Does eliminating benefit eligibility requirements improve unemployed job search and labour market outcomes?

ABSTRACT

Benefit eligibility requirements intend to incentivize the unemployed to find work more quickly. Our results, in an Australian context, suggest that those subjected to benefit eligibility requirements, despite searching at least as hard, take longer to find employment. Moreover, they spend less time in employment in the first twelve months and, if employed, have jobs with lower wages and fewer hours compared to otherwise similar unemployed without benefit eligibility requirements. Our findings are consistent with cognitive theories that emphasize that benefit eligibility requirements externalize job search motivation and increase stress, both of which reduce employment search effectiveness.

KEYWORDS:

• Benefit eligibility requirements
• job search outcomes
• trust experiments
• propensity score matching
• job search stress/motivation

I. Introduction

Local governments are increasingly using ‘trust experiments’ that waive job search requirements, compliance obligations and sanctions normally imposed on the unemployed; in favour of unconditional trust in the efforts and initiatives of the unemployed (Groot, Muffels, and Verlaat 2019). Proponents of trust experiments argue that relieving the unemployed of stresses surrounding Benefit Eligibility Requirements (BERs) enhances their ensuing labour market outcomes.

Inspired by labour supply and search theory and early experimental evidence (e.g. Meyer 1995), BERs were introduced to induce the unemployed to (a) search harder, supposedly leading to higher probability and/or quality of future employment, or (b) reduce their reservation wage, purportedly leading to higher probability of future employment at the expense of its quality. However, cognitive theories predict different outcomes. For example, Conservation of Resources theory (COR) argues that the threat of losing resources (e.g. unemployment benefits) causes stress and fatigue, which reduces job search effectiveness leading to lower re-employment quality (Hobfoll 1989; Lim et al. 2016). Further, Self Determination Theory (SDT) argues that external search motivation (e.g. emanating from BERs) – as opposed to more autonomous motivation – compromises goal achievement (Deci and Ryan 2000). Van Hooft, Wanberg, and Van Hoye (2013) and Koen et al. (2016) confirm that external motivation reduces job search engagement and makes job search haphazard, respectively. Both effects compromise the concomitant employment probability and quality (Koen et al. 2016; Gerards and Welters 2020; Caliendo, Tatsiramos, and Uhlendorff 2013).

Both labour supply and search theories and cognitive theories agree that the threat of repercussions of BER non-compliance provides the unemployed additional job search motivation i.e. BERs increase search intensity, but cognitive theories diverge regarding predicted effects on employment probability and quality. We empirically investigate how BERs affect search intensity, employment probability and quality outcomes in Australia (where BERs are called ‘mutual obligations’), where we find the exogenous variation and rich set of covariates we exploit in our estimation strategy.

Australia introduced Mutual Obligations (MOs) in 1997 to reduce welfare dependency. MOs consist of compulsory attendance to activities including job search training sessions, intensive assistance sessions (one-on-one job search support) and Work-for-the-Dole (the default activity).¹ Penalties for breaching MOs vary from periods of partial to periods of complete withdrawal of income support (Davidson and Whiteford 2012). Both the threat of income loss and having to spend time on compulsory activities are sources of stress and reduced autonomy to job seekers (e.g. Saunders 2007). Using difference-in-differences techniques, Richardson (2002) and Lim (2008) find modest increases in job seekers exiting social assistance schemes to which the (threat of) BERs applies.

To our knowledge, we are the first to assess the effects of Australian BERs on job search intensity and a range of employment outcomes. We use all waves of the Household, Income and Labour Dynamics in Australia (HILDA) (spanning 2001–2019) in a propensity score matching (PSM) analysis. PSM requires variations in BER (not explained by other exogenous variables) across similar unemployed people. These arise for two reasons in our data. First, failure – particularly immediately after the introduction of or eligibility changes to BERs – of case managers to implement BERs uniformly (Richardson 2003; Borland 2014) implies dissimilar treatment for similar persons. As Richardson (2003, 90) puts it: ‘ … MOI requirements were not strictly enforced’. Second, policy changes in BER applicability during 2001–2018 (e.g. extending BER applicability across age cohorts) imply dissimilar treatment for similar persons across time.

II. Materials and methods

The HILDA includes extensive information on labour market dynamics and income, allowing inclusion of an array of covariates including personality traits, and, exploiting the panel character, several lagged versions of time-variant variables. We focus on the unemployed, looking for work, aged 15–65 (sample size: 6,253).

The independent variable is exposure to MOs. Respondents were asked whether the public employment services require activities such as Work-for-the-Dole or job search training. We create a dummy equalling 1 if any such activities were undertaken at the time of the survey (time ‘t’) and 0 otherwise. We measure job search intensity at ‘t’ as hours spent in job search in the survey week. Job search outcomes are (1) ‘time-to-employment’ and (2) ‘time-in-employment’ (both are counts of months between ‘t’ and ‘t + 12ʹ), and (3) the employment status at ‘t + 12ʹ (1 if employed). For those employed at ‘t + 12ʹ we observe job quality (not for other jobs, if any, between ‘t’ and ‘t + 12ʹ). Job quality indicators are the hourly and weekly gross wage and the number of hours worked.

