Randomised control trials of unemployment benefits: An example from Central Europe

John Micklewright, Gyula Nagy

30 April 2008



In their authoritative review of unemployment in OECD countries, Stephen Nickell and his colleagues have argued that the way in which unemployment benefits are administered is crucial in determining the extent to which generous benefit systems actually influence levels of unemployment. A story of high levels of benefit encouraging high levels of unemployment is too simple. If claimants’ efforts to find jobs are closely monitored, with sanctions if search for work is considered inadequate, then benefits can be generous without this encouraging the unemployed to remain unemployed.

One way of assessing changes to policy on the administration of benefits is to conduct a randomised control trial (RCT), in which claimants are randomly allocated to a ‘treatment’ group whose claims are administered according to the new policy that is being considered and to a ‘control’ group where administration continues as before. RCTs – also known as ‘field’ experiments – of this sort offer the hope of a clear answer to the question as to the effect of a proposed policy change. They have seen quite a lot of use in the United States but have been rarely used in Europe.

Together with the Hungarian Ministry of Labour, we designed a RCT of a possible change to the administration of unemployment insurance (benefit secured by previous work history) in Hungary. To our knowledge, this was the first RCT of any benefit programme in Central Europe, and one of only few anywhere in Europe. Systems of unemployment benefit in Central and Eastern Europe have a relatively short history. The absence of open unemployment in planned economies meant that income support for people searching for work did not exist prior to the 1990s. The debate about the behavioural impact of the new benefit systems has been considerable but has focused on benefit levels and lengths of entitlements. As economies contracted sharply in the early 1990s, the administration of benefits concentrated on delivery of payments to claimants. The subsequent recovery, and hence greater availability of jobs, has prompted more consideration of the monitoring of the search for work by people receiving benefit.

A randomised control trial

Our experiment was fairly modest in scope. The ‘treatment’ was more frequent attendance at the local employment office – every three weeks – plus questioning about search for work – employers contacted, jobs applied for, etc. The control group had to attend only every three months and faced no questions. It is worth emphasising that this regime of very light monitoring for the control group was the norm at the time of the experiment. The experiment lasted for four months and involved 2,000 benefit recipients at employment offices in different parts of Hungary.

For women in their 20s and for men of all ages we found no significant impact. But for women aged 30+ the treatment did seem to reduce the time they stayed unemployed and to increase the speed at which they returned to work. Figure 1 shows the proportion of women of this age in each group – treatment and control – who were still unemployed at each point. The dashed line for the treatment group falls below the solid line for the control group and the difference is statistically significant. A quarter of the control group had stopped claiming benefit after 102 days but among the treatment group a quarter had gone by only 85 days. (The majority of both groups were still unemployed when the experiment ended, reflecting the low level of the outflow from the unemployment register.)

Figure 1 Proportion of women over 30 still unemployed

There are three alternative explanations for the results. First, claimants of all ages and both sexes raise their efforts to find work as a result of treatment. But the effort of men and younger women is already high and the marginal return to their additional effort to find work is zero. Second, the monitoring of search for work under the treatment is not binding for men and younger women and no additional effort results – their effort is already high. Third, treatment does in practice bring increased frequency of contact with the employment office but this of itself does not result in greater search for work among the men and younger women. Only the women aged 30+ take advantage of the increased access to information on vacancies and only these women experience discomfort from the questioning and perceive a threat of sanctions if they do not increase their efforts to find work. We could not conclude with certainty which of these explanations is correct. But on balance, other evidence suggested that the third explanation was the most likely.

Hungarian policy concerns

Since our experiment was conducted, the Hungarian government has changed aspects of the administration of unemployment benefit in ways that partly reflect some of its features. What has been the impact? Labour Force Survey data appear to show a greater proportion of benefit recipients who report that they are searching for jobs than before the change in policy. The rise has been particularly notable for women. While it is impossible to conclude that these changes have been a result of the policy change, it does underline the ‘live’ nature of policy surrounding benefit administration in Central Europe and the possible differences between men and women uncovered by our experiment.

Finally, it is worth emphasizing that the majority of the unemployed in Hungary do not receive any unemployment benefit at all, on account of their employment history, their length of unemployment, or the level of other incomes in the household. So a focus on the regime for administering benefits is only looking at part of the picture. As in other OECD countries, Hungarian policy makers should also be concerned about the adequacy of coverage of unemployment benefit systems.


Micklewright, J and G Nagy (2008). ‘Job Search Monitoring and Unemployment Duration: Evidence from a Randomised Control Trial’ CEPR Discussion Paper 6711.
Nickell, S, L Nunziata, and W Ochel (2005). ‘Unemployment in the OECD since the 1960s. What do we know?’ Economic Journal, 115: 1-27.



Topics:  Labour markets

Tags:  unemployment, benefits, Hungary

Professor of Social Statistics and Policy Analysis, University of Southampton and CEPR Research Fellow

Associate Professor of Labour Economics, Corvinus University of Budapest