Econ. Gov. (2004) 5: 53–75 DOI: 10.1007/s10101-002-0052-0

c Springer-Verlag 2004


State competition in higher education: A race to the top, or a race to the bottom? Michael A. Bailey, Mark Carl Rom, Matthew M. Taylor Department of Government, Georgetown University, ICC, Suite 681, 37th & O Sts., N.W., Washington, D.C. 20057, USA (e-mail: [email protected]) Received: August 2001 / Accepted: May 2002

Abstract. How does competition affect higher education? This paper explores this question for public and private universities. Theory indicates that competition can push higher education policy in one of two different directions. On the one hand, competition may increase spending. For states, this would occur if states treat higher education as “developmental;” for private universities this would occur if they view spending as a means to attract students and prestige. On the other hand, competition may decrease spending if states treat higher education spending as “redistributive,” and competition may decrease spending by private schools if lower spending enhances their ability to attract students with low tuition. To determine which of these perspectives is most valid, we examine higher education policy choices in the 1980s and 1990s. We find that states appear to act as if higher education funding is “redistributive” while private schools appear to compete more on the basis of tuition than spending. These results demonstrate the important effects competition and governance structure have on higher education. Key words: Education, Educational Finance, State and Local Budget and Expenditures, Welfare and Poverty JEL Classification Numbers: I2, I22, H72, I3

54

M.A. Bailey et al.

“Education is an investment. The trouble is, they [universities] don’t run it like an investment over there . . . they run it like welfare . . . “ Governor O.T. Early in Moo (Smiley 1995)

Introduction

Educational policy is a highly salient national political issue. Former President George H. W. Bush wanted to be known as the “Education President.” One of President Bill Clinton’s early accomplishments was the Schools to Work Opportunity Act, which sought to increase student achievement and prepare young people for post-secondary education and careers (Hughes et al. 2001). President George W. Bush, like his father before him, made educational reform a centerpiece of his campaign, pledging to establish national standards and to expand school choice. Nevertheless, education policy continues mainly to be a state and local responsibility. At the end of the 1990s, the federal government paid for less than ten percent of all public primary and secondary education expenses. (Given the large size of the private school sector, the overall contribution of the national government to K-12 education was substantially less than ten percent). The federal government plays an even smaller financial role in postsecondary education, with the vast majority of college and university funding coming from either private sources or state treasuries. An interesting aspect of the dispersed nature of education policymaking is the extent to which competition affects higher education policy. The states operate, after all, within a federal system that allows and even encourages the states to compete with each other for capital and labor. How does such interstate competition affect public involvement in higher education? Similarly, private schools compete intensely for students and prestige. How does this competition affect their policies? This paper investigates these questions for state higher education policy choice in the 1980s and 1990s. We build from the empirical structure Bailey and Rom (2001) use to test for evidence of state competition on welfare policy. Many empirical tests for state competition look for evidence that policies in neighboring states affect a state’s policies. But this approach does not adequately control for non-competitive sources of covariation in state policies. Therefore, we specify additional conditions that are consistent with the various forms of state competition and test for those conditions. To determine whether broad governance structure – public or private – affects this aspect of policymaking, we also assess whether competition over public education takes the same form, and proceeds in the same direction, as private education competition. This paper proceeds as follows. Section 1 examines the literature on state policy choices and higher education policymaking. Section 2 develops hypotheses about state competition in higher education. Section 3 discusses data and Sect. 4 presents results and discusses implications. We conclude in Section 5.

State competition in higher education: A race to the top, or a race to the bottom?

55

1. State policy choices and higher education This section summarizes four literatures which are relevant to understanding state higher education policy choices, particularly as they inform this project. Educational policy choice The states make many policy choices for higher education, from admissions criteria to zoology faculty. An extensive literature discusses the merits and demerits of various policies, and an even more expansive body of opinion resides in the minds of students, parents, staff, faculty and administrators. There is, however, a paucity of empirical research on how and why states choose their higher education policies. Mintrom (1997a; also Mintrom and Vergari 1996, 1997) shows “how policy entrepreneurs have helped stimulate the diffusion of state school-choice plans by collecting and generating evidence of the workability of this innovation” (Mintrom 1997a: 41). Hearn et al. (1996) investigate the tuition pricing and discounting policies of the states; in a cross-sectional analysis the authors find that region, social and economic resources and, to a much lesser extent, post-secondary governance arrangements, each had a policy impact (see also Johnstone 1992). Hearn and Griswold (1994) find that higher education governance structures have a modest impact on policy innovation. McLendon et al. (2001) also find that regional patterns of policy choice exist for post-secondary financing and regulatory innovations. The most powerful research has been conducted by Lowry (2000, 2001). His investigation focuses on the idea that political control over universities will lead to lower tuitions and broader enrollments (policies favored by elected politicians); in contrast, university control over policy will lead to higher tuitions and smaller enrollments (policies favored by faculty and university administrators). Lowry’s cross-sectional analyses find that the governance structures of public universities influence effective tuition rates and enrollments in the predicted ways. Tuition rates are lower, and enrollment rates are higher, in states with a single, regulatory, coordinating board than in states with multiple governing boards. Knowing the effects of governance structures does not, however, tell us how competition affects higher education policies. Policy diffusion The diffusion literature focuses on the likelihood that states will adopt the policy innovations implemented by other states. The conventional view is that states are more likely to adopt some innovation if their neighbors have previously adopted it (Walker 1969; Gray 1973; Berry and Berry 1990, 1999). This literature has been used to examine the willingness of the states to adopt new policies and programs for education (McLendon et al. 2001; Hearn and Griswold 1994; Hearn et al. 1996; Mintrom 1997b; Mintrom and Vergari 1996, 1997). These studies have typically found evidence of regional effects, with a state being more likely to adopt a policy if its geographical neighbors have already adopted it. McLendon et al. (2001), for

