-search assistance services. The incentives in the Job Network's price structure did encourage investment in long-term unemployed people, but at a low level. Providers were encouraged to concentrate on low-cost job-search assistance approaches that improved the short-term job prospects of those long-term unemployed people closest to employment; they were discouraged from undertaking more expensive interventions for the most disadvantaged.

4. Evidence from Evaluations of Employment Assistance

This part of the paper reviews official and independent evaluations of different types of employment assistance programs to assess whether the shift towards job-search assistance improved employment outcomes for recipients of unemployment payments.

Two approaches are commonly used to measure the difference employment assistance programs make to their participants' employment prospects or earnings: microeconomic and macroeconomic evaluations. Microeconomic evaluations seek to measure the net impact of a program--the change in the dependent variable attributable to the program--by comparing outcomes attained by treatment and control groups. Macroeconomic evaluations compare trends in employment or unemployment in different countries or regions that have adopted different employment-assistance strategies (Schmid 1996; OECD 2005). The vast majority of Australian evaluations are microeconomic studies, since these are finer grained, yielding estimates of the deadweight costs of a program as well as factors contributing to its net impact, including 'threat' or 'compliance' effects (where people leave income support to avoid participating in a program), improvements in work capacity and 'attachment' or 'lock in' effects (where a program reduces employment by delaying job search). The main advantage of macroeconomic studies is that they seek to take account of displacement and substitution effects (the extent to which the employment of other workers is affected when program participants obtain jobs), as well as any improvements in the overall efficiency of the labour market that may arise from enhanced participation among unemployed people.

Employment assistance programs are typically found to have modest impacts on the job prospects of unemployed people (OECD 2005). Even modest impacts can, nevertheless, substantially reduce the number of unemployment payment recipients if they are sustained over time. A recent official UK review of employment participation policy estimated that if annual flows from unemployment payments to jobs were raised by just five percentage points then, all else equal, the number of recipients would fall by hall over five years (Freud 2007).

4.1 Official Estimates of Gross Employment Outcomes and Cost Effectiveness

The only consistent trend estimates of the employment impacts of Australian employment assistance programs come from official Post Program Monitoring data (see, for examples, DEEWR 2008c; DEWR 2003a). Based on a survey conducted by the Department of Employment, these provide estimates of the average proportion of program participants who are in employment three months after leaving a program. Figure 4 compares these estimates for the main programs for long-term unemployed people--the Working Nation programs and the highest level of assistance available in the Job Network (Intensive or Customised Assistance).


There was an upward trend in flows to employment, from less than 30 per cent in the early 1990s to almost 45 per cent by 2008. These data should be used with care, since they do not measure the impact of the programs. No account is taken of the counterfactual: the outcomes that would be obtained without the programs. For example, the improvement in job outcomes from the late 1990s may be due to an improving labour market over this period. The average annual growth rate in employment from the end of the recession in 1992 to 1999 was 1.9 per cent, but this rose to 2.3 per cent from 2000 to 2008 despite a mild downturn at the turn of the century, (v) Further, most of the jobs obtained by former Job Network participants were part-time, which left many on part-rate income support (DEWRSB 2001b). With the shift towards lower-cost forms of labour market assistance from 1997, the real unit cost of employment assistance sharply declined and has continued to fall. The government uses a measure of the average cost per employment outcome in a given year to assess the cost effectiveness of labour market assistance. This is the average unit cost of all programs divided by the proportion of participants who were in employment three months after leaving a program (taken from the abovementioned Post Program Monitoring data). Figure 5 shows that this statistic fell sharply following the 1997 changes and continued to decline gradually thereafter. The Department of Employment attributes this to a combination of lower unit costs and higher employment outcomes (DEEWR 2008a).


This measure is regularly used in official program evaluations and appears to have considerable impact on public policy. On the face of it, these data support the work first approach that was pursued from 1997. For the reasons outlined above, this is not a valid measure of program effectiveness. To assess this properly we need to capture the net employment impact of a program, or the difference it makes to the employment prospects of participants.

4.2 Estimates of Net Employment Impacts

Techniques to evaluate the net impacts of employment programs have advanced considerably since the early 1990s. These studies typically compare either employment outcomes or changes in earnings between samples of program participants and non-participants with similar labour market characteristics. The main challenge is to specify a control group accurately. This is usually attempted in one of two ways: by randomly assigning individuals to the program, or by identifying a group of non-participants--a matched comparison group-with similar employment prospects to those of the treatment group, based on characteristics such as age, sex, education level and unemployment duration. The main drawback of the matched comparison group approach is selection bias: those selected into a program may have unobserved characteristics (such as motivation) that are not present to the same degree among the non-participants, yet they affect employment outcomes (Schmid 1996; DEWR 2006a). Random assignment, although widely regarded as a more accurate approach, is commonly used in program evaluations in the United States but is rarely used in Australia or European countries. One problem with random assignment is that the control group is deliberately denied assistance, which raises ethical issues and could also undermine activation policies. Following decades of refinement, nevertheless the matched comparison group approach that is typically used in Australian studies yields similar results to those obtained from random assignment methods (Card, Kluve and Weber 2009).

This part of the paper examines evidence from microeconomic evaluations of three types of employment assistance program: (i) job search assistance, (ii) training and (iii) work experience for long-term unemployed people. My focus here is on the ranking of these three program types according to their employment impacts (which work best), rather than the magnitude of the outcomes (how much difference they make). Inadequate access to data--either administrative data or panel surveys with sufficiently large samples of the population of interest--has hampered the involvement of academic researchers in this field. Most net impact studies of Australian employment programs were either conducted or commissioned by the Department of Employment. The main exceptions were a handful of studies that used the Survey of Employment and Unemployment Patterns (SEUP) undertaken by the Australian Bureau of Statistics in the mid1990s. Unfortunately, data from subsequent panel surveys commissioned by the government for program evaluation purposes have not yet been made widely available to researchers, although hopefully this situation will change. (vi)

Table 2 compares the results of the main studies of the net impact of programs for long-term unemployed people, all of which used the matched comparison group approach. Results are presented for four Working Nation programs operating in 1995-96 (Jobtrain, Jobskills, Jobstart and New Work Opportunities) and three subsequent programs (Intensive Assistance, Customised Assistance and Work for the Dole) in 2002 and 2005. Jobtrain provided short vocational training courses (typically six months or less); Jobstart partially subsidised the wages of jobseekers for six months in private employment; Jobskills and New Work Opportunities fully subsidised six months employment in the community and public sector; Work for the Dole provided six months unpaid work experience in the community sector; and Intensive and Customised Assistance within the Job Network, which were offered for up to 12 months and six months respectively, can be regarded as intensive job search assistance programs. (vii) All were targeted towards long-term unemployed people and those assessed as most at risk of long-term unemployment.

The results come mainly from official evaluations. The net employment impact refers to the difference in employment outcomes (in percentage points) between treatment and control groups at a point in time. The main innovation in the methodology of these studies, introduced from about 2003, was a change in the timing of measurement of employment outcomes for the treatment group from six months after leaving the program to 12 months after commencement. Although on the face of it this difference in timing should not matter (since the programs generally lasted for six months), this change improved validity because it enabled evaluators to account for the attachment effects of a program. This had the effect of reducing measured outcomes, because the effect of the delay in job search induced by program participation was captured. (viii) It follows that the earlier (pre-Job Network) studies probably over-estimated net employment impacts (Dockery and Webster 2002). Unfortunately I have not been able to locate estimates of the impacts of these earlier programs using the new methodology.

