A Meta-analysis on Labour Market Deregulations and Employment Performance: No Consensus Around the IMF-OECD Consensus

ABSTRACT

The so-called ‘IMF-OECD consensus’ suggests that labour market deregulations increase employment and reduce unemployment. This paper presents a meta-analysis of research on this topic based on MAER-NET guidelines. We examine the relation between Employment Protection Legislation indexes on one hand, and employment and unemployment on the other. Among 53 academic papers published between 1990 and 2019, only 28 per cent support the consensus view, while the remaining 72 per cent report results that are ambiguous (21 per cent) or contrary to the consensus (51 per cent). The decline in support for the consensus view is particularly evident in the last decade. Our results are independent of the citations of papers examined, the impact factor of journals and the techniques used. A FAT-PET meta-regression model confirms these outcomes.

KEYWORDS:

• Labour market
• employment protection legislation
• unemployment
• meta-analysis
• meta-regression

JEL CLASSIFICATION:

• B5
• C83
• E24
• J48
• K31

Acknowledgements

This paper develops a piece by Emiliano Brancaccio used in a debate with Stefano Scarpetta on 27 June 2019 at the University of Siena (Italy) as part of the INET young scholars activities and STOREP Conference. We are grateful to Antonella Stirati, Fabrizio Amendola, Dean Baker, Enrico Bellino, Tito Boeri, Luigi Cavallaro, Roberto Ciccone, Simon Deakin, Giovanni Dosi, Pietro Garibaldi, David Howell, Alessandro Nuvolari, Andrea Roventini, John Schmitt, Prabirjit Sarkar, Stefano Scarpetta, Per Skedinger, Federico Tamagni and two anonymous referees for their useful insights. The usual disclaimers apply.

Disclosure Statement

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

Notes

1 The choice to carry out a check on the impact factor does not mean that we share the prevailing practice of evaluating publications on the basis of strict bibliometric criteria or journal rankings (on this point see Brancaccio and Garbellini 2018).

2 We will not adjust t-statistics for the degrees of freedom (like Kemper 2016) since our sample is made up of studies analyzing longitudinal data and using a relatively high number of observations (so that the degrees of freedom in the single regression is not a concern).

3 In case of a statistically significant publication bias (${\text{β}}_1\ne 0$), the FAT-PET-PEESE model must be estimated (see Stanley and Doucouliagos 2012). As we shall see, this is not our case. Then, we do not discuss FAT-PET-PEESE model any further.

4 Results available from the authors upon request.