اثر تأمین اجتماعی، بهداشت، جمعیت شناسی و فن آوری در دوران بازنشستگی

This article studies the determinants of the labor force participation of the elderly and investigates the factors that may account for the increase in retirement in the second half of the last century. We develop a lifecycle general equilibrium model with endogenous retirement that embeds Social Security legislation and Medicare. Individuals are ex ante heterogeneous with respect to their preferences for leisure and face uncertainty about labor productivity, health status and out-of-pocket medical expenses. The model is calibrated to the U.S. economy in 2000 and is able to reproduce very closely the retirement behavior of the American population. It reproduces the peaks in the distribution of Social Security applications at ages 62 and 65 and the observed facts that low earners and unhealthy individuals retire earlier. It also matches very closely the increase in retirement from 1950 to 2000. Changes in Social Security policy – which became much more generous – and the introduction of Medicare account for most of the expansion of retirement. In contrast, the isolated impact of the increase in longevity was a delaying of retirement.

The reduction of the participation of elderly people in the labor force was one of most remarkable economic changes of the last century, particularly in the second half. In 1950, 46% of men aged 65 and over in the United States were working, but only 16.5% were in 2000. Just four out of every ten 66-year-old males were retired in 1950, but fifty years later almost seven out of ten were out of the labor force. This phenomenon is hardly exclusive to the United States. Blondal and Scarpetta (1999) and Gruber and Wise (1999) provide evidence that the workforce participation of the elderly population has declined in many countries of the OECD. Currently, more than 50% of workers choose to retire at the age of 62, when they first become eligible for early retirement benefits under social security, although at a reduced level. In 1950, in contrast, there was no legal early retirement age and the minimum and normal retirement age coincided at 65. The decision to retire, and to do it early, is influenced by anumber of factors in addition to age and the rules of social security such as health status, income, preference for leisure, etc.1 This article develops and calibrates a lifecycle general equilibrium model with heterogeneous agents to study the determinants of the labor force participation of elderly American males and to investigate the factors that may account for the changes in retirement between 1950 and 2000. We focus on the role of Social Security, health status and the introduction of Medicare, demographic factors (associated with higher longevity) and changes in the age–efficiency profile. Our artificial economy is populated by agents who live for a realistic number of periods, have preferences over consumption and leisure and choose at each period whether to stay in the labor force or retire. Agents split their working hours between home production and working in the market. Individuals are ex ante heterogeneous with respect to their preferences for leisure and face uncertainty about their labor productivity, their health status and their out-of-pocket medical expenses. Agents can accumulate a single risk-free asset, which takes the form of capital. Savings may be precautionary and allow partial insurance against the idiosyncratic shocks. In addition, we model the U.S. Social Security system in detail and also allow agents to decide when to start collecting retirement benefits regardless of their employment status. This is consistent with the empirical evidence in Rust and Phelan (1997) and Benítez-Silva and Heiland (2008), who show that a large number of agents claim benefits while continuing to work, mainly among the nearly elderly ones. We also take into account Medicare, which was introduced in 1965 and constitutes a federal health insurance program that provides subsidized health insurance coverage to virtually every American over age 65. Medicare provides generous insurance against medical expenditures shocks and could induce earlier retirement because of the limited need to accumulate precautionary savings. Conversely, because of eligibility requirements, it also encourages the delay of retirement until 65. The model is calibrated to the U.S. economy in 2000, our benchmark year, and is able to reproduce very closely the retirement behavior of the American population. In particular, the model reproduces the peaks in the distribution of Social Security applications at ages 62 and 65 and the observation that unhealthy and poor individuals retire earlier. The model is then simulated considering the changes in Social Security, Medicare, age–efficiency profile and demography between 1950 and 2000. We find that the simulated labor force participation of older individuals increases to levels similar to those in the data. We show that the incentives implied by the institutional factors concerning Social Security and Medicare legislation are very effective in influencing retirement behavior. For instance, a counterfactual experiment in which all parameters were kept at their 2000 values, but the rules of Social Security were changed to those of 1950, finds that the retirement rate drops for every age group. More importantly, the retirement peak at age 62 disappears, as in 1950 when there was no early retirement benefits. This article extends and improves the previous literature in many aspects. Our model is