I investigate the immediate effects of the introduction of a large-scale child benefit programme on the labour supply of household members in Poland. Due to non-random eligibility and the universal character of the programme, standard evaluation estimators may be inconsistent. In order to address this issue, I propose an approach that combines difference-in-difference (DID) propensity score based methods with the covariate balancing propensity score (CBPS) approach developed by Imai and Ratkovic [2014]. The DID estimators exploit the time dimension to uncover the causal effect of interest. The CBPS method is expected to significantly reduce the bias resulting from systematic differences between treated and untreated subpopulations. I also account for potential heterogeneity among households by focusing on comparisons between locally defined subpopulations of individuals, which jointly provide a comprehensive view on the overall impact. I find that on average previously employed mothers maintain their labour supply although there are heterogeneous weak responses depending on the age of the youngest children. Additionally, mothers who did not work before the introduction of the programme are even less likely to do so having received the benefit. The fathers’ labour supply remained mostly unaffected by the programme, with the exception of previously unemployed fathers, who tend to work more often having received the benefits. This finding may suggest that the programme strengthens the traditional division of household roles, with the male being the main earner.