¹ Estévez-Abe, Margarita, Iversen, Torben, and Soskice, David, “Social Protection and the Formation of Skills,” in Hall, Peter and Soskice, David, eds., Varieties of Capitalism (Oxford:Oxford University Press, 2001Google Scholar).

² Boix, Carles, Political Parties, Growth, and Equality: Conservative and Social Democratic Strategies in the World Economy (New York:Cambridge University Press, 1998CrossRefGoogle Scholar).

³ Lowry, Robert, “Governmental Structure, Trustee Selection, and Public University Prices and Spending: Multiple Means to Similar Ends,” American Journal of Political Science 45 (October 2001CrossRefGoogle Scholar).

⁴ This concept of a trilemma resembles that in Iversen and Wren, whose trilemma involves broad macroeconomic outcomes. Iversen, Torben and Wren, Anne, “Equality, Employment, and Budgetary Restraint: The Trilemma of the Service Economy,” World Politics 50 (July 1998Google Scholar). In the social sciences, the theoretical concept of a trilemma dates back to the JVlundel Ts “unholy trinity”of capital openness, autonomous monetary policy, and fixed exchange rates. Mundell, Robert A., International Economics (New York Macmillan, 1968Google Scholar).

⁵ A further assumption underpins the trilemma: that the quality of the public good remains constant (that is, that per person spending remains constant). I alter this assumption in the next section.

⁶ One other configuration is possible: an inexpensive, elite, and partially private higher education system. Such a system does not exist in any fully industrialized state but is fairly common in low-income states. However, most middle-income countries, for example Mexico, Brazil, and Turkey, have highly subsidized higher education along the lines of the elite model, as did high-income states in 1950.

⁷ Gross enrollment rates include multiple entries into higher education by the same individual and thus are a more appropriate index of overall budgetary strain on the government than net rates.

⁸ This data is taken from Education at a Glance 2005 (Paris:OECD, 2005Google Scholar).

⁹ The section entitled “Testing the Partisan Theory of Higher Education,” employs dynamic data on public spending on tertiary education and enrollment levels that enable us to separate out the sequencing and the causal impact of variables. However because of data availability problems, these dynamic tests do not incorporate data on the share of private spending.

¹⁰ As a robustness check, I conducted two-stage least squares (2SLS) regression of private spending on public cost and enrollment, using other government spending and tertiary attainment in the adult population as instruments. While these instruments are not ideal, 2SLS regression produces stronger and more robust results than OLS, assuaging concerns about endogeneity.

¹¹ Using the HC3 technique developed in Long, J. Scott and Ervin, Laurie, “Using Heteroskedasticity Consistent Standard Errors in the Linear Regression Model,” The American Statistician 54 (August 2000Google Scholar).

¹² Trow, Martin A., Problems in the Transitionfrom Elite to Mass Higher Education (Berkeley:Carnegie Commission on Higher Education, 1973Google Scholar).

¹³ See also Nugent, Michael, The Transformation of the Student Career: University Study in Germany, the Netherlands, and Sweden (New York:Routledge, 2004Google Scholar); and Teichler, Ulrich, “Structures of Higher Education Systems in Europe,” in Gellert, Claudius, ed., Higher Education in Europe (London:Jessica Kingsley Publishers, 1993Google Scholar).

¹⁴ Trow, Martin A., “Reflections on the Transition from Elite to Mass to Universal Access,” in Forest, James J. and Altbach, Philip G., eds., International Handbook of Higher Education (Dordrecht:Springer, 2006Google Scholar).

¹⁵ Acemoglu, Daron, “Why Do New Technologies Complement Skills? Directed Technical Change and Wage Inequality,” Quarterly Journal of Economics, 113 (November 1998Google Scholar). See also Goldin, Claudia and Katz, Lawrence, “Long-Run Changes in the U.S. Wage Structure,” (Manuscript, Department of Economics, Harvard University, 2007Google Scholar).

¹⁶ Thus, skilled workers always receive higher income than unskilled workers.

¹⁷ I introduce the wealth parameter in order to distinguish among skilled (unskilled) workers. This could be considered as income from nonlabor market sources such as capital gains or wealth bequests.

¹⁸ I do not distinguish between types of higher education that might produce different returns, e.g., prestigious universities, technical colleges, and degree subjects. Doing so would complicate the model substantially, without clear wage effects. Further, the benefits of higher education are solely private in the model. While there may be social benefits to higher education, most labor economists find higher education has substantially lower social returns than other education. See George Psacharopoulos and Harry Patrinos, “Returns to Investment in Education” Policy Research Working Paper 2881 (Washington, D.C.: World Bank, 2002).

