SOCIAL SECURITY systems in Latin American countries mobilize significant quantities of resources.1 In some countries, such as Argentina, Costa Rica, and Uruguay, the revenue and expenditure of social security systems account for a large share of general government revenue and expenditure, and their share of gross domestic product (GDP) approximates the shares of some member countries of the Organization for Economic Cooperation and Development (OECD). The coverage of the social security system in these countries is broad, and the level of benefits is generous. In most of the other countries of the region, social security systems do not represent as large a share of GDP but have the potential to do so, because coverage in general is low and could be expanded and because the pension component of the system is not yet mature.


SOCIAL SECURITY systems in Latin American countries mobilize significant quantities of resources.1 In some countries, such as Argentina, Costa Rica, and Uruguay, the revenue and expenditure of social security systems account for a large share of general government revenue and expenditure, and their share of gross domestic product (GDP) approximates the shares of some member countries of the Organization for Economic Cooperation and Development (OECD). The coverage of the social security system in these countries is broad, and the level of benefits is generous. In most of the other countries of the region, social security systems do not represent as large a share of GDP but have the potential to do so, because coverage in general is low and could be expanded and because the pension component of the system is not yet mature.

SOCIAL SECURITY systems in Latin American countries mobilize significant quantities of resources.1 In some countries, such as Argentina, Costa Rica, and Uruguay, the revenue and expenditure of social security systems account for a large share of general government revenue and expenditure, and their share of gross domestic product (GDP) approximates the shares of some member countries of the Organization for Economic Cooperation and Development (OECD). The coverage of the social security system in these countries is broad, and the level of benefits is generous. In most of the other countries of the region, social security systems do not represent as large a share of GDP but have the potential to do so, because coverage in general is low and could be expanded and because the pension component of the system is not yet mature.

Many problems and issues of financial, economic, and administrative policy arise in connection with the functioning of social security systems in Latin America; this paper surveys the most important of these issues. Section I describes some basic features of Latin American social security systems and prepares the ground for the more extensive Section II, which surveys the principal financial, economic, and administrative issues facing these systems. Section III presents conclusions and summarizes the paper’s findings.

I. Basic Features

The relative size and degree of development of the social security system vary substantially among Latin American countries. In Argentina, Chile, Costa Rica, and Uruguay, the ratio of social security expenditure to GDP approached or exceeded 10 percent and accounted for between 25 percent and 40 percent of general government expenditure in 1978-83. At the other extreme, in Guatemala and other Central American countries, the ratio of total social security expenditures to GDP has been on the order of 2 percent or less (Table 1).

Each system’s size can be related to different features and reflects important differences in the economic structure of each country as well as the political environment.2 Of particular importance is the degree of coverage of the population (Table 2). For example, in Uruguay, the pension scheme is estimated to cover about 81 percent of the labor force (economically active population), and the sickness-maternity system about 68 percent. By contrast, the corresponding figures for Guatemala are 33 percent and 14 percent, and coverage is even lower in some other Central American countries.3 Differences across countries in the number of risks covered and the relative generosity of benefits per contributor are of lesser importance. All Latin American countries have old-age pension schemes, sickness-maternity health plans, and disability schemes. Some countries have no family allowance systems, and only a few have unemployment plans (Mesa-Lago (1985, p. 10)).

Table 1.

Total Expenditure of Social Security System as a Percentage of GDP in Selected Latin American Countries, 1975 and 1978-83

article image
Sources: International Labor Office (ILO) (1985,1988); International Monetary Fund, International Financial Statistics (IFS); and Fund staff calculations.
Table 2.

Coverage of Social Security Programs in Selected Latin American Countries, 1980

article image
Source: Mesa-Lago (1985).

The size of social security is also related to its importance in total expenditure of the pension scheme; thus, the pension scheme accounted for 80 percent of total benefit expenditure in Uruguay in 1980, compared with 5 percent in Guatemala. An exception to this relationship is Costa Rica, where the pension scheme, although it has grown in importance, accounted for only 15 percent of total benefit expenditure in 1980.

The relationship between the size of the whole system and the importance of the pension component is probably attributable to the degree of maturity of the pension scheme. The pension schemes of Uruguay, Chile, and Argentina have been in place longer than those of the Central American countries, so that the share of the population at or above retirement age that would be eligible for a pension is greater in the former group of countries than in the latter. Finally, the older systems often permitted retirement at an early age and after a short period of participation in the plan.

The substantial variation in coverage among Latin American countries is related to differences in the size of the organized sector of the economy. In countries where coverage is low, a substantial share of the labor force will typically be found in the unorganized rural sector, with a high proportion of either self-employed or unpaid family workers. These groups are difficult to include in conventional social security programs financed through salary-based employer-employee contributions.

Those countries where the social security programs are the largest in relation to GDP introduced their programs much earlier than other countries. An important feature of certain of these countries’ programs is the degree of stratification of the system. Certain occupational groups participated in social security programs early on, and other groups were gradually included, not in the original plans but in entirely different and separately administered schemes. Substantial inequalities existed across plans, and these have been the object of reform in both Chile and Uruguay in recent years.

Social security programs in Latin America, as in most countries, are financed mainly through payroll taxes. Typically, the employer’s share is larger than that of the employee. Other taxes are an important source of finance in several countries; for example, in Argentina, a substantial share of revenues from the value-added tax (VAT) is earmarked for the social security system. In addition, transfers from the central government or other levels of government are important in Chile and Uruguay. Neither of these countries can finance social security expenditures from payroll taxes alone, and without other sources of revenue the revenue shortfall would be substantial. The share in total revenue of income from capital is in general low, although it contributes more to revenues in some Central American countries where the pension schemes are relatively young and have been able to accumulate reserves.

Social security expenditures relative to GDP in the group of Latin American countries in the 1970s and early 1980s did not evolve uniformly, although expenditures as a percentage of GDP in the countries with larger systems showed some tendency to increase (Table 1). Taking a somewhat longer perspective, the expenditure ratio has risen since the 1960s. One motive for this secular trend has been the substantial increase in life expectancy that has taken place since 1960 in most countries, which has increased both pension expenditures and the demand for medical services.