There were two main independent academic studies. Stromback and Dockery (2000) used data from the abovementioned SEUP. Their main contribution to Australian research in this field was the use of hazard models designed to capture net impacts better over time. In order to compare their results with the official studies, Stromback and Dockery also sought to replicate the methodology of official studies to produce point in time estimates of program effects (DEETYA 1997a). The hazard models yielded more variable impacts over time, but the overall ranking of program effectiveness remained constant. Stromback and Dockery also used richer controls for individual characteristics than in the official evaluations to account for selection bias, but these yielded similar results to the official studies. In the absence of studies using random assignment techniques to evaluate the same programs, it is difficult to judge the extent to which these studies overcame the hurdle of selection bias.

In the second study Borland and Tseng (2004) assessed the impact of the pilot Work for the Dole scheme in 1998. In contrast to the official evaluations of this program, they found that it reduced the probability of employment, mainly due to attachment effects. A major difference between these studies was that Boland and Tseng used a longitudinal data set drawn from administrative data, while the official studies used surveys of the labour market status of a sample of employment program participants and non-participants developed for the purpose of the evaluations. It is difficult to compare the validity and robustness of these studies in the absence of more detailed information on these data sources. (ix)

Given the differences in research methods, target groups and program content between the various studies, it is difficult to draw many firm conclusions from this comparison. My main interest is in the relative impacts of different programs, and with a few exceptions these are consistent across the different evaluations.

* Apart from the Jobstart program, the impacts of the programs on employment in the short term were modest--they improved employment prospects by well below 20 percentage points. As discussed above, this is consistent with findings in other OECD countries.

* The most expensive programs were those providing fully subsidised paid work experience in the public and community sector (NWO and Jobskills). This was appropriate, since these programs were more tightly targeted towards long-term unemployed people than either Jobstart or the Job Network programs were (Intensive Assistance and Customised Assistance).

* Ranking the programs in terms of average short-term employment outcomes, wage subsidies in the private sector appear to be the most effective, followed by job search assistance (Intensive and Customised Assistance), while vocational training and the other work experience programs (both paid and unpaid) had significantly poorer outcomes. (x)

* There was considerable variance between the results for unpaid work experience (Work for the Dole) between official studies (suggesting modest employment gains) and the single academic study (suggesting a significant reduction in employment prospects). International studies of similar programs generally find either low or negative employment effects (Johri et al. 2004; Brock, Butler and Long. 1993).

Given the limited number of Australian evaluations, I compare these results with those of a meta-analysis of evaluations of European employment programs, which mainly used the matched comparison group methodology and also focused on short-term employment outcomes (Kluve 2006). While institutional factors and differences in evaluation methodology make international comparisons difficult, I am mainly interested in the relative impacts of different types of programs rather than absolute values.

Kluve (2006) compared the results from 95 studies of 137 programs across Europe, most of which were conducted after 2000. The programs were mainly targeted towards long-term unemployed people and others considered relatively disadvantaged. He divided the programs into four types: vocational training, direct employment (mainly fully subsidised jobs in the public and community sectors), private sector incentive schemes (wage subsidies) and services and sanctions (the latter corresponds to our job search assistance category). He grouped the outcomes of these studies into three levels: negative effect, zero effect and positive effect on employment. Overall, 55 per cent of studies found a positive effect, 24 per cent round a zero effect and 21 per cent a negative effect. As in the Australian studies discussed above, both positive and negative effects were modest--generally well below 20 percentage points. Kluve then explored the relationship between these three sets of outcomes and four sets of independent variables: the type of program, the study design, the institutional context (for example, employment dismissal legislation and benefit replacement rates) and the economic conditions in the country at the time (for example, the unemployment rate). His main conclusion was that it was the type of the program, rather than the other variables listed above, that had the most significant and consistent impact on employment.

Kluve's findings regarding the impact of program type on employment outcomes are presented in Table 3. It should be noted that he varied his method to produce six separate simulations, and these data are from the first iteration; the ranking of programs is consistent throughout. The table shows the average difference in the probability of a positive average employment outcome between the program types listed and vocational training programs, which Kluve took as his base case program.

The findings are broadly consistent with the Australian studies. (xi) Private sector wage subsidies and services and sanctions performed relatively strongly, while training and subsidised employment in the public and community sectors recorded the weakest outcomes. There are differences in the rankings of programs which may be due to national circumstances or to differences in methodology (for example, Kluve did not take into account the magnitude of program impacts beyond identifying whether they were statistically significant). Unlike the Australian evaluations, Kluve found that services and sanctions performed better on average than private sector wage subsidies did (possibly because he ignored the magnitude of employment impacts), and that training programs performed better than subsidised employment in the public and community sectors (possibly because the attachment effects of the latter programs were greater in overseas programs).

5. Weaknesses and Gaps in the Evaluation Research

Do the above results mean that policy makers and employment service providers should abandon other forms of assistance and concentrate on wage subsidies and job search? Clearly, this would not be a sensible policy response. What works for one group of unemployed people in a particular institutional context may not work for another group or as part of a different policy mix (Hasluck and Green 2007). Evaluations that focus on the average impacts of specific programs at a single point in time have major limitations as guides for public policy. In the following discussion, I raise three problems.

* Because they focus on average employment outcomes, they tell us little about the effectiveness of the programs in assisting the most disadvantaged jobseekers.

* Because they focus on short-term employment outcomes, they do not tell us about a program's longer-term effects.

* Because they focus on the effects of a single program or intervention, they tell us little about the impact of the broader policy environment--for example, whether a different sequence of employment assistance would be more effective--and the impact of different benefit rules on the effectiveness of employment programs.

5.1 Heterogeneity

Average measures of program performance disguise the range of outcomes achieved by different groups. Bitler et al. 2003 critiqued evaluations of Welfare to Work programs in the United States on this basis. Then, most studies of the effects of the Welfare to Work policy--designed to reduce the incidence of the receipt of welfare among low-income sole parents--reported that, on average, the policy increased employment and incomes and reduced poverty and welfare reliance among the target group. Bitler, Gelbach and Hoynes (2003) undertook a distributional analysis of the effects of the policy in Connecticut and revealed a range of outcomes for income-support recipients that were masked by the average results reported in other studies. They concluded that the policy widened the dispersion of incomes among low-income sole parents. A substantial minority obtained employment following participation in Welfare to Work programs, increased their earnings and benefited from increases in earned income tax credits. A smaller minority of more economically and socially disadvantaged sole parents was unable to obtain paid work and was adversely affected by welfare payment time limits; they experienced substantial declines in income.

A further problem with evaluations confined to population-average impacts, is that they imply that all outcomes are equal. For example, an employment outcome achieved for a short-term unemployed person who might otherwise remain unemployed for one month is treated as equal in value to one achieved by a long-term unemployed person who would otherwise remain out of work for another 12 months. A comprehensive assessment of the fiscal costs and benefits of investment in labour market assistance (let alone social benefits) should take account of the opportunity cost, in unemployment payments and other services, of leaving a substantial minority of harder-to-place jobseekers on income support for prolonged periods.