¹⁹ Halsey, A. H., “Trends in Access and Equity in Higher Education,” Oxford Review of Education 19, no. 2 (1993Google Scholar); and Barr, Nicholas, “Higher Education Funding,” Oxford Review of Economic Policy 20, no. 4 (2004Google Scholar). I adjust this assumption in an extension following the basic model.

²⁰ I denote the altered level of period-one subsidization when public cost is held constant p.

²¹ This reverses the Meltzer-Richard model, wherein the poor demand high levels of public spending since they receive public goods but pay lower taxes than the rich. Meltzer, Allan and Richard, Steven, “A Rational Theory of the Size of government,” Journal of Political Economy 89 (October 1981CrossRefGoogle Scholar).

²² The w has s = 2/3, p =1/2; the mass public model has s = 2/3, p = 1, and the elite model has s = l/3, p = l. The precise values of the parameters are chosen because of their resemblance to the empirical patterns in the section entitled “A Trilemma in Higher Education.” The lines are upward sloping since utility increases in income.

²³ A relevant parallel emerges in Iversen and Wren (fn. 4), where service productivity cannot be increased to resolve their trilemma because doing so reduces service quality. The section entitled “Higher Education in England, Sweden, and Germany,” examines the quality decline in English higher education in the 1980s, noting the strong political pressure on Labour to enact reforms.

²⁴ Despite Cusack and Beramendi s well-taken reminder that political economists should pay closer attention to the structure of tax systems, for technical reasons most redistributive models rely on linear tax schedules. See Thomas Cusack and Pablo Beramendi, “Taxing Work,” European Journal of Political Research AS (January 2006); and Peter Lindert, “Why the Welfare State Looks Like a Free Lunch,” Working Paper 9869 (Cambridge, Mass.: National Bureau of Economic Research, 2003). Modeling progressive tax rates requires using quadratic tax schedules with a linear and squared component, which complicate models substantially. The solution I adopt is to retain only the squared component, which de Donder and Hindriks show is the median voter's preferred tax schedule for approximately lognormal income distributions. See de Donder, Philippe and Hindriks, Jean, “The Politics of Progressive Income Taxation with Incentive Effects,” Journal of Public Economics 87 (October 2003CrossRefGoogle Scholar).

²⁵ The squared coefficient of variation is equal to the variance divided by the squared mean and is a popular measure of inequality related to the Theil index. See Sala-i-Martin, Xavier, “The 'Disturbing' Rise of Global Income Inequality,” Working Paper 8904 (Cambridge, Mass.:National Bureau of Economic Research, 2002CrossRefGoogle Scholar).

²⁶ On the impact of income inequality on access, see Halsey (fn.19), and Usher, Alex and Cervenan, Amy, Global Higher Education Rankings 2005 (Toronto:Educational Policy Institute, 2005Google Scholar).

²⁷ Given the similarity between subsidization and quality, this also applies to preferences over quality.

²⁸ Here I compare equation 8b to the left column of equation 3, which refers to moving from elite to mass public systems.

²⁹ It is instructive though, to think through the implications of the median-voter assumption here. Since preferences over higher education spending are not monotonic with respect to income, it is unlikely that parties would converge to the voter with median income if university spending were the only dimension of voting. Instead, a rich-poor alliance could prove numerically dominant and as such, spending would be limited.

³⁰ This setup approximates that used in Iversen, Torben and Soskice, David, “Electoral Institutions, Parties, and the Politics of Class: Why Some Democracies Redistribute More than Others,” American Political Science Review 100 (May 2006CrossRefGoogle Scholar).

³¹ There is no clear partisan pattern for enrollment, however, since both rich and poor groups prefer increasing enrollment under income independence and that cuts across standard partisan cleavages.

³² Gross enrollment data is from http://www.uis.unesco.org/statsen/statistics/indicators/i_pages/ indic_2.htm, and from http://devdata.worldbank.org/edstats/ (accessed July 12, 2007). Tertiary spending in absolute amounts of national currency is from http://www.uis.unesco.org/pagesen/DBEx-pLevel.asp, which also provides information on absolute spending on other levels of education and thus permits the development of a relative spending indicator (accessed July 12,2007).

³³ The average number of observations per state is over twelve, with some states having a maximum of seventeen observations once lags are used.

³⁴ Cusack, Thomas and Engelhardt, Lutz, “The PGL File Collection: File Structures and Procedures,” Data Manual (Berlin:Wissenschaftszentrum Berlin fur Sozialforschung, 2002Google Scholar).

