¹ Przeworski, Adam and Limongi, Fernando, “Political Regimes and Economic Growth,” Journal of Economic Perspectives 7 (Summer 1993CrossRefGoogle Scholar).

² See Alesina, Alberto and Perotti, Roberto, “The Political Economy of Growth: A Critical Survey of the Recent Literature,” World Bank Economic Review 8 (September 1994CrossRefGoogle Scholar); Geddes, Barbara, “The Politics of Economic Liberalization,” Latin American Research Review 30, no. 2 (1995Google Scholar); Haggard, Stephan, Pathwaysfromthe Periphery: The Politics of Growth in the Newly Industrialized Countries (Ithaca, N. Y.: Cornell University Press, 1990Google Scholar); Haggard, Stephan and Kaufman, Robert, “Institutions and Economic Adjustment,” in Haggard, and Kaufman, , eds., The Politics of Economic Adjustment (Princeton: Princeton University Press, 1992Google Scholar); Schneider, Ben Ross, “Democratic Consolidation: Some Broad Comparisons and Sweeping Arguments,” Latin American Research Review 30, no. 2 (1995Google Scholar).

³ See, for instance, Carlos, Luiz Pereira, Bresser “Economic Reforms and Economic Growth: Efficiency and Politics in Latin America,” in Carlos, Luiz Pereira, Bresser, Maravall, Jose Maria, and Przeworski, Adam, Economic Reforms in New Democracies: A Social-Democratic Approach (New York: Cambridge University Press, 1993Google Scholar); and Fishlow, Albert, “The Latin American State,” Journal of Economic Perspectives 4 (Summer 1990CrossRefGoogle Scholar). John Williamson reports that revenue concerns were present in the tax reform proposals in the ten Latin American countries he examined. Richard M. Bird stresses the concern with increasing revenue in recent tax reforms in Latin America. The World Bank has recently recognized the importance of increased tax revenue in stabilization and reform programs for developing countries. Wiliamson, , ed., Latin American Adjustment: How Much Happened? (Washington, D.C.: Institute of International Economics, 1990Google Scholar); Bird, , “Tax Reform in Latin America: A Review of Some Recent Experiences,” Latin American Research Review 27, no. 1 (1992Google Scholar); Bank, World, World Develop ment Report 1988 (Washington, D.C.: World Bank, 1988CrossRefGoogle Scholar); idem, Lessons of Tax Reform (Washington, D.C.: World Bank, 1991Google Scholar).

⁴ Przeworski, Adam, The State and the Economy under Capitalism (Chur, Switzerland: Harwood Academic Publishers, 1990Google Scholar); Findlay, Ronald, “The New Political Economy: Its Explanatory Power for LDCs,” Economics and Politics 2 (July 1990CrossRefGoogle Scholar); Olson, Mancur, “Autocracy, Democracy, and Prosperity,” Zeckhauser, Richard J., ed., Strategy and Choice (Cambridge: MIT Press, 1991Google Scholar).

⁵ Mancur Olson is an exception since his results hold even if democracies are composed of groups with different levels of income, seeking redistribution in their favor. Olson (fn. 4).

⁴ See De Schweinitz, Karl, Industrialization and Democracy (New York: Free Press, 1964Google Scholar); and Galenson, Walter, “Introduction,” in Galenson, , ed., Labor and Economic Development (New York: John Wiley, 1959Google Scholar).

⁷ De Schweinitz (fn. 6).

⁸ This electoral dynamic does not require organized labor as a large sector of the population. It can be sustained in any system where labor constitutes a significant voting bloc or by any organization or political party that demands higher levels of social services and government expenditure.

⁹ The idea that democracy undermines growth because it lowers the rate of investment is, of course, quite widespread. See Przeworski and Limongi (fn. 1) for a critical discussion of various hypotheses. I singled out this formulation because of the explicit role of taxation in the negative connection between democracy and growth. Note that De Schweinitz (fn. 6) does not consider the potential benefits of an increased amount of resources under the control of the state. His discussion implies that the fundamental role of the state in economic development is to affect the distribution of resources in order to place a disproportionate share of income in the hands of those who are considered to have a higher propensity to save. Taxation is detrimental to growth because it interferes with this process. However, this argument fails if, as Ragnar Nurkse has argued, private savings in developing countries are not forthcoming at the rate necessary to sustain reasonable levels of economic growth. Nurkse, , “Growth in Under-Developed Countries: Some International Aspects of the Problem of Economic Development,” American Economic Review 42 (May 1952Google Scholar).

¹⁰ Bates, Robert H., “A Political Scientist Looks at Tax Reform,” in Gillis, Malcolm, ed., Tax Reform in Developing Countries (Durham, N.C.: Duke University Press, 1989), 479Google Scholar.

¹¹ Levi, Margaret, Of Rule and Revenue (Berkeley: University of California Press, 1988Google Scholar).

¹² Haggard (fn. 2), 262.

¹³ Ibid.

¹⁴ The idea of an “unconstrained dictator” underlies several analyses of policy performance under different political regimes. For examples, see Oszlak, Oscar, “Public Policies and Political Regimes in Latin America,” International Social Science Journal 38, no. 2 (1986Google Scholar); and Most, Benjamin A., “Author itarianisms and the Growth of the State in Latin America: An Assessment of Their Impact on Argentine Public Policy, 1930–1970,” Comparative Political Studies 13 (July 1980CrossRefGoogle Scholar). For critical discussions, see Sloan, John W. and Tedin, Kent T., “The Consequences of Regime Type for Public Policy Outcomes,” Comparative Political Studies 20 (April 1987CrossRefGoogle Scholar); Sloan, John W., “The Policy Capabilities of Democratic Regimes in Latin America,” Latin American Research Review 24, no. 2 (1989Google Scholar); and Remmer, Karen L., “Evaluating the Policy Impact of Military Regimes in Latin America,” Latin American Research Review 13, no. 2 (1978Google Scholar).

¹⁵ Ames, Barry, Political Survival: Politicians and Public Policy in Latin America (Berkeley: University of California Press, 1987Google Scholar). One could argue that the issue is one of relative, not absolute, autonomy, that dictatorships are relatively more autonomous than democratic regimes and that, for this reason, they will be relatively more able to extract revenues from society. However, once the existence of such a limit is accepted, one needs to demonstrate that the degree of autonomy enjoyed by authoritarian governments enables them to pursue specific policies that democratic governments would not have been able to pursue. In other words, one has to demonstrate that the nature of the groups and coalitions that support authoritarian governments allows them a greater degree of freedom with respect to taxation (and other policies). But this can hardly be shown in general. As the history of many Latin American countries suggests, authoritarian governments that depend essentially on the support of only a few groups, say, large industrialists and landowners, are unlikely to be able to collect much revenue since these are precisely the biggest potential taxpayers. See, for example, Best, M. H., “Political Power and Tax Revenues in Central America,” Journal of Development Economics 3 (March 1976CrossRefGoogle Scholar); and Gallo, Carmenza, Taxes and State Power: Political Instability in Bolivia, 1900–1950 (Philadelphia: Temple University Press, 1991Google Scholar).

¹⁶ Bird, Richard M., “The Administrative Dimensions of Tax Reform in Developing Countries,” in Gillis, Malcolm, ed., Tax Reform in Developing Countries (Durham, N.C.: Duke University Press, 1989Google Scholar).

¹⁷ Levi (fn. 11).

¹⁸ The importance of this factor is also emphasized by Bates (fn. 10).

¹⁹ Another way to put it is that the legitimacy of democratic governments increases compliance with taxation, thus reducing enforcement costs and increasing revenue. According to Levi, quasi-voluntary compliance is one aspect of what is usually labeled legitimacy. In fact, as she argues, it provides micro-foundations to the concept of legitimacy by separating analytically the elements that rest on ideology and norms and the elements that rest on strategic calculations and rational behavior. Levi (fn. 11), 54.

²⁰ Political variables are usually disregarded in both empirical and theoretical analyses of taxation. See Walter Hettich and Stanley Winer for a review of the models ofgovernment assumed in the main approaches to public finance. As they point out with respect to “optimal taxation,” the most recent, sophisticated, and promising approach in the area of public finance, “political behavior is simply ignored.” Hettich and Winer, “Blueprints and Pathways: The Shifting Foundation ofTax Reform,” National Tax Journal 38 (December 1985), 440.

²¹ Musgrave, Richard, Fiscal Systems (New Haven: Yale University Press, 1969Google Scholar).

²² See Bahl, Roy W., “A Regression Approach to Tax Effort and Tax Ratio Analysis,” IMF Staff Papers 18 (November 1971Google Scholar); Chelliah, Raja J., “Trends in Taxation in Developing Countries,” IMF Staff Papers 18 (July 1971Google Scholar); Chelliah, Raja J., Baas, Hessel J., and Kelly, Margaret R., “Tax Ratios andTax Ef fort in Developing Countries, 1969–71,” IMF Staff Papers 22 (March 1975Google Scholar); Leuthold, Jane H., “Tax Shares in Developing Economies: A Panel Study,” Journal of Development Economics 35 (January 1991CrossRefGoogle Scholar); Lotz, Jorgen R. and Morss, Elliott R., “Measuring 'Tax Effort' in Developing Countries,” IMF Staff Papers 14 (November 1967Google Scholar); and Tait, Alan A., Gratz, Wilfrid L. M., and Eichengreen, Barry J., “International Comparisons of Taxation for Selected Developing Countries, 1972–76,” IMF StaffPapers 26 (March 1979Google Scholar).

²³ Musgravc (fn. 21).

²⁴ See Tarschys, Daniel, “Tributes, Tariffs, Taxes and Trade: The Changing Sources of Government Revenue,” British Journal of Political Science 18 (January 1988CrossRefGoogle Scholar); and Ardant, Gabriel, “Financial Policy and Economic Infrastructure of Modern States and Nations,” in Tilly, Charles, ed., The Formation of National States in Western Europe (Princeton: Princeton University Press, 1975Google Scholar).

²⁵ See Snider, Lewis W., “The Political Performance of Third World Governments and the Debt Crisis,” American Political Science Review 84 (December 1990CrossRefGoogle Scholar).

²⁶ Chelliah (fh. 22), 295.

²⁷ In December 1992, Brazilian newspapers reported that Petrobras, the state oil monopoly, was the rise's largest debtor.

²⁸ Unfortunately, there are no comparative data on state-owned enterprises that cover as many countries as the ones included in this analysis. Such an exclusion need not bias the results reported here since a large public sector may either require capital outlays, thus exerting pressures for higher taxation, or provide revenue for the government, thus easing the pressure of taxation.

²⁹ For an example concerning Venezuela, see Mann, Arthur J., “On the Utility of Tax Performance Indices: Notes with Reference to Venezuela's Tax System,” Bulletinfor International Fiscal Documentation 31 (October 1977Google Scholar).

³⁰ Cukierman, Alex, Edwards, Sebastian, and Tabellini, Guido, “Seigniorage and Political Instability,” American Economic Review 82 (June 1992Google Scholar). See also Edwards, Sebastian and Tabellini, Guido, “Explaining Fiscal Policies and Inflation in Developing Counties? Journal ofInternational Money and Finance 10 (March 1991Google Scholar); and idem, “Political Instability, Political Weakness and the Inflation Tax,” NBER Working Paper, no. 3721 (Cambridge: National Bureau of Economic Research, 1991).

³¹ This is the “hazard rate” produced by a model of the survival of chief executives in office as a function of length of tenure, economic growth, and the number of accumulated changes in chief executive positions experienced in a country since 1950, assuming a Weibull distribution (p = 1.12, with confidence interval from 1.02 to 1.21). Alternative specifications of the survival function do not change the results. Chief executives are presidents in presidential democracies, prime ministers in the parliamentary and mixed systems, and whoever is the effective ruler in dictatorships. The latter can be designated explicitly as the dictator, or can bear such titles as the head of the militaryjunta, president, leader of the ruling party, executor of the state of emergency, or king. Chief executives were coded on the basis of the work of DaGraca, John V., Heads of State and Government (New York: New York University Press, 1985CrossRefGoogle Scholar); Bienen, Henry and van de Walle, Nicolas, Of Time and Power: Leadership Duration in the Modem World (Stanford: Stanford University Press, 1991Google Scholar); and Banks, Arthur S., Cross-National Time-Series Data Archive (Center for Social Analysis, SUNY at Binghamton, 1993Google Scholar), magnetic tape. One limitation of this indicator of discount rate is that it is entirely backward-looking. An ideal opera-tionalization of this factor would also consider the government's perceptions of its ability to alter its future chances of survival. I owe this observation to Jon Elster.

³² North, Douglass C., Structure and Change in Economic History (New York: Norton, 1981), 28–29Google Scholar

³³ This should be true of any type of elections, even the ones that occur under dictatorships. Authoritarian governments who subject themselves to popular elections do so in an attempt to legitimize their rule, to enlarge their base of support, and to consolidate their hold on power. They are, thus, also dependent on resources controlled by voters and, at least to some degree, are subject to the uncertainties inherent to electoral competition. Note that the frequency of elections is surprisingly high in dictatorships: information on 135 countries since 1950 indicates that presidential elections were held in 5 percent of the years under authoritarianism and 6 percent of the years under a democracy. On the one hand, legislative elections were more frequent under democracies than under dictatorships: 27 percent against 18.1 percent. On the other hand, 107 of the 231 presidential elections held since 1950 took place in democracies and 124 in dictatorships; of the 890 legislative elections during the same period, 457 were held in democracies and 433 in dictatorships. See Jose Antonio Cheibub and Adam Prze-worski, “Democracy, Elections and Accountability for Economic Outcomes,” in Bernard Manin, Susan Stokes, and Adam Przeworski, eds., Democracy andAccountability (New York: Cambridge University Press, forthcoming).

³⁴ Again, there is no reason for this to happen only in one type of regime. The dependence of new governments on external actors should be similar in both authoritarian and democratic regimes. Regime transition was more frequent in dictatorships than in democracies: of the eighty-nine transitions since 1950, forty-nine represented the death of a dictatorship. Changes in the executive, in turn, were more frequent in democracies: 60 percent of all changes since 1950 occurred in democracies. See Cheibub and Przeworski (fn. 33).

³⁵ Guido Tabellini, “International Tax Comparisons Reconsidered” (Manuscript, Fiscal Affairs Department, International Monetary Fund, 1985).

³⁶ Ibid. In his study Tabellini also includes the noninterest level of public expenditures as an explanatory variable when estimating the level of taxation for sixty-one developing countries. This, however, introduces a serious simultaneity bias into the estimation (p. 17). The government's fiscal situation also depends on alternative sources of revenue such as nontax and seigniorage revenue. The inclusion of these variables in the analysis, however, may lead to multicollinearity since the same factors that affect the level of tax revenue probably also influence nontax and seigniorage revenue. To avoid this problem, one could followTabellini's recommendation and use total revenue, defined as the sum of tax, capital, and seigniorage revenue, as the dependent variable. Here, however, we want to know something about the government's willingness and ability to tax, as opposed to its willingness and ability to extract revenue from society, regardless of its form. This distinction is particularly relevant given that the increasing of tax revenue seems to be one of the paths out of the fiscal crises that have troubled many developing countries in the past decade. Moreover, tax revenue represents 83 percent of the total revenue in the countries for which data are available.

³⁷ This equation was estimated with information produced by the International Monetary Fund, the most comprehensive set of comparative data on taxation available for researchers. Data on levels of taxation are available for 108 countries from as early as 1970 through 1990, amounting to 1613 observations. Details about the coverage are presented in the appendix. World Bank, World Data 1994: World Bank Indicators on CD-ROM (Washington, D.C.: World Bank, 1994). A replication data set is available at the author's web site, at http://www.ssc.upenn.edu/polisci/faculty/CheibubJose.html.

³⁸ p was computed as l-'Ad, where d is the Durbin-Watson statistic produced by the OLS estimation. The number of observations varies due to the availability of data on the independent variables and the loss of observations in estimating models with autocorrelated errors.

³⁹ The results do not change in any significant way ifwe modify the definitions of the variables. The GDP share of imports (or exports) alone is positively associated with taxation. Using the cumulative changes of chief executive as a direct indicator of the government's discount rate also does not change anything. Finally, no substantive change occurs if we use other hazard functions (e.g., proportional, estimated only with the gowth of per capita income or with regional dummy variables).

⁴⁰ See Alvarez, Mike et al., “Classifying Political Regimes,” Studies in International Comparative Developmtnt 31 (Summer 1996Google Scholar).

⁴¹ If we exclude OECD countries, taxes are still higher in democracies than in dictatorships, although the difference is considerably smaller: 18.8 percent against 17.7 percent.

⁴² Heckman, James J., “The Microeconomic Evaluation of Social Programs and Economic Institutions,” in Chung-Hua Series ofLectures by Invited Eminent Economists, no. 14 (Taipei: The Institute of Economics, Academia Sinica, 1988Google Scholar).

⁴³ These numbers are in line with what has been found by others. Seymour Martin Lipset was probably the first to provide statistical evidence for the connection between economic development and democracy, a finding that has proven to be quite robust. Lipset, , “Some Social Requisites of Democracy. Economic Development and Political Legitimacy,” American Political Science Review 53 (March 1959CrossRefGoogle Scholar). See also Przeworski, Adam and Limongi, Fernando, “Modernization: Theories and Facts,” World Politics 49 (January 1997CrossRefGoogle Scholar). Empirical support is also strong for the relationship between economic development and levels of taxation. See Hinrichs, H. H., A General Theory of Tax Structure Change during Economic Development (Cambridge: Harvard Law School International Tax Program, 1966Google Scholar); and Tanzi, Vito, “Quantitative Characteristics of the Tax Systems of Developing Countries,” in Newbery, David and Stern, Nicholas, eds., The Theory of Taxationfor Developing Countries (New York: Oxford University Press, 1987Google Scholar).

⁴⁴ Heckman (fh. 42), 17–19.

⁴⁵ Table 3 suggests that some unobserved factor besides the level of economic development may be operating with respect to taxation and political regimes. Although the mix of regimes is different at each level of per capita income, and taxes are higher at each level, taxes are higher in dictatorships when per capita income is low, and higher in democracies when per capita income is high. Although it would be desirable to attempt to specify and observe the factors that account for this pattern, this is not necessary for the more limited purpose of assessing whether political regimes have any causal impact on the level of taxation.

⁴⁶ James J. Heckman, “The Common Structure of Statistical Models of Truncation, Sample Selection, and Limited Dependent Variables and a Simple Estimator for Such Models,” Annals ofEconomic and SocialMeasurement 5, no. 4 (1976).

⁴⁷ Table 3 includes the main variables identified in the cross-national literature on the determinants of political regimes. See Bollen, Kenneth A., “Political Democracy and the Timing of Development,” American Sociological Review 44 (August 1979CrossRefGoogle Scholar); Bollen, Kenneth A. and Jackman, Robert W., “Economic and Noneconomic Determinants of Political Democracy in the 1960s,” Research in Political Sociology (1985Google Scholar); Cutright, Phillips, “National Political Development: Measurement and Analysis,” American Sociological Review 28 (April 1963CrossRefGoogle Scholar); Lipset, Seymour Martin, Seon, Kyoung-Ryung, and Torres, John Charles, “A Comparative Analysis of the Social Requisites of Democracy,” InternationalSocial Science Journal 136 (1993Google Scholar); Neubauer, D. E., “Some Conditions of Democracy,” American Political Science Review 61 (December 1967CrossRefGoogle Scholar); and Vanhanen, Tatu, Tie Process of Democratization (New York: Crane Russack, 1992Google Scholar). Although the concern here is not with the estimation of the determinants of regimes per se, it is worth noting that the results reported in Table 3 conform with the main findings of this literature: we are more likely to observe authoritarian regimes when countries experienced transitions to authoritarianism in the past (STRA), when the number of democratic regimes in a country's own region increases (ODRP), and when countries have a higher Moslem population (MOSLEM). Democracies, in turn, are more likely to be observed in countries that have experienced at least one democratic period (STRD), that are ethnically homogeneous (ETHNIC), that were not independent as of 1950 (NEWC), that were British colonies if not independent as of 1950 (BRITCOL), and that have a large Catholic population (CATH). Together, these estimates predict correctly 90 percent of the democratic regimes observed in the sample and 95 percent of the dictatorships.

⁴⁸ Because transition years are coded as the regime that emerges in that year, TRANS under democracies indicates a new democracy.

⁴⁹ Adam Przeworski and Fernando Limongi, “Selection, Counterfactuals and Comparisons” (Manuscript, New York University, 1996).

⁵⁰ When all variables are assumed to be exogenous, we calculate the figures in Table 5 by computing t_jit, = x_itβ_j, where j = 0 (democracies) or 1 (dictatorships), x_it is the vector of variables in equation 2 for country i at year t, and β_j. is the vector of coefficients for these variables under each regime, as reported in Table 4. When some variables x′ in x are assumed to be endogenous to regimes, we compute t_jit = x_it β_j + x′_jβ_j., where x now represents a subset of the variables in equation 2 and x′_j. is the average for each regime of the variables in equation 2 that are considered endogenous with respect to regime.

⁵¹ Alvarez et al. (fh. 40) is the source for REGIME, ELECTION, ADM, TRANS, STRA, STRD, ODRP, NEWC, and BRITCOL. World Bank (fn. 37) is the source for TAX, AGRIC, MINERAL, DEBT, and GRANTS. Penn World Tables, release 5.6, is the source for LEVEL and TRADE. The World in Figures: Editorial Information Compiled by The Economist (Boston: G. K. Hall, 1988), and Vanhanen (fh. 47) are the sources for ETHNIC. Lehykon Panstw S'wiata 1993–94 (Warsaw: Real Press, 1993) is the source for MOSLEM and CATH.