Official evaluations of the net impacts of Job Network services have focused on average employment outcomes. No estimates are provided of the distribution of net employment impacts among jobseekers. Significantly, a recent official employment impact study of Customised Assistance found that most of the program's impact was concentrated in the first two months of this six-month program (DEWR 2006a, p. 15). This suggests that among their long-term unemployed clients Job Network providers may have focused their efforts on those they considered as being the most likely to secure employment quickly. That is, providers undertook their own evaluation of each job seeker's distance from employment and were able to observe characteristics (such as motivation) that were not observed by Centrelink when it used the Job Seeker Classification Instrument to assess the individual's job prospects and determine the rewards available to the provider for placing them in a job. To explore this question further, Figure 6 shows official data on the distribution of Job Seeker Account expenditure (DEEWR 2008a). Although Job Seeker Account funds were notionally tied to individual jobseekers and calibrated to duration of unemployment and other measures of labour market disadvantage (via the Job Seeker Classification Instrument), providers were allowed to pool the funds available to them and redistribute them according to their assessment of the needs of each job seeker. Figure 6 compares the distribution of notional allocations (according to the level of credits deposited in the Account for each job seeker) with that of actual expenditure by providers.


This graph shows that providers redirected funds from jobseekers assessed by Centrelink as more disadvantaged towards those assessed as less disadvantaged. The official evaluation of the APM model of the Job Network from 2003 to 2006 noted that: 'providers reported that the most disadvantaged clients were less likely to receive assistance because the provider believed they were "too far from work" to benefit from jobseeker Account assistance' and that: 'some [providers] considered that the fee structure provided a weak incentive for Job Network providers to take risks with the very longest-term unemployed Intensive Support clients' (DEEWR 2008a, p. 104).

This echoes the concerns about creaming and parking that were raised in the first official evaluation of the Job Network in 2001 (see above). A system where providers are largely paid according to employment outcomes can give rise to perverse incentives. It is vital that evaluators understand the effects of these incentives on the assistance provided to different subgroups of jobseekers, and their subsequent impact on employment outcomes. These cannot be captured by average net impact measures for an entire program or intervention.

Programs that appear to work well on average, such as job search assistance, may not work so well for the most disadvantaged unemployed people. To understand these differences, evaluations should ideally measure the distribution of net impacts across the treatment group, especially among those whose job prospects are, prima facie, the weakest. In theory this should be feasible using the net impact methodologies previously adopted in Australian evaluations. In practice it would test the limits of validity and reliability of these approaches, especially in a policy environment where employment programs are almost universally provided to broad target groups (such as long-term unemployed people) and it is therefore increasingly difficult to specify a control group. Larger survey samples might also be required. Another way to test and perhaps improve the validity of the matched comparison group approach, and to measure creaming and parking directly, would be to ask employment consultants (employment service providers) to rank samples of both the treatment and control groups according to their employability and then describe the assistance provided to each. (xii) Measures of the cost effectiveness of employment programs should take account of the opportunity cost of failing to assist hard to place unemployed people, or at least the additional cost of future income support payments. A model for this more comprehensive cost-benefit analysis was outlined in Piggott and Chapman (1995). (xiii)

5.2 Timing

Studies of the net employment impacts of Australian labour market programs largely measured employment outcomes in the short to medium term (for example, 12 months after commencing a program), although some studies compared the employment outcomes of treatment and control groups over longer periods (for example, DEETYA 1997b; Stromback and Dockery 2000; DEWR 2006a). The latter studies show that program net impacts vary substantially over time. For example, the net impacts of wage subsidy schemes declined over time whereas those of job search assistance were more consistent (DEETYA 1997b). (xiv) Of particular concern here is the 'carousel effect' where unemployed people return to income support following temporary or casual employment. Official estimates indicate that approximately 38 per cent of unemployed people who left income support after participating in Job Network assistance during 2003-04 returned to income support within 12 months (DEEWR 2008a, p. 130).

Longer-term outcomes are especially relevant when comparing programs with different attachment effects. While job search programs intensify job search, most other programs (including most training and work experience programs) delay it (OECD 2005). The issue for policy is whether the weaker short-term outcomes of these programs are compensated for by better results in later periods. Evaluations will be biased in favour of job search programs if longer-term outcomes are not taken into account. The difference between short and longer-term impacts appears to be greatest for vocational training programs. A recent study of employment program evaluations by Card, Kluve and Weber (2009) using a similar methodology to that employed in Kluve (2006) divided the estimated employment impacts into short-term effects (within one year of leaving a program) and medium-term impacts (more than two years). The study also incorporated evaluations of programs in the United States. Its findings in regard to shorter-term employment impacts (less than two years) were similar to those of Kluve (2006), but the performance of vocational training programs improved relative to that of job search assistance over the medium term (Card, Kluve and Weber 2009, p. 18).

Evaluations of employment programs should compare their net impacts over at least a two-year period, and longer where feasible. (xv) This is not substantially longer than some previous Australian evaluations, and should be feasible, although more costly to undertake. The implications for policy are that investment in some programs with significant attachment effects (especially those that improve employment-relevant skills) may be worthwhile despite low or negative short-term employment impacts. This has implications for the design of optimal fee structures for outcomes-based funding programs such as the Job Network. For example, to overcome short-termism in provider investment in disadvantaged jobseekers, the new employment assistance purchasing model in the United Kingdom rewards employment outcomes that are sustained for at least 12 months (Freud 2007).

5.3 Institutional Context--Effects of Activity Requirements and Interactions between Programs

Most studies of the employment impacts of specific programs attempt to isolate their impacts from those of the surrounding employment assistance and social security infrastructure. Schmid (1996) argued that these relationships between individual programs and their institutional context should be taken into account in a systematic way when evaluating labour market assistance. He proposed a target-based approach to evaluation, in which targets are set for the attainment of employment and other outcomes for particular groups and an array of studies using both quantitative and qualitative methods are used to assess the effectiveness of the suite of programs in meeting the targets. He urged researchers to understand better what forms of assistance are actually provided to unemployed people, rather than treating the program as a black box for evaluation purposes. In this way, we may gain a better understanding of the institutional factors that bear on the outcomes of a program and of biases that might distort the outcomes of employment impact studies.

Official and independent evaluations of Australian employment programs have used a combination of quantitative and qualitative methods (including net impact studies and surveys of employment service providers, unemployed people and employers) to understand better how programs are delivered and experienced on the ground (DEEWR 2008a; Eardley, Abello and McDonald 2001). Less attention has been paid to the dynamic interactions between different employment programs and between employment programs and benefit rules. I explore two related issues below:

* the employment impacts of activity requirements;

* the integration and sequencing of job search assistance and help to overcome barriers to work.

(a) Activation

The extension and intensification of activity requirements underpinned reform of employment assistance in Australia over the last two decades, especially the establishment of the Active Participation Model of the Job Network. Activity requirements and compliance systems are likely to have considerable impact on both the design and effectiveness of employment assistance programs. Indeed, it is difficult to disentangle activation or compliance effects from the employment impacts of the programs themselves. (xvi)

There is evidence to suggest that the imposition of activity requirements where none previously existed, or that they were not strictly enforced, substantially increased exits from income support to employment (Wells 2001; Dolton and O'Neill 1996; Van den Berg and Van der Klaauw 2001). This view is supported by two Australian studies. In 1991, Newstart interviews were trialled for Australians who had received unemployment payments for 12 months. The interviews focused on activity agreements which included active job search and referrals to training and work experience programs where appropriate. Previously, long-term unemployed people faced the same job search requirements as other unemployment payment recipients, but in the absence of regular in-depth interviews with the Commonwealth Employment Service they were not rigorously enforced. On average, among those who did not receive further assistance from labour market programs (such as vocational training) long-term unemployed people who participated in Newstart interviews were 8 per cent more likely to be employed six months later than were a similar group who were not interviewed (xvii) (Kalisch 1991).