³⁵ Budge, Ian, Klingemann, Hans-Dieter, Vblkens, Andrea, Bara, Judith, and Tanenbaum, Eric, Mapping Policy Preferences: Estimatesfor Parties, Electors, and Governments, 1945—1998 (Oxford:Oxford University Press, 2001Google Scholar).

³⁶ Data on streaming come from the International Association of Universities database on Higher Education Systems at http://www.unesco.org/iau/onlinedatabases/index.html (accessed July 12, 2007). Measures of the proportion of students in vocational streams produce similar results to the streaming dummy.

³⁷ Carey, David and Tchilinguirian, Harry, “Average Effective Tax Rates on Capital, Labor, and Consumption,” Working Paper 258 (Paris:OECD, 2000CrossRefGoogle Scholar). I employ their estimates for labor and net capital taxation for 1980-85, 1986-90, and 1991-97. Results are similar using consumption taxation instead of (or combined with) labor taxation or replacing net capital taxation with gross capital taxation. Tax regressivity is simply the inverse of tax progressivity.

³⁸ Evelyne Huber, Charles Ragin, John D. Stephens, David Brady, and Jason Beckfield, Comparative Welfare States Data Set (Northwestern University, University of North Carolina, Duke University, and Indiana University, 2004).

³⁹ I replicate the regressions in Table 3 using a standardized indicator of absolute tertiary spending where each observation has the country mean subtracted and is divided by the country standard deviation. The results are extremely similar in direction and statistical significance and also produce the same threshold effects as using the relative tertiary spending indicator.

⁴⁰ Using code provided by Thomas Brambor, William Clark, and Golder, Matt, “Understanding Interaction Models: Improving Empirical Analyses,” Political Analysis 14 (Winter 2006Google Scholar).

⁴¹ Models 2 and 6 are also replicated using afixedeffects regression with autocorrelated error terms. The findings for the effects of partisanship and the interactive variable are similar in magnitude and statistical significance to those in Table 2. This holds whether lagged dependent variables are retained or omitted.

⁴² I omit the inequality and regressivity variables, which pertain to the tax structure, since I focus on enrollment and not on spending. Including them does not alter the significance of the partisan effect on enrollment but sharply reduces the number of observations. Breaking the subsample into four groups, defined by inequality and previous enrollment, produces similar results to Table 2, with the partisan pattern robust and stronger in high inequality states, but also produces very small subsamples.

⁴³ Since this variable encompasses both subsidization and quality, and potentially could also include the overall public cost of higher education, it does not directly test the trilemma as did Table 1. The estimates for partisanship are essentially identical if this variable is excluded from the analysis.

⁴⁴ I focus on England, rather than the U.K., since Scotland and Wales have independent university systems.

⁴⁵ Chitty, Clyde, Education Policy in Britain (Basingstoke, U.K.,:Palgrave, 2004Google Scholar)

⁴⁶ Layard, Richard, King, Jonathan, and Moser, Claus, The Impact of Robbins (London:Penguin, 1969Google Scholar).

⁴⁷ Gordon, Peter, Aldrich, Richard, and Dean, Dennis, Education and Policy in England in the Twentieth Century (London:Woburn Press, 1991), 238Google Scholar.

⁴⁸ The Robbins Report (London:British Ministry of Education, 1963Google ScholarPubMed).

⁴⁹ Quoted in Silver, Harold, Higher Education and Opinion Making in Twentieth Century England (London:Woburn Press, 2003), 188Google Scholar.

⁵⁰ Bognador, Vernon, “The Politics of Comprehensive Education,” Oxford Review of Education 3, no. 2 (1977Google Scholar).

⁵¹ Cox, C. B. and Dyson, A. E., The Black Papers on Education (London:Davis-Poynter, 1971Google Scholar).

⁵² Chitty (fn. 45).

⁵³ Lawton, Dennis, The Tory Mind on Education (London:Falmer Press, 1994Google Scholar).

⁵⁴ Williams, Gareth, “The Higher Education Market in the United Kingdom” in Teixeira, Pedro, Jongbloed, Ben, Dill, David, and Amaral, Alberto, eds., Markets in Higher Education: Rhetoric orReality? (Norwell Mass.,:Kluwer, 2004), 248Google Scholar.

⁵⁵ Chitty (fn. 45).

⁵⁶ The transition to tuition fees in other 'early expanders' was also undertaken by center-left parties, e.g., Australia in 1989 and New Zealand in 1990. In both cases, as enrollment has expanded beyond 50 percent, the Labor parties of both states have moved away from fees and toward increased funding.