The lack of a uniform pattern also characterizes the evolution of the ratio of revenue to GDP in the 1970s and early 1980s (Table 3). Revenue increased in Costa Rica, where the system was still expanding at a significant rate, and in Panama, but it declined in Uruguay. In view of the lack of a pronounced trend in either revenue or expenditure, it is not surprising that the deficit did not evolve in the same direction in most countries. Nonetheless, although comparable data for the period since 1983 are lacking, the available information suggests that in some countries the deficit has tended to grow.4

II. Issues

There are several principal financial, economic, and administrative issues that must be considered in a survey of these systems.

Pay-As-You-Go Versus Funded Pension Systems

The larger pension plans in Latin America are now financed on a pay-as-you-go basis, in which current contributions pay for current benefits, with any shortfall being met either out of reserves or by means of a transfer from other government resources. The original pension plans were funded at the outset but suffered a de capitalization of their reserves that created strong pressure for the adoption of pay-as-you-go financing. One reason for this decapitalization was the erosion of the real value of financial investments, mainly government bonds, that took place when nominal returns failed to keep pace with inflation. Another reason was the extension of social security systems to encompass groups that were poor actuarial risks, and the introduction of benefits that could not be financed by payroll contributions from the intended beneficiaries (for example, minimum pensions for low-income workers). Finally, a substantial share of reserves was allocated to public investment projects— public housing, for example—where the real rate of return proved to be low or even negative.5

Table 3.

Total Revenue of Social Security System as a Percentage of GDP in Selected Latin American Countries, 1975 and 1978-83

article image
Source: ILO (1985, 1988); International Monetary Fund, IFS, and Fund staff calculations.Note: Central government contributions are excluded from revenues.

Would there be advantages for countries with well-developed systems, such as Uruguay, to convert their pension systems to funded systems? Should the countries whose systems are just now expanding and maturing eschew the pay-as-you-go method? Many economists have argued that pay-as-you-go plans depress saving and capital formation, but the extensive debate on this issue is inconclusive.6 Nevertheless, among other considerations, the adoption of a pay-as-you-go system can obscure the long-run financial implications of a pension program, particularly when a country’s demographic structure is changing.

The government that introduces a pension scheme commits itself to expenditures both now and in the long run. If the pension scheme is financed on a pay-as-you-go basis, and if an increase in the dependency ratio is projected, then contribution rates will have to increase to maintain the average pension as a given proportion of the average wage.7 If a pension program relates the size of the pension to the number of years in which participants make contributions, expenditures can be expected to rise rapidly in the initial years of the program.

An advantage of a fully funded system is that the possible unpleasant surprise of an increase in the contribution rate is avoided. A related argument is that if it is difficult to change social security contribution rates once they are set, then they should be set high enough initially to generate sufficient resources for the expenditure programs over a long period. To put it another way, a funded system conveys the right kinds of signals about the future costs of a pension program.*8 The relevance of these arguments depends on the rapidity with which changes in rates are required.

An important consideration in the Latin American setting is the impact of inflation on a funded system, A high and variable rate of inflation will substantially increase the variance of the rate of return to the fund’s reserves unless its investments are indexed. Furthermore, an unexpected surge in inflation could wipe out the real value of the reserves. The problem of inflation is mitigated by the development of sophisticated financial instruments and may not be a serious problem in a stable financial environment. Nonetheless, a pay-as-you-go system is not vulnerable to inflation in the same way.

These various considerations do not constitute a strong case for a funded system in most Latin American countries. The argument that a pay-as-you-go system depresses saving is irrelevant if confidence in the pension system is low, because consumption will not be stimulated if individuals do not expect to have their contributions returned to them in the form of pensions. In any case, pay-as-you-go systems have proved to be politically attractive.

If a country opts for a fully or partially funded system but uses the system’s reserves simply to finance additional current expenditure by the government, any possible advantages of the choice are lost. The appropriate policy for the government to adopt will depend, however, on how the private sector reacts to a funded scheme. If it views its contributions as saving, albeit forced, then the social security system’s savings could substitute for other forms of saving, leaving private consumption unchanged. In this case, any additional expenditure financed by the surplus has an expansionary impact on the economy. Unless the stance of fiscal policy before the introduction of the program was overly tight, this use of the surplus would be inappropriate policy. If the stance was right, then the central government should act as if the surplus were not available to finance expenditure increases or tax reductions. One appealing option would be to invest the reserves in the central government debt and thereby reduce the deficit of the consolidated central government.9

If contributions to a funded system are viewed as a tax, then the introduction of the program should give rise to some offsetting reduction in taxation or increase in expenditure unless the stance of fiscal policy before the introduction of the program was too loose. If the rate of investment in the economy is deemed to be too low, the reserves should finance public investments chosen for their social rate of return.

Sources of Financing

Payroll taxation has traditionally been viewed as an equitable means of financing social security because it relates an individual’s cost of participation in social security to the benefits he will receive in a manner analogous to the relationship between the value of premiums in a private pension plan and average benefits. This view, however, presupposes that the effective incidence of payroll taxation is on the insured; that is, that the introduction or increase in a payroll tax does not affect employers’ labor costs but reduces employees’ pay net of both the employer and employee portions of the tax. In any case, the relationship between the value of contributions made by, or on behalf of, different individuals and the expected value of future benefits can be tenuous. For example, often when social security programs are introduced they are immediately extended to the current generation of retirees, who cannot have made any contributions.

A related argument in favor of using payroll taxes to finance social security, either in part or in full, is that their use imposes financial discipline on the social security system. An undue expansion in benefits will be checked because the burden of the contributions is effectively borne by the contributors. Finally, the use of the payroll tax has been justified on the grounds of administrative ease.

In many Latin American countries, however, exclusive reliance on payroll taxation is simply not feasible for a social security system that aspires to broad coverage. The base of a payroll tax is effectively limited to the organized sector, and, as has been noted, this sector is often relatively small.

The assumption that the incidence of the payroll tax falls effectively on the insured is also open to question. Because of institutional rigidities in wage setting and oligopolistic market structures in the organized sector, the tax could simply be passed on to consumers. An example of institutional rigidity is the regulation in Mexico’s social security laws that employers of workers receiving the minimum wage are obliged to pay both the employer’s and the employee’s share of payroll taxes (see Wilson (1985)). A study of social security and other government programs in Chile revealed that prices were determined by average cost plus a markup, suggesting that payroll costs are simply passed on to the consumer (cited in Wilson (1985, p. 262)).

