Lifetime income and old age mortality risk in Italy over two decades

This paper examines the association between lifetime income and old age mortality risk, referred to as the income–mortality gradient, in Italy during the 1980s and 1990s. We extend the literature by estimating the income–mortality gradient using Cox proportional hazard models, where the positions of the knots in the spline function for income are determined by the data, and, most importantly, by providing empirical evidence for Italy on the evolution of the income–mortality gradient between the 1980s and the 1990s. In addition, we investigate the importance of controlling for regional differences. For this purpose, we use data drawn from an administrative pension archive held by the main Italian social security institution and proxy individual lifetime income with the amount of individual pension benefit.We find that the shape of the income–mortality gradient is characterized by two discontinuities (knots) for males and one discontinuity for females. For both genders, the estimated associations between income and mortality risk are negative and stronger at higher income levels.Results for the change in the income–mortality gradient between the 1980s and 1990s are sensitive to inclusion in the model of controls for regions. Without controlling for regional differences, the income-mortality gradient widens between the 1980s and the 1990s due to a rather surprising positive gradient for males whose income is below the 66th percentile in the 1980s. One possible explanation for this result could be the presence of manual workers with relatively high salary and high mortality during the 1980s in the industrialized Northwest. After controlling for regional differences, we find no empirical evidence of a change in the income–mortality gradient, including the position of the knots, over time.