|
|
|
Map of South America |
Email this citationCIAO DATE: 06/02
Globalization, Domestic Politics and Welfare Spending in Latin America: A Time-Series Cross-Setion Analysis, 1973-1997.
Robert R. Kaufman and Alex Segura-Ubiergo
2000
"Has globalization gone too far?" This question — the title of a recent book by Dani Rodrik (1997) — has been asked for over a century in Latin America. The issues it raises, however, have acquired special force in the last 25 years, as once-closed import-substituting economies have been transformed by structural reforms that have linked them far more closely to international trade and capital markets. As in other parts of the world, the specific effects of this transformation on Latin American societies remain unclear. Nevertheless, it seems quite apparent that it has brought about important modifications in the balance of political power and has altered the margins of choice available to domestic governments.
In this paper, we examine one of the most controversial aspects of this economic opening: its impact on governments' fiscal commitments to social security, health and education. Many have argued that the new era of neoliberal reforms has undermined the thin protections that states in the region had provided to at least some of their citizens during earlier periods of ISI. Whether or not this is the case, the central challenge going forward is whether shattered welfare systems can be reconstructed and expanded in ways that will shield citizens exposed to new market forces and enable them to compete effectively in the new era of "globalization."
We explore these issues through an analysis of changes in social security transfers, health, and education expenditures in a time-series cross-sectional analysis in 14 Latin American countries from 1973 to 1997. We examine three sets of issues. First, we want to know whether there is evidence that integration into global markets has in fact constrained social spending. On this question, we draw heavily on the distinction drawn by Geoffrey Garrett (1999a, 1999b) between an "efficiency" hypothesis which posits that increasing exposure to international competition will induce governments to roll back social expenditures, and a "compensation" hypothesis which emphasizes incentives to invest in "human capital" and to respond to political demands for protection against risk. Within this context, we then turn to an examination of the extent to which such outcomes might be influenced by two additional sets of domestic political and institutional factors: the balance of partisan power, and the electoral pressures of democratic institutions.
Consistent with Garrett's findings for a larger global sample, we show that trade integration has a consistently negative effect on social spending, and that this is compounded significantly by openness to capital markets. This is the strongest and most robust finding in our study. Against at least some of the studies of OECD countries, moreover, partisan politics appears to make little difference; left-oriented governments do not spend significantly more or less than others. Democracy also has only a minimal and rather inconsistent effect. When spending measures are disaggregated, however, we do find that globalization and electoral participation have a more complex impact. The negative impact of international economic integration operates mainly through social security transfers, while electoral participation has a strong and positive effect on health and education expenditures.
One reason our study is distinctive, we believe, is that it deploys broader measures of social spending than are found in most other LDC samples, and that these are examined on an annual basis over a relatively long period of time. In our analysis of these data, we use a pooled time-series error-correction model, estimated through Ordinary Least Squares with panel-corrected standard errors to correct for panel heteroskedasticity and spatial correlation; a lagged dependent variable to model the time dynamics and correct for serial correlation; and country and time dummies to control for fixed effects. This methodological procedure establishes a high threshold for estimating conventional levels of significance (Beck and Katz 1995, 1996). Such estimates are more reliable in the sense that the estimation of the standard errors is more efficient and consistent (if we accept, as we do in this paper, Beck and Katz's claims) 1 .
We present our analysis in the following steps. In the first section, we outline the main theoretical arguments about how globalization and domestic politics might influence social spending in Latin America. The second section discusses the variables and the models that are used in the analysis. In the third section, we present our findings for changes in aggregate social spending. In the fourth, we show the impact of economic and political variables when spending is disaggregated into transfers on the one hand, and health and education expenditures on the other. The last section is the conclusion.
I. The argument: globalization, domestic politics, and welfare spending in Latin America
The efficiency and compensation hypotheses.
Contending hypotheses about the effects of globalization on welfare spending constitute our point of departure. As Garrett (1999a, 1999b) has noted, there are two quite contradictory sets of arguments that cannot be resolved without empirical research. Each offers quite different propositions about the preferences and power of labor and capital, and about the economic and political options which governments face.
The "efficiency hypothesis" rests on the assumption that high levels of welfarespending reduce competitiveness in global markets. This effect operates through multiple channels. Welfare spending increases the cost of labor and reduces the competitiveness of exports. Fiscal expenditures also put upward pressure on interest rates, which in turn crowd out private investment and contribute to an appreciation of the exchange rate. As business groups become increasingly exposed to international competition, therefore, they can be expected to press governments to reduce social expenditures. Integration into capital markets would presumably compound this pressure, since it increases the exit opportunities available to asset holders.
At the same time, we might also expect a decline in labor's capacity to resistreductions in social spending. Although the Hecksher-Ohlin theorem implies increased demand for low-skill labor, the abundant factor in most poor countries (Rogowski 1989), workers' bargaining strength in Latin America is undermined by several factors. First, unions have been based in the public sector and in import-substituting industries, both of which have been seriously weakened by trade liberalization. Moreover, the large pool of underemployed rural and informal sector workers creates a slack in the labor market that cannot be reduced quickly (Rudra 1999). Even in an era of increased labor migration, finally, capitalists have greater exit options than workers, and are thus in a better position to close their plants or relocate as the price of labor increases (Rodrik 1997, 46). The greater the exposure to trade competition, therefore, the stronger the incentives for governments to cut back welfare expenditures, or reduce the rate of increase.
The "compensation" counter-argument posits just the opposite effect. It focuses on the role of the welfare state as a mechanism for offsetting the social costs of international integration and for contributing to the development of "human capital." It is inspired by the strong empirical relationship between economic openness and large public sectors in the OECD countries (Cameron 1978, Katzenstein 1985). It is important to be cautious about the causal inferences that can be drawn from this link, since the expansion of European welfare states was influenced by a variety of long-term factors which predated the current age of globalization. Nevertheless, even in the contemporary period, both Garrett (1999a ) and Rodrik (1997) find some connection between openness and the size of government in both LDCs and developed countries.
There are at least two reasons why this might be so (each of which suggests that the term "compensation hypothesis" may be somewhat misleading). First, the economic uncertainty and social dislocations associated with integration into international markets can be politically destabilizing. Incumbent governments, and conceivably private capitalists as well, have an incentive to use increased welfare spending to ward off such threats. Second, to the extent that social spending enhances the skill levels and productivity of the labor force, it provides a collective good to the business sector. The more competition this sector faces, therefore, the more it might welcome — or even press for — these expenditures.
Domestic politics.
Whether governments respond to the incentives for "efficiency" or "compensation" strategies may also depend on the means citizens have to mobilize around economic interests and to hold governments accountable. This implies that two additional sets of political and institutional factors may also influence social spending as economies become more open. One of these concerns the balance of power among interest groups and party organizations. In the OECD countries, strong unions and social democratic governments have often been the paramount forces behind the expansion of welfare systems. Conceivably, they are also important forces for resisting cutbacks, although this is a matter of some dispute in the OECD cases (Pierson 1996, Hicks 1999).
In Latin America, as in other LDCs, unions are notoriously weak, at least when compared to their counterparts in Western Europe; moreover, cross-nationaldifferences within the region are extremely difficult to measure systematically. A recent study by Nita Rudra (1999) attempts to circumvent this measurement problem by focusing on variations in labor market conditions as a proxy for bargaining power of organized labor. In a global sample of LDCs, she finds that social security spending varies positively with the ratio of skilled to unskilled labor, and negatively with the pool of "surplus" labor. On the other hand, we still lack more direct and reliable indicators of organizational strength (membership, cohesion, etc.) that characterize studies of the OECD.
In this work, we take a somewhat more direct approach to this problem by examining the relative balance of partisan power. Specifically, we hypothesize that social spending is more likely to be sustained under presidents who have been elected with the support of unions and/or popularly-oriented political parties.
Finally, within the Latin American and LDC context, we need to ask more explicitly whether democracy itself "makes a difference." This has been a matter of some dispute in the literature on economic and social reform. One view is that the general distinction does not have much explanatory significance and that it is more important to focus on more specific features of constitutional design, party systems, and partisan politics (Haggard and Kaufman 1995). An alternative perspective rests on a relatively straightforward theoretical point: democratic rulers face pressures from a mass electorate to deliver social services, and are thus more likely than authoritarians to respond to demands for "compensation." In fact, a recent study by Brown and Hunter (1999) provides some support for this hypothesis; they find that democratic regimes in Latin America are more inclined than autocracies to respond to political pressures for social spending. Their work, however, does not address the question of how this relationship might be affected by integration into international markets.
The Latin American sample.
These issues and related questions have received considerable attention in quantitative studies of OECD countries, and more recently, they have been explored in global samples (Garrett 1999a, Rodrik 1997, Rudra 1999). Our Latin American sample cannot draw on the refined data sets available in the OECD and it lacks the wide empirical scope of the broader samples. 2 On the other hand, focusing on the countries of the region does have a number of advantages.
First, unlike many other LDCs and transition economies, many countries of Latin America have long had welfare systems modeled along European lines, with defined-benefit pension plans, health services, and family allowances. By the 1920s, the groundwork for these systems had been established in Argentina, Uruguay, Chile, and Brazil. During the 1930s and 1940s, a second wave of countries followed suit, including Costa Rica, Mexico, Venezuela, Panama, and Colombia (Mesa-Lago 1978, 1989).
Of course, coverage was limited and highly unequal, especially in the poorer countries of the region. Nevertheless, by the 1980s, Mesa-Lago (1989, 41) estimates that coverage reached from 62 to 96 percent of the economically-active population in at least five countries (Uruguay, Argentina, Chile, Brazil, and Costa Rica), and from 45 to 53 percent of the active population in Panama, Mexico, and Venezuela. Social safety nets were thin and tattered by developed-country standards. Nevertheless, at the onset of Latin America's "great transformation," they constituted an important part of the "social contract" connecting citizens with the state. As in the OECD, therefore, ensuing pressures to expand or contract the public financing for such programs can be presumed to have an important bearing on the social organization of the political system. 3
The Latin American sample is also interesting because of the political transformations that swept the region over the past two decades. Latin American countries were among the first non-European states to join the "third wave" of democratization. The fact that these political transitions occurred more or less concurrently with economic openings gives special salience to the question of whether democracies can mitigate the potential negative effects of globalization. The first of the regime changes occurred in Ecuador during the late 1970s, and then spread to other countries during the 1980s and early 1990s. On average, this means that these "new democracies" have been in place longer than those of other regions, including the transition economies — which gives us a longer time span with which to evaluate their effects. On the other hand, substantial differences in the timing of the transitions and the current age of the democratic regimes makes meaningful intra-regional comparisons possible.
