Social Security Incentives and Retirement Decisions in Italy: An Empirical Insight

by Luca Spataro

Abstract
Understanding how individuals decide the timing of retirement has become a critical question for Italy, where tightening reforms of the National
Social Security (SS from now) system have already been introduced in early 90s and other reforms are likely to be implemented in the next future.
Yet, studies assessing the determinants of retirement are very recent and results are still contradictory, so that very little can be said about the
effectiveness of any reforms in improving activity rates.
While there appears a general consensus on the generosity of the SS system to be the leading explanation of the sizeable exit rates from labor
force at early ages occurred in Italy so far, there is still little evidence on the degree of sensitivity of workers to such incentives. Clearly, this issue is
preliminary and necessary to the implementation of whatever reform aiming at encouraging the postponement of retirement and, in general, at
improving the activity rates.
Moreover, the pronounced spike in retirement at the “normal age” (age 60) documented in previous works is an empirical puzzle whose nature
requires further investigation. Intuition suggests that the spike has much to do with retirement rules, that is with eligibility constraints, and with a
significant preference for leisure. However to my knowledge no studies have explicitly dealt with this issue.

The aim of the present work is twofold: on the one hand, to give a contribution to better understand the nature of the trade-off connected to
choice of retiring; on the other hand, to provide an empirical analysis in order to investigate the responsiveness of Italian male employees to the
early retirement incentives provided by the Social Security system and to propose an explanation to the to the puzzling spike of retirement at age 60.

As for the former goal, I propose some new measures of Social Security incentives for early exits from labor force: the first of them is called the
Marginal Cost of Retirement (MCR). Coherently with economic theory, such parameter captures the trade-off comprised in the choice of delaying
retirement by one year. Furthermore, a straightforward extension of the MCR allows to obtain a more forward looking measure of SS wealth
accruals, referred to as the Minimum Cost Value (MCV). Finally, following the same strategy, I present an extension of the already existing Accruals,
which I refer to as the Minimum Tax Value (MTV). I also compare the time pattern of such measures with those adopted in previous studies
(Accruals, Option Value, Peak Value).
From a methodological point of view, my starting point are two important lessons given by the recent literature on retirement and carried out for
other countries.
First, as several authors have pointed out, retirement is likely to be a decision undertaken by rational, forward looking individuals who, in order
to maximize their lifetime utility, typically contrast present SS wealth accumulation opportunities with those at some time in the future. Starting
from the work by Stock and Wise (1990) on the Option Value of retirement, many works have shared this assumption and assessed its relevance by
both estimating structural forms (like Rust (1989), Gustman and Steinmeier (1986) and Rust and Phelan (1997)) and reduced forms of retirement
decisions (Lumsdaine, Stock and Wise (1992) and, for Italy, Brugiavini and Peracchi (2001)). The reduced form approaches are most directly related
to the methodology adopted in the present paper. Second, some recent studies (Krueger and Pischke (1992) Coile and Gruber (2000) and Chan and
Stevens (2001)), have highlighted the importance of the identification problem of the retirement incentive effects estimated by reduced forms. In
other words, these authors stress the difficulty to disentangle the role of earnings from the “pure” SS incentives, since the latter are a (non linear)
function of the former and, in turn, the former are likely to be endogenous to unobserved tastes for retirement.
For this reason I carry out a panel analysis by applying a discrete-time duration model on the male employees sample drawn from the Bank of
Italy Survey on Households’ Income and Wealth (SHIW).
In particular I discriminate between Public Sector and Private Sector employees and control for the role played by other socio-demographic
variables and by eligibility rules so as to explain the determinants of retirement choices occurred since late 1980s through the mid-1990s. Notice that
such period is particularly interesting in that the 1993 and 1995 SS reforms, which have tightened eligibility rules and cut SS benefits, provide an
exogenous source of variability in the data and, thus, should help getting robust statistics.
The paper proceeds as follows. I begin with the institutional features of the Italian SS system prior to and after 1990s reforms respectively. Next,
after sketching the main findings of previous related literature in Italy, I describe the meaning of the new measures I use for capturing the early
retirement incentives. In the next sections I describe the empirical strategy adopted for obtaining the panel, contributions and wages necessary for
computations and present the main features of the SS incentives prior to and after reforms. Finally, I present the econometric results and concluding
remarks.
As for the SS incentives to early retirement, results show that:
1)It is necessary to discriminate between Public and Private Sector, since eligibility rules and benefit computations (more favourable for the
latter) and reform impacts have been substantially different for these categories of workers. I do find evidence of strong incentives to early
retirement for Public Sector employees before reforms. In fact for such workers SSW was substantially decreasing with respect to age; moreover, the
one-year dynamic incentives (MCR, Accruals) confirm the relevant “tax” levied on the prosecution to work provided by SS system prior to reforms:
in particular the median MCR was abundantly lower than the current wage (about 29% of current salary) and decreasing for almost all ages. As for
the Private Sector workers, calculations show that the contribution requirement for retirement eligibility has been binding for the majority of Private
Sector workers, which has been probably understated in previous works: however, also such workers enjoyed strong incentives to retire once the
normal age (age 60) had been reached. However, some of the forward-looking dynamic measures (particularly the negative values of both RMCV
and MTV for both Sectors) reveal the presence of incentives for postponing retirement which may have been contrasting (or even offsetting) the
impact of the former incentives.
2) Early 1990s reforms have reduced the median SSW by about 20% and the median Peak Value by almost one percentage point less; however,
among eligible workers, Public Sector employees appear to have borne most of the burden of reform cuts, with higher penalizations on early
retirement. There also comes out that younger cohorts have been and will be more dramatically in the future the net-payers for the rebalance of the
Italian SS system. In fact, both the generous pre-reform benefits and the reduction of the forward-looking incentives to postpone retirement (RMCV
and MTV) occurred in the transition phase of reforms has provided the older cohorts with a substantial incentives for exiting the labour force.
Empirical estimates can be summarized as follows:
2) It is not the absolute value of SSW that matters in retirement choices, but, rather, the RR ratio. Such parameter is by far the most important
(in terms of Likelihood gains) among the whole set of SS measure incentives analyzed.
3) Neither short sight (i.e. one-year Accruals/Costs) or more forward looking dynamic measures of SS incentives turn out to be significant in the
original specification adopted for computations. In fact only the Option Value measure presents coefficients with univocally correct signs; however,
such coefficients are significant only when the Option Value are computed according to previous estimations. Among other forward-looking
measures, also the MTV coefficients present the correct sign and are significantly different from zero. As a consequence, I can conclude that male
Italian employees appear to have been aware of and exploited the incentives they were provided with for early retirement: short sight measures seem
to have played a prevailing role in determining the timing of retirement, even though evidence of (some) forward-looking behavior cannot be
statistically rejected.
4) The age 60 spike does appear to be significant only when associated to the first year of eligibility: in other words, the analysis brings evidence
of binding eligibility constraints rather than “rules of thumb” applying by the normal of retirement.
 5) As for other socio-economic factors, family composition and coordination in retirement choices within the household does seem to play a role
for Public Sector employees only; on the other hand, Private Sector employees turn out to be particularly concerned with the achievement of
eligibility and of a “safe” standard of living: in fact, for the latter category of workers house-ownership before retirement seems to be an especially
relevant goal.

JEL Nos. H55, J26