Hiring subsidies for people with a disability: Evidence from a small‐scale social field experiment

• Published date: 01 June 2017
• Date: 01 June 2017
• DOI: http://doi.org/10.1111/j.1564-913X.2014.00025.x
• Author: Lukas KAUER,Eva DEUCHERT

International Labour Review, Vol. 156 (2017), No. 2

Copyright © The authors 2017

Journal compilation © International Labour Organization 2017

* Center for Disability and Integration, School of Economics and Political Science, Uni-

versity of St Gallen, emails: eva.deuchert@unisg.ch and lukas.kauer@unisg.ch. The authors wish

to acknowledge the nancial support provided by the Ernst Göhner Foundation and the excellent

research assistance provided by Karina Schindler. They would like to thank all of the participants

and coordinators at the rehabilitation centres and the local disability insurance ofce. They are

also grateful for the helpful comments they received from Michael Lechner, Per Johansson,

Reto Foellmi and participants in the 4th Biennial Conference of the American Society of Health

Economists, the international conference “Field Experiments in Policy Evaluation” held in Nur-

emberg, the 3rd Workshop “Arbeitsmarkt und Sozialpolitik” held in Dresden, the 6th “Maastricht

Behavioral and Experimental Economics Symposium”, and the 12th IZA/SOLE Transatlantic

Meeting of Labor Economists.

Responsibility for opinions expressed in signed articles rests solely with their authors, and

publication does not constitute an endorsement by the ILO.

Hiring subsidies for people

with a disability: Evidence from

a small-scale social eld experiment

Eva DEUCHERT* and Lukas KAUER*

Abstract. The effectiveness of hiring subsidies for people with disabilities remains

unclear due to potential free-rider, substitution and signalling effects. The authors

propose a novel evaluation approach wherein it is randomly decided whether or

not job applications disclose the subsidy to potential employers. Based on call-back

rates for interviews, the subsidy is found to be ineffective or even counterproductive

in a group of adolescents having completed their vocational training programme.

However, the negative signalling effect seems to be much weaker in a group of cli-

ents of job-coaching services who acquired their disability during their working life.

H

iring subsidies are common in many industrialized countries (Marx, 2001;

Katz, 1996). These programmes are designed to reduce labour costs

and to stimulate employment for disadvantaged groups (such as young people,

welfare recipients, or people with disabilities). As discussed by Neumark

(2011), however, the effectiveness and efciency of these programmes crucially

depend on whether they generate free-rider effects (where hiring would have

taken place even without the subsidy), substitution effects (if employment of

the targeted group increases at the expense of employment in other, non-tar-

geted groups), and signalling effects (if employers perceive hiring subsidies as

a signal for lower productivity).

International Labour Review270

Since hiring subsidies carry considerable costs and can even be harm-

ful, it is necessary to evaluate them carefully. In this article, we empirically

evaluate the effectiveness of a hiring subsidy scheme targeted at people with

disabilities. From a policy perspective, this is an extremely important topic:

the number of people with disabilities is high – on average across the OECD

14 per cent of the working-age population classify themselves as disabled – and

this group is particularly disadvantaged in the labour market (OECD, 2010).

Many countries seek to improve their employment prospects by implement-

ing subsidy schemes (OECD, 20 03). For such schemes to be effective, poten-

tial employers need to be informed about the hiring subsidy (particularly if

not all persons with a disability are eligible). However, broaching the possibil-

ity of hiring subsidies during the job application process inevitably means that

the applicant must disclose the disability. The signalling effect is thus likely to

occur, particularly if employers associate disability with lower productivity.

There is increasing academic interest in the effectiveness of employment

promotion measures for people with disabilities,1 but there is very limited em-

pirical evidence on the effectiveness of hiring subsidies targeting these people.

The literature mostly focuses on the Danish “Flexjob” scheme, which provides

salary reimbursements to rms that employ long-term disabled persons whose

work capacity is permanently reduced. Gupta and Larsen (2008 and 2010) es-

timate the effect of this scheme by using the introduction of the scheme in

1998 as a quasi-experiment. They nd a sizeable employment effect, but dif-

ferent versions of their article demonstrate how difcult it is to nd a suitable

control group of individuals who are not targeted by hiring subsidies. Gupta,

Larsen and Thomsen (2013) focus on rms and use a change of the reimburse-

ment structure that applies to government rms only (but not to other public-

sector employers, such as municipal and regional rms). Unfortunately, this

study provides little descriptive evidence on the scheme’s overall effective-

ness – mainly because the assumption of parallel trends between the targeted

and control groups is likely to be violated. Indeed, this research demonstrates

that it is inherently difcult to nd convincing empirical strategies to predict

the counterfactual outcomes that would occur if people with disabilities were

not receiving subsidies.

The empirical literature on hiring subsidies for the unemployed and wel-

fare recipients proposes three different methods to solve this problem. First,

the “gold standard” is to randomize eligibility for a subsidy ex ante. This is

typically done in voucher experiments, where a randomized proportion of the

target population receives a voucher for a wage subsidy, which can be handed

over to an employer to be cashed in from a government agency (e.g. Bell and

Orr, 1994; Burtless, 1985; Dubin and Rivers, 1993; Galasso, Ravallion and Sal-

via, 2004; Woodbury and Spiegelman, 1987). Second, in situations where eld

experiments are not feasible, natural experimental designs are used for ex post

1

See, for example, Malo and Muñoz-Bullón (20 06); Malo and Pagán (2014); Humer, Wuellrich

and Zweimüller (2007); Lalive, Wuellrich and Zweimüller (2013).