¹ Schiavo-Campo, Salvatore, de Tommaso, Giulio, and Mukherjee, Amitabha, An International Statistical Survey of Government Employment and Pay (Washington, D.C.: World Bank, 1997)Google Scholar.

² Alesina, Alberto, Danninger, Stephan, and Rostagno, Massimo, “Redistribution through Public Employment: The Case of Italy,” National Bureau of Economic Research, Working Paper 7387 (1999).Google Scholar

³ Alesina, Alberto, Baqir, Reza, and Easterly, William, “Redistributive Public Employment,” National Bureau of Economic Research, Working Paper 6746 (1998), 10Google Scholar.

⁴ Wagner, Adolf, Finanzwissenschaft, 3d ed. (Leipzig: C. F. Winter, 1883)Google Scholar, excerpted in Musgrave, Richard A. and Peacock, Alan T., Classics in the Theory of Public Finance (London: MacMillan, 1958)CrossRefGoogle Scholar.

⁵ See Schiavo-Campo, Salvatore, de Tommaso, Giulio, and Mukherjee, Amitabha, Government Employment and Pay: A Global and Regional Perspective (Washington, D.C.: World Bank, 1997), viiiGoogle Scholar; the authors find a close and significant positive relationship between the share of government employment and national income. See also Heller, Peter S. and Tait, Alan A., Government Employment and Pay: Some International Comparisons (Washington, D.C.: IMF, 1984)Google Scholar. They do not observe a clear association between income and public employment among the OECD countries taken separately (in this small-N estimation of data for the early 1990s). Using a different dataset on public employment in 1963–83, however, Thomas Cusack, Ton Notermans, and Martin Rein do; see Cusack, , Notermans, , and Rein, , “Political-Economic Aspects of Public Employment,” European Journal of Political Research 17 (July 1989)CrossRefGoogle Scholar. Aart Kraay and Caroline Van Rijckeghem find that government employment increases with urbanization and education levels; see Kraay, and Rijckeghem, Van, “Employment and Wages in the Public Sector: A Cross Country Study,” IMF Working Paper (Washington, D.C.: IMF, 1995)Google Scholar.

⁶ Alesina, Baqir, and Easterly (fn. 3).

⁷ Cameron, David, “The Expansion of the Public Economy: A Comparative Analysis,” American Political Science Review 72, no. 4 (1978)CrossRefGoogle Scholar.

⁸ Rodrik, Dani, “What Drives Public Employment?” National Bureau of Economic Research, Working Paper 6141 (August 1997)Google Scholar.

⁹ Much of the literature on the growth of the state has focused on increases in public spending. Many of the same issues arise in explaining growth of public employment, but some additional issues are unique to the latter and require separate consideration. This article focuses in particular on ways that the strike potential and voting of public employees (and the regional populations they serve) affect the fiscal games played by central and subnational governments.

¹⁰ Meltzer, Allan H. and Richard, Scott F., “A Rational Theory of the Size of Government” Journal of Political Economy 89 (October 1981)CrossRefGoogle Scholar.

¹¹ Sanchita Saxena reports average levels of public sector employment per thousand inhabitants for Argentina in 1996 of twenty-seven for the five high-income provinces, forty-nine for the fourteen middle-income provinces, and seventy-four for the five poorest provinces; see Saxena, , “Variations in Reform among the Sub-National Units in Argentina and India” (Manuscript, Department of Political Science, UCLA, 2001)Google Scholar.

¹² Alesina, Baqir, and Easterly (fn. 3).

¹³ Schiavo-Campo, de Tommaso, and Mukherjee (fn. 5).

¹⁴ In Argentina provincial governments in the 1990s “issued IOU's in lieu of paying wages”; Schwartz, Gerd and Liuksila, Claire, “Argentina,” in Ter-Minassian, Theresa, ed., Fiscal Federalism in Theory and Practice (Washington, D.C.: IMF, 1997)Google Scholar; see also Saiegh, Sebastian and Tommasi, Mariano, “The Dark Side of Federalism” (Manuscript, Department of Politics, New York University, New York, 2001)Google Scholar. The International Monetary Fund reports that Brazilian states accumulated payroll arrears in the 1990s; IMF, Brazil: Selected Issues and Statistical Appendix (Washington, D.C.: IMF, 2001), 156Google Scholar.

¹⁵ Public sector wage arrears also decrease the reservation wage in the economy and set a new social norm, which induces the private sector to follow suit. Ultimately, the practice of not paying the contracted wage may spread to the whole economy; see Earle, John and Sabirianova, Klara, “Equilibrium Wage Arrears: A Theoretical and Empirical Analysis of Institutional Lock-In,” IZA DP no.196 (Bonn: IZA, 2000)Google Scholar. All this increases the fiscal and political pressures on the central government.