Finally, the limited cross-national scope of our sample is partly offset by the quality and reliability of the data that can be compiled. This is particularly important with respect to the measurement of the dependent variable. Our coverage of social spending contains a number of problems that we will discuss below; but it does add some important dimensions to other studies of LDCs. The aggregate measures of government spending used in Rodrik's (1997) and Garrett's (1999a) global samples are imperfect substitutes for social spending, at least in Latin America. The simple correlation between central government expenditures and social spending as percentages of GDP is high (.81), but those with welfare expenditures per capita and as a proportion of government spending are only .51 and .31 respectively. Spending data in several other important studies have not included health and education expenditures (Rudra 1999), or cover a more limited period of time (Brown and Hunter 1999).
Like these other measures, the validity of our data is compromised by the fact that some of the most serious problems of LDC welfare systems involve defects in the delivery of services, rather than financing per se.(Huber 1996, Nelson 2000) For this reason, all types of spending measures are very imperfect proxies for the actual payoffs which citizens receive. It is plausible to assume, however, that even relatively efficient delivery systems will require significant financial commitments from the public sector. The types of spending measures used in this study and others provide at least a rough indication of the resources governments are prepared to devote to welfare needs. We turn now to a more direct discussion of the variables and the models used in this analysis.
II. The variables and the models
The Variables
Social Spending. As discussed, our measures of social spending encompass health care, education, and social security transfers. These have been collected from annual issues of the IMF's Government Financial Statistics, and corrected where necessary for inflation and changes in currency. Expenditures are standardized per capita (in 1995 USD), and as percentages of public spending and GDP. We present findings for each specification of the dependent variable, since each arguably represents somewhat different kinds of welfare effort. As a percentage of GDP and government expenditures, social spending reflects the extent to which governments are willing to commit national resources, and their budgetary priorities. Welfare dollars per capita measures the value of the resources potentially available to the population. Given these differences, we thought it best to examine the effects of our independent variables on all three specifications of spending.
Unfortunately, as in most other large-n studies that include LDCs, the data that is available on an annual basis is confined to central government spending only. 4 This presents a serious problem for our analysis, since a number of federal systems began in the late 1980s to shift some responsibility for social programs to state governments. We deal with this problem in a number of ways, although none are fully satisfactory. First, it should be noted that our data set extends back into the early 1970s, while the main impulses for fiscal decentralization did not occur until the 1990s. Even then, federal governments retained responsibility for pensions and many social services. To some extent, moreover, the introduction of country dummies into the model should correct for some of the specific differences among the relatively few highly decentralized federal systems such as Brazil and Argentina and the other countries in the sample. Indeed, we find no important differences in the results of our model when we exclude Brazil and Argentina, the two countries which experienced the most extensive fiscal decentralization during the late 1980s and early 1990s.
Region-wide, trends in our measures of spending present sharp contrasts during the decades of the 1980s and the 1990s. Expenditures dropped sharply in the first period — a pattern consistent with the "efficiency hypothesis." On the other hand, they rose substantially during the 1990s (ECLAC 1999, 97), a trend supportive of the "compensation hypothesis". For our purposes, however, it is important to emphasize that the rate of change varied considerably from one country to the next. Countries such as Costa Rica and Uruguay managed to maintain relatively high levels of social spending during the 1980s, for example. During the upward trend of the 1990s, annual rates of change varied from a low of minus 1.7 percent in Honduras to 22 percent in Peru, and even by the end of the decade, spending in Honduras, El Salvador, Guatemala, Nicaragua and Venezuela remained below pre-1980 highs. (ECLAC 1999, 100). Over the period covered by our data, the average annual change in spending per capita was $7.30, whereas the standard deviation was $52.39. The changes as a percentage of the budget and of GDP averaged .16 and .08 percent respectively, while the standard deviations were 4.82 and 1.31. So, there is enormous variation in the dependent variable.
Globalization: Exposure to international markets is measured in two ways. Following conventional practices in most of the literature, trade integration is calculated as imports+exports/GDP. For openness to international capital markets, we use an index of capital account liberalization developed by Morley, Machado, and Pettinato (1999) which reflects the sectoral control of foreign investment, limits on profit and interest repatriation, controls on external credits by national borrowers, and capital outflows. Given the conceptual problems and data limitations associated with alternative measures of openness such as capital flows or interest rate convergence, this is the most viable alternative. 5 A negative coefficient for changes in trade integration or capital openness would support the efficiency hypothesis. A positive sign would support the compensation hypothesis.
Democracy. Following Alvarez, Cheibub, Limongi and Przeworski (1991), we use a dichotomous measure of democracy, based on the Polity III data set of Keith and Gurr (1996). Countries were ranked as democratic when they received a score or at least six on a 10-point scale, and authoritarian if they received less than six. As noted in the preceding section, we would expect democratically-elected governments to have a positive effect on changes in welfare spending as their countries become more integrated into the international economy. 6
Electoral participation. Since electoral pressures may an important incentive for governments to deliver social services, we include voter turnout in presidential elections as a percentage of the voting age population. For non-election years, we use the figures for the preceding election, on the assumption that that will affect expectations about the coming one. Finally, although we are primarily interested in whether turnout affects the behavior of democratic governments, we also code this variable for non-democratic regimes (such as Mexico) which have also held regularly-scheduled elections to see whether reliance on electoral legitimation constrains authoritarian rulers as well. Governments elected with high turnouts, we would expect, would be more likely to increase social spending.
Left-oriented presidents. To gauge the relative balance of partisan power, finally, we have coded all democratic heads of state in terms the political orientation of their party base. Presidents are coded as left-oriented if they come from parties with close historical links with labor unions (for example, the Peronists in Argentina or Accion Democratica in Venezuela), and/or if their parties have long-standing programmatic orientations toward "the popular sector" (for example, the MNR in Bolivia, or the PLN in Costa Rica). We have also coded a number of autocratic regimes as "left-oriented," based on characterizations in the general literature. An example is the military regime that took power in Peru in 1968. As with the voting turnout variable, we have more confidence in the validity of this coding in democratic regimes. Nevertheless it is of interest to see whether "left orientations" determine behavior independently of regime type.
Control Variables. For our controls, we have attempted to track closely a variety of social and economic factors used in studies of social spending in OECD countries. These include: GDP per capita; growth in GDP; central government expenditures/GDP; "age dependency" ratios; and government debt. The last of these can be expected to constrain spending, while the others should have a positive impact. A fuller description is provided in the appendix.
As a proxy for changing labor market conditions, we use annual measures of industrial value added, as a percent of GDP. Most countries in the region experienced substantial "deindustrialization" over the course of the last several decades, in part as a result of globalization and in part as a result of the macroeconomic crises. Downward changes in industrial value-added imply the expansion of the size of the informal sector workforce, which generally remains uncovered by welfare benefits. It may also reflect a decline in the bargaining leverage of industrially-based unions (Rudra 1999). For both reasons, we would expect downward changes in industrial value-added to be directly associated with declines in social spending.
Since declines in social spending may be due to the economic crises which have hit Latin America rather than to globalization per se, we also include two "shock" variables. These are entered as dummies, based on each country's average inflation and GDP growth over the time period. Any year that inflation rises above the standard deviation is scored a 1; all others are zero. GDP shocks are recorded as 1 for any year in which growth drops one standard deviation below the average. Decade dummies are used to account for the important differences in regional and international conditions over the course of our time period. The first covers the years prior to the debt crisis, from 1973 to 1981. The second extends from 1982 to 1990, which were generally marked by economic recession and painful structural adjustments. The last covers the period of economic recovery that took place during the first half of the 1990s. 7
Finally, country dummies are included in all of the specifications of the model. Among other things, these account for fixed effects that might impact a country's economic openness and/or welfare spending over the long run: size of the population and territory, demographic characteristic, long-term political history, etc.
The Model
Our construction of the time-series model takes into account the important theoretical distinction between variables measured as levels of a given property and those measured in terms of rates of change. (Garrett, Huber and Stephens). Cross-national differences in the size of the welfare state, Garrett (1999a) argues, are likely to be invariant over time, because they are influenced by historical factors at work over long periods, or by structural conditions that change only slowly. The causes of such differences are best assessed statistically through cross-sectional analyses in which the key explanatory variables (openness, left strength, etc.) are expressed as long-term properties of the system. In this connection, we mention in passing that, in contrast to the OECD cases, cross-sectional OLS regressions show no significant relationship between openness and the size of government in Latin America. Some countries with open economies, such as Panama, do have large governments, but many other small, open Central American societies do not. These results, moreover, are unaffected by controls for GDP and democracy.
In this paper, however, we are interested in changes in social spending, which are presumably influenced more directly by short-term political pressures and by the contemporaneous and dynamic economic process of globalization. We use an error-correction model which is well-suited for just such a purpose. This is discussed at greater length in appendix 1. The generic version can be specified as: 8
DYi,t = Da + Yi,t-1 b 0 + DXi,t-1 b k + Xi,t-1 b j + Tl + e i,t
Where Yi,t is welfare expenditures in country i during year t, X is a vector of independent variables, D is a vector of country dummy variables or fixed effects, and T is a vector of time effects. Specifications of the dependent variable are measured as first-differences, and the independent variables include the lagged level of welfare expenditures, the lagged level of each independent variable, and the yearly changes (D) in the independent variables.
As Beck and Katz (1996), Huber (1998), Iverssen and Cussack (2000) and others have indicated, this type of model is based on the idea that the dependent and independent variables are in a long-run equilibrium relationship, but that there are also important short-term or temporary effects that may have a long-run impact. As noted above, the "Dvariables" on the right hand side of the equation measure annual first-difference changes. Their overall impact on spending depends on the magnitude of the regression coefficient (b k ) associated with the change variable and the extent to which the change persists over time, which in turn depends on the coefficient of the lagged dependent variable (f). In other words, if a ten percent change in trade is sustained in subsequent years, that will have a larger effect than if the change is subsequently reversed. The "levels" variables measure the long-term equilibrium relation between the independent and dependent variables; the value of each levels variable provides a mechanism for estimating its permanent effect on spending. (see appendix). The country dummies, in turn, absorb effects that are the products of long-term structural and historical factors.
As discussed in the introduction, we have also taken particular care to deal with the most common problems that affect time-series cross-sectional models. We have followed the methodology suggested by Beck and Katz (1995, 1996) whereby the use of Ordinary Least Squares with panel-corrected standard errors deals with the problem of panel heteroskedasticity and spatial correlation, and the lagged dependent variable corrects for serial correlation.
We use of a set of country dummy variables and time dummies to control for country-specific and time-specific fixed effects. The use of fixed effects is becoming the norm in panel studies of the welfare state (see Rodrik 1997 and Garret 1999b), and is particularly important in our model, since most variables vary more across units than over time. Finally, the use of panel-corrected standard errors usually produces rather conservative results, since it tends to increase the standard errors of the estimates. Moreover, the inclusion of dummy variables tends to deflate the statistical significance of the other regressors (Sayrs 1989). As we shall discuss below, this carries some risk that causal hypotheses will be rejected prematurely. On the other hand, it also increases our confidence that results which do emerge as significant are not the consequence of unsound statistical assumptions or inappropriate econometric methods. 9
III. Results for aggregate social spending
The estimates for changes in aggregate levels of spending are shown in Table 1 below. To enhance the clarity of presentation, we do not include country dummies in the table. Overall, the models explain between about 30 and 45 percent of the variance in social spending, a reasonably good fit. The r-squared is only 22 percent for central government spending, but this is not a major focus of this study.
