Downward Wage Rigidity: a Micro-Level Empirical Analysis for France in the 90s

We use microdata to examine the existence of downward wage rigidity in France, during the second half of the 90s. We use three annual datasets : a sample of the Déclarations Annuelles de Données Sociales, the Fiscal Income Survey and the Labour Force Survey. The first two sources, compiled on behalf of the fiscal administration, are reputed to be more accurate than the last one, subject to the traditional limitations of household surveys. Indeed, we show a tendency to underdeclaration as well as massive rounding to be present in the LFS wage measure. According to standard economic analysis, wages reflect individual productivity. All mechanisms reducing the transmission of productivity shocks to wages induce by definition wage rigidity. In practice, the presence and extent of wage rigidity must be deduced from the comparison between the actual wage evolution and some « reference evolution » observed in the absence of rigidity. A classical identifying assumption holds that downward wage rigidity transforms negative variations in the reference distribution into zero variations in the actual distribution, thus inducing a spike at zero in the latter. In French data, only the LFS exhibits such a spike, which seems to stem solely from reporting errors. We investigate the presence of more complex downward wage rigidity, by testing for the symmetry of the response of wages to positive and negative productivity shocks. For that purpose, we match the employee files with a firm level dataset (Bénéfices Réels Normaux). Results suggest that wages adjust less completely in the case of negative shocks. Although this asymmetry is shown to decrease with the initial wage level, wage rigidity cannot be reduced to the sole presence of a minimum wage. In particular, it is also shown to be higher for executives, and increasing in the local rate of unemployment.