¹⁶ It would be more realistic to assume that they seek to maximize the probability of reelection, but the results are likely to be similar, and assuming vote-share maximization avoids considerable additional technical complexity.

¹⁷ In Russia many governors were initially appointed in 1991, but elections had spread gradually to all regions by 1996–97. In all, there was an expectation from the start that leaders would soon need to face elections.

¹⁸ Of course, we abstract here from the often complicated institutional settings in which policy is made at both subnational and central levels. Elaborating the model in a richer institutional context is left for future work.

¹⁹ Not surprisingly, the governor will fire all partisans of his rival and set m_ri = 0, as we prove below. From this, we derive the prediction of a temporary drop in public employment when the leadership turns over.

²⁰ In the model the central government moves first. All that is required for this to be realistic is that regional governments are able to adjust their levels of employment after they know how large their central transfers will be. It seems to us a reasonable assumption that governments can at least adjust the public workforce upward—hiring new workers—relatively quickly. We believe it would be possible— and probably easier—to derive predictions similar to ours if the model were complicated to permit the central government to increase transfers after public employment levels are set (this would create an additional element of moral hazard that regional governments could exploit). However, incorporating this additional stage in the game would render the algebra far more complex while yielding essentially similar insights.

²¹ We assume that workers have a threshold level of arrears at which they will be willing to strike, and most workers' thresholds are clustered together, leading to an explosion of strike readiness as arrears reach some intermediate range. No one will strike over a one-hour or a one-day wage delay, almost no one will strike over a one-week delay, but as arrears get above two or three months, the number willing to strike increases rapidly. The increase in strike risk as arrears rise will also be faster above a certain point because of the spreading knowledge of workers that their own wage experience is not isolated or accidental.

²² This implies that there will be some positive level of arrears.

²³ Square brackets denote the function operator; round brackets are for multiplication.

²⁴ We use asterisks to represent equilibrium values.

²⁵ Note that our theory does not imply that regional governors actually want strikes to break out. On the contrary, they suffer a loss of utility in the event of strike equal to S − σ_i > 0. The point is that they deliberately create a risk of strike, knowing that this will elicit additional transfers from the center. Although they choose to gamble in this way, they would prefer to avoid an actual strike.

²⁶ See, e.g., Rose, Richard, ed., Public Employment in Western Nations (New York: Cambridge University Press, 1985)Google Scholar; and Schiavo-Campo, de Tommaso, and Mukherjee (fn. 1).

²⁷ Schiavo-Campo, de Tommaso, and Mukherjee (fn. 1), 47.

²⁸ “Social protection” includes organizations providing social assistance to the disabled, elderly, large families, orphans, and so on. “Public administration” covers employees in government administration, as well as various police forces (mostly regional and local police, traffic police) and firefighters.

²⁹ See, for example, Schiavo-Campo, de Tommaso, and Mukherjee (fn. 5); Alesina, Baqir, and Easterly (fn. 3).

³⁰ Figures for education, social protection, and health and sport are from Freinkman, Lev, Treisman, Daniel, and Titov, Stepan, SubnationalBudgeting in Russia: Preempting a Potential Crisis (Washington, D.C.: World Bank, 1999)Google Scholar, Table A10. The 1992 figure for administration is calculated from figures in Sinelnikov, Sergei, Byudzhetny Krizis v Rossii: 1985–1995 g. (Moscow: Yevrasia, 1995)Google Scholar, Table 5.1; the figures given are 42 billion and 64 billion rubles for state administration spending in the federal and subnational budgets, respectively. The 1996 figure is from Rossii, Goskomstat, Rossiisky statistichesky yezhegodnik 1999 (Moscow: Goskomstat Rossii, 1999), 492Google Scholar.

³¹ As of 1997, only 0.8 percent of general education schools were in the nonstate sector (more than one-third of them in Moscow), and only 0.2 percent of all school students attended them; Glavatskaya, N., Moldavsky, A., and Lopatmkov, L., eds., Reformirovanie Nekotorykh Otraslei Sotsialnoi Sphery Rossii (Reform of several branchs of the Russian social sector) (Moscow: IEPP, 1999), 150Google Scholar.

³² Goskomstat Rossii (fn. 30), 187.

³³ We have adjusted Goskomstat figures to correct for a change in classification in 1996 of firefighters as public administration employees. For lack of a better alternative, we increased the public administration figures for previous years by an amount equal to the firefighter total as of 1996. If the number of firefighters increased in 1992–95, as in other branches of public administration, our figure for employment growth will be an underestimate.

³⁴ Cheasty, A. and Davis, J., “Fiscal Transition in Countries of the Former Soviet Union: An Interim Assessment,” MOCT-MOST 6, no. 3 (1976)Google Scholar.