If the payroll tax is not effectively borne by the contributor, then it must increase the cost of labor. In turn, this will have some impact on employment, on the assumption that labor and other factors of production are to some extent substitutable. The shift away from payroll taxes to value-added taxation that took place in Uruguay and Argentina in 1978-79 was in part prompted by the view that high rates of payroll taxation were depressing the level of employment. A recent study calculated a range of estimates for the impact on employment in Mexican manufacturing that would be generated by a switch from payroll to value-added taxation. This impact ranged from 1.7 percent to 12.5 percent (Wilson (1985, pp. 267-69)).

If the burden of the payroll tax does not fall on those who benefit from the tax, then there is no reason, on purely economic grounds, that social security programs should not be financed from general revenues. Moreover, there may be a case for reducing payroll tax rates if an alternative financing source is available.

Financial Implications of Demographic Trends

The social security systems of OECD countries have had to confront the financial implications of an increase in the dependency ratio brought about by the increased life expectancy of older persons, the declining birthrate, and the systems’ maturation. Similar developments are evident in Uruguay, and declining mortality rates in Chile and Costa Rica have contributed to an increase in the dependency ratio in these countries as well. Further increases in the dependency ratio in these countries could create serious financial pressures because pension expenditures are already a significant share of GDP. If life expectancy increases rapidly in those countries where it is now low, a substantial increase in pension expenditure is likely to result. At the same time, the aging of the population could require a substantial increase in medical expenditure for the treatment of the degenerative diseases associated with old age. In consequence, it is useful to consider what would happen to the composition of the population—and to the dependency ratio—if the present trends of a falling birthrate and an increasing life expectancy were to continue.

The World Bank’s population projections for virtually all countries are based on the assumption that these trends will continue.10 By computing the ratio of persons at or above a standard retirement age to the number of persons of working age, a crude proxy for the dependency ratio may be obtained. The ratio thus derived cannot take into account changes in the rate of labor force participation, but it gives a rough indication of the impact of demographic change on the financial requirements of a pension system.

The base projections of the World Bank do not show any great variation in the ratio of the retirement-age to working-age populations between 1980 and 2000 for any of the Latin American countries, with the possible exceptions of Chile and Uruguay (Table 4). Thus, increases in the rate of growth of the elderly population, owing to increased longevity, and decreases in the rate of growth of the working-age population, stemming from a falling birthrate, would by themselves have little impact on the dependency ratio in the foreseeable future. Beyond the year 2000, however, the impact of the projected decline in the birthrate on the size of the working-age population begins to make itself felt. The increase is particularly marked in some of the Central American countries, as well as in Colombia and Venezuela. Moreover, the combination of a relatively rapid increase in the elderly population of these countries and the extension of coverage to the bulk of the population could cause expenditures to increase enormously.

Table 4.

Projections of Dependency Ratio in Selected Latin American Countries, 1980-2030

(In percent)

article image
Sources: Vu (1985), and Fund staff calculations.Note: The dependency ratio is calculated by expressing the population aged 60 or more as a percentage of the population aged between 20 and 59.

Calculated by weighting each country’s dependency ratio by its share of the group’s total projected population.

Alternatives to Public Pension and Health Insurance Systems and the Experience of Chile

In 1979, Chile implemented reforms to its pension system that were intended to replace the old public pension system with what could be termed a compulsory private sector savings plan, under which the state’s role would be reduced to one of regulation. Under the new scheme the employer’s contribution has been eliminated, and individuals are required to deposit 10 percent of their salary in one of a group of eligible financial institutions—the Administradoras de Fondos de Pensiones, or AFPs—that the individuals can choose themselves. These funds, which are required to specialize in the management of the savings entrusted to them, restrict their investments to a range of financial securities,11

Each participant in the savings plan has a personal account with his chosen fund, and his share of the net earnings of the fund—its investment income net of commissions—accrues to that account. Pensions are determined by the capitalized value of each person’s contributions to the plan; at retirement, each participant may choose to invest the accumulated value of his contributions and their earnings in a life annuity purchased from an insurance company. The annuity is expressed in real terms—unidades de fomenlo—which are linked to the consumer price index.12 The reforms also provided for a minimum pension for persons with at least twenty years of work experience. If the accumulated capital in an individual’s account is insufficient to provide this minimum pension, the government makes up the difference.13

The measures implemented in 1979 included a provision that allowed Chileans covered by the old public system to switch to the new scheme at any time between 1981 and 1986. Subsequently, the deadline for transfer was extended indefinitely. Persons opting for the new system would have their accounts in the new scheme credited at the time they retired by an amount related to their accrued contributions under the old public system.14 After December 1982, all new entrants to the labor market except the self-employed, whose participation was voluntary, were required to join the new scheme. To foster competition, and hence more efficient management, participants in the new scheme were allowed to switch from one fund to another.

The reforms also included increases in the retirement age, to 60 for women and 65 for men, which strengthened the finances of the old system. Its contribution rates, however, were not changed. Because the percentage of income involuntarily saved under the new system did reflect the new retirement ages, it was less than the contribution rate of the old system, and a substantial incentive to switch to the new system was created. A majority of contributors in the old system switched to the new.15

The government plays an important regulatory role in the new system by overseeing the performance and functioning of the private funds, but it also ensures that their contributors receive a minimum rate of return. This rate of return is defined as 50 percent of the average rate of return for all the funds or the average rate less 2 percent, whichever is less.

The method of determining an individual contributor’s pension differs radically between the two systems. Under the old system, an individual’s pension was determined by the number of years he contributed and his average salary in the period immediately before retirement. Pensions were adjusted from time to time but were not subject to any automatic indexation rule, and their real value fluctuated substantially in the twenty years before the reform was instituted (Arellano (1985, p. 79)).

Under the new system, the value of a pension will be determined by a contributor’s savings and the average rate of return of his fund, subject to a minimum rate of return determined by the government. Consequently, there is no guarantee that pensions will bear any given relationship to a contributor’s earnings, nor that a pension will maintain its value in real terms over the retirement period. The funds conceivably may not be able to achieve positive real rates of return, so the government might be compelled to intervene.