⁵⁷ Robert Stevens, From University to Vni (London: Politicos, 2004). The opposition of Labour backbenchers may seem surprising given the progressive nature of tuition fees. Some of the opposition may have been a function of ideological inertia—a deep dislike of private money in the public sector. However, there was considerable concern that fees would inhibit enrollment by poorer students. An alternative explanation of this opposition to fees is that as higher education passed the 50 percent threshold, it would be Labour's core constituency that would benefit from public spending on higher education.

⁵⁸ Williams (fn. 54), 257.

⁵⁹ Nugent (fn.13), 29.

⁶⁰ Elzinga, Aant, “Universities, Research and the Transformation of the State in Sweden,” in Rothblatt, Sheldon and Wittrock, Bjorn, eds., The European andAmerican University since 1800 (London:Cambridge University Press, 1993Google Scholar).

⁶¹ Salerno, Carlo, “Higher Education in Sweden,” CHEPS Higher Education Monitor Country Re-fort (October 2002Google Scholar).

⁶² The Swedish Way towards a Learning Society (Swedish Ministry of Education and Science, 1993).

⁶³ Nugent (fn. 13), 38.

⁶⁴ Bauer, Marianne, Askling, Berit, Marton, Susan, and Marton, Ference, Transforming Universities (London:Jessica Kingsley, 1999Google Scholar).

⁶⁵ Economic Survey: Sweden (Paris: OECD, 1993).

⁶⁶ Salerno (fn. 61). The 25-4 plan allowed citizens over twenty-five to enter higher education provided they had completed four years of employment.

⁶⁷ Johansson, Maria, “The Swedish Model for Capital Expenditure in Higher Education from a Lifelong Learning Perspective” (Ph.D. diss., Stockholm University, 2004Google Scholar).

⁶⁸ Reuterberg, S. and Svensson, A., “Financial Aid and Recruitment to Higher Education in Sweden,” Studies in Higher Education 19, no. 1 (1994), 34Google Scholar.

⁶⁹ Fritzell, A., The Current Swedish Model of University Governance: Background and Description (Stockholm:National Agency for Higher Education, 1998Google Scholar)

⁷⁰ Salerno (fn. 61).

⁷¹ Bauer et al. (fn. 64) ,128.

⁷² Marton, Susan, The Mind of the State: The Politics of University Autonomy in Sweden, 1968-1998 (Goteborg:BAS, 2000Google Scholar)

⁷³ Quoted in Nugent (fn. 13), 53.

⁷⁴ Marton(fn. 72).

⁷⁵ The Guardian (London), “Tuition Fees Gain Allure in Cash-hit European Campuses,” October 13,2003.

⁷⁶ Fischer-Appelt, Peter, “The University: Past, Present and Future,” in Muller, Steven, ed., Universities in the Twenty-First Century, vol. 2 (New York:Berghahn, 1996), 5Google Scholar.

⁷⁷ Hufner, Martin, “Governance and Funding of Higher Education in Germany,” Higher Education in Europe, 28 (June 2003CrossRefGoogle Scholar).

⁷⁸ Konow, Gerhard, “Planning, Financing and Accountability of German Universities”Google Scholar in Muller (fn. 76), 50.

⁷⁹ Nugent (fn. 13), 93.

⁸⁰ The association of university presidents, the Westdeutsche Rektorkonferenz, was against building even a single new university, see Nugent (fn. 13), 98.

⁸¹ Onestini, Cesare, Federalism and Lander Autonomy: The Higher Education Policy Network in the Federal Republic of Germany, 1948-1998 (New York:Routledge, 2002Google Scholar).

⁸² From 1995, students were charged the regular interest rate. Lingens, Hans, German Higher Education: Issues and Challenges (Bloomington, Ind.:Phi Delta Kappa International Studies in Education, 1998), 47Google Scholar.

⁸³ Hufner (fn. 77).

⁸⁴ Ostermann, Hanna, “The Higher Education Debate in Germany,” German Politics 11, no. 1 (2002CrossRefGoogle Scholar).

⁸⁵ Nugent (fn. 13), 107.

⁸⁶ Hufner (fn. 77).

⁸⁷ Windholf, Paul, Expansion and Structural Change: Higher Education in Germany, the United States, andJapan, 1870-1990 (London:Westview Press, 1997), 191Google Scholar.

⁸⁸ For example, the head of the OECD education division. Andreas Schleicher, “The Economics of Knowledge: Why Education is Key for Europe's Success,” Lisbon Council Policy Brief (2006).

⁸⁹ Fuhr, Christoph, The German Education System Since 1945 (Bonn:Inter Nationes, 1997), 183Google Scholar.

⁹⁰ Financial Times (London) “Germany's Corridors of Learning Start to Chart Independent Path,” October 25,2006.