Not surprisingly, when entered as a control variable, central government spending did have a substantial impact on social spending; as the size of the central government increases, so does welfare spending per capita and as a percent of GDP. Somewhat paradoxically, social spending as a share of the government budget is inversely related to total spending. This, however, may reflect efforts to shift priorities toward welfare as the overall size of the state is reduced. Even during the austere 1980s, for example, the fiscal share of welfare spending increased in one-half the countries in our sample. We also note that, consistent with Wagner's Law, the sign of GDP per capita is positive across all specifications of the model. Many of the other control variables do not reach 0.1 significance levels, but they generally go in the expected direction; and as we shall see in the next section, they have stronger effects when social spending is disaggregated into social security and health/education (see Tables 2 and 3).
The control for (de) industrialization is also strong and significant for welfare spending per capita and as a percent of public expenditures. The signs of the coefficient are positive because both the independent and dependent variables change in the same direction: the larger the decline in industrial value-added, the larger the decline in social spending.
Perhaps the biggest surprise with respect to the control variables is that the dummy for inflation shocks has a positive impact on social spending, an impact which becomes even stronger when we look specifically at health and education expenditures. We do not have a good explanation for this, but again, it may indicate that governments may be more inclined than expected to protect the share of social spending during periods of austerity.
When we turn to the substantive variables highlighted in the general discussion of globalization, the most striking finding is the strong, negative effect of trade openings on changes in aggregate social spending. This relationship held across all specifications of the dependent variable and was robust against a battery of controls used in all the estimates which preceded those shown in Table 1. 10
It is noteworthy that the effects of trade integration are independent of economic shocks, of (de)industrialization, and of the "conjunctural" circumstances of the "lost decade" of the 1980s and the mild recoveries of the 1990s. 11 This suggests that the main causal mechanisms through which trade affects social spending are connected to long-term shifts in the preferences and relative power of business interests exposed to international competition.
The results in the table show that for every 10 percent increase in trade and all else held constant, welfare spending could be expected to decline by over $14 per capita, and that its share of the budget and of GDP would decline by 1.2 and 0.3 percent respectively. Since trade ratios in the region increased by about half (from about 20 to 30 percent of GDP) during the 1980s and 1990s, the cumulative impact of such changes was potentially quite high. A fifty percent increase in trade implied a decline in social spending of about 70 dollars per capita, almost 27 percent of the regional average during the 1973-1997 period (see Table A1). The estimates for the decline as share of social spending in government budgets and in the GDP would be about 6 percent and 1.5 percent respectively.
The opening to capital markets, taken by itself, tended to have a positive relation with social spending — a finding also reported by Quinn (1997) for OECD countries, and in Garrett's (1999a) study of global spending. When interacted with trade openings, however, the effects are negative and significant for two of the three measures of social spending. As Garrett found in his cross-sectional global study of government spending, capital mobility tended to compound the negative effects that trade openings had on changes in social spending.
In Figure 1, we present the findings of a counter-factual analysis which parallels the one used by Garrett (1999a), with welfare/capita in place of aggregate government spending. The results are virtually identical. With all other variables held at their means, per capita social spending is estimated to decline by $25 in conditions of high and sustained trade openings and high capital mobility, but by only about $1 when capital mobility is low. As we shall see, this finding is modified somewhat by the analysis of the disaggregated measures of spending. Nevertheless, as a first pass, it provides substantial support for the "efficiency hypothesis."
Note as well that the main effects come from trade openings, rather than capital mobility. Differences in spending are substantial as we move from across the rows from low to high trade, while the extent of capital mobility has scarcely any impact when trade openings are limited. One implication is that the main pressure for reducing expenditures comes from producer interests exposed to competition, rather than from a more general concern with establishing "credibility" in international financial markets. For producers, increases in welfare expenses imply higher payroll taxes, which directly impacts their bottom line. They thus have an incentive to lobby directly against expanding welfare commitments. (Frieden 1991) Liquid asset holders worry more about aggregate macroeconomic "fundamentals" than about welfare expenditures or labor costs per se. And although they might become concerned if fiscal or current accounts deficits grow too large, they may also stand to profit from high interest rates.
The "null findings" for the other substantive variables are also of considerable interest. Against much of the literature on the OECD countries, we find no support for the proposition that "left-oriented" governments have an impact on social spending, however that variable is specified. In Latin America, of course, it is far more difficult to locate governments on a "left-right" dimension, so this finding might simply reflect a measurement problem. On the other hand, more substantive factors might also be at work. The organized labor base of these governments is typically far weaker and more dependent than is the case in the developed countries, and the economic room for maneuver is much narrower. In contrast to the parliamentary regimes of Europe, moreover, the ability of left-oriented governments to provide compensatory social assistance in Latin America may be constrained by factors typical of presidential regimes such as divided government or party indiscipline. (Linz 1994, Tsebelis 1995).
Democratic regimes also have very little predictable impact on social spending. Surprisingly, the coefficients for democracy in Table 1 are negative, and were actually significant in several preliminary estimates of the model. We also found no stable interaction effects between democracy and any of the other independent variables used in the model. In some estimates, democracies did appear more likely than autocracies to protect social spending during periods of inflation; and in some, they were less constrained by growth in GDP per capita, a finding consistent with Brown and Hunter's (1999). Unlike our findings for trade, however, these effects were highly sensitive to the other variables included in the model, and were not robust across alternative specifications of social spending. In contrast to Brown and Hunter, moreover, we found no robust interaction effects between democracy and age dependency, once globalization variables were entered into the model.
The limited effects of regime type may be due to the fact that transitions were relatively recent and democratic institutions were not fully consolidated. Whatever the reason, however, we find no sharp or consistent differences in the way democracies and autocracies deal with the constraints that trade integration places on welfare spending.
Of all of the "political variables" we examine in this study, electoral turnout comes closest to producing the expected impacts on aggregate social spending. The effects of the "levels" measure on spending per capita, and of Dturnout on its share of the budget are both positive and significant. The causal inferences drawn from this relationship, however, would require careful interpretation. In the large majority of the cases, turnout was measured in elections held under democratic regimes, but our measure also includes turnout in authoritarian regimes, such as Mexico or Brazil, where semi-competitive elections were also an important source of legitimation. Thus, the effects of participation are attributable less to regime type per se, than to the incentive which rulers have to exchange social spending for popular support or acquiescence.
IV. Welfare spending disaggregated: pensions, and health&education
In this section, we examine effects of the globalization and political variables on social security payments and on combined expenditures on health and education. Although these measures have typically been combined in analytical overviews of social spending in Latin America (Eclac 1999), there are reasons to expect that they are influenced by different political logics.
On the one hand, there are several reasons why social security expenditures might be most susceptible to the "efficiency" pressures of trade integration. First, such expenditures presumably have the most direct and transparent impact on the costs of labor; although the largest pension systems had built up huge deficits which had to be covered from public revenues (Mesa-Lago1989), payroll taxes still constituted an important component of the financing. In the second place, the restricted scope and inequities in pension coverage limited the number of stakeholders in the system. Thus, with the exception of some of the very large systems such in Uruguay or Costa Rica, cutbacks are less likely to generate wide popular protest than they have been in many European countries (see Pierson 1996, Esping-Andersen 1996) for discussion of the European cases).
Conceivably, the political constraints and opportunities are different in the case of health and education (H&E) expenditures. Although health insurance is also sometimes a component of the wage bill, these expenditures generally have a smaller direct impact on labor costs; indeed, from the point of view of employers, they may have more substantial payoffs as "human capital" investments. There is also a greater likelihood of strong political opposition to cutbacks in these areas. Teachers and health-worker unions are typically among the strongest and most militant in Latin America. And despite the severe inadequacy of social service delivery systems, such cutbacks are also likely to be felt by relatively broad sectors of the population.
In short, as Latin American economies become more integrated into global markets, incumbent governments may face stronger political incentives to protect H&E expenditures than those for social security. In fact, the simple correlation (Pearson's r= -0.52) between these measures as a percent of the budget does imply a rather sharp tradeoff. Thus, particularly in an era of "hard budget constraints," governments appear to be under considerable pressure to establish priorities.
Tables 2 and 3 show how expenditures in social security and H&E are affected by the variables used in the general model. The estimates are generally very consistent with our expectations. Let us look first at the control variables. The effects of most of these are about the same as in the general model, but there are some interesting exceptions. First, in the estimates for social security (Table 2), the effects of age dependency show up as much more uniformly and strongly positive. It is unclear whether this effect is due to non-discretionary entitlement spending, or to political pressure from pensioners (Brown and Hunter 1999), but the finding goes very much in the expected direction. Somewhat more surprising is that in Table 3, H&E spending is negatively related to age dependency. As suggested above, this may reflect the tradeoffs which governments encounter as they seek to contain the overall growth in government spending.
The decline in industry also has a very different effect on the two types of spending. The effects of deindustrialization are very strong on all three measures of social security expenditures. As noted, this may reflect declining coverage as more workers move into the informal sector, and/or a weakening in the leverage of industrial unions. Conversely, changes in industrial value-added have no significant effect on H&E expenditures, conceivably because the beneficiaries depend less on formal-sector employment.
There are also interesting differences with respect to inflation shocks. Table 3 shows that the positive relation found in the general model works almost entirely through the effects on H&E. This indicates that governments act to protect those areas of social spending most subject to popular political pressures. Again, however, there are no consistently significant interaction effects between shocks and regime types. Democracies are not consistently more likely than autocracies to protect health and education expenditures during periods of very high inflation.
For our purposes, however, the most interesting findings are the differences in the effects of the globalization and political variables on the two types of social spending. As noted, these work very much in the direction suggested above. Table 2 shows that the impact of the trade variable which we saw in the general model works primarily through its effects on pensions and transfers. Trade openings have a uniformly negative effect on social security spending, and this effect is compounded by capital account liberalization.
The democracy coefficients are not significant; but in contrast to what we see in Table 3, the signs are consistently negative. While we cannot draw firm conclusions, this might reflect "Bismarkian" forms of cooptation on the part of autocratic regimes. Highly stratified distributions of pension benefits constitute a source of rents that can be used by authoritarian rulers to consolidate their bases of support. In Brazil, for example the most extensive expansions of the social security system in fact occurred under the Vargas dictatorship and the post-1964 military regimes. (Malloy 1979)
In the models for health and education expenditures (Table 3), we see a very different picture. First, trade liberalization has a negative impact on only one specification of spending, H&E per capita, and the only interaction term that approaches significance goes counter to the efficiency hypothesis. Interestingly, changes in capital account liberalization are positively related to all three measures of H&E spending, which provides at least modest support for the "compensation" argument.
