In this paper, the labour supply of married women in Poland is studied using the Polish Labour Force Survey of February 1993 on three competing econometric models, viz. Tobit, Three Regime Tobit and Heckit. It is shown that the choice of model has an influence on conclusions drawn. In particular, the Tobit model - which has been widely applied in empirical labour supply analysis - turns out to be a strikingly bad predictor. Further, there is disagreement between the models regarding the wage elasticity of hours supplied. The preferred model for predicting hours of work is the Heckit model. This shows that an increase in the wage rate will induce employees to reduce their number of hours worked. On the other hand, more women will join the labour market after a rise in the wage. The regression models support the common view that young wives, those who have been unemployed before, or those who have low educational attainments are ceteris paribus much more excluded from effective labour supply than the average person. In addition, wives with young children tend to have both lower participation rates and work shorter hours. The exclusion of a large part of especially the young female population from the labour market may have serious economic costs to Poland because part of its potentially most dynamic human capital is left untrained and unemployed.