Faculteit der Economische Wetenschappen en Econometric 05348

799s 036 * 1 i Serie research memoranda 1 h Dismissal through Disability

Wolter Hassink Jan van Ours Ceert Ridder

Research Memorandum 1995-36 October 1995

applied research team

vrije Universiteit amsterdam Dismissal through Disability

October 26, 1995

Wolter Hassink* Jan van Ours** Geert Riddere**

Abstract

Ifa finn wants to reduce its worllforce, it may dismiss some of its workers. Alternatively, it may make workers eligible for disability benefits. Upon examination these workers formally satisfy the conditions for disability enrolment. Because these conditions allow for a rather liberal interpretation of disability, these workers could have stayed on their job had they not become redundant. We use data on Dutch firms to show that at the end of the 1980s about 10 percent of the observed inflow into disability were in fact dismissals.

&code: D21,J63 Keywords: dismissals, disability.

Department of Economics, Vrije Universiteit, Amsterdam and Tinbergen Institute, The ** Tinbergen Institute and Erasmus University Rotterdam, The Netherlands *** Department of Econometrics, Vrije Universiteit, Amsterdam and Tinbergen Institute, The Netherlands

Mailing address: Jan van Ours, Tinbergen Institute, Erasmus University Rotterdam, P.O. Box 1738,3000 DR Rotterdam, The Netherlands; Tel: (+3 1) - 10 408892 1; Fax: (+3 1) - 10 4527347.

We thank the Organisation for Labour Market Research (OSA) in The Hague, The Netherlands for permission to use their data and for financial support. We are grateful to Philip de Jong for his comments on a previous version.

Table 4 Estimation results bivariate Tobit regression, reduced and structural forma)

Reduced form Structural fom

Dismissal rate

W90 - YA- Y8A8 990

L89 Start-up 2 300 days Age 2 50 Constant

Q1 0.048 (10.32)"' 71 - 0.053 (27.03)"' Disability rate

W, - W, Y& - y8&8 990 f88 L, Start-up 2 300 days Age 2 50 Part-time Bad conditions Illness 1988 Constant

Pv 0.085 (0.82) h - 0.095 (2.96)" X$ b, 8 11.95" N 225 225 * Statistically significant from zero at the 10% level. ** Statistically significant from zero at the 5% level. *** Statistically significant from zero at the 1% level. a) Absolute t-values are in parentheses; N is the number of observations used to estimate the model; 4,k=1,2, is the standard mor of the regression of the reduced form estimates; $, k=1,2, is the standard mor of the regression of the structural fom; p, is the correlation coefficient in the error structure of the reduced fcnm; is a Chi-square test on overidentification with 5 degrees of freedom. b) We experimented with additional variables in the structural disability equation: 1) inclusion of Age 2 50 gives h = 0.0937 (2.89)" and &4; = 11.84"; 2) inclusion of b9gives h = 0.063 (1.86)' and z4; = 2.46. l i