In the second place, whereas electoral participation does not affect pension spending, it is positive in all three specifications of H&E spending, and is highly significant in two of them. Without further research, we cannot say whether this is actually the result of generalized popular pressure, or whether it somehow reflects the power of the strong unions in the health and education sectors. The results do, however, provide strong reinforcement for our hunch that the findings already reported in Table 1 did not occur by chance. With respect to health and education, the impact of domestic political pressures appear to offset the race-to-the-bottom effects of globalization.
V. Conclusions
Like most statistical studies, our findings leave open a variety of questions, many of which can only be answered by more qualitative research. In the first section and at various points in the rest of the paper, we have discussed a number of explanations for the relations we have found between globalization, political pressures, and welfare expenditures. In many instances, however, we cannot be sure about which of a number of causal mechanisms actually affect outcomes. This puzzle has just been raised with respect to the relation between participation and H&E spending, but it extends to other relations as well. This is perhaps most important with respect to the impact of trade openings on pension spending. To what extent do the constraints on spending reflect the pressure of producer interests, and to what extent might it be due to indifference or even opposition from numerous sectors of the population that receive no significant benefits from the pension system?
Causal puzzles are also raised by some of the null or unexpected findings in the study. The apparently positive effects of capital account liberalization in at least some of the models has no very convincing explanation. And we cannot tell whether the ambiguous findings with respect to democracy are due to incomplete consolidation, defects in electoral systems or constitutional design, or to the Bismarkian strategies of authoritarian regimes.
To answer such questions we need closer analysis of the organization of social service systems, who has benefited from them in the past, and who stands to gain or lose from changes. As noted briefly in the introduction, some of the most vexing challenges of "second phase" welfare reforms have less to do with the amount of financing, than with the way financing is allocated and with how delivery systems are organized. To understand how politics affects the outcome of such challenges requires a close examination of bargaining and decision-making processes, research that is often most usefully conducted through case studies and small-N comparisons.
If our analysis cannot provide conclusive answers to such questions, however, it does provide a frame of reference that might orient future research. Three sets of conclusions are of particular importance. The first concerns the contending arguments about the effects of trade integration on the welfare state: the overwhelming weight of evidence favors the efficiency over the compensation hypothesis. Even if the precise causal mechanisms are unclear, we can infer from our findings that trade integration does change power resources in ways that lead to a reduction in pensions and other transfers, the components of social spending that provide the most direct protections from vulnerability to market forces. Conversely, even in the case of health and education, there is no evidence that expansion of trade encourages states to enlarge the size of their welfare commitments. Indeed, if anything, trade openings have a negative impact on these components of social spending as well.
Integration into global capital markets has a more ambiguous effect. On the one hand, it does appear to encourage increases (or discourage decreases) in spending on health and education, possibly as a way to upgrade the quality of the labor force available to foreign investors. But capital account liberalization also compounds the negative effects of trade openings on social security expenditures. Presumably this is because it increases the leverage of traded goods producers with interests in containing the cost of labor. As their economies become more closely linked to capital markets, they can make more credible threats to liquidate their assets and shift them elsewhere.
A second conclusion is that it is important to distinguish among different types of social spending. The distinctions used in this study follow those conventionally used in studies of the U.N. Economic Commission on Latin America and the Caribbean (1999). Social security transfers are, as just noted, most relevant to the efficiency and compensation hypotheses, since they both add to the wage bill and offer the most direct protection against market forces. Health and education expenditures arguably involve longer term investments in "human capital" and are likely to have a greater long-term impact on the distribution of income.
The point that is clearly indicated in this study, however, is that these categories of social spending are influenced by very different sets of political and economic factors. The good news is that the pronounced constraints that globalization appears to place on social security transfers do not extend to spending on health and education. Possibly because the health and education sectors encompass a wider set of stakeholders, decisions on spending in these areas appear to reflect a very different political logic, much more connected to electoral competition and political participation. Although qualitative case studies of social sector reforms are more concerned with restructuring than with expenditures, it should be noted that they also consistently show distinctions between the politics of pension reform and the politically more difficult challenges related to the restructuring of social service sectors (Nelson 1998, 2000).
We have not attempted to explore the politics of social sector reform in this paper. To maintain the clarity of our presentation, moreover, we have also left for later efforts to analyze differences between expenditures in health and those in education. Given our findings so far, however, an exploration of these questions constitutes a logical next step for further research.
Our third set of conclusions returns to the question of how democratic regimes in Latin America have affected social spending. There was some fragmentary evidence that democracies might have "mattered" (for example, during inflation shocks), but the findings were far too unstable to reject the "null hypothesis" that democracy had no effect. As noted, moreover, many other estimates for democracy were negative and, at times, significant.
It is quite possible, of course, that these results are partly due to measurement error. But when viewed in conjunction with the more robust findings on electoral turnout, they also suggest a more substantive point. On the one hand, we cannot assume that transitions to democracy, in and of themselves, will provide sufficient conditions for dealing with the welfare challenges of globalization. This does not mean that democratic politics are unimportant. What is likely to count most, however, at least in the areas of health and education, is the extent to which citizens enter the political process, and the mechanisms that enable them to hold governments accountable. To assess the chances for the construction of viable welfare states, we will have to pay careful attention to the organization of civil society and to the design of electoral and representative institutions.
Technical Appendix
Model Description
The error correction model is given by:
DYi,t = a + DXi,t-1 b k + f(Yi,t - Xi,t-1 U) + e i,t (1)
Where, in our case, Yi,t is welfare expenditures in country i during year t, D is the first differences operator, X is a vector of independent variables and e i,t is a white noise error term. The model describes a short-term equilibrium relationship given by DYi,t = a + DXi,t-1 b k + e i,t and a term f (Yi,t-1 - Xi,t-1 U), which measures the deviation from this short-term equilibrium relationship. [Note that e i,t-1 = (Yi,t-1 - Xi,t-1 U)]. Equation 1 shows that, first, a one-off change in Xi,t-1 produces a contemporary change in Yi,t. This short-term effect is determined by the k-dimensional vector of regressors b k. Furthermore, when the impact of Xi,t-1 on Yi,t throws the model off its long run equilibrium (given by the cointegrating vector Y*i,t-1 = X*i,t-1 U, where the "*" indicates equilibrium)), the discrepancy or "error" (Yi,t-1 - Xi,t-1 U) is corrected at a yearly rate of f.
One way to show in a more intuitive way how to interpret the different short term and long term coefficients is to transform equation (1) through a simple mathematical operation:
Let b j be defined as -(f U), where both parameters f and U come from equation 1, then it follows that U = bj /-f. Equation 1 can therefore be rewritten as:
DYi,t = a + Yi,t-1 f + DXi,t-1 b k + Xi,t-1 b j + e i,t (2)Equation 2 is then estimated through OLS. The interpretation of the coefficients is then as follows. The regression coefficient for an independent level variable is a measure of the long run equilibrium relationship between a vector of cointegrated independent variables (i.e. sharing the same long run trend) and the dependent variable. As noted above, the long run equilibrium relationship is given by Y*i,t-1 = X*i,t-1 U. The parameter font face="Symbol">U (which measures this long run equilibrium relationship) is not directly observable from equation (2), but can be found by dividing b j by -f (see above).
On the other hand, the importance of the short term effects DXi,t-1 depends on the size of b k and on how long the effects of changes in Xi,t-1 persist through time. A one-off change in Xi,t-1 produces an immediate (contemporary) change in Yi,t that is measured by b k. If at time t there is a change in Xi,t in the opposite direction to the change in Xi,t-1, then there are no more effects. But if the change in Xi,t-1 is sustained, then the impact will continue in subsequent periods and can be measured by DXi,t-1(1+f)t, where t is the number of periods after the initial change. Thus, for example, 3 years after the initial change DXi,t-1, the effect will be DXi,t-1(1+f)3. Since 0 < f < -1, the smaller the value of f, the longer the sustained changes in X will persist through time.
Even though error correction models are quite popular in a wide variety of situations, they are only internally consistent if the variables are cointegrated [i.e. they share a common trend]. [Greene 2000:793, Banerjee et al 1993].
In our specific case, based on the Dickey Fuller test (in its three versions: simple, with a trend variable included, and the augmented version), each measure of welfare effort (our dependent variable) has a unit root (i.e. shows nonstationarity). In addition, most of our economic variables (i.e GDP per capita, the age dependency ratio, trade, capital mobility and the value added of industry) have a unit root. 12 Thus, regressions in "levels" may be affected by spurious associations. Since all these variables are integrated of order 1 [I(1)] [i.e. after taking first differences each series becomes stationary], some researchers would prefer to use regressions in first differences. Although this is a possible solution, it abandons all efforts to model the long run dynamics and, it is not necessary when the variables are cointegrated. In other words, if the variables are in a long run equilibrium relationship [they move together along a long-term trend due to an underlying causal process], it is possible to use a model that uses both level and change variables to capture short and long run dynamics. Results from Dickey Fuller tests 13 suggest that the errors in all our regressions are stationary. We may therefore conclude that the independent variables are co-integrated and consequently that the error correction model is appropriate. 14
Interpretation of Results
(A) Long Term Relationships
The coefficient from the level variables indicates that two variables vary along a common trend line and that this variation can be shown to be nonrandom. The interpretation of the long term effects should take the form of a "one unit increase produces a change in the long run target value of the dependent variable". As noted above, the vector of cointegrated variables is given by X, and the long term effects are given by the vector b j. In particular, b j = {b1, b2 ,b3, b4, b5, b6}.
WELFAREi,t= b0 + b1GDPCAPi,t + b2AGEi,t + b3PUBGDPi,t + b4TRADEi,t + b5CAPITALi,t + b6TURNOUTi,t
Where the subscript i indicates the country and the subscript t the year of the observation.
Each b measures the long run equilibrium relationship between each independent variable and welfare expenditures. However, since we have estimated equation 2 (see above), each b is not directly observable. To "unveil" the true value of each b we need to divide it by -f (see explanation above). For example, if our purpose is to understand how countries that have larger public sectors, all else being equal, also have in the long run higher levels of social spending per capita, we would need to divide b3 (3.3566) by -f(0.2396) . This yields a long run relationship between public spending and social spending per capita of 3.3566/0.2396 = 14.009. (see Table 2). Hence, if a country like Bolivia (with an average level of public spending as a percentage of GDP of about 17%) were to increase the size of the public sector to reach the levels of Chile (with an average of about 27%), this would raise the long term predicted value of social spending by about $140 per capita.