related to Imrohoroglu et al. (1995), Huggett and Ventura (1999), Nishiyama and Smetters (2007), and Rojas and Urrutia (2008). These models provide a framework rich enough to deal with all the factors that potentially affect the retirement decision. Furthermore, this structure allows us to model more accurately the dynamic structure of a social security system. In these papers, however, the retirement decision is exogenous in contrast to our model. Conde-Ruiz and Galasso (2003) endogenize retirement, but in a purely theoretical political-economy framework with no quantitative analysis. French (2005) estimates a partial equilibrium lifecycle model of retirement behavior in which health and wages are uncertain. He uses the model to simulate the impact on the labor supply of modifications to Social Security legislation. Diaz-Gimenez and Diaz-Savavedra (2009) use an overlapping generational model with an endogenous retirement decision to study pension system reform in Spain. Our model has many features in common with theirs; but as we study the American economy, the calibration and institutional details of the model are obviously very different as are the experiments we run. Finally, in Kopecky (2011) whereas the decision to leave the labor force is endogenous as in our article, hours worked are fixed in every period and there is no social security in the model, which plays an important role in our case.2 As for the channels we emphasize as affecting retirement behavior, the importance of higher Social Security benefits has been investigated in a number of articles using a variety of estimation methods.3 Nevertheless, this literature has not come to a consensus. In fact, whereas Gustman and Steinmeier (1986) and Rust and Phelan (1997) have found that Social Security benefits have had a strong negative effect on male labor supply, Burtless (1986), Stock and Wise (1990) and Krueger and Pischke (1992) concluded that it had little effect. These results suggest that either there are problems associated with the methods that have been used to investigate this relationship,4 or there are other explanations that must be taken into consideration.5 In this article, we bring together, in a single model, different explanations for the decision to retire.The impact of health status and Medicare on retirement has also been investigated by Rust and Phelan (1997) and French and Jones (2004, 2011). However, they do not study the evolution of retirement over the last decades, which is a major goal of this paper. Additionally, none of these articles include home production, which is an important factor for the model to be able to reproduce the pattern of consumption over the lifecycle (Aguiar and Hurst, 2005).6 Regarding the impact of the rise in longevity on the decision to leave the labor force, Kalemli-Ozcan and Weil (2010) show that an exogenous decrease in the probability of death, which allows people to better plan saving for old age, generates a longer retirement life. In contrast, Bloom et al. (2007) show that, depending on social security provisions, improvements in life expectancy may induce people to remain in the labor force to increase savings for old age. Our simulations show that the latter effect dominates. Finally, technology change may modify age–earnings profiles and hence the decision to leave the labor force, as shown by Ferreira and Pessôa (2007).7 In fact, Heckman et al. (2003) provide evidence that older workers have become less productive relative to younger workers over the second part of the last century. This trend could induce people to work more intensively in the first part of their productive life, increase savings and retire earlier.8 The article is organized as follows. Some retirement facts are presented in Section 2. The model is presented in Section 3 and the calibration procedures and data are presented in Section 4. In Section 5 results are presented and discussed; Section 6 concludes.

In this paper, we have studied a lifecycle economy in which individuals optimally pick the time to leave the labor force. The model mimics relevant features of the American economy and takes special care in the calibration of the Social Security system. Simulations were able to match very closely the retirement profile of American men as well as the changes in retirement from 1950 to 2000. The model was able to reproduce the differences in the retirement profiles between low and high productivity individuals and between the healthy and unhealthy. The match of the peak of the distribution of social security applications at age 62 was particularly good. The model suggests that the changes in Social Security over the second part of the last century and the introduction of Medicare account for most of the changes in the retirement profile by age observed in the period. In contrast, we found that the isolated effect of demography would be to postpone the date that workers leave the labor force. Hence, the increase in longevity observed in the last century cannot explain the reduction in the labor force participation of the elderly. The impact of technology factors represented in the change in the age–efficiency profile was mixed and small. In addition, we found that the elimination of the Delayed Retirement Credit has a small impact on retirement but would drastically increase Social Security applications at the age of 65. Hence, the incentive of the DRC has played an important role in inducing workers to postpone claiming benefits beyond the normal retirement age and thus ease the pressure on the Social Security system.