The more stable the financial environment, the more successful management of the funds is likely to be, and the existence of the new system is also likely to create additional incentives for just such an environment. Nonetheless, although 1981-86 was not a stable period, the funds earned rates of return that were high in real terms.

Earlier in Section II, a funded system was compared with a pay-as-you-go system as a generator of savings, but neither system was found to be conclusively superior to the other. Many economists would conclude that Chile’s new system should be preferred because of its ability to generate savings. Nonetheless, it remains critical that the contributions to funds not be used to finance additional current expenditure by the government.

Table 5.

Social Security Administration Expenditures as a Percentage of Total Expenditure in Selected Latin American and OECD Countries, 1980

article image
Source: ILO (1985, 1988).



The new system has been touted as more efficient than the old, although the existence of substantial economies of scale in the administration of public savings plans could mean that a unified public system would be potentially less costly than the new privatized system. The old system, however, was far from unified, and in 1980 administrative expenditure was estimated at about 6 percent of total expenditure by the Chilean social security system, a high figure by the standards of industrial countries but below average for Latin American countries (Table 5) see also “Administrative Issues,” below).16 Private sector operations would presumably have more incentive to minimize their operating costs than would a public bureaucracy, and the government’s guarantee of a minimum rate of return to the contributors is not a guarantee that it will rescue a fund in financial difficulty. Hence, the regulation of the industry does not appear to create a disincentive to cost minimization.

Nonetheless, it is not clear that the new system is less costly than the old. One observer has drawn attention to the fund’s apparently high level of promotional and advertising expenditures in its first few years of operation.17 In addition, the coexistence of the two systems must substantially increase the amount of resources devoted to the administration of savings plans.

The switch to a regulated private and funded system has had the effect of substantially increasing the deficit of the financial operations of the consolidated central government because revenues of the public social security system fell with the transfer of contributors to the private system—although the old system still had to pay the pensions of persons who had retired under that system—and because the central government now had to pay the bono de reconocimiento of each transferee to the new system as he retired.18

Should the increase in the deficit resulting from the reforms, which has been estimated at about 5 percent of GDP in 1983, be regarded as an expansionary shift in the stance of fiscal policy (Yáñez (1985))? To the extent that payroll tax contributions to the public social security system are simply replaced by contributions to a private and compulsory savings plan, while the public social security system goes on making pension payments to retired contributors to the system, the increase in the deficit would not be expansionary. The private funds could invest their revenue in public debt, and the increase in the public sector deficit would be offset by an increase in the savings of the private sector, with no increase in expenditure either for investment or consumption. Only if the savings entrusted to the funds leads to an increase in expenditure—for example, by being invested in private securities that give rise to an increase in investment—would the increase in the deficit be associated with an expansionary shift in the fiscal stance.

The reforms of 1980-81 also affected the health and medical com-lonent of the social security system. Under the previous system public hospitals, whose services were largely free and which catered mainly to the blue-collar worker, existed side by side with a voucher system, whose participants could choose among a variety of health care providers. The social security system financed 50-70 percent of the cost of the vouchers, with users financing the remainder; the social security system’s costs were financed through contributions of 4 percent of workers’ taxable earned income. (Foxley, Aninat, and Arellano (1979, p. 106)).

The new system is also financed by a payroll tax, whose rate was raised to 6 percent in 1983 and, subsequently, to 7 percent. Individuals who do not wish to rely on the services of public hospitals, however, can opt to become members of providential health institutions (Instituciones de Salud Previsional, ISAPRES), which receive the 7 percent contribution from their membership. Each ISAPRES accepts only persons whose annual income is above a certain minimum. At the end of 1984 this minimum varied from 40,000 pesos to 120,000 pesos and was several times the average income of the contributors remaining with the old system.

These changes in the health component of the system have affected considerably fewer people than the reforms made in the pension component. At the end of 1983, some 108,000 persons were registered in some 15 different ISAPRES. The new system has some interesting implications. In particular, it has reduced the revenue of the public system, which no longer receives the contributions of the individuals who registered with the ISAPRES. Health expenditures have also been reduced; however, because persons registered at the ISAPRES were among the better off, they were undoubtedly subsidizing the group of persons remaining with the public system. The result is that the redis-tributive role of the health component has been lessened. (Arellano (1985, pp. 188-91)).

Effects of Social Security on Income Distribution

Two characteristics of Latin American social security systems are particularly relevant in any assessment of the effects of social security on income distribution; their coverage, which is often quite low, and the financing role played by transfers from the central government, in the form of either earmarked revenues or general budgetary support.

When coverage is not universal, the incidence of the social security system depends critically on the financing source and its incidence.19 There is typically a correlation between the degree of urbanization and industrialization and the extent of coverage. In general, those excluded from coverage are in the rural and unorganized urban sectors, where the incidence of extreme poverty is the highest. Those covered often form a relatively privileged group.20

Under these circumstances, the incidence of the system could be broadly neutral if social security expenditures were financed entirely by payroll taxes whose effective (regardless of legal) incidence fell on employees, and if an individual’s benefits were related to his contributions. As noted previously, however, there is reason to believe that the consumer bears at least part of the burden of the incidence of payroll taxes. Thus, uninsured persons could be paying, in the form of higher prices, for the benefits that accrue to a relatively privileged minority. Among the insured, some transfer of income from one income class to another could also take place. If the system has a minimum pension, then those who are better off subsidize those who are worse off: this provision has a progressive incidence. Conversely, a cap on the absolute value of contributions has a regressive incidence if no similar cap affects pensions.

By contrast, if coverage is universal or at least does not exclude those at the lowest end of the income scale, then the incidence of the payroll tax is less important. If employers pass on the tax in higher prices, the group bearing the effective incidence of the tax, the general population, is also benefiting from the expenditures the tax finances.

When coverage is limited and social security expenditure is financed at least in part out of general revenues, then its effective incidence is likely to be regressive, even if the employed contributors bear the burden of the payroll tax component of the system’s revenues. The general revenue component is partially borne by the insured, who benefit from the system’s expenditures; the greater burden of general revenues is borne by the uninsured, who do not. But the uninsured are likely to be among the relatively poor. It can be tentatively concluded that the incidence of the social security system is regressive in most of Central America, and also in relatively affluent countries such as Colombia and Venezuela, because coverage in these countries is low.