[4] Short term effects. When trying to evaluate short-term effects, we need to separate two situations. One in which the increase is sustained and one in which it is not. If it is a sustained change, then it will have an effect in the "medium" run, a period of 5-7 years see figure 2 below). The length of the effect depends on the parameter of the lagged dependent variable f. In other words, an important characteristic of error correction models is that the effects of a sustained change in one of the independent variables depends not only on the magnitude of the effect (which is captured by the regression coefficient of the corresponding change variable) but also on how long it persists (which depends on the coefficient of the lagged dependent variable). An example may be useful to illustrate these effects. For instance, a one standard deviation increase in trade (i.e. an increase of 35% percent in the ratio of imports plus exports to GDP) would situate Argentina, which has an average ratio of 15.3%, at the level of Chile, which has an average ratio of 52.2. The first year, this increase would reduce welfare spending per capita by $50, welfare spending as percentage of public spending would go down 4.2 percent points, and welfare spending as a percentage of GDP would be reduced by 1.25 points. Furthermore, this effect would continue over subsequent years and would decay at a rate of (1 + f), where f is the parameter of the lagged dependent variable. Thus, if, in order to keep the discussion simple, we look at welfare spending per capita only, the second year there would be a reduction of 0.77*50= 38.50; the third year the reduction would be 0.772*50= 29.26, the fourth year 0.773*50=22.8 and so forth.
Figure 2 below simulates the effects of a one standard deviation increase on welfare expenditures (i.e. an increase of 35%). It can be seen that the effect dies out slowly, and that the cumulative impact is remarkable (e.g. after five years the cumulative effect is about $172 dollars). To be sure, two possible arguments can be made against this simulation. The first is that a 35% increase in trade is not realistic. The second is that, even if such increase was to occur, it would be temporary and would therefore be offset by reductions to bring back the economy to its "natural" rate of openness. A closer examination of the evidence shows, however, that there are indeed in our sample, a significant number of cases in which trade as a percentage of GDP grew from one year to the next by about 30% or more (Ecuador 1974, Costa Rica 1981, Dominican Republic 1984 and 1993, etc.). Furthermore, some of these yearly increases were followed by subsequent decreases that would partially reduce the trend (eg., Costa Rica reduced trade by 27% between 1982-1985 following an increase of 28% in 1981). But many other big increases were not followed by subsequent similar reductions (e.g. trade grew by 39.81% in the Dominican Republic in 1993 and between 1994 and 1997 only 5% of this increase was reduced).
Figure 2: Yearly Reductions in Welfare Spending Per Capita if Argentina Raised and Sustained Trade Openness to the Average Level of Chile.
TABLE A1: Three Measures of Welfare Effort and Public Spending in Latin America (Average 1973-1997).
| Countries | WELFCAP | WELFPUB | WELFGDP | PUBGDP |
| ARGENTINA | 539.15 [125.69] |
47.85 [9.45] |
6.86 [1.23] |
14.68 [2.79] |
| BOLIVIA | 56.30 [19.33] |
38.17 [19.33] |
6.20 [2.15] |
16.72 [5.82] |
| BRAZIL | 395.01 [89.99] |
42.38 [7.00] |
9.93 [1.72] |
24.39 [7.23] |
| CHILE | 392.32 [89.13] |
54.31 [7.17] |
14.61 [2.78] |
27.24 [5.33] |
| COSTA RICA | 320.24 [52.43] |
59.14 [3.62] |
13.45 [1.56] |
22.82 [2.90] |
| D. REPUBLIC | 58.38 [10.71] |
28.37 [3.91] |
4.28 [0.69] |
15.19 [2.15] |
| EL SALVADOR | 66.57 [19.41] |
30.03 [4.29] |
4.24 [0.90] |
14.16 [2.62] |
| MEXICO | 200.71 [37.03] |
36.96 [12.10] |
6.40 [0.96] |
18.86 [5.38] |
| URUGUAY | 839.90 [307.47] |
63.32 [7.73] |
16.91 [4.22] |
26.43 [3.83] |
| ECUADOR | 69.89 [13.68] |
35.25 [3.13] |
4.80 [0.72] |
13.93 [1.61] |
| GUATEMALA | 42.03 [9.03] |
26.81 [5.66] |
2.95 [0.61] |
10.59 [1.44] |
| PARAGUAY | 67.24 [19.84] |
37.68 [6.57] |
4.03 [0.90] |
10.59 [1.41] |
| PERU | 116.53 [20.80] |
24.08 [3.08] |
4.30 [0.65] |
16.70 [2.10] |
| VENEZUELA | 303.0 [38.28] |
32.56 [3.01] |
7.65 [0.85] |
22.27 [3.62] |
| TOTAL | 263.16 [249.16] |
41.52 [13.42] |
8.39 [4.94] |
19.05 [7.74] |
Note: Standard deviations in parentheses.
Table A2: Individual Country Regressions of Changes in Social Spending per Capita in 14 Latin American Countries 1973-1997
| ARG | BOL | BRA | CHIL | CR | DR | ELS | MEX | URU | ECU | GUA | PAR | PER | VEN | |
| DGDP | 2.89 | 1.28*** | .72 | 1.33 | 3.12 | .84 | .61*** | 1.37*** | 6.92* | .75* | -.422 | .59*** | .79 | 2.73 |
| DTRADE | -6.86 | -1.34*** | -9.1** | -2.74*** | -1.97*** | .001 | .16 | -2.62*** | -6.06*** | -.40 | 1.01 | -.77*** | -2.3*** | - 3.76*** |
| DCAPITAL | -21.5 | 17.07 | 124.7 | 119.6*** | 2.16 | -12.3 | .035 | 301*** | 1488*** | 88.7 | -13.9 | -32.8 | -153 | 1394 |
| DEMOCRACY | 45.51 | 3.85 | 7.48 | 10.69 | Dropped | 8.14 | .90 | dropped | 57.12 | 3.72 | dropped | 8.32 | 10.4 | dropped |
| CONSTANT | -27.37 | -1.94 | 9.76 | 4.13 | -2.84 | -8.8 | -1.2 | 2.09 | -16.6 | -6.6 | 1.38 | -.407 | -9.9 | 4.28 |
| R-square | .146 | .458 | .294 | .5157 | .28 | .165 | .125 | .6454 | .2433 | .153 | .1629 | .69 | .567 | .354 |
| F | .67 | 12.56 | 4.79 | 34.01 | 5.58 | .89 | 0.79 | .19.41 | 9.35 | 1.75 | 1.21 | 13.4 | 3.87 | 2.02 |
| Sig F | .618 | .0001 | .0099 | .0000 | .0064 | .491 | .549 | .0000 | .0002 | .204 | .3475 | .0002 | .049 | .177 |
| Durbin Watson | 2.03 | 1.98 | 2.18 | 1.61 | 2.84 | 1.64 | 2.16 | 1.92 | 2.58 | 2.42 | 2.40 | 1.92 | 2.06 | 1.70 |
| N | 24 | 20 | 21 | 24 | 23 | 24 | 23 | 24 | 24 | 17 | 16 | 18 | 13 | 13 |
Notes: (*) = p < .15 ; (**) =.05 < p < .1; (***) = p < .05. "dropped" indicates that the computer program (STATA) automatically dropped the variable due to insufficient variation. All the variables except democracy express yearly rates of change. In addition, DGDP has been lagged one year. Country codes are as follows: ARG= ARGENTINA; BOL= BOLIVIA; BRA=BRAZIL; CHIL=CHILE; CR= COSTA RICA; DR= DOMINICAN REPUBLIC; ELS= EL SALVADOR; MEX= MEXICO; PAN= PANAMA; URU=URUGUAY; ECU=ECUADOR; GUA=GUATEMALA; NIC=NICARAGUA; PAR=PARAGUAY; PER=PERU; VEN=VENEZUELA.
Table A3: Individual Country Regressions of Changes in Social Spending as a Percentage of Public Spending in 14 Latin American Countries 1973-1997.
| ARG | BOL | BRA | CHIL | CR | DR | ELS | MEX | URU | ECU | GUA | PAR | PER | VEN | |
| DGDP | .001 | .008*** | -.0008 | .0003 | -.0009 | .003 | -.0002 | -.002 | .0009 | .0004 | -.006 | .-.00001 | -.0004 | -.0024 |
| DTRADE | -.010*** | -.008 | -.0024 | -.0011 | .0002 | -.0009* | .0014 | -.004*** | -.001* | -.0006 | .002 | -.002** | -.0018 | -.002** |
| DOPENNESS | -.030 | .594* | .34 | .058* | .341*** | .0917 | -.067*** | .454*** | .30*** | .026 | -.019 | .156 | -.042 | -1.69 |
| DEMOCRACY | .004 | .0002 | -.008 | -.006 | dropped | .022 | .0178 | Dropped | .0067 | .005 | dropped | .017 | .015 | dropped |
| CONSTANT | .005 | -.027 | .001 | .014 | .0007 | -.034 | -.010* | .012 | -.003 | -.006 | .015 | .006 | -.008 | -.011 |
| R-square | .303 | .364 | .056 | .052 | .238 | .189 | .246 | .306 | .192 | .169 | .194 | .3367 | .3558 | .5856 |
| F | 2.54 | 5.7 | .061 | 2.65 | 2.06 | 1.83 | 4.41 | 5.59 | 7.61 | 0.89 | 3.27 | 4.49 | 1.27 | 7.82 |
| Sig F | .07 | .005 | .6582 | .064 | .1398 | .164 | .011 | .006 | .0008 | .497 | .058 | .017 | .355 | .0071 |
| Durbin Watson | 1.51 | 2.26 | 2.29 | 1.83 | 2.5 | 2.19 | 1.92 | 1.36 | 2.66 | 2.60 | 2.98 | 1.48 | 2.14 | 2.28 |
| N | 24 | 20 | 21 | 24 | 23 | 24 | 23 | 24 | 24 | 17 | 16 | 18 | 13 | 13 |
Table A4: Individual Country Regressions of Changes in Social Spending as a Percentage of GDP in 14 Latin American Countries 1973-1997.
| ARG | BOL | BRA | CHIL | CR | DR | ELS | MEX | URU | ECU | GUA | PAR | PER | VEN | |
| D GDP | .0003 | .001** | .0001 | .0003 | .0006 | .0004 | .0001 | .0003*** | .0009 | .0006 | -.0004 | .0001 | -.0004 | .0004 |
| DTRADE | -.0005 | -.0012*** | -.0048 | -.0009 | -.0007*** | .00002 | .00001 | -.0004*** | -.001* | -.0003 | .0005 | -.0004*** | -.0018 | -.001*** |
| DOPENNESS | .0012 | .0218 | -.52 | .027 | .0099 | -.0068 | .0015 | .0814*** | .303*** | .026 | -.0005 | -.024 | -.042 | .283 |
| DEMOCRACY | .0046 | .0039 | .022 | -.002 | dropped | .006* | -.0006 | Dropped | .006 | .005 | dropped | .004 | .015 | dropped |
| CONSTANT | -.003 | -.0015 | -.001 | -.0004 | .00029 | -.007 | -.0002 | -.0004 | -.003 | -.006 | .0013 | .0006 | -.008 | .0019 |
| R-square | .1032 | .36 | .09 | .136 | .218 | .188 | .047 | .5332 | .192 | .169 | .147 | .6427 | .422 | .458 |
| F | .58 | 9.89 | .77 | 15.79 | 5.61 | 1.02 | .37 | 14.24 | 7.61 | .89 | 1.76 | 9.53 | 1.27 | 2.17 |
| Sig F | .682 | .0004 | .558 | .0000 | .0063 | .4202 | .824 | .0000 | .0008 | .497 | .208 | .0008 | .355 | .1614 |
| Durbin Watson | 2.32 | 1.91 | 1.94 | 1.64 | 2.73 | 1.92 | 1.45 | 1.54 | 2.66 | 2.6 | 2.4 | 1.99 | 2.14 | 1.73 |
| N | 24 | 20 | 21 | 24 | 23 | 24 | 23 | 24 | 24 | 17 | 16 | 18 | 13 | 13 |
Table A5: Sign and Statistical Significance of the Main Theoretical Variables in Fourteen Individual Country Regressions.
| WELFCAP | WELFPUB | WELFGPD | |
|
DGDP Positive and Significant Positive but not Significant Negative and Significant Negative and Not significant |
6 7 0 1 |
1 5 0 8 |
2 10 0 2 |
|
DTRADE Positive and Significant Positive but not Significant Negative and Significant Negative and Not significant |
0 3 9 2 |
0 3 6 5 |
0 3 6 5 |
|
DCAPITAL Positive and Significant Positive but not Significant Negative and Significant Negative and Not significant |
3 6 0 5 |
5 4 1 4 |
2 7 0 5 |
| DEMOCRACY Positive and Significant Positive but not Significant Negative and Significant Negative and Not significant |
0 10 0 0 |
0 8 0 2 |
1 7 0 2 |
Source:Tables A2-A5.