³⁵ In Russia in 1996 there was 1 doctor for every 219 inhabitants; Goskomstat Rossii (fn. 30), 588. In the U.K. the figure was 610 (in 1993), in Germany 286 (in 1998), in France 330 (in 1997), in the U.S. 358 (in 1995), and in Japan 518 (in 1996); figures from World Health Organization statistical database (www-nt.who.int/whosis/statistics; accessed June 2001).

³⁶ Russian figure from Goskomstat Rossii, Rossiisky Statistichesky Yezhegodnik 1998 (Moscow: Goskomstat Rossii, 1998)Google Scholar; data for other countries are calculated from figures provided by UNESCO Institute for Statistics (http://unescostat.unesco.org/), Canada and U.S. 1995, U.K., and Germany, 1996.

³⁷ A recent ILO study summarizing trends in public employment around the world concludes: “In general, where there has been a decrease in current total employment compared to the 1990 level, the decrease in public employment has been more marked; and where there has been an increase in total employment the increase in public sector employment has been less marked. It is the countries in transition which have registered the steepest declines in public sector employment”; Hammouya, M., “Statistics on Public Sector Employment: Methodology, Structures and Trends,” Working Paper no. 144 (Geneva: ILO, 2000), 23Google Scholar. In Poland the absolute levels remained stable, but against the background of an increase in total employment.

³⁸ Goskomstat Rossii (fn. 36).

³⁹ Rossii, Goskomstat, Trud i zanyatost v Rossii 1999 (Moscow: Goskomstat Rossii, 1999), 96Google Scholar.

⁴⁰ Ibid., 98. It is noteworthy that science, the one subsector of public employment that experienced a sharp drop during the 1990s, was the one that remained almost entirely federal.

⁴¹ This represents the share of total employment of education, culture, art; health care and sport; social protection; and public administration. Ingushetia and Chechnya represent abnormal cases, since the ongoing hostilities may have resulted in a larger federal state presence. We do not include them in our regression analysis in any case, because of lack of data from earlier years.

⁴² Geddes, Barbara, Politkan's Dilemma: Building State Capacity in Latin America (Berkeley: University of California Press, 1994)Google Scholar.

⁴³ For more details on the rationales and practice of transfer programs, see Daniel, Ttasman, After the Deluge: Regional Crises and Political Consolidation in Russia (Ann Arbor: University of Michigan Press, 1999)Google Scholar. A main channel of transfers was the Fund for the Financial Support of Regions, which was supposed to aid “needy” and “especially needy” regions. In practice, addressing social need appears to have become a more important motive in determining the fund allocations as the decade progressed.

⁴⁴ Panel-corrected standard errors are more accurate than standard errors computed by the Parks FGLS method for data of the kind analyzed in this paper; Beck, Nathaniel and Katz, Jonathan N., “What to Do (and Not to Do) with Time-Series-Cross-Section Data,” American Political Science Review 89 (September 1995)CrossRefGoogle Scholar. In STATA, we used the operation xtpcse.

⁴⁵ Beck, Nathaniel and Katz, Jonathan N., “Nuisance vs. Substance: Specifying and Estimating Time-Series-Cross-Section Models,” Political Analysis 6 (1996)CrossRefGoogle Scholar.

⁴⁶ These groupings were necessitated by data availability.

⁴⁷ This should not be viewed as a test of Wagner's Law per se, since the correlation between these two variables might reduce the estimated significance of each. And we also include regional revenues in the regressions, which should correlate with per capita GRP. The source of these data is Goskomstat. To minimize problems of endogeneity, we use data from as close to the beginning of the period as possible. The urbanization rate is for 1992. GRP was available only from 1994, so we use figures for that year.

⁴⁸ Alesina, Baqir, and Easterly (fn. 3).

⁴⁹ In including the age structure controls, we follow previous studies such as Alesina Baqir, and Easterly (fn. 3).

⁵⁰ Freinkman, Treisman, and Titov (fn. 30).

⁵¹ Ibid.

⁵² Ibid. The median varied between 16 and 25 percent in the years 1992–97.

⁵³ Transfers and loans might be endogenous to public employment. We discuss this possibility below.

⁵⁴ These included McFaul, Michael and Petrov, Nikolai, eds., Politicheskiy Almanakh Rossii 1997 (Moscow: Moscow Carnegie Center, 1998)Google Scholar; McFaul, Michael, Petrov, Nikolai, and Ryabov, Andrei, Rossia nakanune Dumskikh Vyborov 1999 goda (Moscow: Moscow Carnegie Center, 1999)Google Scholar; Institute for East-West Studies, Handbook of Regional Executives (New York: IEWS, 1999)Google Scholar.

⁵⁵ Anecdotal evidence supports the finding that newly elected governors are prone to cut public employment. In July 2001 the newly elected governor of Primorsky Krai, Sergey Darkin, announced a proposed reduction of 15–20 percent in regional administration personnel; Rosbusinessconsultingvieb site (www.rbc.ru, accessed July 13, 2001)Google Scholar.