Similarly, the extension of job search requirements to sole parents with school age children and people with a partial capacity to work under the Welfare to Work policy appears to have increased exits from income support. The official evaluation of Welfare to Work used a natural experiment approach to compare exits from income support to employment over a six-month period after the Welfare to Work policy was introduced in 2006, with the equivalent data one year prior to the change. It found that exits from income support (mainly to employment) rose by 12 percentage points after the policy change for those sole parents diverted to Newstart Allowance, and by 11 percentage points for those who remained on the higher Parenting Payment Single payment but who nevertheless faced identical activity requirements. Exits to employment among those with a partial work capacity rose by 6 percentage points (from a very low base of 4 per cent) DEEWR 2009a, pp. 36-37; 49). The above comparison of outcomes for sole parents suggests that the employment impact did not come mainly from the shift to the lower Newstart Allowance payment; rather, it came from the new activity requirements and the employment programs to which sole parents were referred. (xviii)

On the other hand, there is evidence that activation measures face diminishing returns where well-developed activity requirements and compliance systems are already in place (OECD 2005, p. 191). For example, the successor to Newstart activity interviews in Australia--the nine-month review interviews for recipients of unemployment payments that were in place from 1996 to 2000--were found to have no significant employment impact (Borland and Wilkins 2003). These interviews were similar to the Newstart interviews which they replaced but they were brought forward three months into the payment spell. As indicated previously, by the late 1990s activity requirements and compliance systems were much stricter than in 1991 when the Newstart interviews were introduced. Activation effects may be weaker for those who face greater hurdles to employment. For example, the introduction of a Job Seeker Diary for unemployed people to report their job search efforts in detail to Centrelink had much weaker employment impacts among long-term unemployed people than short-term unemployed people (Borland and Tseng 2003). Similarly, review interviews with people unemployed for five years or more were found to slightly reduce future hours of employment (Breunig et al. 2003).

Australian and international evidence suggests that the effects of more intensive activity requirements are strongest when few or no requirements previously applied, or they were not strictly enforced, and were weaker for the most disadvantaged jobseekers. In program evaluations it is difficult to separate out the employment impacts of activation (compliance effects) per se from the effect of the programs themselves on an individual's job prospects (program effects). Studies that take advantage of variations in the timing or targeting of activity requirements and their enforcement are likely to provide the best estimates of activation effects, as in the official evaluation of the Australian Welfare to Work policy (DEEWR 2009a). One reason for this is that in a mature system of activity requirements, it is difficult to identify a suitable comparison group that was not affected by activity requirements or compliance measures.

(b) Sequencing

Another key weakness of standard net impact studies of individual programs is that they do not take account of the interactions between different employment programs, for example where a sequence of employment assistance is offered. This is particularly relevant for people facing steep barriers to employment, who may not benefit much from job search assistance alone. This is illustrated by the British experience with the New Deal for long-term unemployed adults (Hasluck and Green 2007). When this program was first introduced in 1998, it focused on the provision of job search assistance by personal advisers from the public employment service. Unlike the equivalent program for young people--the New Deal for Young People (NDYP)--there was no automatic progression to a set of programs (called New Deal Options) to overcome barriers to work if the participant was unsuccessful in securing a job within the first three to four months of the program. When an official evaluation revealed that the program was much less effective than the NDYP, at first the job search requirements and contacts with advisers were intensified. This made little difference to employment outcomes. Personal advisers informed the evaluators that the profile of participants had become more disadvantaged and that job search assistance alone was not sufficient to get them into employment. The program was then restructured along similar lines to the NDYP to provide for a gateway period of four months of assessment, employment counselling and intensive job search, followed by an Intensive Activity Period designed to overcome barriers to work for those still unemployed at that stage. This included work-focused training and temporary work placements extending for 13 to 26 weeks. Following these changes, the job entry rate from the program doubled to 30 per cent.

Job search assistance and more substantial capacity-building can either be offered in sequence (as with the NDYP) or these two components can be integrated into a single package of support and offered simultaneously. Both strategies have been used in Australia to assist unemployed people who require a range of interventions beyond traditional employment programs, including mental health services, stabilisation of housing, family support and addiction counselling (Perkins 2007; Killacky Jackson and McGorry 2008). Evaluations of employment programs should aim to account for the effects of the sequencing and integration of programs. The right sequence or combination of job-search and capacity-building programs can enhance the employment outcomes of both. This takes us beyond a dichotomy between work first and capacity-building approaches, towards consideration of combined approaches (Berlin 2002). Experimental and quasi-experimental methods--using random assignment or comparisons of outcomes from different programs assisting similar target groups--could be used for this purpose. This can be done without unfairly denying people access to assistance in cases where a new scheme is being trialled either locally or nationally for a specific group of unemployed people. An example is the British pathfinder evaluations for the New Deal program for young unemployed people, where new programs were trialled in some regions first and the evaluators compared outcomes for the treatment group in those regions with those of control groups in both those and similar regions (Blundell et al. 2003).

6. Conclusion

Along with many other OECD countries, Australia has a problem with entrenched long-term unemployment. Even before the economic downturn of 2009, the number of long-term recipients of unemployment payments stood at over 300 thousand people. This was well above the number two decades earlier, despite robust employment growth and an unemployment rate that plummeted to 30-year lows. A likely reason for this is that as unemployment fell, the profile of recipients of unemployment payments became more disadvantaged. From 1990 to 2008, the proportion of unemployment payment recipients on these payments for over two years had risen from 16 per cent to 43 per cent and the proportion with over five year's duration had risen from 5 per cent to 23 per cent. The profile of participants in the main employment programs for long-term unemployed people also became more disadvantaged, with a higher incidence of Indigenous peoples, people of mature age, people with disabilities, and people with social barriers to work such as homelessness, addictions or mental illness.

Australian governments have experimented with a range of labour market programs to reduce long-term unemployment. Following the introduction of Job Network in 1997 there was a shift away from investment in programs such as work experience and training that emphasised preparation for employment, and towards job search assistance within a work first framework. The structure of financial incentives for Job Network providers favoured investment in low-cost interventions such as job-search assistance that yielded quick employment outcomes.

Although the average unit cost of labour market assistance declined sharply, average employment outcomes for unemployed people improved. This outcome, however, could be attributed to the improvement in labour market conditions from the late 1990s onwards. More sophisticated net impact evaluations--which seek to measure the difference that employment programs make to people's job prospects--have reported broadly consistent findings on the relative short-term employment impacts of different types of employment programs. The range of short-term program impacts is typically low, ranging from negative impacts to a 20 per cent increase in the probability of employment at around 12 months after commencement in the program. Overall, the most effective programs appear to be wage subsidies for temporary mainstream jobs and job-search assistance. Fully subsidised employment in the public sector and vocational training programs appear to be less effective in the short term due in part to attachment effects. Apart from the abandonment of wage subsidy schemes, these findings appear to support the reorientation of employment assistance from capacity-building assistance towards a work first approach.

There are, however, three weaknesses in these evaluations that caution against an over reliance on them as a guide to policy.

(i) Heterogeneity: The average impacts measured in these studies may mask a diversity of employment outcomes for different groups, including weaker outcomes for the most disadvantaged. This is pertinent to the evaluation of programs such as the Job Network which encourage providers to focus their efforts on those who are closer to employment.

