Pension recipients line up in front of the INP payment central (8:30 am).

There is growing concern about how the social security systems will evolve in face of growing costs produced by aging population. Several alternatives have been offered. From modifying the existing plans, by increasing contribution rates and retirement age, to replacing the system with individual retirement accounts.

Many countries have already introduced reforms. One of the oldest reformers is Chile that replaced the old unfunded-defined benefit system with a private- fully funded - defined contribution system. The system has been in operation for 19 years allowing researchers to analyze long run effects. It's design makes it ideal for the analysis of labor market impacts because it allows the coexistence of both (old and new) schemes permitting the identification of a control group. This limits the need to discretionary simulate what workers would have done if they had stayed in the old system.

The research objective is to use the Chilean social security reform to analyze how the work incentives were modified and how this affected labor supply, specifically of old workers. Special emphasis will be put in analyzing emerging gender inequalities and in identifying potential biases against specific group of workers.

The findings will be relevant for countries that have already implemented reforms and for future reformers. It will be innovative since most studies are centered on macro-economic and financial effects, but none has analyzed empirically the impacts on labor market outcomes. This has been the result of data unavailability, which will be surmounted by using a new dataset that will soon be available, and by using the existing cross section data in a different way.

Bi-annual cross-section household (CASEN) surveys allow us to compare the behavior of workers in both systems and draw conclusions on to how much of the difference is due to the reform and observable and unobservable characteristics. Labor market participation equations are constructed for workers in the new system and used to predict participation (or hours, sector, etc) of the other group, assigning the difference to the above mentioned factors. Pseudo-panel techniques and out of the sample simulations will also be used to reduce the problem of missing information.

But ideally the analysis should use time series on both types of workers, to control for working history differences, that will determine the outcome of the reform Furthermore one can actually observe the changes in labor market participation of workers as the reform took place and how participation/retirement differs between those that stayed in the old system and those that switched.

My conviction that using time series data is the correct way to assess the impact of social security reform in worker participation decisions and that the analysis of such effects are of vital importance to all countries that are in phase of reforming their social security system encouraged me to spend the summer in Chile. The goal was to learn more about the system by interviewing authorities of both old and new systems. Additionally, my intention was to find the data sets.

From the interviews I learned that there is an emerging discomfort with the new privatized system. It appears as if it has not met the expectations of the workers by not being able to provide equivalent levels of pensions to the old system. Moreover, the expected benefit of having a fully funded system is that the government does not have to cover the costs of the system but as the system stands now, it is extracting even more fiscal resources by increasing the amount of workers without enough funds to meet the minimum pension guarantee. Some of the interviewers made the point that the fiscal social security deficit was bound to keep increasing and that a new reform of the system is imminent.

I found that it is nonexistent as of today. The research director of the Finance Ministry, Mr. Arenas, informed me of a joint project of the World Bank, the Finance Ministry of Chile and the Social Security Intendance to collect a representative data set of about 20000 individual contributors to the system. The estimated date of completion is march 2001. I was offered access to the data for my own research if I collaborate in their project of social security coverage by economic sectors. (Obviously I agreed to the offer). Unfortunately this data set does not contain information on non-affiliates to the new system and therefore no cross-system comparisons are possible, but relevant information on the labor supply impacts of the reform can be deduced by using standard time series models of labor supply. Conclusions can also be drawn on how pension funds affect retirement, participation and sector decisions. Additionally, gender biases and unintended redistribution within and between generations can be estimated.

The Reform and Expected Effects in Labor Supply:

My interview with the director of the Benefits of Concession Department at INP.

In May 1981 the Chilean Government instituted a radical reform of its system of old-aged pensions, transferring responsibility for their provision from a public scheme based on unfunded defined benefit formulas to a privately administered, fully funded, mandatory, defined contribution system. The need for such reform was evident from the increasing costs of the old system as the population aged and the contributions were insufficient to cover pensions. The inequality introduced by the different payment schemes as well as the inability to provide the planned benefits and coverage made the reform imminent. Additionally, the nonuniversal and complex rules for determining future benefits brought about two other problems: uncertainty and inefficiency. The link between benefits and contributions was so weak that workers viewed the contributions as pure taxes, which introduced distortions to the labor market.

Measures were introduced even before 1981: In 1979, the pensioning by years of service was eliminated. There was also unification in the indexation of pensions through out the 30 social security institutions. In 1980 these 30 institutions were merged into one publicly administered pension system (INP). This reduced distortions and unfairness, particularly, by eliminating occupation and sector specific benefits, requirements and CPI adjustments. The system still remained regressive: as only the last few years were included in the base wage creating a bias against those with flatter age-earning profiles; the use of a flat actuarial factor also introduced a bias against those with shorter life expectancy; and the use of a maximum number of 35 years in the pension calculation implied a bias against those with longer working life. All these characteristics were typical of low wage workers.

The new privatized mandatory saving plan had all dependent workers in the covered sector contribute 10% of their monthly taxable wage on regulated intermediaries called pension funds administrators (AFP). At retirement age, 65 for men and 60 for women, the accumulated sum can be used to: purchase indexed annuities from a life insurance company, phased withdrawals from the AFP or a combination of both. Since 1988, early retirement is allowed.

The transition was expected to be gradual, predicting that the old system would disappear in 50 years. Pensioners from the old system continued with the same rules. Switching to the new system was allowed only until 1986. Those who changed and had contributed at least 12 months during the last 5 years received a "recognition bond". Since 1983 all new entrants to the labor force had to join the new system.

Me (Andrea Tokman) working at the office provided by the ILO in Santiago.

By changing incentives, taxes and other rules, the social security reform is bound to generate important indirect effects. The induced changes in the implicit labor tax on the covered sector affects the choice of moving between systems and sectors (covered vs. uncovered), the amount of hours worked, participation rates and retirement age. Additionally, observed differences in labor participation of older workers would be due to different pensioning requirements. The old system required a proof of employment termination before being awarded a pension, whereas the new system does not. This higher independence between benefits received and retirement implies that their participation decisions have a lower substitution effect, thus inducing to higher participation rates, lower job rotation, and lower incentives for firms to employ them since they no longer know how much time they will remain at work.

The new system's requirement of 20 years of contributions to be eligible for minimum pension motivates low wage workers to work in the covered sectors only until they are eligible, and then turn to uncovered sectors or underreporting of income because their minimum pension is guaranteed, and any additional contribution is pure tax.

With regard to the effects on females, there were some positive aspects of the old system. Such as: equal pension for more expected years of life; no requirement of years of contribution for minimum pension, a factor of particular importance for women given their lower labor market attachment; and calculus of pension at age 60 when they are at a higher portion of their age-earning profile (as compared to age 65 for men). Additionally, the new system perpetuates gender discrimination in wages and employment, as the pension is a function of what they had been able to accumulate during their working life and by charging fixed costs in their contributions. However, the new system benefits women by making the survivors' pensions sex- independent.