Table A.6: Codings for Left-oriented Presidents
| Country | Presidents | Period |
| Argentina | Isabel M. de Perón (Peronist Party) Carlos S. Menem (Peronist Party) |
1974-1975 1990-1997 |
| Bolivia | Siles Suazo (MNR) Paz Estensoro (MNR) Paz Zamora (MNR) Gonzalo Sánchez (MNR) |
1983-1985 1985-1989 1989-1994 1995-1997 |
| Brazil | Fernando H. Cardoso (PSDB) | 1994 |
| Chile | Salvador Allende (Socialist Party of Chile) | 1973 1990-1997 |
| Costa Rica | Daniel Oduber Quirós (PLN) Luis Alberto Monge (PLN) Óscar Arias (PLN) |
1973-1977 1982-1989 1994-1997 |
| Dominican Republic | Antonio Guzmán Fernández (PRD) Salvador Jorge Blanco (PRD) Peña Gómez (PRD) |
1979-1982 1983-1986 1997 |
| Mexico | Echeverría (PRI) | 1973-1976 |
| Uruguay | Sanguinetti (Colorado Party) Sanguinetti(Colorado Party) |
1985-1989 1995-1997 |
| Ecuador | Rodriguez Lara/ (*) Poveda/Duran/Franco(**) Rodrigo Borja (Democratic Left) |
1973-1976 1976-1979 1989-1990 |
| Peru | Alan García (APRA) | 1985-1986 |
| Venezuela | Carlos Andres Perez (AD) Lusinchi (AD) |
1974-1978 1984-1988 |
Note: Only the years in which the dependent variable (welfare expenditures) was available have been coded. If a left-oriented president takes office between January and March, that year is coded as "left". If the president, however, takes office between July and December, the year is coded "not left".
(*) Left-oriented military government; (**) left-oriented military junta.
Description of Variables
| NAME | DESCRIPTION, MEASUREMENTAND SOURCE |
| WELFCAP 15 | Welfare expenditures percapita. Welfare expenditures include public expenditures in health care,education and social security programmes. Measured in 1995 constant USdollars. Source: Government FinanceStatistics (IMF) various issues. |
| WELFPUB | Welfare expenditures as apercentage of central government spending. |
| WELFGDP | Welfare expenditures as apercentage of GDP |
| PUBGDP(Public Spending/GDP) | Central governmentspending as percentage of GDP |
| GDPCAP | GDP per capita in 1995 constant US dollars.Source: WDI |
| DGDP | Annual real growth of GDPSource: World DevelopmentIndicators, 2000 (WDI) |
| DEPENDENTS | Age Dependency Ratio.Measures the number of dependents over the working age population. The agedependency ratio is calculated as the ratio of dependents — the populationunder age 15 and above age 65‹ to the working age population — those aged15-64. For example, 7 means that there are 7 dependents for every 10 workingage people. Source: WDI |
| DEBT | Debt Service as a Percentageof Central Government Current Revenue. Public and publicly guaranteed debtservice (PPG) is the sum of principal repayments and interest actually paidon long-term obligations of public debtors and long-term private obligationsguaranteed by a public entity. Source: WDI |
| GDPSHOCK | A dummy variable coded 1in those years in which economic growth was one standard deviation below themean, and zero for the rest of the years. |
| CPISHOCK | A dummy variable coded 1in those years in which inflation was one standard deviation above the mean,and zero for the rest of the years. |
| TRADE | Imports plus exports as apercentage of GDP.Source: WDI |
| CAPITAL | Measures the degree ofliberalization or freedom from government intervention or distortion tocapital mobility. The values have been normalized to be between zero and one,with one being perfectly free from distortion (no legal restrictions to theflow of capital). We have multiplied the index by 100 to facilitateinterpretation in terms of percentages. Source: Morley, Samuel;Machado, Roberto; and Pettinato Stefano. 1999. "Indexes of Structural Reformin Latin America". SerieReformas Económicas #12. Santiagode Chile: Economic Commission for Latin America and the Caribbean. |
| DEMOCRACY | Dummy variable with avalue of 1 in democratic years and zero in nondemocratic years. It has been constructed from the variable"DEMOC" in Keith and Gurr's dataset. DEMOC provides a democracy score on a0-10 scale (0=low democracy; 10=high democracy). If DEMOC<6, DEMOCRACY=0;if DEMOC=6 or greater, DEMOCRACY=1. The values for 1995, 1996, and 1997 comefrom an update of this study called Polity98. Source: Jaggers, Keith and Gurr, Ted Robert. 1996. Polity III: Regime Type and PoliticalAuthority, 1800-1994. Computer File. Ann Arbor, MI: Inter-UniversityConsortium for Political and Social Research. Codings are based on annualobservations from 1800 to 1994 of 20 historical countries and 157contemporary countries encompassing all independent members of theinternational system with populations of greater than 500,000 in the early1990s. The values for 1995, 1996, and 1997 come from an update of this studycalled Polity98. |
| LEFT | Dummy variable coded 1for years in which a left-oriented president was in office and zerootherwise.Source: See table on p. 45 |
| VOTER TURNOUT | Voter turnout incongressional elections as a percentage of the voting age population. Codingis as follows. If there have not been congressional elections after a periodif six years, then each year between elections is coded as zero (unless thereare presidential elections, in which case the value of the presidentialelection is used instead of the congressional one.Source: InternationalInstitute for Democracy and Electoral Assistance (IDEA) |
| INDUSTRY | Value added of industryas a percentage of GDP Source: WDI. |
Bibliography
Alvarez, Michael, Jose Antonio Cheibub, Fernando Limongi and Adam Przeworski. 1996. "Classifying Political Regimes". Studies in Comparative International Development 31:3-36
Banerjee, Anindya., Dolado, Juan., Galbraith, John and David Henry. 1993. Co-Integration, Error Correction, and the Econometric Analysis of Non-stationary Data. Oxford: Oxford University Press.
Boix, Carles. 1999. "Democracy, Development and the Public Sector". University of Chicago, unpublished paper.
Beck, Nathaniel. 1991. "Comparing Dynamic Specifications: The Case of Presidential Approval". Political Analysis 3: 51-87.
Beck, Nathaniel. 1992. "The Methodology of Cointegration". Political Analysis 4: 237-254.
Beck, Nathaniel and Jonathan Katz. 1995. "What to do (and not to do) with Time-Series Cross-Section Data". American Political Science Review 89 (3): 634-647.
Beck, Nathaniel and Jonathan Katz. 1996. "Nuisance versus Substance: Specifying and Estimating Time-Series Cross-Section Models". Political Analysis 6 (July): 1-37.
Brown, David and Wendy Hunter. 1999. "Democracy and Social Spending in Latin America, 1980-92". American Political Science Review 93 (4): 779-790.
Cameron. David. 1978. "The Expansion of the Public Economy: A Comparative Analysis". American Political Science Review 72: 1243-1261
Economic and Social Commission on Latin America and the Caribbean (ECLAC) 1999. Social Panorama of Latin America
Esping-Andersen, Gosta. 1991. The Three Worlds of Welfare Capitalism. Princeton, NJ: Princeton University Press.
Esping-Andersen, Gosta (ed.) 1996. Welfare States in Transition: National Adaptations in Global Economies. London: Sage Publications
Frieden, Jeffry 1991. Debt, Development, and Democracy. Princeton University Press.
Garret, Geoffrey. 1995. ŒCapital Mobility, Trade, and the Domestic Politics of Economic Policy". International Organization 49 (4), p.682
Garret, Geoffrey. 1999a. "Globalization and Government Spending Around the World". Paper presented at the 1999 Meeting of the Political Science Association, Atlanta GA, Sept 1-5.
Garret, Geoffrey and Deborah Mitchell. 1999b. "Globalization and the Welfare State". Mimeo, Yale University.
Granato, Jim. 1991. "An Agenda for Econometric Model Building". Political Analysis 3:123-154
Greene, William. 2000. Econometric Analysis. New Jersey: Prentice Hall
Haggard, Stephan and Robert R. Kaufman. 1995. The Political Economy of Democratic Transitions. Princeton: Princeton University Press.
Hicks, Alexander and Dwane Swank. 1992. "Politics, Institutions and Welfare Spending in Industrialized Democracies, 1960-1982". American Political Science Review 86: 658-74.
Hicks, Alexander 1999. Social Democracy and Welfare Capitalism: A Century of Income Security Politics. Cornell University Press.
Huber, Evelyne. 1996. "Options for Social Policy in Latin America: Neoliberal versus Social Democratic Models." In Gosta Esping-Andersen (ed.) Welfare States in Transition: national Adaptations in Global Economies. London: Sage Publications
Huber, Evelyn and John Stephens. Development and Crisis of the Welfare State: Parties and Policies in Global Markets. University of Chicago Press, forthcoming.
Huber, John. 1998. "How Does Party Instability Affect Political Performance? Portfolio Volatility and Health Care Cost Containment in Parliamentary Democracies". American Political Science Review 92(3): 577-591
International Monetary Fund (Various Years). International Financial Statistics. Washington DC
International Monetary Fund (Various Years). Government Finance Statistics. Washington DC
Institute for Democracy and Electoral Assistance (IDEA). 1997. Voter Turnout from 1945 to 1997: A Global Report on Political Participation. Stockholm: International Institute for Democracy and Electoral Assistance.
Iverssen, Torben and Thomas Cusack. 2000. "The Causes of Welfare State Expansion: Deindustrialization or Globalization?" World Politics 52: 313-349.
Johnston, J. 1972. Econometric Methods. New York: MacGraw-Hill.