⁵⁶ We treat the additional standard deviation in transfers as a one-year increase.

⁵⁷ Lev Freinkman and Michael Haney found that greater federal transfers were also associated with larger regional subsidies to inefficient enterprises and housing; see Freinkman, and Haney, , “What Affects the Propensity to Subsidize: Determinants of Budget Subsidies and Transfers Financed by the Russian Regional Governments in 1992–1995,” Policy Research Working Paper 1818 (Washington, D.C.: World Bank, 1997)Google Scholar.

⁵⁸ The regressions were of the form: (1) pe_t, = a + b₁pe_{t –1} + b₂trans_{t –1} + b₃d, and (2) trans_t = a + b₁pe_{t –1} +b₂trans_{t –1} + b₃d, where trans ≡ federal transfers and loans; d is a vector of year dummies; and b₃ is a vector of coefficients on the year dummies.

⁵⁹ This suggests that current-year changes are particularly important; these correlate negatively with lagged transfers and loans, which explains the insignificant result when the change variable is omitted.

⁶⁰ In fact, not only was public employment highly insignificant in a regression to explain transfers and loans, but its sign was also negative.

⁶¹ This may be due in part to the fact that little is often required of them, and they can work privately on the side while drawing a public salary.

⁶² Goskomstat Rossii (fn. 39), 345.

⁶³ The health care sector also has a very high propensity for collective protest, which is partly neutralized by strong moral pressures on medical personnel not to strike and public outrage when fatalities can be linked to strikes. Nevertheless, cases of collective action such as strikes or hunger strikes of unpaid doctors abound.

⁶⁴ Gimpelson, Vladimir, “Politics of Labor Market Adjustment: The Case of Russia,” in Kornai, Janos, Haggard, Stephan, and Kaufman, Robert, Reforming the State: Fiscal and Welfare Reform in Post- Socialist Countries (New York: Cambridge University Press, 2001)Google Scholar.

⁶⁵ Treisman (fn. 43), chap. 4.

⁶⁶ Byuleten Obshchestvennogo Monitoringa, no.2 (Moscow: VCIOM, 1996), 88Google Scholar.

⁶⁷ Institute for East-West Studies, Russian Regional Report 5 (February 24, 2000)Google Scholar.

⁶⁸ Glavatskaya, Moldavsky, and Lopatnikov (fn. 31), 176.

⁶⁹ See, for example, “Uchitelya poterpyat,” Ekspert, May 15, 2000, 14Google Scholar; for a report on the massive misuse of such transfers, see Segodnya, March 13, 1997.

⁷⁰ “You Are Talented, so Find Money!” Kommersant, December 17, 1999Google Scholar.

⁷¹ Radio Free Europe/Radio Liberty, RFE/RL Newsline 2, no. 2361, pt. I, December 9, 1998Google Scholar.

⁷² Rosbusinessconsulting (www.rbc.ru, accessed February 3, 2000)Google Scholar.

⁷³ For another argument that many of the roots of current problems in Russia's political economy lie in the poor state of center-region relations, see OECD, Economic Surveys: Russian Federation (Paris: OECD, 2000)Google Scholar.

⁷⁴ For Russian figures, see Goskomstat Rossii (fn. 30), 213,588; for U.S. and U.K. figures, see World Health Organization (fn. 35).

⁷⁵ This is based on the increase in employment in all regions except Moscow.

⁷⁶ Russian Economic Trends 8, no. 4 (1999), 109Google Scholar.

⁷⁷ If the central politician is confident that he has other stable bases of electoral support, he may be more willing to risk the consequences of regional strikes. If he has other leverage over governors, he may be able to intimidate them into quiescence. The familiar fiscal games may well reemerge, however, as a new election approaches or if the president's popularity slips significantly. Interlevel fiscal games are also likely to be less common in periods of economic growth, when private sector opportunities attract public sector workers who are not well rewarded. Such opportunities may, however, attract skilled and diligent workers while leaving patronage recipients in place, so the quality of public services may not improve with reductions in overstaffing.

⁷⁸ We are grateful to Robert Kaufman and Joan Nelson for calling our attention to similarity between Russia and large Latin American countries in this regard.

⁷⁹ Saxena(fn. 11).

⁸⁰ Ibid.

⁸¹ Tommasi, Mariano, Saiegh, Sebastian, and Sanguinetti, Pablo, “Fiscal Federalism in Argentina: Policies, Politics, and Institutional Reform,” Economia 1 (Spring 2001)Google Scholar.

⁸² New York Times, November 18, 2001, A3Google Scholar.

⁸³ International Monetary Fund (fn. 14), 156.

⁸⁴ Rodrik(fn. 8).

⁸⁵ Schiavo-Campo, deTommaso, and Mukherjee (fn. 5).