Two recent studies (cited in Mesa-Lago (1983)) of countries where coverage is relatively broad by the standards of the region have found that the incidence of social security was either neutral or slightly progressive. A study of the Costa Rican system as of 1973 found that, even with 50 percent coverage, the system’s incidence was slightly progressive. The progressivity of the incidence of the system’s health component, which was found to be distributed quite evenly regardless of income level, offset the regressivity of a ceiling on income subject to contributions and the regressive impact of less than universal coverage. (The degree of coverage has increased substantially since 1973; see Wilner-Green (1977).) A study of Chile in the early 1970s found that the incidence of the system was broadly neutral for the insured group (Foxley, Aninat, and Arellano (1979)). But for the population as a whole, the incidence could be regressive.

Problems Posed by an Inflationary Environment

An inflationary environment complicates the administration of social security and the social security system itself. As an important component of general government, the social security system can play an important role in sustaining or generating inflationary pressures, or in resisting these pressures, depending on its indexation policies.

Consider first the impact of inflation on the financial balance of a social security system whose sole revenue source is the payroll tax. A revenue system that is neutral with respect to inflation is defined here to mean one in which a change in the rate of inflation does not affect the ratio of revenues to GDP. For neutrality to hold, the following set of conditions is sufficient: first, the built-in elasticity of tax revenues with respect to the tax base—the wage bill—must be unity; second, the elasticity of the wage bill with respect to GDP must also be unity. If the second condition holds, the built-in elasticity of payroll tax revenues with respect to the wage bill will be unity when the tax is a fixed percentage of wages, with no ceiling on contributions and no lag in collection.

Lags in collection reduce the ratio of taxes to their base when the rate of inflation increases because the base is raised by inflation before the yield of the tax is raised (Tanzi (1977)). In practice there will always be some lag, although in the case of the payroll tax it should be relatively short if the degree of contributors’ compliance is sufficiently high. However, the inflationary process can create strong incentives for enterprises to delay their tax remittances unless the penalty for doing so varies directly with the rate of inflation and with the period of the delay.

If the penalty for late remittances cannot be adjusted promptly when inflation accelerates, it should be set at a very high level; otherwise, the acceleration of inflation can create serious problems for social security finances, even if the tax is a fixed percentage of wage income and there is no ceiling on taxable income.21 If the financial position of enterprises deteriorates along with the acceleration of inflation, the incentive to delay remittances will be all the greater, especially if the compliance of contributors is not strong to begin with. Thus, in some circumstances accelerating inflation can significantly reduce the ratio of social security revenue to GDP.

The ratio of the social security system’s expenditure to GDP will depend on the indexation policy the system follows. Assuming that the indexation formula relies on the consumer price index (CPI), that the CPI increases at the same rate as the GDP deflator, and that the elasticity of real expenditures with respect to real GDP is unity, the ratio of expenditure to GDP will vary directly with the frequency of adjustment, inversely with the lag in the adjustment, and inversely with the rate of inflation. In other words, accelerating inflation reduces the ratio more if the adjustment lag is long and if the frequency of adjustment is low. For example, with an initial ratio of expenditure to GDP of 10 percent, an increase in the rate of inflation from zero to 50 percent reduces the expenditure ratio to 8.7 percent even when expenditure is adjusted quarterly with a quarterly lag. When adjustment takes place once instead of four times a year, the ratio falls to 7.5 percent. Unless the adjustment lag is zero and the adjustment is continuous, an increase in the rate of inflation will lower the ratio of expenditure to GDP, even with full indexation.22

The choice of an automatic or discretionary approach to the indexation of the pension component of social security expenditures depends to some extent on the index used. A policy of full indexation to the CPI implies that one social group receives special protection from shocks to the economic system that entail a decline in real income; for example, when the economy suffers a substantial terms of trade loss. In such circumstances, this kind of indexation may forestall the adjustment in real incomes necessary to restore external balance, and it is worth adding that pensioners in Latin America can be a privileged group.

In consequence, a policy of indexation to the CPI may be misguided. Indexing pensions on wages would be a better rule for financial stabilization. But could a policy of complete discretion be even better? The argument for discretion is that no single rule would always be appropriate. If real wages are rising at an unsustainably rapid pace, then wage-based indexation would not be appropriate. Yet it may be argued that the long-term functioning of a social security system may require that contributors be able to expect a certain stability in their benefit entitlements. The periodic adjustment of pensions in line with wages can thus be seen as the expression of an implicit contract that pensions will normally be related to earned incomes, even if pensioners will not always be protected from declines in their real income.

Administrative Issues

Evasion of payment of social security contributions appears to have been significant in several Latin American countries, although it is difficult to ascertain its extent. It is likely to be more prevalent in the rural and unorganized urban sectors because enterprises in these sectors typically do not keep accounting records and because, given the small size of these enterprises, administrative surveillance is often not cost effective. In addition, evasion is likely to be more prevalent in systems in which there is little or no relation between a participant’s contributions (including contributions made on his behalf by his employer) and the benefits he enjoys. For example, the incentive for workers to collude with employers to evade contributions would be greater if pensions were not related to the value of contributions, or if the contributions financed medical care and payments for income lost while workers were sick.23

The temptation to evade may increase in periods of recession, and it must certainly increase when the rate of inflation increases if penalties for evasion are a fixed percentage of the value of contributions not remitted. Finally, the rate of the payroll tax itself is related to the incentive to evade.

Clearly, evasion will be a problem if the social security institution cannot command the administrative means to ensure a reasonable degree of compliance. In a system in which benefits do depend on an individual’s contributions, the incentive to cheat is probably greater when no adequate records of an individual’s contributions are maintained. Keeping such records is imperative, not just to reduce evasion but also to ensure that benefits are fairly evaluated.

Substitution of the VAT for the payroll tax has been seen as a remedy for the problem of compliance because with the credit-invoice system of the VAT there is no incentive to underinvoice at either the wholesale or manufacturing stage of production. The VAT is considerably more difficult to administer than the payroll tax, however, and if it does not fully replace the payroll tax there is no saving from the dismantling of the administrative apparatus of the payroll tax (McClure (1981)).