Katzenstein, Peter. 1985. Small States in World Markets. Ithaca: Cornell University Press
Keohane, Robert and Helen Milner (eds.). 1996. Internationalization and Domestic Politics. New York: Cambridge University Press Kmenta, J. 1971. Elements of Econometrics. New York: MacMillan
Krishnakumar, Jaya and Elvezio Ronchetti (eds). 2000. Panel Data Econometrics: Future Directions. Amsterdam: Elsevier.
Levin Andrew, and C.F. Lin. 1993. "Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties". Discussion Paper 92-93, Department of Economics, University of California, San Diego.
Linz, Juan J. 1994. "Presidential or Parliamentary Democracy: Does it Make a Difference?" in Juan J. Linz and Arturo Valenzuela, eds., The Failure of Presidential Democracy: The Case of Latin America. The Johns Hopkins University Press.
Maddala, G.S. 1971. Econometrics. New York: McGraw-Hill.
Maddala, G.S. 1998. "Recent Developments in Dynamic Econometric Modelling: A Personal Viewpoint". Political Analysis 7: 59-87.
Malloy, James M. 1979. The Politics of Social Security in Brazil. University of Pittsburgh Press.
Mesa-Lago, Carmelo. 1978. Social Security in Latin America: Pressure Groups, Stratification, and Inequality. University of Pittsburgh Press.
Mesa-Lago, Carmelo. 1989. Ascent to Bankruptcy: Financing Social Security in Latin America. University of Pittsburgh Press.
Samuel A. Morely, Roberto Machado, and Stefano Pettinato. 1999. "Indexes of Structural Reform in Latin America," ECLAC Economic Development Division, LC/L.1166, January.
Nelson, Joan M. 1997. "Social costs, Social-Sector Reforms, and Politics in Post-Communist Transformations." In Joan M. Nelson, Charles Tilly, and Lee Walker (eds.) Transforming Post-Communist Political Economies. Washington, D.C.: National Academy Press.
Nelson, Joan M. 2000. "The Politics of Pension and Health-Care Delivery: Reforms in Hungary and Poland," in Janos Kornai, Stephan Haggard, and Robert R. Kaufman, eds., Reforming the State: Fiscal and Welfare Reform in Post-Socialist Countries. Cambridge University Press.
Ostrom, Charles and Reneé Smith. 1992. "Error Correction, Attitude Persistence, and Executive Rewards and Punishments: A Behavioral Theory of Presidential Approval". Political Analysis 4: 127-181.
Palacios, Robert and Montserrat Pallares-Miralles. 2000. "International Patterns of Pension Provision." Manuscript, World Bank.
Pfaller, Alfred et al. 1991. Can the Welfare State Compete? London: Macmillan.
Pierson, Paul. 1991. Beyond the Welfare State? Cambridge: Polity Press
Pierson, Paul 1996. "The New Politics of the Welfare State." World Politics. 48,2: 143-179.
Quinn, Dennis. 1997. "The Correlates of Change in International Financial Regulation" American Political Science Review 91: 531-552.
Rodrik, Dani. 1997. Has Globalization Gone Too Far? Washington DC: Institute for International Economics.
Rogowski, Ronald 1989. Commerce and Coalitions. Princeton: Princeton University Press.
Rudra, Nita. 1999. "Globalization and the Decline of the Welfare State in Less Developed Countries. University of Southern California, Unpublished document.
Sayrs. Lois. 1989. Pooled Time Series Analysis. London: Sage Publications.
Tsebelis, George 1995. "Decision-making in Political Systems: Veto Players in Presidentialism, Parliamentarism, Multicameralism and Multipartyism," British Journal of Political Science 25, 89-325.
World Bank (Various Years). World Development Report. New York: Oxford University Press.
Tables and Figures
Table 1: Changes in Welfare Expenditures in 14 Latin American Countries, 1973- 1997
| WELFCAP | WELFPUB | WELFGPD | PUBGDP | |
| GDPCAP | 0.0189 (1.294) |
0.0016** (2.155) |
0.0001876 (1.006) |
0.00118*** (3.082) |
| DGDP | 2.003** (1.955) |
0.0246 (0.271) |
0.0105 (0.498) |
-0.0393 (-0.844) |
| DEPENDENTS | 0.2384 (0.499) |
0.0521 (1.152) |
0.00204 (0.223) |
0.0528** (1.990) |
| DEBT | -2.4909* (-1.621) |
-0.0339 (-0.286) |
-0.0245 (-0.831) |
0.0600 (1.031) |
| GDP SHOCK |
8.3809 (0.625) |
-0.1014 (-0.087) |
0.00362 (0.015) |
-0.9847 (-1.424) |
| INFLATION SHOCK |
2.6644 (0.175) |
1.4199 (1.250) |
0.06774 (0.283) |
0.8206 (1.106) |
| PUBLIC SPENDING/GDP |
3.657*** (3.145) |
-0.03989 (-0.489) |
0.1207*** (4.366) |
NA |
| DPUBLIC SPENDING/GDP | 10.8296*** (8.583) |
-0.5341*** (-4.986) |
0.2868*** (11.405) |
NA |
| INDUSTRY | 1.6520* (1.794) |
0.1051 (1.346) |
0.0199 (1.151) |
-0.0904** (-2.116) |
| D INDUSTRY | 2.5184** (2.123) |
0.1896** (1.947) |
0.0300 (1.296) |
0.0282 |
| 1970s DECADE DUMMY |
-53.92*** (-4.509) |
-2.3070** (-2.137) |
-0.6066*** (-2.600) |
0.8774 (1.603) |
| 1980s DECADE DUMMY |
-48.11*** (-4.642) |
-2.5919*** (-2.655) |
-0.7043*** (-3.321 |
0.2926 (0.642) |
| TRADE | 0.3102 (0.584) |
0.0184 (0.320) |
-0.0046 (-0.416) |
-0.014 (-0.475) |
| DTRADE | -1.4259*** (-3.428) |
-0.1205*** (-2.966) |
-0.0356*** (-3.869) |
-0.0283 (-1.326) |
| CAPITAL MOBILITY | 0.4042 (0.932) |
0.1062*** (2.448) |
0.01518** (1.886) |
0.0243 (1.353) |
| D CAPITAL MOBILITY | -0.076 (-0.176) |
0.0686* (1.808) |
0.0072 (0.968) |
0.0120 (0.638) |
| TRADE*CAPITAL | -0.0152** (-1.966) |
-0.00122* (-1.653) |
-0.0001936 (-1.369) |
0.00009 (0.250) |
|
D
TRADE*CAPITAL |
-0.0488 (-0.983) |
0.00031 (0.050) |
0.00049 (0.351) |
0.0093*** (2.552) |
| DEMOCRACY | -3.6010 (-0.329) |
-0.888 (-0.860) |
-0.1649 (-0.761) |
0.1621 (0.341) |
| LEFT | -2.029 (-0.220) |
-0.6263 (-0.852) |
-0.1482 (-0.847) |
-0.4292 (-1.085) |
| VOTER TURNOUT | 0.4502* (1.711) |
0.0278 (1.447) |
0.0051 (1.152) |
-0.0005 (-0.076) |
| D VOTER TURNOUT | 0.2786 (0.773) |
0.0433* (1.679) |
0.0041 (0.622) |
0.0033 (0.315) |
| Lagged WELFARE |
-0.2335** (-2.347) |
-0.2754*** (-4.756) |
-0.2704*** (-3.712) |
-0.2995*** (-5.306) |
| R-squared | 0.4150 | 0.3208 | 0.4704 | 0.2699 |
| Wald Chi2 |
170.57 | 212.73 | 312.06 | 123.56 |
| Prob>Chi2 | 0.0000 | 0.000 | 0.000 | 0.0000 |
| Observations | 255 | 255 | 255 | 276 |
All level variables have been lagged one year. T-ratios in parentheses. "*"= 0.05 < p < 0.1; "**"= 0.01 < p < 0.05; "***"= p < 0.01. Model estimated through OLS with panel corrected standard errors as implemented by STATA ["xtpcse" specifying the pairwise option and suppressing the constant]. Country and time fixed-effects included, coefficients for the country dummies are not shown to save space. [The country dummy for Uruguay and the decade dummy for the 90s has been excluded].
Table 2: Changes in Social Security Expenditures in 14 Latin American Countries, 1973- 1997.
| Social Sec. Per capita | Social Sec. as % Gov. Spending | Social Sec. as % of GDP |
|
| GDPCAP | 0.0103 (0.948) |
0.00075 (1.131) |
0.00013 (0.798) |
| DGDP | 1.7806** (1.886) |
0.0881 (1.142) |
0.0145 (0.730) |
| DEPENDENTS | 0.4877 (1.066) |
0.1010*** (2.347) |
0.0162 (1.548) |
| DEBT | -2.2428 (-1.499) |
-0.0835 (-0.660) |
-0.0366 (-1.107) |
| GDP SHOCK |
10.24 (0.860) |
0.9022 (0.912) |
0.0841 (0.357) |
| INFLATION SHOCK |
-2.39 (-0.183) |
0.0263 (0.026) |
-0.1897 (-0.894) |
| PUBLIC SPENDING/GDP |
3.3566*** (3.494) |
0.0549 (0.681) |
0.1016*** (4.210) |
| DPUBLIC SPENDING/GDP | 7.7821*** (6.788) |
-0.1339 (-1.268) |
0.1833*** (6.983) |
| INDUSTRY | 1.6249** (2.021) |
0.1181* (1.835) |
0.0286* (1.826) |
| DINDUSTRY | 2.3086** (2.266) |
0.1565** (1.939) |
0.0386* (1.863) |
| 1970s DECADE DUMMY |
-55.99*** (-5.215) |
-3.3711*** (-3.069) |
-0.9005*** (-3.457) |
| 1980S DECADE DUMMY |
-48.60*** (-5.191) |
-2.9356*** (-3.190) |
-0.8101*** (-3.617) |
| TRADE | 0.4014 (0.705) |
0.0333 (0.582) |
-0.0056 (-0.428) |
| DTRADE | -1.3077*** (-2.996) |
-0.1311*** (-2.971) |
-0.03467*** (-3.153) |
| CAPITAL MOBILITY | 0.5275 (1.141) |
0.1008** (2.276) |
0.0116 (1.258) |
|
D CAPITAL MOBILITY |
-0.2989 (-0.824) |
0.0156 (0.452) |
< td>>-0.0027|
| TRADE*CAPITAL | -0.0170** (-1.959) |
-0.00138* (-1.688) |
-0.00018 (-0.979) |
| DTRADE*CAPITAL | -0.0327 (-0.543) |
0.0038 (0.468) |
0.00003 (0.016) |
| DEMOCRACY | -5.95 (-0.556) |
-1.346 (-1.305) |
-0.2102 (-0.867) |
| LEFT | -5.97 (-0.627) |
-0.6780 (-0.922) |
-0.2210 (-1.080) |
| VOTER TURNOUT | 0.3453 (1.436) |
0.0221 (1.347) |
0.0036 (0.865) |
|
D VOTER TURNOUT |
0.0453 (0.134) |
0.0158 (0.680) |
-0.0010 (-0.170) |
| Lagged WELFARE |
-0.2396*** | -0.3400*** | -0.3157*** |
| R-square | 0.3475 | 0.2843 | 0.3568 |
| Wald Chi2 |
186.06 | 141.55 | 152.25 |
| Prob>Chi2 | 0.0000 | 0.0000 | 0.0000 |
| Observations | 255 | 255 | 255 |
All level variables have been lagged one year. T-ratios in parentheses. "*"= 0.05 < p < 0.1; "**"= 0.01 < p < 0.05; "***"= p < 0.01. Model estimated through OLS with panel corrected standard errors as implemented by STATA ["xtpcse" specifying the pairwise option and suppressing the constant]. Country and time fixed-effects included, coefficients for the country dummies are not shown to save space. [The country dummy for Uruguay and the decade dummy for the 90s has been excluded].