(ii) Timing: Most of the official studies of the effects of Australian programs focus on short-term outcomes (up to 12 months). If measurement is limited to the short term, this can bias results towards programs such as job-search assistance that achieve quick outcomes with minimal attachment effects, and against capacity-building programs such as vocational training or work experience which delay job search now in order to improve future employment prospects. There is clear evidence from overseas studies to suggest that well-designed training programs boost employment prospects over the long term.

(iii) Institutional context: Researchers need to get inside the black box that sits between program inputs and outcomes in order to understand the dynamic interactions between activity requirements and program participation, and the effects of different sequences of employment assistance. The latter is likely to be particularly important when assessing the effectiveness of assistance for more disadvantaged jobseekers.

The progressive intensification of activity requirements coupled with job search assistance over the past decade and a half has helped to reduce reliance on unemployment payments at a low cost to the government. There is a risk, however, that an over-reliance on a work first strategy will leave the most disadvantaged unemployed people behind. British policy makers concluded in the late 1990s that activation and job-search assistance alone could only reduce long-term reliance on unemployment payments up to a point. Beyond that, new investment to improve the employment capacity of long-term unemployed people was required. The policy response was to add value to activation by investing in capacity-building programs for long-term and disadvantaged unemployed people through the various New Deals for unemployed people. These programs targeted the most expensive capacity-building assistance towards those least likely to secure employment through low-cost interventions alone (Wells 2001; Finn 2001). Australia may have reached that point by the early 2000s, but given the absence of detailed time series data on the profiles and labour market transitions of income support recipients, and the limited investment in capacity-building programs over the last 15 years, this is hard to assess. A further impediment to more investment for the more disadvantaged jobseekers has been a reliance by policy makers on measures of cost effectiveness that ignore the opportunity cost of failing to assist this group.

The evaluations surveyed in this article caution against a return to large-scale investment in poorly targeted capacity-building programs such as job guarantees, and the use of vocational training or Work for the Dole programs as default schemes for activation purposes. Instead, interventions should be targeted towards those most likely to benefit; engagement with the labour market should be maintained and providers should be encouraged to target sustained employment outcomes.

6.1 Recent Reforms

I conclude with brief comments on reforms to employment assistance by the Rudd government. An official discussion paper on that government's reforms to employment assistance opened with the following statement: 'The Job Network is no longer suited to a labour market characterised by lower unemployment, widespread skill shortages and a growing proportion of jobseekers who are highly disadvantaged and long-term unemployed (O'Connor 2008a).

Since its election, the government has retained the basic program infrastructure that it inherited, but with two significant changes. First, a large-scale vocational training program for jobless people, Productivity Places, was introduced, offering 92 thousand places for jobless people in 2008-09 (DEEWR 2009e). Employment service providers were encouraged to refer people to the program by the relaxation of requirements to place participants into jobs quickly and by rewarding providers whenever a job seeker obtained employment after participating in Productivity Places. This is a sensible departure from the work first approach. Its effectiveness is likely to depend on how well the training is tailored to the needs of unemployed people (Barnett and Spoehr 2008; ACOSS 2007) A potential weakness of the program from this standpoint is that it is now administered by the state governments, that have no direct interest in reducing reliance on Federal unemployment payments.

Second, the Job Network was replaced by the Job Services Australia (JSA) program. This retains its key features (outcomes-based funding, regular tender rounds, star ratings and a job-seeker Account), but in response to widespread criticisms of the inflexibility of the Job Network service continuum, detailed administrative oversight of providers and the escalation of penalties for jobseekers, JSA providers have more flexibility to tailor assistance to individual needs, and the compliance regime has been reformed (O'Connor, 2008a). The biggest constraint on flexibility is limited resources, since the overall budget for the program was reduced compared with Job Network. This continues the drive to lower costs per employment outcome that commenced 15 years earlier with the introduction of the Job Network. In the JSA system, scarce resources are rationed by using assessment tools to direct jobseekers into four streams of support, rather than a fixed continuum of service in which the level of assistance escalates as people are unemployed for longer periods. This directs more assistance towards those at greatest risk of long-term unemployment, especially those people with social barriers to work who were previously referred to the Personal Support Program.

This comes, however, at the expense of investment in those already unemployed for a long term. After 12 months unemployment, jobseekers enter a work experience phase for which the notional allocation towards the costs of overcoming barriers to work is typically lower; it is usually $500 for the remaining duration of the unemployment spell. Providers are allocated funds to interview this group on a two-monthly basis. This very low level of investment in long-term unemployed people contrasts with the widespread emphasis in OECD countries on intensive help for this group, and the Job Network system where the highest level of support (Customised Assistance) was mainly reserved for them. At a time when a growing proportion of unemployed people are out of work for a very long time, this could prove to be the Achilles heel of the new regime.


Australian Bureau of Statistics (ABS), Labour Force, Australia, various years, cat. 6202.0.

ACOSS (2004), The Obligation is Mutual, Australian Council of Social Service, Sydney.

ACOSS (2005), Submission to Senate Community Affairs Committee Inquiry into the Welfare to Work Bill, Australian Council of Social Service, Sydney.

ACOSS (2007), The Role of Education and Training in Welfare to Work, Australian Council of Social Service, Sydney.

Altman, J., Gray, M. and Levitus, R. (2005), Policy Issues for the CDEP Scheme in Rural and Remote Australia, Centre for Aboriginal Economic Policy Research, Discussion Paper No. 271/2005, Canberra.

Altman, J. and Jordan, K. (2008), The Untimely Abolition of the Community Development Employment Program, submission to Senate Community Affairs Committee Inquiry into the Family Assistance and Other Legislation Amendment (2008 Budget and Other Measures) Bill 2008, Centre for Aboriginal Economic Policy Research, Canberra.

Barnett, K. and Spoehr, J. (2008), Complex Not Simple: The Vocational Education and Training Pathway from Welfare to Work, National Council for Vocational Education Research, Research Report.

Berlin, G. (2002), What Works in Welfare Reform, Manpower Development Research Corporation, Washington.

Bitler, M., Gelbach, J. and Hoynes, H. (2003), What Mean Impacts Miss, National Bureau of Economic Research Working Papers No. 10121, Washington.

Blundell, R., Costa-Dias, M., Meghir, K. and Van Reenen, J. (2003), Evaluating the Employment Effects of a Mandatory Job Search Program, University College London Discussion Papers in Economics No. 03-05, London.

Bond, K. and Wang, J. (2001), Income Support and Related Statistics, a 10-year Compendium 1989-1999, Department of Family and Community Services Occasional Papers No. 1, Canberra.

Borland, J. and Tseng, Y. (2003), How Do Administrative Arrangements Affect Exit from Unemployment Payments? The Case of the Job Seeker Diary in Australia, Melbourne Institute Working Papers No. 27/03, Melbourne Institute of Applied Economic and Social Research, Melbourne.

Borland, J. and Tseng, Y. (2004), Does 'Work for the Dole' Work? Melbourne Institute Working Papers No. 14/04, Melbourne Institute of Applied Economic and Social Research, Melbourne.

Borland, J. and Wilkins, R. (2003), Effects of Activity Tests on Exit from Payments--The 9-Month Intensive Review, Melbourne Institute Working Papers No. 25/03, Melbourne Institute of Applied Economic and Social Research, Melbourne.

Breunig R., Cobb-Clark, D., Dunlop, Y. and Terrill, M. (2003), 'Assisting the Long-term Unemployed: Results from a Randomised Trial', Economic Record, vol. 79, pp. 84-102.