Tardy remittances have also been a problem. Because of inadequate penalties for late remittances, in some countries business enterprises have enjoyed a loan from the social security regime at a negative rate of interest. Either periodic adjustments of the penalty for late payment or a constant, very high, penalty rate is necessary to resolve this problem when the rate of inflation is high and variable. Finally, late remittances are tacitly encouraged if the government is late in making its own contributions to the social security system, or in establishing and settling claims to pensions.

The stratification of social security schemes in the countries that first introduced social security must have contributed to substantial administrative inefficiency in the past. Because of the large number of plans, administrative resources were duplicated. Moreover, the lack of coordinated administration was responsible for serious inequities, including the nontransferability of years of participation under one scheme to another and the possibility of an individual’s receiving two pensions at excessively generous terms because benefits were not coordinated when contributory service in one plan was transferred from another. The reform measures implemented recently in Chile and Uruguay have substantially increased the unification of those systems. Nonetheless, the process is not yet complete, and there is evidence that some advantages of a unified system have yet to be exploited fully (Mesa-Lago (1985, p. 199)).

Administrative expenditures in Latin American countries are high in relation to those of OECD member countries if the ratio of administrative expenditures to total expenditures is taken as the index of relative cost (Table 5). In Bolivia and Ecuador, this index is high by any standard. Economies of scale in program administration could explain the lower cost of the OECD programs if these programs were in general larger than programs in Latin America. However, Denmark and Norway, both with relatively small programs, have cost indices lower than any Latin American country.24 Mesa-Lago (1985) has drawn attention to the lavishness of social security administrations’ headquarters and has argued that payrolls and benefits of the administrative staff have typically been excessive.

The health expenditure component of social security systems, particularly the medical services component, presents special administrative problems. In many countries, parallel hospital systems exist: one for social security contributors, another—operated by the ministry of health—for the rest of the population not relying on private clinics. The social security medical program typically covers a minority of the population—mainly blue- and white-collar workers in the cities. The rural population and indigent city dwellers in large part are covered by the system operated by the ministry of health. Despite its smaller coverage, the social security medical care system usually has a larger budget than the medical care component of the ministry of health, and its per capita expenditure is much larger (Ugalde (1985, p. 111)).

In some countries the services of the two institutions have not been coordinated, leading to a considerable duplication of services; an example is two hospitals serving the same community where one would suffice, so that in many countries the occupancy rates of hospital beds have been as low as 50 percent (Mesa-Lago (1985, p. 26)). Efforts to improve the coordination of services have in general met with little success. One source of difficulty has been the preference of contributors to social security for a separate hospital system because of the poorer quality of service in the public system;25 however, not all countries face this type of problem. In Costa Rica, for example, the Ministry of Health is responsible for public health programs and for medical care in rural areas not covered by the social security system, and the social security system is responsible for the bulk of medical care services. The two services are being gradually integrated.

The concentration of social security’s resources in the cities and the financial constraints imposed on the budgets of health ministries imply that a significant proportion of the rural population has limited or no access to modern medical care. These inequalities cannot be eliminated by an extension of the coverage of the social security system at its present level of benefits per capita. Such an extension would require an enormous increase in the financial resources mobilized by the system.

III. Conclusions

The social security systems of some Latin American countries are large enough that their financial imbalances can have serious macro-economic consequences. These imbalances reflect the influence of the erosion of the payroll tax base because of falling employment, as well as the secular influence of an aging population on the finances of the system’s pension component. To forestall the emergence of more serious financial disequilibria, several countries have had to allow a marked erosion in the level of social security expenditure in real terms. This reduction has been accomplished by various means, including less than full and timely indexation of pension benefits and increases in minimum retirement ages.

The social security systems of most Latin American countries now absorb only a small share of GDP, essentially because the coverage of the systems is low, even though expenditure per participant can be quite high. However, as the pension component of these systems matures—as a greater share of the elderly population, having contributed for a longer time, becomes entitled to more benefits—the share of expenditure in GDP can be expected to grow, even without a marked expansion in the systems’ coverage. The reported accounting surplus in the pension system in some of these countries most likely disguises an actuarial deficit. Future benefits promised under current legislation cannot be financed out of the system’s income from reserves and contributions at current rates of contribution by the system’s participants. An additional problem confronts these countries: any significant expansion of the system will not be feasible at the current level of expenditure per participant, so there will be a trade-off between the generosity of benefits and the degree of coverage of the system.

Although demographic forces have contributed to the upward pressure on social security expenditures in the last twenty-five years, their influence should abate in most countries in the near future because no marked increase in the ratio of elderly to working-age population is projected to take place in the next twenty years. In the more distant future, however, a further decline in the birthrate and some further increase in life expectancy could push the dependency ratio up again, putting pressure on pension expenditure. At the same time, the increase in life expectancy and demands for high-technology medical care will create pressure on medical expenses.

Funding Versus Pay-As-You-Go

No country in the region has a fully funded pension system except Chile. Pay-as-you-go systems may reduce national saving and capital formation, but evidence to support this view is not conclusive. Funded systems can be successful in a stable financial environment, but they are vulnerable to both inflationary shocks and misguided investment policies. Their reserves should not normally be used to finance increases in government expenditure or tax cuts, so that the introduction or expansion of a funded system should reduce the deficit of the consolidated central government. An exception to this rule can be made if the private sector views its contributions to the system as a tax, rather than as a kind of forced saving.

Revenue Sources and Distributional Considerations

The main but not exclusive source of finance for Latin American social security systems is the payroll tax. The base of this levy is considerably less broad than it is in most OECD countries, so that exclusive reliance on it would impose a stringent limit on the size of the system. The payroll tax is a relatively simple levy to administer, and if its effective burden rests on participants in the social security system it has the additional advantage that those who benefit from the system pay for it.

When coverage is less than universal, as is the case in virtually all Latin American countries, and the system is partially financed out of general revenues, the incidence of the system is probably regressive because persons excluded from coverage will bear part of the burden of the taxes that finance it. This conclusion is strengthened if the burden of the payroll tax is shifted forward to the consumer.

