OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 57, 4 (1995) 0305-9049 $3.00 WHYDOWOMENMA~EDTO UNEMPLOYED MEN HAVE LOW PARTICIPATION RATES? Gianna Giannelli and John Micklewrightt I. INTRODUCTION An 'added-worker' effect would result in women married to unemployed men having higher labour force participation rates than women married to employed men. But in a variety of countries the opposite is the case - see Table 1. Participation in employment for wives of the unemployed is signifi- cantly lower than for wives of working men at the 1 percent level in seven of the 11 countries and at the 5 percent level in one other. In no country is participation by wives of the unemployed significantly higher. Of course, there are possible reasons for the pattern in Table 1 that are consistent with the independent operation of an added worker effect. Charac- teristics associated with low participation may be more common among wives of the unemployed, such as low productivity and residence in areas of low labour demand. Conditions under which unemployment benefits are paid to the husband may also affect the wife's labour supply. A disincentive effect for married women may be expected if benefits paid to their husbands are means-tested on family income. Disincentives to family labour supply that may stem from the benefit system have been the focus of considerable attention in the UK where the most important unemployment benefit (in terms of numbers of recipients at anyone time) is means-tested on family income (Dilnot and Kell, 1987; Kell and Wright, 1990; Garcia, 1991: Davies et al.; 1992; Pudney and Thomas, 1992). It is tempting to speculate whether participation rates in other countries in Table 1 reflect their benefit systems. All unemployment benefit is subject to a family means-test in Australia. (In other countries there is a 'first- tWe thank Christian Dustrnann and Eckhard Wurzel for help in clarifying details of the German unemployment benefit scheme, Uschi Jaenichen for advice on the SEP data, and Peter Blossfeld for providing access to these data through his research project at EUI. We are par- ticularly grateful to Costas Meghir and Geert Ridder for help with the fixed-effects model of Section 3 and to Gotz Rohwer for help with the models of Section IV (which were estimated with his TDA package (Rohwer, 1993)) and more generally for his efforts in pushing us into being less conservative in our approach to the data. A referee made useful suggestions. 471 o Blackwell Publishers Ltd. 1995. Published by Blackwell Publishers, lOll Cowley Road, Oxford OX4 UF, UK & 238 Main Street, Cambridge. MA 02142, USA.