A Micro-Econometric Evaluation of the UK Work Programme
Total Page:16
File Type:pdf, Size:1020Kb
'I, Daniel Blake' revisited: A micro-econometric evaluation of the UK Work Programme Danula K. Gamage∗ Pedro S. Martinsy Queen Mary University of London Queen Mary University of London CRED & NovaSBE & IZA June 19, 2017 Work in Progress Abstract Although many countries are making greater use of public-private partnerships in em- ployment services, there are few detailed econometric analysis of their effects, in contrast to a large body of small-sample or qualitative case studies. This paper contributes to this literature by examining the case of the UK Work Programme, drawing on popula- tion data of all nearly two-million participants between 2011 and 2016. We also exploit the original structure of the programme to disentangle the impact of different provider and jobseeker characteristics from business cycle, cohort, regional and time-in-programme effects. Moreover, we consider both transitions to employment and transitions out of unemployment. Our main results indicate considerable differences in performance across providers and across jobseeker profiles. The latter results suggest that, by changing the incentive structure offered to providers, the government could obtain better results at the same cost. Keywords: Public employment services, job search, public policy evaluation. JEL Codes: J64, J68, J22. ∗Corresponding author. Email: [email protected], Address: School of Business and Management, Queen Mary, University of London, Mile End Road, London E1 4NS, United Kingdom. yEmail: [email protected]. Address: School of Business and Management, Queen Mary, University of London, Mile End Road, London E1 4NS, United Kingdom. Web: http://webspace.qmul.ac.uk/pmartins 1 1 Introduction Focusing on the individual case of a fictional elderly widower, the award-winning film 'I, Daniel Blake' portraits a negative facade of UK welfare-to-work programmes over the last years. In this case, 'Daniel Blake' is incorrectly denied welfare payments and required to actively search for jobs, with ultimately tragic consequences. This film has contributed to the lively debate on the merits and demerits of welfare programmes including their activation dimensions, as evidenced in the recent general elections in the UK. This paper amounts to an additional contribution to this debate, although adopting a very different methodology. We study econometrically the Work Programme (WP henceforth), the leading employment services or activation initiative conducted in the UK since 2011. Amongst other distinguishing features, the WP was delivered through contracts with private and non- for-profit providers, with a total value of almost 5 billion. These providers were tasked to complement the activation efforts of jobcentres (public employment services), in particular with jobseekers that are considered harder to help and more distant from a full return to the labour market. The WP was also characterised by payments to providers conditional on the results achieved and on the jobseeker profiles. In this context, our analysis seeks to understand to what extent an assignment to the WP leads to transitions out of unemployment and or to employment. It also seeks to compare the effectiveness of different providers and the impact of the programme across groups of jobseekers (and contrast this impact with the financial incentives offered by the Department for Work and Pensions). More broadly, we are also interested in shedding light on the potential role of the WP in the resilience of the UK labour market since 2011. Our analysis draws on DWP population micro data of over 1.9 million individual partici- pants between 2011 and 2016 and their outcomes, in particular transitions out of unemploy- ment and transitions into employment. The data is based on generally small groups of job- seekers defined by the month when the individual is referred to the WP (from June 2011 until December 2016), the region ("contract package area") where the individual lives (18 different regions), the provider of the services (15 different organisations), and the profile ("payment group") of the jobseeker (10 different profiles, corresponding to 3 different groups of jobseekers - JSA, ESA and IS/IB). Except for the cohorts initiated since 2015, all cohorts are followed monthly for 24 months, leading to a data set of 510,078 cohort/month/profile/region/provider 2 observations. The econometric analysis is based on the estimation of equations of different outcomes on several groups of fixed effects (start months, current months, months in WP, regions, providers, and jobseeker profiles), resulting in a total of about 150 regressors. Provider and region fixed effects can be identified simultaneously as there are multiple providers per region and providers operate in multiple regions. The same applies to the three sets of month fixed effects, as multiple cohorts are observed at any given month, cohorts are followed for up to 24 months and multiple cohorts are observed at any given duration of their spell. Our preliminary findings indicate that, on average, 10% (30%) of a cohort is employed (off benefits) by the 12th month in the WP. These rates increase to 21% (39%) by the 24th month. We also find that provider, jobseeker characteristics and number of months in the WP play a significant role in explaining the variation in the job outcomes (transition into employment). Indeed, we find considerable variation across providers, which we use to establish their relative performance, a finding that may be of interest for DWP. These results underpins the results established by Dorsett and Lucchino, 2016 which examines the variation in programme service delivery among prime providers. Moreover, the differences in job outcomes across jobseeker groups that we uncover suggest that the DWP payment structures could be redesigned to increase job outcomes for the same expenditure and or to obtain the same job outcomes while reducing expenditure. 2 Literature Review Most European Union countries, employment related policies comprise of passive and active labour market dimensions. The passive element engages all transfer that help and support the unemployed or facilitates (early) retirement such as unemployment benefit, unemployment social assistance or early retirement support. Moreover, active element comprise of those poli- cies, targeted at specific groups of unemployed and aims at facilitating their (re)insertion to the labour market such as counselling, monitoring of job search efforts, training and employ- ment subsidies, among others. There is an extensive theoretical literature on the effects of activation strategies using a standard job search model. The underlying principle of activation strategies is that it increases 3 the cost of unemployment that results in an increase in the reservation wage1. Ultimately resulting in greater job search effort and transition out of unemployment. This principle have been confirm by empirical studies but many questions the destination state for these individ- uals who unemployment. More stringent activation policies may demand greater job search effort from the unemployed and they would prefer to move into a non-subsidised unemploy- ment state and resulting in low/no search effort. On the contrary decrease in reservation wages may lead to higher job acceptance among the unemployed resulting in transition into employment. These two potential effect are further documented empirically. The most com- mon definition of activation strategies is that of the OECD is "bringing more people into the effective labour force, to counteract the potentially negative effect of unemployment and related benefits on work incentives by enforcing their conditionality on active job search and participation in measures to improve employability and to manage employment service and other labour market measures so that they effectively promote and assist the return to work" (Martin, 2014). However, even within this definition, it is common to consider that activation strategies are, generally of twofold. On one hand, a monitoring and sanctions procedure which can be identifies as the "stick" and on the other the counselling, job assistance and ALMP which is the "carrot". In the implementation of activation strategies both concepts go hand in hand. However, there are some studies that were able to evaluate the impact of only on the two dimensions. One of the first and of the few that managed to distinguish between monitoring and counselling is Klepinger et al., 2002, who studied the impact of the tighter search requirement (a doubling of the standard number of employer contacts per month), monitoring (verification of work-search contacts) and counselling (participation in a job search workshop) on total UI paid and UI insurance, following a 1994 random experiment in Maryland. They find that both more demanding search requirement and monitoring lead to a decrease of around 6% in the UI payments and a decrease of between 6 - 7% in the number of weeks of UI. Counselling also reduced total UI payment and unemployment duration. Klepinger et al., 2002 also examined the impact of employment and earnings and found that little impact on both. Similar results were found for other countries, in particular in the UK (Dolton and O'Neill, 2002) and Australia (Borland and Tseng, 2007). 1Refers to the wage that that makes an unemployed indifferent between remain unemployed or to accept the job offer. 4 Another strand of literature examines the transition out of unemployment and transi- tion into employment and finds that the former is accompanied by the