STRUCTURAL EMPIRICAL EVALUATION OF JOB SEARCH MONITORING

• Published date: 01 May 2019
• Date: 01 May 2019
• DOI: http://doi.org/10.1111/iere.12376
• Author: Bas Klaauw,Gerard J. Berg

INTERNATIONAL ECONOMIC REVIEW

Vol. 60, No. 2, May 2019 DOI: 10.1111/iere.12376

STRUCTURAL EMPIRICAL EVALUATION OF JOB SEARCH MONITORING∗

BYGERARD J. VAN DEN BERG AND BAS VAN DER KLAAUW 1

University of Bristol, UK, IFAU-Uppsala, IZA, ZEW, CEPR, and CESifo; VU University

Amsterdam, The Netherlands, Tinbergen Institute, and CEPR

To evaluate search effort monitoring of unemployed workers, it is important to take account of post-

unemployment wages and job-to-job mobility. We structurally estimate a model with search channels, using

a controlled trial in which monitoring is randomized. The data include registers and survey data on search

behavior. We find that the opportunity to move to better-paid jobs in employment reduces the extent to which

monitoring induces substitution toward formal search channels in unemployment. Job mobility compensates for

adverse long-run effects of monitoring on wages. We examine counterfactual policies against moral hazard, like

reemployment bonuses and changes of the benefits path.

1. INTRODUCTION

Generous unemployment benefits schemes are potentially subject to moral hazard: Unem-

ployed workers reduce search effort and increase their reservation wage. This may reduce their

transition rate to work. Policymakers have become interested in approaches to counteract this

by using alternative policy measures (e.g., OECD, 2007). The monitoring of job search behavior

(including the threat of punitive benefit reductions) and the payment of reemployment bonuses

are examples of such policy measures. These measures, as well as the simple policy device of

lowering unemployment benefits, have the disadvantage that they tend to reduce the quality

of the post-unemployment jobs (e.g., Acemoglu and Shimer, 2000). The relative merits of the

various policy measures can only be assessed empirically. In the absence of a large range of

randomized controlled trials, such an assessment involves counterfactual evaluations.

This article provides a structural analysis based on a randomized social experiment of a

monitoring scheme for unemployed workers. Structural analysis aims at uncovering policy-

invariant parameters characterizing individual preferences and boundary conditions resulting

from labor market imperfections and policy constraints. Our data include post-unemployment

outcomes like wages and job durations. The observation of such outcomes is important because

it enables us to address the extent to which policies against moral hazard have detrimental effects

on them. Moreover, it enables us to address the importance of two tools that the individual

has at his disposal to mitigate the reduction of his expected present value that these policies

cause. First, the individual may substitute search effort away from the search channels that

are not monitored to the channel that is monitored. Channel substitution may be beneficial

from the individual point of view, and it may help to prevent a low-quality job match, but it

tends to reinforce the moral hazard problem. The second tool is on-the-job search. Job-to-job

∗Manuscript received May 2015; revised February 2018.

1We thank the editor, Holger Sieg, three anonymous referees, Peter Fredriksson, Olof ˚

Aslund, and participants in

conferences in Amsterdam, Bonn, and Uppsala and seminars in Bergen (Belgium) and Louvain-la-Neuve for helpful

comments. We are grateful to Regioplan, the Dutch National Institute for Social Security (LISV), and the Netherlands

Employees Insurance Agency (Uitvoeringsinstituut Werknemers Verzekeringen, or UWV), for providing the data.

We thank in particular Ger Homburg from Regioplan and Hans Doodeman, Ralf Pieloor, and Michel Moerenhout

from UWV for their help and for useful suggestions. Please address correspondence to: Gerard J. van den Berg,

Department of Economics, University of Bristol, The Priory Road Complex, Priory Road, Clifton, Bristol, BS8 1TU,

United Kingdom (GB). E-mail: gjvdberg@xs4all.nl.