An unemployed person’s characteristics and circumstances drive MO applicability and labour market outcomes simultaneously. Hence, PSM must create pairs of unemployed persons that are similar in all relevant personal characteristics and circumstances, except for exogenously determined MO applicability differences (i.e. the treatment).

Table A1 (supplementary appendix) shows separately for the unemployed with and without MOs the (differences in) means (before matching) for the covariates included in the PSM analysis to satisfy conditional independence. As we match respondents across time, we include nine covariates to control for changes in labour market environment. Following Gerards and Welters (2020) and Caliendo, Gielen, and Mahlstedt (2015), we apply PSM using the Epanechnikov kernel and a bandwidth of 0.06. The mean and median standardized biases in Table 1 summarize successful matching; all are well below the recommended 3–5% range (Caliendo and Kopeinig 2008). Tables A2, A3 and Figure A1 (supplementary appendix) contain detailed indicators of the successful matching, the propensity score estimates, the propensity score distribution, respectively.

Table 1. Matching estimates of MOs on search intensity and job outcomes

	ATT	Se	N	n treated	n untreated	Off support	Mean Bias	Median Bias
Search intensity at ‘t’	0.378	0.264	6,253	1,208	5,041	4	1.1	0.7
Job search outcomes
Months-to-employment between ‘t’ and ‘t + 12’	0.511**	0.214	3,594	731	2,861	2	1.3	1.1
Months-in-employment between ‘t’ and ‘t + 12’	−0.484**	0.206	3,594	731	2,861	2	1.3	1.1
Employed at ‘t + 12’	−0.020	0.023	3,594	731	2,861	2	1.3	1.1
Job quality variables
Wage (log hourly wage)	−0.066**	0.033	2,418	371	2,043	4	1.8	1.6
Wage (log weekly wage)	−0.137***	0.047	2,418	371	2,043	4	1.8	1.6
Number of hours worked	−1.710*	0.880	2,418	371	2,043	4	1.8	1.6

Note: Propensity scores separately calculated for all matching estimates. Coefficients are average treatment effects on the treated. We use the Epanechnikov kernel, common support, bandwidth 0.06, robust standard errors (499 bootstraps) clustered on the individual. *** p < 0.01, ** p < 0.05, * p < 0.1.

III. Results

Table 1 shows that the unemployed subjected to MOs sustain (if not increase) their search intensity, take longer to find employment and spend less time in employment within the first twelve months since the identification of the MO. Although MOs do not affect employment probabilities twelve months later, if a job is secured, those with MOs are in lower quality jobs in terms of hourly wage, hours worked and (their product) weekly wage, than otherwise identical unemployed without MOs. Sensitivity and attrition selectivity analyses (supplementary Tables A4–A5) show the robustness of these results and no systematic attrition. Further, following Clarke, Méndez, and Sepúlveda (2020) and Augsburg et al. (2015), we present p-values corrected for multiple hypothesis testing (Romano and Wolf 2005; Clarke, Romano, and Wolf 2019) in supplementary Tables A6–A7. Testing the three job search outcomes at once (Table A6) and the three job quality outcomes at once (Table A7), all effects remain significant. Thus, our findings survive these demanding corrections.

IV. Discussion

MOs may cause lock-in effects (time allocated to fulfiling MOs is time unavailable for job search) or reductions in reservation wages. However, the combination of sustained search intensity, longer ‘time-to-employment’ and less ‘time-in-employment’ rules out both these explanations for our findings. In contrast, COR and SDT can explain sustained job search intensity and poorer job search outcomes, highlighting the adverse effect of MOs on job search quality. Based on these theories and related empirical evidence, we hypothesize that the unemployed subject to MOs, exhibit lower job search quality, explaining the negative labour market outcomes we observe. Future research should study both job search intensity and quality effects of BERs and extend our first investigation, looking for institutional settings enabling the use of more advanced identification strategies.

V. Conclusions

Using Australian panel data, we find evidence suggesting that the unemployed subjected to MOs sustain their job search intensity, yet take longer to become re-employed and spend less time in employment compared to otherwise identical unemployed searching without MOs. If they find employment, those with MOs are in comparatively lower quality jobs. Our findings accord with COR and SDT alluding that MOs are an external motivator and stressor that lowers job search quality.

As breaching MOs is punishable by (partial) withdrawal of income support, MOs constitute a stress inducing eligibility requirement that advocates of ‘trust experiments’ argue should be eliminated to improve labour market outcomes. Our results accord with this view that supports more unconditional trust relations with benefit recipients.

Disclosure of potential conflicts of interest

No potential conflict of interest was reported by the author(s).

Acknowledgments

* This paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Project was initiated and is funded by the Australian Government Department of Social Services (DSS) and is managed by the Melbourne Institute of Applied Economic and Social Research (Melbourne Institute). The findings and views reported in this article, however, are those of the authors and should not be attributed to either DSS or the Melbourne Institute. We thank the Editor and an anonymous reviewer for their comments. We thank the participants of Maastricht University’s Learning & Work poster session for their helpful comments and Dr. Daniel Grainger for proofreading our manuscript. We also thank Britta Augsburg, Didier Fouarge and especially Damian Clarke for their kind assistance with some of the Stata coding.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

¹ See Davidson and Whiteford (2012) for details.