Table 3: Changes in Public Health and Education Expenditures in 14 Latin American Countries, 1973- 1997.
| Health+Educ Per capita |
Health + Educ. % Gov. Spending |
Health+Educ % GDP | |
| GDPCAP | 0.0174*** (0.0174) |
0.00145*** (3.389) |
0.0001 (1.475) |
| DGDP | 0.0864 (0.228) |
-0.0887 (-1.329) |
-0.0076 (-0.611) |
| DEPENDENTS | -0.3798* (-1.616) |
-0.0205 (-0.544) |
-0.0141* (-1.884) |
| DEBT | -0.0114 (-0.016) |
0.1113 (0.929) |
0.0254 (0.991) |
| GDP SHOCK |
-2.0427 (-0.376) |
-0.9177 (-1.015) |
-0.0558 (-0.327) |
| INFLATION SHOCK |
4.2922 (0.786) |
1.3898** (2.168) |
0.1949 (1.582) |
| PUBLIC SPENDING/GDP |
0.8420** (1.991) |
-0.1652*** (-2.573) |
0.0537*** (3.503) |
| D PUBLIC SPENDING/GDP | 3.1274*** (5.138) |
-0.3806*** (-4.255) |
0.1113*** (5.749) |
| INDUSTRY | 0.0310 (0.097) |
0.0177 (0.358) |
-0.0044 (-0.480) |
| DINDUSTRY | 0.1879 (0.399) |
0.0545 (0.786) |
-0.0052 (-0.387) |
| 1970s DECADE DUMMY |
2.0376 (0.331) |
0.8896 (0.850) |
0.2486 (1.227) |
| 1980S DECADE DUMMY |
-3.3922 (-0.727) |
-0.1824 (-0.231) |
-0.0260 (-0.179) |
| TRADE | -0.4178 (-1.722) |
-0.0039 (-0.089) |
-0.0039 (-0.470) |
| DTRADE | -0.1022 (-0.458) |
0.0256 (0.650) |
0.0004 (0.053) |
| CAPITAL MOBILITY | -0.3158* (-1.724) |
0.0255 (0.870) |
0.0032 (0.572) |
| DCAPITAL MOBILITY | 0.1845 (0.795) |
0.0537* (1.651) |
0.0093 (1.573) |
| TRADE*CAPITAL | 0.0055 (1.421) |
0.00004 (0.064) |
0.00003 (0.257) |
| DTRADE*CAPITAL | -0.0171 (-0.388) |
-0.0035 (-0.466) |
0.0003 (0.218) |
| DEMOCRACY | 0.2077 (0.047) |
0.1609 (0.208) |
-0.0768 (-0.557) |
| LEFT | 3.1259 (0.874) |
-0.0717 (-0.111) |
0.0245 (0.179) |
| VOTER TURNOUT | 0.1361*** (2.383) |
0.0109 (1.139) |
0.0028 (1.568) |
|
D VOTER TURNOUT |
0.1992*** (2.640) |
0.0194* (1.850) |
0.0037* (1.882) |
| Lagged WELFARE |
-0.4159*** (0.0859) |
-0.4479*** | -0.4789*** (-4.535) |
| R-square | 0.3885 | 0.3235 | 0.3804 |
| Wald Chi2 |
1216.28 | 112.59 | 196.50 |
| Prob>Chi2 | 0.0000 | 0.0000 | 0.0000 |
| Observations | 260 | 260 | 260 |
All level variables have been lagged one year. T-ratios in parentheses. "*"= 0.05 < p < 0.1; "**"= 0.01 < p < 0.05; "***"= p < 0.01. Model estimated through OLS with panel corrected standard errors as implemented by STATA ["xtpcse" specifying the pairwise option and suppressing the constant]. Country and time fixed-effects included, coefficients for the country dummies are not shown to save space. [The country dummy for Uruguay and the decade dummy for the 90s has been excluded].
Figures
Figure 1: Predicted Changes in Welfare Expenditures per Capita at Different levels of Trade and Capital Mobility.
| High Trade | Low Trade | |
| High Capital Mobility | -25 [-49, 3] |
20.97 [0.7, 40] |
| Low Capital Mobility | -0.97 [-20, 17] |
19.45 [-3, 41.5] |
Note: High and Low values correspond to the 80th and 20th percentiles, respectively [Both levels and first differences of each variable were used]. All other variables have been set at their mean values. The coefficient of the interaction term (TRADE*CAPITAL) was significant at a 0.1 level or better in the original regression with Panel Corrected Standard Errors. Similar results could be obtained using welfare expenditures as a percentage of public spending. However, there was no significant interaction when the dependent variable was welfare spending as a percentage of GDP. In order to facilitate estimation a trimmed model has been used. Particular care was taken to make sure that the coefficients of theoretical interest did not change substantially from the original regression equation. The variables included were welfcap, gdpcap,trade,trade, capital, capital, trade*capital, trade*capital, public spending/gdp, public spending/gdp,voter turnount, industry plus the usual set of country and decade dummies.
Endnotes
Note 1: Maddala (1998) has raised some questions about the validity of some of the recommendations made by Beck and Katz. However, since Beck and Katz's methodology is becoming the most widely accepted technique to estimate pooled time-series models in the political science literature, we have decided to follow their recommendations. Using other alternative methods (e.g. Generalized Least Squares) did not produce any significant changes in the estimates of our variables of substantive interest. In fact, alternative methods of estimation such as Generalized Least Squares (which Maddala argues is more efficient) usually raised the levels of statistical significance of most variables. Back.
Note 2: The following countries have been included in the analysis: Argentina, Bolivia, Brazil, Chile, Costa Rica, Dominican Republic, El Salvador, Mexico, Uruguay, Ecuador, Guatemala, Paraguay, Venezuela, and Peru. For three additional countries (Colombia, Panama, and Nicaragua), we could obtain data on public spending, but not on social spending. These are included in the estimates for government spending, but are omitted from the analysis of social security, health, and education. Back.
Note 3: During the 1990s, pension spending as a percent of GDP reached 15 percent in Uruguay — a larger percentage than in every OECD country except Italy. Argentina, Brazil, Chile, Nicaragua and Panama ranged from 4 to about 6 percent of GDP, figures roughly comparable to Australia (4.6), Canada (5.4), Iceland (5.7), Ireland (7.1), and New Zealand (6.2). (Palacios and Pallares-Miralles 2000:29,42) Back.
Note 4: Garrett (1999a) uses general government consumption expenditures, but this is based on cross-sectional averages and does not include transfer payments. Back.
Note 5: see Garrett's discussion of these issues (1999a, 10-15). Back.
Note 6: We have also tried other specifications of democracy such as (1) using a continuous measure, (2) changing the cutting point from 6 to 7 or 5. We did not see any big changes in the results Back.
Note 7: We also ran the regressions substituting year dummies for decade dummies. This did not significantly affect the results. Back.
Note 8: This model is equivalent to the one described by Beck and Katz (1996) in which the authors explain the importance of separating short-term from long-term effects in dynamic models (see appendix 1). Back.
Note 9: The failure to address these technical problems has called the findings of a number of earlier studies into question. For example, in a replication of Hicks and Swank's (1992, 643) influential study of OECD spending, only 4 of 13 political and institutional variables reach conventional levels of significance when panel corrected standard errors are used (Beck and Katz). Hunter and Brown (1999) do use panel corrected standard errors, but estimate their model without country-specific dummies that control for fixed effects. They are forced to do so by the small number of time-points in their data set, but their estimates are consequently based on the assumption each country has the same intercept. Furthermore, since fixed effects usually absorb cross-national variance that is not explained by any other regressors, their exclusion increases the probability that the omission of other relevant factors (e.g. trade and capital mobility) may be biasing the results. Back.
Note 10: We have also used robust regression to correct for the possible effects of outliers. Moreover, a sequential series of regressions, excluding one country at a time, shows that the results are not driven by any given country. Individual country time series regressions are shown in Tables A1 through A4. Back.
Note 11: The impact of trade liberalization can also be presumed to work through its effects on deindustrialization. However, a simple path analysis shows that the coefficient of the direct effect of trade is almost twice as large as the path that goes through deindustrialization. Back.
Note 12: The political variables (i.e. Left, democracy and voter turnout are stationary). Back.
Note 13: Conventional tests for cointegration and unit roots are in fact not appropriate when dealing with TSCS data. Though some authors have proposed alternative tests [e.g. Levin and Lin (1993)], tests for unit roots and cointegration in the context of panel data remain a rather undertheorized field. Back.
Note 14: Like any other dynamic model, a correct estimation of the error-correction model also requires that no significant amount of serial correlation be left in the model. This is particularly important when a lagged dependent variable is included. If present, serial correlation might then lead not only to biased, but also to inconsistent estimates (i.e. estimates that do not converge toward their true values as the sample size increases). We have used two different tests for serial correlation (Durbin's h test and a Lagrange multiplier test). Based on the results for these two tests, we fail to reject the null hypothesis of no serial correlation, and can therefore conclude that our parameter estimates are not biased by serial correlation. Back.
Note 15: The construction of the welfare variables was as follows. First, from the Government Finance Statistics of the International Monetary Fund (various issues) we obtained data on expenditures on health care, education and social security in local currency units. We added together these three concepts and divided the result by the overall public budget of the consolidated central government. This gave us the share of social spending in the public budget (i.e. variable WELFPUB). Then, from the International Finance Statistics (IMF) we obtained the value of the Gross Domestic Product in local currency units. Second, we computed the ratio of public expenditures to GDP (i.e. PUBGDP variable). Third, we multiplied WELFPUB (welfare spending as a percentage of public spending) by PUBGDP (public spending as a percentage of GDP). This way we obtained a new variable that measured social spending as a percentage of GDP (i.e. WELFGDP). Finally, for each year, we multiplied WELFGDP (welfare spending as a percentage of GDP) by GDP in constant 1995US dollars (which we had obtained from the International Finance Statistics of the IMF) and divided it by each year's population levels. The final result was a series of welfare spending per capita in constant 1995 USD (i.e. WELFCAP), which makes social spending comparable across cases. Back.