Brock, T., Butler, D. and Long, D. (1993), Unpaid Work Experience and Welfare Recipients, Manpower Development Research Corporation, Washington. Unpublished report accessed on 1/11/10 at

Bruttel, O. and Sol, E. (2006), 'Work First as a European Model', Policy and Politics, vol. 34, pp. 69-89.

Butterworth, P., Fairweather, K., Anstey, K. and Windsor, T. (2006), 'Hopelessness Demoralisation and Suicidal Behaviour, the Backdrop to Welfare Reform in Australia', Australia and New Zealand Journal of Psychiatry, vol. 40, pp. 648-656.

Calmfors, L. Forslund, A. and Hemstr5m, M. (2002), Does Active Labour Market Policy Work? Lessons from the Swedish Experiences, CESifo Working Papers No. 675. Ifo Institute for Economic Research, University of Munich.

Carcillo, S. and Grubb, D. (2006), From Inactivity to Work: The Role of Active Labour Market Policies, OECD Social Employment and Migration Working Papers No. 36, Organisation for Economic Cooperation and Development, Paris.

Card, D., Kluve, J. and Weber, A. (2009), Active Labour Market Policy Evaluations, a Meta Analysis, IZA Discussion Papers No. 4002, Institute for the Study of Labor, Bonn.

Chapman, B. (1994), Long-term Unemployment, the Case for Policy Reform, Centre for Economic Policy Research Discussion Papers, No. 302, Australian National University, Canberra.

DEETYA (1996a), Post Program Monitoring Survey, Department of Employment, Education, Training and Youth Affairs, Canberra.

DEETYA (1996b), Annual Report, 1995-96, Department of Employment, Education, Training and Youth Affairs, Canberra.

DEETYA (1997a), Working Nation Evaluation--Labour Market Elements, Department of Employment, Education, Training and Youth Affairs, Canberra.

DEETYA (1997b), The Net Impact of Labour Market Programs, Department of Employment, Education, Training and Youth Affairs, Canberra.

DEEWR (2008a), The APM Evaluation, Department of Education, Employment and Workplace Relations, Canberra.

DEEWR (2008b), Submission by the Department of Education, Employment and Workplace Relations to the Senate Education, Employment and Workplace Relations Committee Inquiry into the Social Security Legislation Amendment (Employment Services Reform) Bill 2008, Parliament House, Canberra.

DEEWR (2008c), Labour Market Assistance Outcomes, Department of Education, Employment and Workplace Relations, Canberra.

DEEWR (2008d), Labour Market Assistance Longer Term Outcomes, Department of Education, Employment and Workplace Relations, Canberra.

DEEWR (2009a), Welfare to Work Evaluation, Department of Education, Employment and Workplace Relations, Canberra.

DEEWR (2009b), Department of Education, Employment and Workplace Relations Response to Senate Estimates, Question No. EW0015-10, Senate Education, Employment and Workplace Relations Committee, Parliament House, Canberra.

DEEWR (2009c), Department of Education, Employment and Workplace Relations, Responses to Questions on Notice EW24_ 10, EW 22_10, Senate Education, Employment and Workplace Relations Committee, Parliament House, Canberra.

DEEWR (2009d), Department of Education, Employment and Workplace Relations, Response to Question on Notice EWO016_ 10, Senate Education, Employment and Workplace Relations Committee, Parliament House, Canberra.

DEEWR (2009e), Annual Report, Department of Education, Employment and Workplace Relations, Canberra.

DEEWR (2009f) Response to Senate Question on Notice No. EW33_10, Disability Employment Services.

DEEWR (2010a), Labour Market Assistance: A Net Impact Study, Department of Education, Employment and Workplace Relations, Canberra.

DEEWR (2010b), Submission to the Independent Review of the New Jobseeker Compliance Framework, Department of Education, Employment and Workplace Relations, Canberra.

DEWR (2003a), Labour Market Assistance Outcomes, Department of Employment and Workplace Relations, Canberra.

DEWR (2003b), Intensive Assistance and Job Search Training, a Net Impact Study, EPPB Report No. 2/2003, Department of Employment and Workplace Relations, Canberra.

DEWR (2004), The Sustainability of Outcomes, Department of Employment and Workplace Relations, Canberra.

DEWR (2006a). Customised Assistance, Job Search Training, and Work for the Dole, a Net Impact Study, Department of Employment and Workplace Relations, Canberra.

DEWR (2006b), Jobseeker Account Evaluation Report, Department of Employment and Workplace Relations, Canberra.

DEWRSB (2000a), Background Paper for OECD Labour Market Review of Australia, Department of Employment, Workplace Relations and Small Business, Canberra.

DEWRSB (2000b), Job Network Evaluation Stage 1, Department of Employment, Workplace Relations and Small Business, Canberra.

DEWRSB (2001a), Job Network Evaluation Stage 3, Department of Employment, Workplace Relations and Small Business, Canberra.

DEWRSB (2001b), What's Driving the Change in Intensive Assistance Outcomes? in Labour Market Assistance Outcomes, Department of Employment, Workplace Relations and Small Business, Canberra.

DEWRSB (2001c), Job Network: A Net Impact Study, Department of Employment, Workplace Relations and Small Business, Canberra.

Disney, J. Buduls, A. and Grant, P. (2010), Impacts of the New Jobseeker Compliance Framework, Department of Education, Employment and Workplace Relations, Canberra.

Dockery, M. and Stromback, T. (2001), 'Devolving Public Employment Services', International Labour Review, vol. 140, pp. 429-451.

Dockery, M. and Webster, E. (2002), 'Long-term Unemployment and Work Deprived Individuals: Issues and Policies', Australian Journal of Labour Economics, vol. 5, pp. 175-194.

Dolton, P. and O'Neill, D. (1996), 'Unemployment Duration and the Restart Effect', Economic Journal, vol. 106, pp. 387-400.

Dorsett, R. (2006), 'The New Deal for Young People: Effect of the Options on the Labour Market Status of Young Men', Labour Economics, vol. 13, pp. 405-422.

Eardley, A. Abello, D. and MacDonald, C. (2001), Is the Job Network Benefiting Disadvantaged Jobseekers? Social Policy Research Centre Discussion Papers No. 111, Social Policy Research Centre, University of New South Wales, Sydney.

FACS (1999), Annual Report, 1998-99, Department of Family and Community Services, Canberra.

FACS (2001), Welfare Reform Pilots, Department of Family and Community Services Occasional Papers No. 5, Canberra.

FACS (2004), Annual Report, 2003-04, Department of Family and Community Services, Canberra.

FAHCSIA (2008a), Annual Report and Portfolio Budget Estimates, Department of Families, Housing, Community Services and Indigenous Affairs, Canberra.

FAHCSIA (2008b), Portfolio Budget Estimates, Department of Families, Housing, Community Services and Indigenous Affairs, Canberra.

Finn, D. (2001), A New Deal for Unemployed Australians?, mimeo, Dusseldorp Skills Forum, Melbourne.

Finn, D. and Schulte, B. (2007), Activation Policies in Great Britain, University of Portsmouth, Portsmouth.

Freeland, J. (1998), 'Employment and Training Policy and Programs', Evatt Papers, vol. 6, No. 2, Evatt Foundation, Sydney.

Freud, D. (2007), Reducing Dependency, Increasing Opportunity. Department for Work and Pensions, Leeds.

