*An apparent resilience of labour-managed firms has been noticed during 2008 economic crisis in France and Italy through lower aggregate lay-offs and lower aggregate number of bankruptcies. If the productivity of French workers cooperatives has been found on average equal or higher than conventional companies depending on industries (Fakhfakh et al., 2012) and employment adjustments lower in time of crisis (Pencavel et al., 2006), the short-term dynamic of labour-managed firms productivity has still not been investigated, especially in reaction to an economic crisis. We make use of a large employee-employer administrative panel data set covering the period 2005-2011 to do a precise study of labour-managed firms resilience and its factors. We tackle more specifically the role of workers self-selection by accounting for job quality and workers intrinsic motivation estimated via a Roy selection model extended with non-pecuniary decision factors (D’Haultfoeuille et al., 2010).

**This paper develops a new criterion for comparing education Systems on the basis of their Equality of Opportunity. By recognizing that comparing societies behind a veil of ignorance is equivalent to ranking lotteries under ambiguity-aversion, we are able to construct a new EOp criterion which is arguably consistent with John Rawls' initial vision. This criterion is further shown to be a modified version of the Lorenz Zonoid, which was initially intendend for mutli-dimensional Inequality. By merging PISA and TIMSS data, we make an empirical application of this criterion, by studying how education systems in different countries widen inequality of opportunity as children climb up the grade years. The main contributions of this paper are threefold: First, it bridges the gap between decision theory and the litterature on Equality of Opportunity; second, it offers a new criterion for Equality of Opportunity which is consistent with John Rawls' vision; and third, it provides an empirical application to education systems using a combination of TIMSS and PISA data.