The narrow base of the tax means that its rates must be high to finance an extensive system; if the tax is not borne by labor but results in higher labor costs, these rates can encourage capital-intensive production methods and reduce employment. If the base of the tax cannot be broadened to finance the expansion of the system, then both equity and efficiency considerations suggest that additional revenue must be sought from a tax or combination of taxes whose incidence is less regressive and more neutral in its impact on factor costs.

Alternatives to Conventional Social Security Programs

Alternatives to the traditional model of social security are possible in Latin America, as the reforms in Chile demonstrate. The new Chilean system preserves some basic features of the old public system but reduces the direct role of the government. Thus, the new pension system obliges the current generation of young adults to make contributions to a kind of individual retirement account managed by a private financial institution under public supervision. If its participants completely trust the new system, it will have the merits of lessening the incentive to evade the making of contributions and of creating a more solid and transparent link between the value of contributions and the value of benefits that contributors will ultimately receive. The successful operation of a Chilean-style plan, however, requires both a well-developed system of financial intermediation and a stable financial environment.

Impact of Inflation

The inflationary environment of Latin America complicates the management of social security systems, particularly of their pension component. Accelerating inflation tends to reduce both revenue and expenditure relative to GDP. A high and volatile rate of inflation can produce sizable swings in the real value of benefits and their share of GDP even if the frequency of adjustment is fairly high and the lag in adjustment short. With a significant lag in revenue collection, and prompt and timely indexation, the deficit could grow as a share of GDP.

Administrative Issues

Finally, administrative expenditures in Latin American social security systems are high by comparison with OECD countries because of excessive staffing levels and bureaucratic competition over the supply of medical services between social security institutions and ministries of health. In some Latin American countries the systems still carry the vestiges of numerous benefit plans and institutions that cover different occupational and social groups, with resultant administrative duplication and inefficiency. In consequence, the potential for cost savings is great.


  • Aaron, Henry J., Economic Effects of Social Security (Washington: The Brookings Institution, 1982).

  • Arellano, Jose Pablo, Politicas Sociales y Desarrollo—Chile 1924-84 (Santiago: CIEPLAN, 1985).

  • Boskin, Michael, Too Many Promises: The Uncertain Future of Social Security (Homewood, Illinois: Dow Jones-Irwin, 1986).

  • Boskin, Michael, Lawrence J. Kotlikoff, and John B, Shoven, “Personal Security Accounts: A Proposal for Fundamental Social Security Reform,” Publication No. 63 (Stanford, California: Center for Economic Policy Research, 1985).

    • Search Google Scholar
    • Export Citation
  • Break, George F., “The Economic Effects of the OASI Program,” in Social Security Financing, ed. by Felicity Skidmore (Cambridge, Massachusetts: MIT Press, 1981), pp. 45- 81.

    • Search Google Scholar
    • Export Citation
  • Foxley, Alejandro, Eduardo Aninat, and Jose “Pablo Arellano, Redistribute Effects of Government Programmes—The Chilean Case (Elmsford, New York: Pergamon Press, 1979).

    • Search Google Scholar
    • Export Citation
  • HalteT, William A., and Richard Hemming, “The Impact of Demographic Change on Social Security Financing,” Staff Papers, International Monetary Fund (Washington), Vol. 34 (September 1987), pp. 471- 502.

    • Search Google Scholar
    • Export Citation
  • International Labor Office, The Cost of Social Security, Eleventh International Inquiry, 1978-80 (Geneva, 1985).

  • International Labor Office,The Cost of Social Security, Twelfth International Inquiry, 1981-83 (Geneva, 1988).

  • International Monetary Fund , International Financial Statistics (Washington, various issues).

  • International Social Security Association, Chapter 3 in Medical Care Under Social Security in Developing Countries, Studies and Research No. 18 (Geneva, 1982).

    • Search Google Scholar
    • Export Citation
  • McClure, Charles E., Jr., “VAT Versus the Payroll Tax,” in Social Security Financing, ed. by Felicity Skidmore (Cambridge, Massachusetts: MIT Press, 1981), pp. 129- 64.

    • Search Google Scholar
    • Export Citation
  • Mesa-Lago, Carmelo, “Social Security and Extreme Poverty in Latin America,” Journal of Development Economics (Amsterdam), Vol. 12 (February-April 1983), pp. 83- 110.

    • Search Google Scholar
    • Export Citation
  • Mesa-Lago, Carmelo, El Desarrollo de la Seguridad Social en America Latina (Santiago: CEPAL, 1985).

  • (1986a), “Comparative Study of the Development of Social Security in Latin America,” International Social Security Review (Geneva), Vol. 39 (1986), pp. 127- 52.

    • Search Google Scholar
    • Export Citation
  • (1986b), “Social Security and Development in Latin America,” CEPALReview (Santiago), No. 28 (April 1986), pp. 135- 50.

  • Shome, Parthasarathi, and Lyn Squire, “Alternative Mechanisms for Financing Social Security,” World Bank Staff Working Paper 625 (Washington, 1983).

    • Search Google Scholar
    • Export Citation
  • Tanzi, Vito, “Inflation, Lags in Collection, and the Real Value of Tax Revenue,” Staff Papers, International Monetary Fund (Washington), Vol. 24 (March 1977), pp. 154- 67.

    • Search Google Scholar
    • Export Citation
  • Ugalde, Antonio, “The Integration of Health Care Programs into a National Health Service,” in The Crisis of Social Security and Health Care—Latin American Experiences and Lessons, ed. by Carmelo Mesa-Lago (Pittsburgh: University of Pittsburgh Press, 1985), pp. 109- 46.

    • Search Google Scholar
    • Export Citation
  • Vu, My T., World Population Projections 1985—Short- and Long-Term Estimates by Age and Sex with Related Demographic Statistics (Baltimore: Johns Hopkins University Press, 1985).

    • Search Google Scholar
    • Export Citation
  • Wallich, Christine, “Savings Mobilization Through Social Security—The Experience of Chile During 1916-77, ” World Bank Staff Working Paper 553 (Washington, 1983).

    • Search Google Scholar
    • Export Citation
  • Wilner-Green, Dianne, “Some Effects of Social Security Programs on the Distribution of Income in Costa Rica” (Ph.D. dissertation; Pittsburgh: University of Pittsburgh, 1977).