Frijters, P. and Gregory, R. (2006), 'From Golden Age to Golden Age: Australia's Great Leap Forward?' Economic Record, vol. 82, pp. 207-224.

Hasluck, C. and Green, A. (2007), What Works for Whom? Warwick Institute for Employment Research, Report for Department for Work and Pensions, Leeds.

Hotz, V., Imbens, G. and Klerman, J. (2000), The Long-term Gains from GAIN:A Re-analysis of the Impacts of the California GAIN Program, National Bureau of Economic Research Working Paper No. W8007, Cambridge, MA.

Johri, R., de Boer, M., Pusch, H., Ramasamy, S. and Wong, K. (2004), Evidence to Date on the Working and Effectiveness of Active Labour Market Policies in New Zealand, New Zealand Department of Labour, Wellington.

Kalisch, D. (1991), Long-term UB Recipients, the Impact of Newstart and Economic Conditions, National Social Policy Conference, Sydney, July.

Killacky, E. Jackson, H. and McGorry, P. (2008), 'Vocational Intervention in First-episode Psychosis: Individual Placement and Support Versus Treatment as Usual', The British Journal of Psychiatry, vol. 193, pp. 114-120.

Kinnear, P., Grant, D. and Oliver, K. (2003), Welfare Reform in Australia, an Evidence-based Approach, Paper for Australian Social Policy Conference July 9-11, University of New South Wales, Sydney.

Kluve, J. (2006), The Effectiveness of European Active Labor Market Policy, IZA Discussion Papers No. 2018, Bonn.

Layard, R., Nickell, S. and Jackman, R. (1991 ), Unemployment, Macroeconomic Performance and the Labour Market, Oxford University Press, Oxford.

Loedemel, I. and Trickey, H. (2001), An Offer You Can't Refuse, Policy Press, London.

Martin, J. (2000), What Works Among Active Labour Market Policies, OECD Economic Studies No. 30/1, pp. 79-113, Organisation for Economic Cooperation and Development, Paris.

O'Connor, B. (2008a), The Future of Employment Services in Australia, Department of Education, Employment and Workplace Relations, Canberra.

O'Connor, B. (2008b), Second Reading Speech, House of Representatives Hansard, 24 September 2008, Minister for Employment Participation.

OECD (1994), The OECD Jobs Study: Facts, Analysis and Strategy, Organisation for Economic Cooperation and Development, Paris.

OECD (2002a), Innovation in Employment Policies, the Australian Way, Organisation for Economic Cooperation and Development, Paris.

OECD (2002b), Education at a Glance 2002, Organisation for Economic Cooperation and Development, Paris.

OECD (2005), Employment Outlook, Organisation for Economic Cooperation and Development, Paris.

OECD (2006), Employment Outlook, Organisation for Economic Cooperation and Development, Paris.

OECD (2007a), Education at a Glance, Organisation for Economic Cooperation and Development, Paris.

OECD (2007b), Employment Outlook, Organisation for Economic Cooperation and Development, Paris.

OECD (2009a), Employment Outlook, Organisation for Economic Cooperation and Development, Paris.

OECD (2009b), Education at a Glance 2009: OECD Indicators, Organisation for Economic Cooperation and Development, Paris

Pearce, D., Disney, J. and Ridout, H. (2002), Making it Work, Independent Review of Breaches and Penalties in the Social Security System, ACOSS Paper No. 124.

Perkins, D. (2007), Making it Work: Promoting Participation of Job Seekers with Multiple Barriers through the Personal Support Program, Brotherhood of St Laurence, Melbourne.

Piggott, J. and Chapman, B. (1995), 'Costing the Job Compact', Economic Record, vol. 71, pp. 3. 13-28.

Preston, H. (1993), 'Long-term Unemployment: A Statistical Summary, Graphs, Tables', Social Security Journal, December, pp. 48-69.

Productivity Commission (2003), Report of the Independent Review of the Job Network, Productivity Commission, Canberra.

Schmid, G. (1996), 'Theory and Methodology of Labour Market Policy and Evaluation', in Schmid, G., O'Reilly, J. and Schomann, K. (eds), International Handbook of Labour Market Policy and Evaluation, Edward Elgar, Cheltenham.

Social Inclusion Board (2010), Social Inclusion in Australia, How Australia is Faring, Department of the Prime Minister and Cabinet, Canberra.

Stromback, T. and Dockery, A. (2000), Labour Market Programs, Unemployment and Employment Hazards, ABS Occasional Papers, cat. 6293.0.00.002, Canberra.

Van den Berg, G. and Van der Klaauw, B. (2001), Counselling and Monitoring of Unemployed Workers, IZA Discussion Papers No. 374.

Vanstone, A. (1996), Reforming Employment Assistance, Ministerial Statement, Minister for Employment Education and Training and Youth Affairs, Canberra.

Warburton, M., Okopu, A. and Vuong, L. (1999), 'Long-term Unemployment: A Statistical Analysis of FACS Customers', Australian Social Policy No. 2, Department of Family and Community Services.

Webster, E. and Harding, D. (2001), 'Outsourcing Public Employment Services: The Australian Experience', Australian Economic Review, vol. 34, pp. 213-242.

Wells, W. (2001), 'From Restart to the New Deal in the UK', in Labour Market Policies and the Public Employment Service, Organisation for Economic Cooperation and Development, Paris.

Wiseman, J. (2001), 'Making Work for Welfare in the US', in Loedemel, I. and Trickey, H., An Offer You Can't Refuse, Policy Press, London.


(i) Long term conventionally refers to 12 months or more. Given the focus of this paper, I rely on social security and employment assistance data rather than Australian Bureau of Statistics data on long-term unemployment, which capture a somewhat different population.

(ii) For descriptions of all of these programs see DEEWR (2008c).

(iii) Note that this assessment is based on the sum of participants in mainstream and specialist programs (subject to data availability). For example, the proportion of participants in mainstream programs identified as having disabilities declined, but when specialist programs are included the overall proportion with disabilities increased. I was unable to find data for the proportion of people of working age lacking Year 10 qualifications, but the proportion lacking Year 12 declined from 43 per cent in 1995 to 32 per cent in 2007 (OECD 2002b; OECD 2009b).

(iv) Data on placement and related activities were not consistently collected by the OECD, and this category includes some work-preparation activities. The vast bulk of employment assistance expenditure in Australia prior to the Job Network was in programs such as work experience. As indicated below, Job Network providers invested only modestly in these forms of assistance.

(v) Author's calculations from ABS, Labour Force, Australia, various years, ABS cat. 6202.0, Australian Bureau of Statistics, Canberra.

(vi) These include the Employment Pathways Survey conducted as part of the evaluation of the Welfare to Work policy (, see 'program evaluation').

(vii) Although in theory they also provided help with barriers to work such as training and work experience, as indicated previously expenditure on these services was minimal in the vast majority of cases.

(viii) This change followed criticism of previous official evaluations from independent researchers, the OECD and the Productivity Commission (Stromback and Dockery 2000; OECD 2002a; Productivity Commission 2003).

(ix) This early version of the program was targeted differently to the later versions that were officially evaluated: towards short-term young unemployed people as distinct from longer-term unemployed people generally. Also, the measure of effectiveness used was the change in probability of leaving income support. These factors alone are unlikely to explain the wide variance in results from the official studies.

(x) Note that (as discussed previously) the pre-1998 net impacts were overstated compared to later studies due to a change in methodology of the official evaluations. It seems unlikely, however, that this would close the large gap between estimated outcomes for Jobstart and later programs.

