This study analyses a practically unnoticed event that characterized the evolution of almost every labour markets in the industrialized countries: the fact that the participation rates of the 55 to 64 years old maie workers decreased sharply over the last 25 years or so. A decrease standing between 14 and 28 percentage points in our sample.

Four types of variables are put forward in order to explain such a fall in participation rates, for five countries (Germany, France, Canada, United Kingdom and Sweden) which experienced somewhat different economie environment and policy changes over the period 1967 to 1987. These variables are the unemployment rate (reverse of employment opportunities), wage growth differentials, social security allowances and specifie mandatory or facultative retirement adjustment measures (income or wealth effects).

The paper contains two different models. On the one hand, the first model looks into the general determinants of labour force participation for this age-sex group using pooled annual time series and cross section data (source = OECD) for the five countries already mentioned over the period 1967-1987, and for a strictly common set of explanatory variables. It is found that a large part of both the international an intertemporal variance in the participation rates of the 55 to 64 years old maie workers responds (negatively) to unemployment rates, labour income per worker, social security allowances, and the varying age of mandatory or facultative retirement.

On the other hand, the second model follows a comparative estimate method (one separate regression for each country) emphasizing the role of each country's own specificities. In that respect, the paper briefly describes the content of important institutional measures that were developed in the past and assesses their relative contribution to the explanation of the participation rates decline by country. These measures range from changes in the age of mandatory retirement (France and Germany), access to preretirement with actuarial penalties (Canada and Sweden), pre-retirement for job losses, to conditional or unconditional pre-retirement for voluntary dismissals. Pre-retirement for job losses occurs where specifie groups of workers both loose their job and qualify under a minimal age condition (France). Pre-retirement for voluntary dismissal also requires a minimum age condition (Germany) but may also be contingent (France and United Kingdom) upon a replacement condition (one worker out, one worker in).

The econometric tools combine corrections for both time and space types of residuals autocorrelation and interacting explanatory variables with country dummies (first model). The robustness of the results is also tested against ordinary least squares, generalized least squares and two units of exchange rates definition for the data on wages.

Given the robustness of the results, we do ask whether the reversai of some measures (e.g. raising the age of mandatory retirement) would have a symmetric effect. Given the high dependency of the participation rates of this age-sex group upon unemployment rates and wage growth, this may be the case, but much so in an environment of low unemployment and slow wage growth as experienced mainly in S weden (at least, up to the end of our period of observation). Thus, social security expenditures and specifie ad hoc measures addressed at workers between the age of 55 to 64 may help regulate labour market transitions in the future but with varying success depending upon the economie environment in which they take place.