879

C

(2018) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

880 VAN DEN BERG AND VAN DER KLAAUW

transitions reduce the importance of the first job accepted after unemployment. A low starting

wage can be mitigated by subsequent wage gains. On-the-job search therefore reduces adverse

post-unemployment effects of policies aimed at fighting moral hazard, while at the same time

it does not stimulate moral hazard during unemployment. The structural model we develop is

consistent with the differences in behavior and labor market outcomes between the treatment

and control groups observed in the data.

By structurally estimating a model that distinguishes between different search channels and

that allows for job-to-job mobility, we can quantify the relative importance of the two above-

mentioned mechanisms. Moreover, we may study their interaction. We show that the extent

of channel substitution depends on job mobility. If it is easy to move to better jobs while in

employment then channel substitution is less strong than otherwise. This is because with a

high job mobility, any job with a low wage can be exchanged quickly for a better job, so the

attractiveness of finding work using all available channels is high. On the other hand, a high

option value of job mobility may decrease the incentive to make search efforts among the

unemployed. If a high job mobility does go along with a high search effort by the unemployed,

then, if actual job mobility is high, the imposition of a minimum threshold for the monitored

search effort level will more often not be binding. All this suggests more in general that it is

relevant to take post-unemployment choices into account when evaluating the policy measures

of the employment office. In the policy evaluation literature, the role of subsequent job mobility

has typically been ignored.

The data concern a sample of relatively skilled individuals from the Netherlands. A subset

of the variables in the data has been analyzed in Van den Berg and Van der Klaauw (2006).

That study consists of a reduced-form analysis of the average treatment effect of the monitoring

program on unemployment durations. It did address the issue of channel substitution, but the

study did not have access to post-unemployment outcomes. The reduced-form results do not

provide evidence for a strong effect of monitoring on the unemployment duration, but they do

indicate that channel substitution takes place. Reduced-form studies cannot extrapolate such

results to individuals in different circumstances. Structural analysis is more amenable to this. In

our present article, we exploit the advantage of structural analysis that it enables counterfactual

policy analysis (see, e.g., Eckstein and Van den Berg, 2007, for a more general discussion of the

advantages of structural analysis, with a focus on unemployment outcomes).

Our study complements the empirical literature on the effectiveness of monitoring unem-

ployed workers (see Johnson and Klepinger, 1994; Dolton and O’Neill, 1996; Gorter and Kalb,

1996; Klepinger et al., 2002; Ashenfelter et al., 2005; Van den Berg and Van der Klaauw, 2006;

McVicar, 2008; Manning, 2009; and Micklewright and Nagy, 2010). All these studies provide

reduced-form analyses, and many are based on randomized social experiments. The evidence

is surveyed in Van den Berg and Van der Klaauw (2006). In general, the effect of monitoring is

stronger if labor market conditions and job prospects are worse.

Evaluation studies on post-unemployment effects of “treatments” during unemployment

include Ham and LaLonde (1996), who examine training programs, Dolton and O’Neill (2002),

who examine a counseling and monitoring program, and Van den Berg and Vikstr ¨

om (2014),

who examine punitive sanctions for individuals who do not comply with monitoring guidelines.

The latter study finds that sanctions on average lead to significantly lower accepted daily wages.

Finally, there is an expanding branch of literature that uses structural models to evaluate active

labor market programs (Adda et al., 2009; Foug`

ere et al., 2009; Wunsch, 2013; Lise et al., 2015;

Cockx et al., 2018; Gautier et al., 2018).

Section 2 provides institutional details for the policy that is evaluated in the social experiment.

It also describes the experiment itself. Section 3 summarizes our data. These are from a range

of registers as well as from a survey among the participants in the experiment. In Section 4, we

develop and analyze the theoretical job search model with multiple job search channels and job-

to-job mobility. We discuss identification of the structural model and we derive the likelihood

function. Section 5 presents the parameter estimates, the evaluation of counterfactual policies,