    • Search Google Scholar
    • Export Citation
  • Wilson, Richard R., “The Impact of Social Security on Employment,” in The Crisis of Social Security and Health Care—Latin American Experiences and Lessons, ed. by Carmelo Mesa-Lago (Pittsburgh: University of Pittsburgh Press, 1985), pp. 245- 84.

    • Search Google Scholar
    • Export Citation
  • Yáñez, José H., “Impacto de las Políticas Previsionales y del Trabajo en el Presupuesto Público: 1979-83,” Estudios de Economía (Santiago), Vol. 12 (August 1985), pp. 143- 67.

    • Search Google Scholar
    • Export Citation

Mr. Mackenzie, Senior Economist in the Government Expenditure Analysis Division of the Fiscal Affairs Department, holds degrees from Dalhousie University, Canada, and Oxford University, He has published articles on public financial and macroeconomic issues. In addition to many colleagues at the Fund, the author thanks Miguel Urrutia and Professor Carmelo Mesa-Lago for their comments on this study.


The International Labor Office (iLO) includes in its coverage of social security the schemes or services that have the following three objectives: the provision of curative or preventive health care; the maintenance of income in the case of involuntary loss of all or a large part of income, including income lost as a result of retirement; and the granting of a supplemental income to persons having family responsibilities. Benefits can be received under the following nine categories: medical care, sickness, unemployment, old-age, employment injury, family, maternity, invalidity, and survivor’s benefit. These definitions were used to compile the data that underlie Table 1. See ILO (1985, pp. 1-2).


This section draws heavily on Mesa-Lago (1985). The main results of this study are summarized in Mesa-Lago (1986a, pp. 127-52, and 1986b).


The estimates of coverage are uncertain for several countries for a variety of reasons, including the lack of accurate records of the number of dependents and some duplication of coverage. See Mesa-Lago (1985, pp. 10-13, and Table 2, p. 270).


An earlier version of the paper contains an appendix discussing developments in Uruguay and is available from the author.


These points are made in a discussion of investment policy in Chile's social security system in Wallich (1983).


Boskin (1986, p. 80) who argues that the pay-as-you-go system in the United States has depressed private saving and capital formation, has nonetheless stated that the “evidence is far from conclusive.” Aaron (1982, p. 51) states that “using the best that economic theory and statistical techniques have to offer, [economists] have produced a series of studies that can be selectively cited by the true believers of conflicting hunches [about the effect of an unfunded social security system on saving] or by people with political agendas that they seek to advance,” This work surveys the empirical problems involved in estimating the impact of social security on saving. A survey of empirical studies and a tabulation of their conclusions may be found in Break (1981).


The dependency ratio is the ratio of the number of pensioners (P) to the number of contributors (C). The total value of pensions in a given period is P times the average value A. Under a pay-as-you-go system, twC PA, where t is the contribution rate and w the average wage. If A iw is a constant, then 1 varies directly with PIC, the dependency ratio.


This argument is discussed in Halter and Hemming (1987).


For example, if the deficit of the central government before consolidating the financial operations of the social security system is 5 percent of GDP and a funded social security program is introduced that initially generates a surplus equal to 2 percent of GDP, the surplus could be invested in government bonds, on the assumption that the government's operations are unaffected by the presence of the social security system, so that the deficit of the consolidated central government declines to 3 percent of GDP. In a closed economy, the private sector would have a surplus of 5 percent before the introduction of the social security system and a surplus of 3 percent afterward.


The 1985 projections are contained in Vu (1985).


A discussion of various aspects of the new scheme can be found in Arellano (1985, Chapter 3).


Individuals retiring under the new system may also pick the programmed retirement option, which entitles them to monthly payments directly from the AFP. This second option does not guarantee a constant real payment to the pensioner, as the first option does. The Chilean system is similar in some respects to a proposed reform of the U.S. social security system. See Boskin, Kotlikoff, and Shoven (1985).


An additional obligatory deduction of about 3.5 percent is used to purchase disability and survivor's insurance from an insurance company.


At an individual's retirement, his account is credited with a bono de recon-ocimiento, a “recognition bond,” by the government.


Arellano (1985) estimates that the number of persons contributing to the old system declined from 1,695,000 in 1980 to 449,000 in 1983, and that the number of individuals contributing to the new system reached 1,106,000 by the end of 1984 (Arellano (1985, p. 145)). Mesa-Lago has estimated that 83 percent of insured persons was in the new system at the end of 1986 (Mesa-Lago, personal correspondence with the author (1987)).


ILO (1988, Table 1). The average for the 14 countries in Latin America shown in Table 5 is 11 percent.


Arellano (1985, p. 170) has estimated that marketing and sales costs amounted to 39.8 percent of total operating costs exclusive of depreciation and amortization in 1982, and 30 percent in 1983. Expenditures in 1981 were even higher because the sales campaigns were launched in that year.


Another measure taken in the reform, but not discussed previously, that had the effect of increasing the deficit was the reduction in the rate of contribution for the family allowance and cesantia (early retirement subsidy) component of social security, which continued to be publicly administered.


The incidence of the system as a whole is said to be regressive if the net transfer it creates increases in proportion to income as income increases. Typically, a system will be neither regressive nor progressive over the entire range of income.


Mesa-Lago (1983) notes that in the countries where social security was pioneered, such as Chile, Argentina, and Uruguay, the first groups to be covered were in general relatively well-off workers in the mining enclave, transport sector, or public sector.


Setting the penalty at a high level can cause problems too: when the rate of inflation is low, a penalty rate high enough to deter late remittances at high rates of inflation will be punitive.


An earlier version of the paper contains an appendix with illustrative calculations of the impact of lags and frequency of adjustment on the ratio of pension expenditure to GDP and is available from the author on request.


This relationship assumes that contributors would not discount future income excessively.


A further possible explanation is that some of the capital equipment used by administrative agencies in Latin America may be as expensive as that used by OECD countries, whereas the value of benefits per capita is much lower, given the much lower per capita income levels in Latin America to which benefits are scaled.


This is discussed in Chapter 3 of International Social Security Association (1982).