(xi) The results are also consistent with OECD analysis of the outcomes of employment programs (Martin 2000), official evaluations of the New Deal for Young People in the United Kingdom, which includes elements of all three program types (Dorsett 2006).

(xii) This assumes that a minimum level of assessment and employment assistance is undertaken by providers for all jobseekers and that the evaluation is measuring the value added by more intensive interventions.

(xiii) An official review of employment participation policies in the United Kingdom recently applied this logic to make the case for greater investment in employment programs for disadvantaged jobseekers (Freud 2007). Evaluators in the New Zealand Department for Social Development also advocated a more comprehensive cost benefits measure (Johri et al. 2004, p. 39).

(xiv) Other official studies traced gross employment or off-benefit impacts (but not the net impacts of the programs) over longer periods. These studies found that short-term employment impacts were usually sustained over periods of up to two years after program participation. These studies did not, however, use control groups, so they do not estimate what the outcomes would have been in the absence of the programs (DEWR 2004; DEEWR 2008c).

(xv) A challenge for evaluators is attrition bias. Successful long-term evaluation studies in the United States suggest that this can be overcome if the research is sufficiently well resourced. For example Hotz, Imbens and Klerman (2000) assessed the impact of Californian welfare to work programs over a nine-year period.

(xvi) A number of early official evaluations attempted to do so by comparing outcomes for unemployed people referred to a program who did not commence, and those of participants in the program, with the outcomes for the control group. The first comparison yielded estimates of a program's referral or compliance effects, while the second yielded estimates of program effects. The magnitude of the compliance effects at times exceeded that of the program effects (DEETYA 1997b; DEWRSB 2001c). Subsequent streamlining of program participation meant that this evaluation method was no longer feasible (DEWR 2006a).

(xvii) This provides an estimate of the impact of the Newstart interviews, as distinct from any programs to which unemployed people were subsequently referred.

(xviii) In regard to sole parents, the evaluation controlled for the impact on exits from income support of the tighter income test applying to Newstart Allowance compared with Parenting Payment Single.

Peter Davidson, Social Policy Research Centre, University of New South Wales.
Table 1: Profile of Participants in Programs for Long-term
Unemployed People

                             1995       1999          2003
                             (Working   (Intensive    (Intensive
                             Nation)    Assistance)   assistance)
Job Compact
and Job Network *                    % of commencements

1 to 2 years unemployed      25         25            17
Over 2 years unemployed      40         39            36
Indigenous                   6          5             7
Disability                   13         17            11
Over 44 years old            12         29            32
Lacking Year 10 education    24         31            26
Sole parent                  3          1             2
Total annual                 224,283    438,500       228,600
  commencements (Job
  Compact and Job
  Network) *

Specialist programs

Personal Support Program     0          2             15
Disability programs (+)      NA         NA            NA.
Indigenous employment        3          1             3
Annual commencements         7,718      15,738        47,500
  (PSP, disability
  programs and IEP)

Annual commencements         232,001    454,238       276,100

                             2008          (average
                             (Customised   employment
                             assistance)   outcomes) (~)
                                           % employed
Job Compact
and Job Network *            % of commencements

1 to 2 years unemployed      19            47
Over 2 years unemployed      41            45
Indigenous                   22            33
Disability                   11            38
Over 44 years old            25            42
Lacking Year 10 education    26            36
Sole parent                  19            52
Total annual                 210,103       47
  commencements (Job
  Compact and Job
  Network) *

Specialist programs

Personal Support Program     19            20
Disability programs (+)      15
Indigenous employment        8             68
Annual commencements         111,957
  (PSP, disability
  programs and IEP)

Annual commencements         322,060

* Job Network commencements refer to Intensive Assistance and
Customised Assistance only.

(#) Commencements in each of the specialist programs as a
percentage of (commencements in that program plus commencements
in the relevant mainstream program, for example Customised

(+) Disability Employment Network and Vocational Rehabilitation
Services specifically for unemployment payment recipients:
program commencements were 44 thousand in 2008.

(~) Percentage employed six months after commencement in
Customised Assistance in 2007-08

Sources: DEEWR (2008a); FAHCSIA (2008a); FAHCSIA (2008b); FACS
(2004; 1999), Annual Reports 2003-04 and 1998-99; DEETYA (1996b);
DEETYA(1997a); DEWRSB (2001); DEWR (2003a, b); DEEWR (2008c, d).

Table 2: Costs, Targeting and Net Employment Impacts of
Australian Programs

                                         % of          Net
                                     participants   employment
                         Unit cost     who were     impact (%)
                         (current     long-term     (official
                         dollars)     unemployed     studies)
(Working Nation)

Jobtrain                 1,173       63             7
Jobstart                 1,263       63             28
Jobskills                7,105       96             11
New Work                 10,009      93             4

2002 (Job Network
and Work for the

Intensive                924         54             6.2
Work for the Dole        2,000       58             4

2005 (Job Network
and Work for the

Customised Assistance    1,078       66             10.1
Work for the Dole        NA          68             7.3

                          impact (%)
(Working Nation)

Jobtrain                 3.6 **
Jobstart                 42.8 **
Jobskills                13:7 **
New Work

2002 (Job Network
and Work for the

Intensive                NA
Work for the Dole        -12.1 ***

2005 (Job Network
and Work for the

Customised Assistance    NA
Work for the Dole        NA

Sources: DEETYA (1997b); Stromback and Dockery (2000); DEWR
(2006a); DEWR (2003a); DEEWR (2008 c, d).

Note: Jobtrain offered short-term vocational training; Jobstart
offered subsidised employment, mainly in the private sector;
Jobskills offered subsidised employment and training, mainly in
the community and public sectors; New Work Opportunities (NOW)
offered fully subsidised employment, mainly in the community and
public sectors; Intensive Assistance and Customised Assistance
offered mainly job-search assistance; and Work for the Dole
offered unpaid work experience in the community and public

Stromback and Dockery (2000) did not distinguish between NWO and

* Percentage point increase in the probability of employment six
months after program participation (Working Nation programs) or
12 months after program commencement (post-Working Nation

** Stromback and Dockery (2000).

*** Borland and Tseng (2004) Change in the probability of leaving
income support 12 months after commencement in the program. Note
that this study assessed the earlier pilot phase of Work for the
Dole in 1998, which had a different target group (mainly
short-term unemployed young people) than it had later on (mainly
long-term unemployed people of various ages).

Apart from Intensive Assistance (which usually lasted for up to12
months), programs generally had a maximum duration of six months.

Table 3: Short-term Average Net Employment Impacts of Programs
in OECD Countries after the Late 1990s

                                   Difference in likelihood
                                   of a significant positive
                                   employment outcome
                                   compared to vocational
                                   training (%)

Direct employment                  -31
Private sector incentive scheme    +28
Services and sanctions             +38
Vocational training                0

Source: Kluve (2006).

Note: 'Direct employment' refers to public sector 'job creation'
schemes; 'private sector incentive scheme' refers to wage subsidy
schemes; 'services and sanctions' refers mainly to job search
COPYRIGHT 2011 National Institute of Labour Studies Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2011 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Contributed Article
Author:Davidson, Peter
Publication:Australian Bulletin of Labour
Geographic Code:8AUST
Date:Dec 1, 2011
Previous Article:'It's a lot of hardwork': the experiences of student-workers in university term-time employment.
Next Article:The 'rise and rise' of new professional groups: mental health professions under Medicare.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters