Skip to main content

The Fiscal Contract: States, Taxes, and Public Services

  • Jeffrey F. Timmons (a1)

Using data from approximately ninety countries, the author shows that the more a state taxes the rich as a percentage of GDP, the more it protects property rights; and the more it taxes the poor, the more it provides basic public services. There is no evidence that states gouge the rich to benefit the poor or vice versa, contrary to state-capture theories. Nor is there any evidence that taxes and spending are unrelated, contrary to state-autonomy models. Instead, states operate much like fiscal contracts, with groups getting what they pay for.

Hide All

1 See North, Douglass C., Growth and Structural Change (New York: W. W. Norton, 1981); Bates, Robert H. and Lien, Da-Hsiang Donald, “A Note on Taxation, Development, and Representative Government,” Politics and Society 14 (March 1985); and Levi, Margaret, Of Rule and Revenue (Berkeley: University of California Press, 1988). For tax compliance, see Andreoni, James, Erard, Brian, and Feinstein, Jonathan, “Tax Compliance,” Journal ofEconomic Literature 36 (June 1998).

2 Important contributions include Weber, Max, Economy and Society, ed. Roth, Guenther and Wittich, Claus (New York: Bedminster Press, 1968); Hinrichs, Harley H., A General Theory of Tax Structure Change during Economic Development (Cambridge: Harvard Law School, 1966); Hintze, Otto, “Military Organization and the Organization of the State,” in Gilbert, F., ed., The Historical Essays of Otto Hintze (New York: Oxford University Press, 1975); Steinmo, Sven, Taxation and Democracy: Swedish, British and American Approaches to Financing the Modern State (New Haven: Yale University Press, 1993); and Lindert, Peter H., Growing Public: Social Spending and Economic Growth since the Eighteenth Century (Cambridge: Cambridge University Press, 2004), data:

3 North, Douglass C. and Weingast, Barry, “Constitutions and Credible Commitments,” Journal of Economic History 49 (December 1989).

4 Kato, Junko, Regressive Taxation and the Welfare State (Cambridge: Cambridge University Press, 2003), 3.

5 Developing countries have generally been characterized by low taxes and low services, but there are some notable exceptions: Jamaica, Chile, South Korea, and some Eastern European countries, for example, have typically collected two to three times as much revenue from regressive taxes as have other developing countries; they have also provided higher levels of services.

6 Musgrave, Richard A. and Musgrave, Peggy B., Public Finance in Theory and Practice, 5th ed. (New York: McGraw-Hill, 1989).

7 Brennan, Geoffrey and Buchanan, James M., The Power to Tax: Analytical Foundations of Fiscal Constitution (New York: Cambridge University Press, 1980).

8 See Olson, Mancur, The Logic of Collective Action (Cambridge: Harvard University Press, 1965).

9 Hibbs, Douglas, “Political Parties and Macroeconomic Policy,” American Political Science Review 7 (December 1977). Among the capture-based theories, partisan models of (democratic) government clearly have the most empirical support. Partisan effects are most notable in terms of transfers, spending, and macroeconomic policies, but less so in terms of taxes. Geoffrey Garrett, for example, found that countries with strong left-wing parties and powerful labor unions actually had more regressive tax systems than right-wing governments with weak labor unions—consistent with thefiscalcontract model and the data presented in Figure 7; Garrett, , Partisan Politics in the Global Economy (Cambridge: Cambridge University Press, 1998).

10 See North (fn. 1); Bates and Lien (fn. 1); and Levi (fn. 1)

11 For evolutionary biology, see Axelrod, David, The Evolution of Cooperation (New York: Basic Books, 1984); and Binmore, Ken G., Playing Fair (Cambridge: MIT Press, 1994). For tax compliance, see Aim, James, McClelland, Gary, and Schulze, William D., “Why Do People Pay Taxes?” Journal of Public Economics 48 (June 1992); Scholz, John T. and Pinney, Neil, “Duty, Fear, and Tax Compliance: The Heuristic Basis of Citizenship Behavior,” American Journal of Political Science 39 (May 1995); Slemrod, Joel, “Trust in Public Finance” (Paper presented at Public Finances and Public Policy in the New Millennium Conference, Munich, January 12-13, 2001); and Andreoni, Erard, and Feinstein (fn. 1).

12 For the sake of simplicity, I assume that these games exist independent of one another and independent of other state/society relations.

13 I drop the subscripts for now since it does not matter which group the state is playing against.

14 The TFT probability can be infinitesimal if the players have long time horizons. Watson, Joel, “Cooperation in the Infinitely Repeated Prisoners' Dilemma with Perturbations,” Games and Economic Behavior 7 (September 1994).

15 Ibid. Watson's results effectively rule out long chains of defections by either party, implying that Quadrants II and III are not sustainable; Quadrant I could be sustainable if the TFT probability is low and discount rates are high.

16 Martin McGuire and Mancur Olson's “encompassing interest” model of the state, for example, assumes that Sc = 0; McGuire, and Olson, , “The Economics of Autocracy and Majority Rule: The Invisible Hand and the Use of Force,” Journal of Economic Literature 34 (June 1996).

17 Since all of the quadrants can be equilibria under fairly reasonable conditions, I am really making an empirical bet that conditions 1-4 hold.

18 According the National Election Studies (NES) Guide to Public Opinion and Electoral Behavior, for example, support for government-sponsored health insurance in the United States is inversely related to income, with 57 percent of the bottom 15 percent of the income distribution preferring it compared with 29 percent among the wealthiest 5 percent (2004), a pattern that has been relatively stable since 1970. Similarly, NES, General Social Surveys, and World Value Surveys consistently show that lower-income groups prefer more state ownership of industry, greater control of prices, and more intervention in markets—all of which infringe on private property rights. See www.umich. edu/~nes/nesguide/nesguide.htm; the Survey Documentation and Analysis Web site, maintained by the Computer-assisted Survey Methods Program at the University of California, Berkeley, http://sda.; and Basanez, Miguel, Inglehart, Ronald, and Moreno, Alejandro, Human Values and Beliefs: A Cross-Cultural Sourcebook Based on the 1990-1993 World Values Survey (Ann Arbor: University of Michigan Press, 1998).

19 It is certainly possible that these groups desire, receive, and pay for other goods besides the ones proxied herein. These are indicators, not the whole story. Education is not chosen, because it is not clear that lower-income groups have an especially strong desire for the provision of education. I presume the rich are willing to pay for their children (for example, Highland Park, Texas) but not for redistribution across districts.

20 Heritage Foundation, Annual Survey of Economic Freedom, 1995-2003, See the appendix for a more detailed definition.

21 Simeon Djankov, Rafael La Porta, Florencio Lopez-de-Silanes, and Andre Shleifer, “Courts: The Lex Mundi Project,” NBER Working Paper, no. 8890 (Cambridge, Mass., April 2002). See the appendix for a more detailed definition.

22 Ibid., 1.

23 Most incidence studies evaluate either the spending or the revenue side of government accounts but not both simultaneously. The Luxembourg Income Studies (LIS) offer the best cross-national data about tax and expenditures, but they exclude consumption taxes. See

24 See Ebrill, Liam P., Keen, Michael J., Bodin, Jean-Paul, and Summers, Victoria P., The Modem VAT (Washington, D.C.: International Monetary Fund, 2001); and Fullerton, Don and Metcalf, Gilbert E., “Tax Incidence,” NBER Working Paper, no. 8829 (Cambridge, Mass., 2002). It is important to note that while regressive taxes typically account for a small fraction of the income of the rich, the dollar amounts can be sizable.

25 To take some examples: until recently, the minimum level of taxable income in Brazil was more than three times per capita income; in Ecuador, Nicaragua, and Guatemala it was about ten times per capita income. See Inter American Development Bank, Economic and Social Progress in Latin America, 1998-1999: Facing Up to Inequality in Latin America (Washington, D.C.: Inter American Development Bank, 1999), 185.

26 In theory, payroll, trade, and social security taxes are regressive to the extent that they can be passed on to consumers and/or labor. In practice, the extent of pass on depends heavily on the elasticities of each factor. Unfortunately, these elasticities vary considerably from country to country and from time period to time period, making it difficult to make broad generalizations. Empirical studies suggest that social security and trade taxes tend to be regressive, while there is no consensus about property taxes. See Ebrill et al. (fn. 24); and Fullerton-and Metcalf (fn. 24).

27 The correlation between revenue from goods and services and revenue from capital is approximately 0.3, suggesting that there is no trade-off between taxing the rich and taxing the poor.

28 Social security taxes are systematically related to outcomes, while trade taxes generally are not. These results and other related material can be found at my Web site,

29 Property taxes are included in the other revenue category. Because they are tiny relative to GDP (about 0.3 percent of GDP in developing countries), changing the assumption about the incidence of property taxes has no meaningful impact on the results (calculation made using IMF data from 1990-99; IMF Tax Revenue Database, Document no. 1338900, Washington, 2000).

30 By focusing on central government revenue and, by extension, performance, I am overlooking all of the subnational action. Excluding subnational data, which is unfortunately scarce, could create some problems, notably measurement error in the revenue variables and, hence, both biased and inconsistent estimates. To try to avoid these problems, I split the sample into centralized and federal countries based on the tax, spending, and legislative authority granted to subnational governments, labeled “author” by Thorsten Beck, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh, “New Tools in Comparative Political Economy: The Database of Political Institutions (2001/2003),” World Bank Economic Review 15 (September 2003). Surprisingly, perhaps, tests of means reveal no statistical difference in revenue collection between centralized and federal regimes, perhaps reflecting heterogeneity among federalisms. The results using nonfederal countries, which account for roughly three-quarters of total observations, are very similar to those reported herein. The results for federal countries alone are also consistent with the reported results, though the significance levels are lower (around 85 percent) and the coefficients are marginally smaller, perhaps reflecting the sample size and attenuation bias. Thoseresultsare available from the author.

31 The spending data is drawn from Linden (fh. 2). Unless otherwise mentioned in the text or data appendix, the remaining data are drawn from the World Development Indicators CD-Rom (Washington, D.C.: World Bank, 2001)

32 Three-year and ten-year averages produced results similar to the five-year averages.

33 The N for each variable is available in Table 2.

34 See, for example, Bradley, David, Huber, Evelyne, Moller, Stephanie, Nielsen, Francois, and Stephens, John D., “Distribution and Redistribution in Postindustrial Democracies,” World Politics 55 (January 2003).

35 Results using seemingly unrelated regression methods (SUR) are similar to those reported below. Zellner, Arnold, “An Efficient Method for Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias,” Journal of American Statistical Association 57 (June 1962).

36 Arellano-Bond specifications using lagged levels and first differences of the variables as instruments were always the right sign and significant for the predictions of the fiscal contract, but I could not find a specification that rejected serial correlation and accepted the validity of the instruments. Panel estimates for infant mortality and measles and DPT immunizations using panel corrected standard errors (PCSE) and a generalized estimating equation (GEE) also supported the fiscal contract; they too were plagued by problems of serial correlation. See my Web site (fn. 28) for these results. The main problem is that my time series is too short to capture within-country changes, which are quite small, especially compared with the cross-country variation. Manuel Arellano and Bond, Stephen, “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations,” Review of Economic Studies 58 (April 1991); Beck, Neal and Katz, Jonathan, “What to Do (and Not to Do) with Time-Series Cross-Section Data,” American Political Science Review 89 (September 1995); Zorn, Chris, “Generalized Estimating Equation Models for Correlated Data: A Review with Applications,” American Journal of Political Science 45 (October 2001).

37 Rogers, William H., “sg17: Regression Standard Errors in Clustered Samples,” Stata Technical Bulletin 13 (1993).

38 See Hinrichs(fn.2).

39 See Weber (fn. 2); and Steinmo (fn. 2). Endogenizing the tax system is devilishly difficult. When the tax variables are placed on the right-hand side and plausible explanatory variables (national income, urban population, industry value added, manufacturing value added, population age structure, country size, democracy, federalism, presidentialism, trade, fuel exports, and so on) are placed on the left-hand side, the regressions typically have an R-squared below 0.5, indicating that most of the explanation is in the error term; see Web site (fn. 28).

40 To address endogeneity from an econometric standpoint requires instrumental variables. Unfortunately, it is difficult to find instruments for the sources of revenue that have a high explanatory value and are not correlated with each other or the disturbance term. Within the OECD, for example, the variable with the highest correlation with consumption taxes is the dependent variable, social spending (corr = 0.75), followed by gross union membership (0.64) and population 65+ (0.56).

41 The theory of the fiscal contract predicts that consumption taxes should be positively related to public services, while capital taxes should have no effect. Since the other revenue category contains social security, which finances medical expenditure in many parts of the place, it should be associated with public health spending.

42 When the other revenue category is broken into its component parts, there are noticeable effects from social security taxes, which are clearly linked to basic public services; see Web site (fn. 28).

43 The average R-squared for the ten base regressions with infant mortality and life expectancy is approximately 0.8 and the inclusion of the revenue variables raises the adjusted R-squared by as much as 0.05. The inclusion of the revenue variables typically raises the R-squared with social spending by 0.60 or more.

44 The 1995 results are slightly more compelling than the other cross-sections. Part of the reason is that the 1995 cross-section contains more countries from Eastern Europe and fewer from Africa.

45 A one standard deviation increase in revenue from goods and services translates into somewhere between one-tenth and one-quarter of a standard deviation decrease in infant mortality in both the linear and log models, depending on the year and control variables.

46 A one standard deviation increase in revenue from goods and services translates into somewhere between one-tenth and one-fifth of a standard deviation increase in life expectancy, depending on the year and control variables.

47 Other revenue is always positive and typically significant with total transfers.

48 Because of space constraints, I present only the formalism check results; the results using eviction are similar.

49 See Lake, David A. and Baum, Matthew A., “The Invisible Hand of Democracy: Political Control and the Provision of Public Services,” Comparative Political Studies 34 (August 2001).

50 For data, see Evelyne Huber, Charles Ragin, John D. Stephens, David Brady, and Jason Beckfield, Comparative Welfare States Data Set, Northwestern University, University of North Carolina, Duke I University and Indiana University,, 2004.

51 Unless otherwise mentioned, all data come from the World Development Indicators CD-Rom-Bank (fn. 31).

* Special thanks to Gary Cox, David Lake, Peter Gourevitch, Pablo Pinto, Eric Magar, Margaret Levi, Peter Lindert, Neal Beck, and three anonymous referees for their assistance along the way. A previous version of this paper was presented at the annual meeting of the American Political Science Association, Chicago, September 2004. All errors are mine.

Recommend this journal

Email your librarian or administrator to recommend adding this journal to your organisation's collection.

World Politics
  • ISSN: 0043-8871
  • EISSN: 1086-3338
  • URL: /core/journals/world-politics
Please enter your name
Please enter a valid email address
Who would you like to send this to? *


Altmetric attention score

Full text views

Total number of HTML views: 0
Total number of PDF views: 0 *
Loading metrics...

Abstract views

Total abstract views: 0 *
Loading metrics...

* Views captured on Cambridge Core between <date>. This data will be updated every 24 hours.

Usage data cannot currently be displayed