56

M.A. Bailey et al.

example, find that an increase in the number of innovations in contiguous states makes it more likely that a state will adopt a regulatory or financial innovation. As in much of the diffusion literature, they find regional effects to be important even after controlling for relevant political, economic and demographic factors. Mooney (2001) rejects the view that diffusion is only regional and unidirectional, and instead demonstrates that there is no single, or simple, story explaining diffusion. He argues that a state may be more likely to adopt a neighbor’s policy if that policy is seen as a success, but a state may be less likely to adopt the policy if it is seen as a failure. Mooney also contends that states may look to their immediate neighbors if policy learning involves local networks, or that states may look to national leaders if the policy network is nationalized. In short, policy diffusion can take more than one path. The diffusion literature is most useful in illuminating the adoption of new policies (e.g., charter schools, lottery-funded scholarships, etc.) by the states. It is less helpful in explaining how states adjust the parameters of existing policies (e.g., tax rates, spending allocations, benefit levels, and program eligibility). Without denying the importance of innovation and emulation, much educational policy consists of adjusting policy parameters through manipulating prices (salaries, tuitions, scholarships) or quantities (enrollments, admission standards, for example.) Because every state has a postsecondary education system, all states, innovative or not, must concern themselves with these parameters. State choices about policy parameters are less about innovation and more about adjusting the parameters to meet the state’s political and policy preferences and constraints.

Policy learning The policy learning literature addresses the questions: From whom do the states learn, and how do they learn from them? Walker (1969) assumed that state officials “make most of their decisions by analogy,” (p. 889) taking cues from states that seem to have successfully addressed the policy problem. The assumption that states learn from their immediate neighbors through a word-of-mouth process remains the most common one (Mooney 2001, p. 104; for a comprehensive review of the social learning literature, see Rogers, 1995). Other scholars suggest that states learn about policy innovations and options from the federal government, policy entrepreneurs, and professional associations (see especially the discussions in Balla 2001; Mooney 2001). While the literature above deals with learning by policymakers, “yardstick competition” theory is an off-shoot that deals with learning by voters, and how it influences politicians (Besley and Case 1995). Two main premises underlie such theory. First, voters lack information about the quality of state leaders. Voters may be told that the policies being implemented are the best possible under the circumstances, but they lack the ability to directly assess these claims. Second, voters do have information about the cost and effectiveness of state policies relative to neighbors, however. Through the media and through contact with people from neighboring states, voters can get a sense of differences and similarities across state lines.

State competition in higher education: A race to the top, or a race to the bottom?

57

The prediction from yardstick competition theory is that states will be sensitive to neighboring states’ policies. For example, a legislator from Arkansas who claims that rising costs are behind tuition hikes risks looking bad if voters sense that tuition rates have declined in Missouri. Even though a typical Arkansas voter may not investigate the determinants of tuition rates, he or she may reasonably ask: If Missouri can lower tuition, why can’t we? Regarding education policy, the voters might be especially sensitive to the idea that their state is not “keeping up” with the neighbors. In yardstick competition theory, state policymakers have an incentive to respond in two ways to the policies of neighboring states when setting policy. First, state policymakers are motivated to follow their neighbors regarding popular policies. When neighboring states offer lower tuition, or more spending, voters will be more likely to believe that their state should also be able to cut tuition or increase spending. Second, given the importance politicians place on economic development, state policymakers might be particularly sensitive to the possibility that they are perceived as policy laggards. The political costs of not “keeping up with the Joneses” might be higher than the risks of being too far out in front of them. This second element is crucial if state competition is to lead to a “race to the top” (RTT) on certain developmental policies. If it is politically damaging to be seen as a straggler compared to one’s neighbors, one would expect that states would be motivated to engage in an educational policy race to the top.

Policy competition The state competition literature can provide more insight into how states set and adjust these policy parameters. One set of insights is optimistic: Tiebout (1956) indicates that competition leads citizens to “vote with their feet” and to locate within the jurisdiction that best matches their policy preferences. States, to retain their citizens, have incentives to adjust their policies to meet the preferences of the citizens. For Tiebout, policy competition improves the efficiency of government and better matches the supply of government services to the demand for them. Peterson (1995), in contrast, emphasizes that states compete in different policy domains and that this competition leads to outcomes that differ across these domains. The key distinction, for Peterson, is between redistributive and developmental policies. As Peterson defines them, Redistributive programs reallocate societal resources from the “haves” to the “have-nots.” They transfer economic resources away from those who have gained the most from economic development to those who have gained the least . . . Developmental policies provide the physical and social infrastructure necessary to facilitate a country’s economic growth . . . The social infrastructure includes institutions that . . . educate the next generation (Peterson 1995 p. 17). The core of Peterson’s argument is that states will avoid redistributive policies and seek developmental ones. He argues that “To avoid becoming a magnet for

58

M.A. Bailey et al.

the poor, the sick, and the needy, states must be cautious that they do not design a system of welfare considerably more lavish than that of the other states” (Peterson 1995, p. 30). States thus have incentives to be less generous than their peers so that they will neither attract nor retain the poor, and so they will not need to set taxes at levels that will be resisted by the more affluent. Competitive pressures induce states to lower their redistributive generosity in order to avoid becoming “welfare magnets” (Peterson and Rom 1990). As a result, the states may engage in a “race to the bottom” (RTB).1 There is evidence that state control over redistribution leads states to be less generous in welfare programs than they would otherwise be (Bailey and Rom 2001). The reader might immediately ask the question: is education policy developmental or is it redistributive?2 This is not easy. On the one hand, public education, at least at the K-12 level, does indeed transfer resources from the more affluent to the less so. On the other hand, though, education serves developmental purposes. In fact, Peterson clearly views education as developmental: “[C]lassifying education as primarily redistributive ignores the fact that investments in education have been routinely shown to be among the best predictors of national economic growth and productivity” (Peterson 1995, p. 65). The developmental nature of education may even be stronger for universities. At least some governments have long sought to use colleges and universities as a tool to increase productive capacity.3 Consider the situation in Arkansas, which is ranked 50th in the nation in average household income and 49th in the number of college degrees per capita (Caillouet 2001, 3A). Arkansas officials estimated that if the state’s college graduation rates were at the national average, the state’s economy would be one third larger (growing from $62 billion to $83 billion). As a result, the state’s Higher Education Department teamed up with the Department of Economic Development to call for an additional $300 million in state spending on higher education (Caillouet 2001, 1A). It has also been demonstrated, moreover, that higher education benefits do not flow disproportionately to the poor; if higher education redistributes resources, it does so mainly to the more affluent (see, for example, Heller 2002). However, higher education may be considered redistributive if it is perceived as being unproductive, and thereby redistributing income from taxpayers to certain classes of state residents. As such, excessive spending on higher education may be viewed as a drag on the state budget that takes money from other needs or causes taxes to be raised. Peterson is less explicit about the impact of state competition on developmental policies or the direction the competition leads. To the extent that states seek to grow

1 Rom (1989) argues that the RTB is neither a race nor does it finish at the bottom. Still, the RTB is a useful shorthand for indicating that competition reduces redistributive program generosity. There is a substantial literature on the possible RTB in welfare policy (Schram and Beer 1999). 2 Rom (1989) contends that most policies have both redistributive and developmental elements. 3 During the 19th century, the Laander in Germany competed openly among themselves (and with France) to attract the best scholars and build the best labs (Rocke 2000). These are the first cases (so far as we know) of governments using universities as a tool for economic development.

State competition in higher education: A race to the top, or a race to the bottom?

59

economically and to compete for scarce resources to promote growth, it would seem that states have incentives to choose policies that will attract capital and labor. To understand how states may compete in developmental policy, it is useful to distinguish between two economic development strategies (Eisinger 1989). Eisinger distinguishes between “supply side” developmental policies (such as tax reductions, deregulation, and loan guarantees) that attempt to lower the prices of goods, and “demand side” policies (such as state research and development grants, technical assistance to small businesses, or grants to increase business innovation) that seek to boost capacity. Supply side policies typically attempt to reduce costs; demand side policies generally seek to increase output. The states can apply these policies towards existing assets (capital and labor) or to bring new assets into the state. Some supply side policies may be susceptible to a variant of the RTB. To make the costs of doing business in a state lower than among its peer states, states might choose to lower taxes, reduce environmental standards, or reduce labor costs, among other cost-reducing policies (Fox 1996; Donahue 1997). Alternately, states could invest in demand side policies by increasing expenditures for infrastructure or research and development. To gain competitive advantage over peer states, the states may engage in a RTT in demand side policies, with each state seeking to invest more than its competitors. A note on terminology is necessary here. The phrases RTB and RTT can have both positive and normative implications. Normatively, the term RTB is often used to designate movement away from an observer’s preferred policy outcome, and RTT implies progress towards those preferences. Hence, those who prefer more generous welfare programs worry that state control over redistribution will lead to a RTB; those who want smaller welfare programs, however, might portray these same consequences as a RTT. In this paper we do not use RTB and RTT normatively (we leave that choice to our readers). Instead, we define competitive outcomes that reduce state spending, diminish access to governmental services, or otherwise lessen state generosity as a RTB. Competitive moves that increase spending, broaden access, or enhance generosity we label RTT.4 Higher education generally involves demand side policies, so to the extent that they treat it as developmental, states may have incentives to RTT on the various policy parameters. States may seek to charge lower tuition rates than their competitors by offering greater subsidies, so that their own bright students remain within the state and students from other states are encouraged to transfer there.5 States may also offer faculty salaries that are higher than in other states, as a way of recruiting the top teachers or researchers. States may also invest in higher education in order to 4 In these terms, policies that have similar goals might nonetheless be alternately considered RTB or RTT. Consider the possibility that states compete to attract professional sports franchises. The states may do so by promising lower tax rates (leading to a RTB) or by pledging to provide infrastructure (implying a RTT). 5 One should distinguish between cutting tuition rates as a demand side policy and cutting taxes as a supply side policy. Lower tuition rates imply that a state is doing more; if total spending on higher education remains the same, lower tuition rates imply greater state financial support for higher education (as state spending must increase to compensate for less tuition revenue). Lower tax rates simply imply that the state is doing less.

60

M.A. Bailey et al.

attract business investment, given that businesses presumably need a well-educated workforce. In summary, the educational policy choice literature and the policy innovation literatures pay relatively little attention to competition. The policy learning literature and policy competition literatures pay more attention to competition and both indicate that if voters and policymakers view educational policy as “developmental”, they will be more sensitive to falling behind their neighbors, while if they view educational policy as “redistributive” or somehow wasteful, they will be more sensitive to spending more than their neighbors. Our goal in the rest of the paper is to assess which perspective best characterizes higher education.

2. Hypotheses and methods Many studies find that state policies can be explained in part by their neighbor’s policies (see, inter alia, Figlio et al. 1999; Peterson and Rom 1990; Rom et al. 1998; Saavedra 1998; Tweedie 1994; Volden 2002). The challenge, though, is figuring out exactly what these findings mean, especially with regard to theories of state competition. State-to-state tracking might be due to competition, but it might be the result of natural spatial autocorrelation due to regional economic, social and political commonalities. In fact, many spatial econometric models that are used in this literature cannot distinguish between these possibilities (see Saavedra 1998, p. 15). As a consequence, what appears to be policy convergence may in fact be policy independence. Our efforts to untangle natural regional covariation from interstate competition are based on the patterns depicted in Figs. 1 and 2. If there is uniform convergence of policies, then policy changes will look something like Fig. 1. Changes in state policy are on the vertical axis and own state minus neighbor spending is on the horizontal axis. When a state is spending more than its neighbors, convergence implies that the state will decrease spending. When a state is spending less than its neighbors, convergence implies the state will increase spending.6 Evidence of convergence does not necessarily imply that states compete, however. Bailey and Rom (2001) analyzed a variety of state welfare programs and found strong evidence of convergence across all programs, including Medicare, a program where state competition is impossible because policy is not controlled by the states. This implies that unmeasured commonalties within geographic areas can induce policy convergence even in the absence of state competition. Bailey and Rom therefore argue that we must look for more specific evidence of RTB. They used the method discussed below and found evidence of a RTB in programs 6 Note, however, that we do not expect to see complete convergence of states over time. The convergence we are looking for is convergence holding all else equal. Variables such as change in preferences (through a change in political party, for example) may lead states to have different policies. The presence of random shocks in individual states in each period, secular changes unrelated to what happens in neighboring states, and the random effects of changing growth or demographic variables throughout the country, which may have different real effects depending on prior levels of each variable, suggests that some divergence will continue to exist over time.

State competition in higher education: A race to the top, or a race to the bottom?

61

Fig. 1. Converging policy changes

controlled by states, while they found no such evidence in programs not controlled by states. The idea is that races to the bottom and top are likely to have particular patterns to their convergence. If there is a race to the bottom, states will be especially wary of spending more than their neighbors. Generous states might fear inflows of resourceconsuming poor people, or their voters might be particularly concerned about erring too high on a policy they do not like. This means that RTB theory implies that there will be asymmetric convergence as in the left-hand side of Fig. 2. In it, convergence occurs in a manner such that states that spend more than their neighbors decrease their spending more precipitously than states that spend less than their neighbors increase their spending. If there is a race to the top, on the other hand, the opposite should occur. That is, low states might fear losing an attractive population (educated students or high technology business, for example) or their voters may be more concerned about erring too low on a policy than too high. This expectation is depicted in the righthand side of Fig. 2 in which states that spend less than their neighbors increase their spending more precipitously than states that spend more than their neighbors decrease their spending. We seek to estimate a model that distinguishes between the two kinds of convergence. One alternative is to use spatial econometrics models in which state policy is a contemporaneous function of all other state policies (Anselin 1988; Besley and Case 1995). For policies set by state governments, however, one can reasonably expect that responses are not contemporaneous, as changes in neighboring states need time to set in motion political responses. If there is such a lag, we can use lagged values of neighbors’ policy choices to explain current state policy. In this case, the neighborhood effects are exogenous and conventional panel methods will suffice. We therefore estimate the following model: ∆Eit = αi + αt + β1 CONVERGEit−1 + β2 COMPETEit−1 + βk Xlit + eit (1)

62

M.A. Bailey et al.

Fig. 2. Race to the bottom (left) and race to the top (right)

where ∆Eit is annual percent change in state support for higher education (variously defined), αi is a state level fixed effect, αt is a time period fixed effect , βk is a 1 × k vector of coefficients and Xlit is a k × 1 vector of state economic, political and demographic control variables. CONVERGE is the policy measure of state i minus the policy measure of its neighbors. A positive value of this variable indicates the state is spending more than its neighbors. COMPETE is the value of CONVERGE for states that are more generous than their neighbors; it captures the differential in slope for generous states relative to less generous states. We use a two-way error component panel data model (also known as a two-way fixed effect model; see Baltagi 1995 and Hsiao 1986). Doing so achieves two purposes. First, it allows us to account for unchanging state and period characteristics: the αi variable accounts for state variables that are constant over time; the αt variable accounts for period variables that are constant across states. For example, one important unchanging factor for states may be the governance structures of university systems. Another may be the ideology of the states, which Erikson et al. (1993) show is steady over time. Second, the two-way error component model allows us to address dependence of observations that arises due to the panel structure of the data. Each observation is not independent, but is correlated across time periods for each state and across states for each time period. Based on the above discussion, we formulate the following hypotheses about the coefficients of interest. First, we expect there to be convergence, implying that β1 < 0. That is, states with higher generosity than their neighbors will tend to lower their generosity, while states with lower generosity than their neighbors (for whom CONVERGE will be negative) will tend to increase their generosity. As discussed above, however, this variable cannot distinguish between competitioninduced convergence and natural regional covariation. Our main hypotheses therefore relate to β2 . For public schools we specify hypotheses from both the developmental and redistributive perspectives; for private schools we specify corresponding hypotheses from race to the bottom and race to the top perspectives. The contrast between results for public and private policies –

State competition in higher education: A race to the top, or a race to the bottom?

63

if any – will reflect differences due to the most important governance question of all: whether a school is public or private. Developmental hypothesis. If states treat higher education as developmental policy, we would expect a race to the top in the sense that they would be more sensitive to spending less or charging more for higher education. This implies that – For tuition variables, β2 < 0. For example, states with tuition that is higher than their neighbors will reduce tuition faster than the states with lower tuition rates raise them. This implies that, even as states are converging, the least generous states are becoming more generous faster than the more generous states are becoming less generous. – For spending variables, β2 > 0. In this case, states that spend less on education will increase their expenditures faster than states spending more will lower their own spending. For private schools, a race to the top would induce the above patterns for β2 . Redistributive hypothesis. If states treat higher education as redistributive policy, we would expect a race to the bottom in the sense that they would be more sensitive to spending more or charging less for higher education. This implies that – For tuition variables, β2 > 0. For example, states with tuition that is lower than their neighbors will raise rates faster than states with higher tuition lower them. – For spending variables, β2 < 0. In this case, states that spend more on education than their neighbors will decrease their expenditures faster than states spending less will increase their spending. For private schools, a race to the bottom would induce the above patterns for β2 . 3. Data and expectations Dependent variables The various dependent variables in the model all measure change in policy as measured in spending of various sorts or tuition. We measure change in both percentage and level terms, although we have a preference for the percentage measure in light of Volden’s (2002) findings that level measures may introduce bias into the estimates.7 The dependent variables measure spending from the perspectives of students and taxpayers. One measure of state willingness to invest in education is how much state money goes to higher education per state resident. (Details on the sources of this and all other variables are available in the Appendix.) Another such measure is state higher education appropriations per student. In addition, we account for the possibility that states desire that a specific amount be spent on higher education 7 The reason is that inflation is a major source of “changes” in spending, but the same level of inflation will induce a greater real level change in states with high levels than in states with low levels. To see this, suppose State A spent $1000 per student and State B spent $500 and that there were 10% annual inflation. If neither state changed its spending from one year to the next, State A’s real spending would drop to $909 (a drop of $91), while State B’s real spending would drop to $455, a drop of only $45.

64

M.A. Bailey et al.

and so use public funds to “top off” private resources by calculating total spending (from public and private sources) per student. How such spending should relate to neighboring state spending depends on the perspective. Developmental logic suggests that states have incentives to seek to “invest” in higher education at rates greater than their neighbors. This investment will allow the schools to hire better faculty and to attract better students as well as to provide superior laboratories, libraries and classrooms. On the other hand, if redistributive logic holds, state politicians may also face pressures to keep a lid on “wasteful” educational expenditures (and a skeptical public might see almost any spending on radical professors and coddled students as wasteful). In order to keep down costs – and thereby keep taxes low, which retains and attracts affluent individuals and business – states may seek to limit spending. Private schools are more constrained in their spending as they (generally) cannot rely on tax dollars to bail them out. But if private schools are competing on quality, and quality is, to some degree at least, associated with increased spending, then we might expect a race to the top in spending. On the other hand, if competition is in terms of cost, then we might expect private schools to race to the bottom on spending. We have restricted our data set to schools with a Carnegie ranking between 11 and 31; this includes all schools with substantial graduate programs and liberal arts colleges deemed to be “selective” in their admissions criteria, for a total of 452 private schools. We also look at more specific forms of public higher education spending. The first is average salary levels. There is a well-defined market for faculty and it may well be that the competition is both based on more information and is more intense, possibly inducing different patterns of behavior. As before, developmental logic suggests a race to the top, while redistributive logic suggests a race to the bottom. The second is spending on research. Development logic would suggest that states want to stay at the cutting edge and, in so doing, would be very concerned about falling behind their neighbors. In redistributive logic, money spent on research would simply be going toward sating faculty curiosity and not much more. For tuition, we use tuition revenues per student. We do not examine the advertised “sticker” price (the nominal tuition price), because a large proportion of students does not pay full tuition (due to fellowships, scholarships, and grants) and because tuition varies from school to school within each state system. The developmental logic for tuition is that states having effective tuition rates that are higher than their neighbors will lower them to become more competitive and that states paying a smaller share of the total cost of education will increase their share of the cost. However, if states view higher education primarily as a redistributive program, then states would raise tuition rates and the proportion of tuition to spending to the level of their neighbors. Again, private schools are more constrained because they cannot rely on tax dollars to cover any deficits they run. This puts pressure on them to set tuition rates high enough to maintain solvency. In addition, university administrators may prefer higher tuition as a way of enlarging their budgets (Lowry 2000). Alternately, private schools risk losing students if, ceteris paribus, tuition rates are higher than at competitive institutions.

State competition in higher education: A race to the top, or a race to the bottom?

65

Independent variables The main independent variables are the COMPETE and CONVERGE variables discussed earlier. They are calculated as the difference between a state and its neighbors. “Neighbor” can be defined in many ways: neighbors may be geographically close or they may be socio-politically similar. We focus on geographic proximity in the belief that neighbors are the best yardsticks most voters have. Note that such effects are plausible even if policy competition is national in scope. Consider an example. Minnesota and Wisconsin might both want to attract students and faculty from around the country, and so of course are interested in tuition, spending, and enrollment in Florida, Texas, and Washington. Still, because both Minnesota and Wisconsin are competing for the same pool of students and faculty, each state must pay close attention to what the other is doing. We estimate models in which COMPETE and CONVERGE are calculated in terms of the average policies of all neighbors and in terms of the population-weighted averages of all neighbors. Since the two methods yield similar results, we report on the average policy model. For private schools, neighbors are defined as all schools in our sample in the same state or neighboring states.8 We also control for other political, demographic and economic factors. For political effects, we use a measure of Democratic Party strength which weights seat shares in the state House and Senate and whether the governor is a Democrat. Because the nature of party cleavages differs across the states and between the south and north in particular (Brown 1995), we also interact Democratic Party strength with an indicator for the eleven southern states of the old Confederacy. This allows Democrats in the south to have a different effect than Democrats in the north. To capture demographic change, we use the percentage of the state that is elderly. It may be that political support for higher education changes as the age composition of the state changes. We also used percentage of population under 18, but found that it had little or no effect on the results so it has been left out. To capture economic change, we use state median income. The general expectation is that as state resources increase – in the form of higher income – state generosity will increase as well (Tweedie 1994). For private schools we have only one independent variable, state median income. The idea is that private schools in states that are booming may have more latitude to increase tuition or spending than private schools in other areas. Our efforts to find additional independent variables were thwarted by limited data availability. However, note that we in fact have hundreds of independent variables, as we have controlled for 452 individual school effects and period effects with the αi and αt parameters. To convert spending into real dollars we use Berry et al. (2000) state level consumer price index (CPI). For cases in which we had data beyond 1995 – the last year in their data set – we created our own extension of the state CPI in which state CPI from 1995 was multiplied by national CPI for 1996 or 1997, ensuring that for each state, its ratio to national CPI was the same as in 1995. Analysis of 8 Private schools, especially the most selective schools, compete nationally as well. We focus on regional competition as one form of competition.

66

M.A. Bailey et al.

the earlier years of the state CPI data indicates that this ratio was very steady over time. In order to ensure that this state level correction is not responsible for any of our results, we also ran the analysis in which we used annual national CPI data for all states. The results were always similar. 4. Results The main results of our estimates are presented in Tables 1 through 4. The most important findings concern two independent variables of central interest, CONVERGE and COMPETE. CONVERGE is significant in virtually every model. This means that state higher education systems that spend more, charge more, offer higher salaries, or fund more research than their neighbors in any given year decrease spending in the following year (all things equal). Conversely, state higher education systems that spend less than their neighbors tend, all things equal, to spend more the next year. The COMPETE variable is statistically significant at the 0.01 level (one-sided) in 9 of the 18 estimates and is statistically significant at the 0.10 level in 11 of the 18 estimates. The sign indicates a race to the bottom or no effect in public education; only for private tuition are there signs of a race to the top. To see in more detail the meaning of the CONVERGE variable, consider public spending per student (Table 1). The net impact of neighbor states on state policy is the coefficient on CONVERGE for states that are less generous than their neighbors and the sum of the coefficients of CONVERGE and COMPETE for the states that are more generous. The coefficients indicate that, all things being equal, a state that spent ten percent per student more than its neighbors in one year will decrease its per student spending by 5.6% the next year. In contrast, a state that spent ten percent less per student than its neighbors will increase per student spending by 1.6% the next year. In other words, the more generous states are moving down toward their peers faster than the less generous states are moving up. This is more consistent with a race to the bottom than a race to the top. We also observe patterns consistent with a RTB in public appropriations per student and public appropriations per capita (Table 1). States that spend more per student, or more per resident, on higher education than their neighbors move down faster than the less generous states move up. The net difference in slopes, however, is smaller than for public spending per student. Interestingly, private per student spending shows a similar pattern to public per student spending (Table 2). All things equal, a school that spent 10% less than its neighbors in one year would increase its per student spending by 3.3% the next year. In contrast, a school that spent 10% more than its neighboring schools in a given year would be expected to decrease its per student spending by 6.6% the following year. While the spending of relatively low-spending private schools is moving upwards toward peer levels, spending by more generous private schools is falling more rapidly. The coefficients for public and private tuition reveal interesting differences (Table 3). A race to the top exists for private tuition, with schools apparently competing to lower tuition rates, while a race to the bottom occurs in public tuition. For private tuition, the schools that are charging higher tuition than their neighbors lower their

48

1986 – 1997

Number of states

Range of years

1986 – 1997

48

528

3.80 (p = 0.00)

528

3.30 (p = 0.00)

F statistic for no state effects

N

0.568 12.85 (p = 0.00)

0.571

−5.49

−0.24

0.59

−0.31

−2.04

0.14

−4.93

13.01 (p = 0.00)

0.622

−140.05

−0.05

832.64

−6.23

−8.82

0.33

−0.36

−0.15

t-stat −2.86

Level Coef.

F statistic for no fixed effects

Adj.

0.625

−5.14

−0.000009

Student/population

R2

−0.18

−0.000003

State budget deficit

R2

0.93

−0.56

0.000892

−0.000772

Median income

Elderly

−0.30

−1.88

−0.000049

−0.000555

Democratic strength – South

−5.62

−0.40

“Compete"

Democratic strength

−4.15

−0.16

t-stat

“Converge"

Coef.

Percent

Public spending per student

−3.75

2.19

4.07

1.32

−1.03

−0.46

−2.60

−4.51

t-stat

1986 – 1997

48

528

2.94 (p = 0.00)

3.77 (p = 0.00)

0.262

0.354

−0.00806

0.00004

0.00475

0.00223

−0.00039

−0.00010

−0.17

−0.17

Coef.

Percent t-stat

−3.05

2.01

3.75

0.94

−1.17

−0.24

−0.86

−3.67

1986 – 1997

48

528

2.89 (p = 0.00)

3.72 (p = 0.00)

0.270

0.361

−43.56

0.24

294.98

10.77

−2.93

−0.34

−0.06

−0.20

Coef.

Level

Public appropriations per student

Table 1. General spending by public universities

3.36

1.43

2.85

2.06

−0.52

−0.23

−2.82

−5.12

t-stat

1986 – 1997

48

528

2.69 (p = 0.00)

3.48 (p = 0.00)

0.232

0.328

0.00696

0.00003

0.00339

0.00345

−0.00019

−0.00005

−0.21

−0.17

Coef.

Percent

t-stat

3.14

1.58

1.86

1.73

−0.58

0.14

−0.78

−4.96

1986 – 1997

48

528

2.84 (p = 0.00)

3.56 (p = 0.00)

0.237

0.332

0.80

0.0034

2.7491

0.3565

−0.0255

0.0034

−0.05

−0.24

Coef.

Level

Public appropriations per capita

State competition in higher education: A race to the top, or a race to the bottom? 67

68

M.A. Bailey et al.

Table 2. Spending by private universities Private spending per student Percent Coef.

Level t-stat

Coef.

t-stat

“Converge"

−0.33

−26.16

−0.44

−10.58

“Compete"

−0.33

−6.85

−0.04

−0.83

0.39

0.04

1.16

State median income

0.00000058

R2

0.421

Adj. R2

0.336

0.206

F statistic for no fixed effects

4.82 (p = 0.00)

3.05 (p = 0.00)

F statistic for no school effects

2.85 (p = 0.00)

2.26 (p = 0.00)

N

3616

3616

Number of schools

452

452

1986 – 1994

1986 – 1994

Range of years

0.307

tuition rates at a much faster pace than schools with lower private tuition raise theirs. Ceteris paribus, private schools charging 10% higher tuition than their peers in a given year would decrease tuition rates by 5.7% the following year, while schools charging 10% lower tuition in a given year would increase tuition by only 2.3%. Note that the ceteris paribus statements here are important: the year effects may be (and in fact generally are) quite large and positive, such that all schools may be expected to increase tuition from one year to the next. Holding that constant, however, the relatively high tuition schools fall more quickly than the relatively low tuition schools rise; or put differently, they may increase more slowly than do the relatively low tuition schools. In contrast, public tuition rates look more like a RTB. When changes in the dependent variable are measured in percentage terms, the results indicate that states with public tuition rates that are 10% higher than their neighbors would see a decline in tuition of only 0.7% the following year. States with 10% lower public tuition rates would see an increase in tuition of 1.4% the following year. This difference in slopes is not statistically significant, however. When changes are measured in level terms, however, there is a statistically significant difference in slopes and the lower tuition charging states increase tuition at a faster rate than the high tuition states lower theirs. In short, it appears that states are more worried about charging too little for higher education than charging too much. Public tuition ratios (Table 3), which measure the size of tuition receipts as compared to total higher education spending, offer mixed results. The COMPETE variable on public tuition ratios is not statistically significant in percentage terms (at the 0.95 confidence level), and shows only a small but significant effect in terms of level. Nonetheless, the coefficients suggest that the pace of convergence is relatively balanced. Other things equal, states with a higher public tuition ratio in a given year

−0.56

−0.00014

–

Number of schools

1986 – 1997

48

Range of years

528

1986 – 1997

–

48

528

2.28 (p = 0.00)

1.36 (p = 0.06)

F statistic for no state effects

Number of states

2.85 (p = 0.00)

1.79 (p = 0.00)

F statistic for no fixed effects

N

0.225

−3.21

0.322

−19.82

−1.74

0.198

−1.96

−0.00506

Student/population

−0.09

−0.95

0.298

−2.03

−0.00004

State budget deficit

−34.68

−0.75

0.45

−1.05

2.51

−6.52

t-stat

Adj. R2

−1.12

−0.00169

Elderly

−3.72

0.48

−0.63

0.15

−0.26

Coef.

Level

R2

0.45

−0.19

0.00020

−0.00040

Median income

Democratic strength – South

Democratic strength

0.94

0.07

−7.42

−0.14

“Converge"

t-stat

“Compete"

Coef.

Percent

Public tuition per student

0.31

−1.91

−2.10

0.28

1.59

−1.14

0.09

−7.05

t-stat

1986 – 1997

–

48

528

1.83 (p = 0.00)

6.73 (p = 0.00)

0.427

0.498

0.00079

−0.00004

−0.00313

0.00057

0.00071

−0.00029

0.01

−0.15

Coef.

Percent

1.77

−7.89

t-stat

Level

1.70 −2.82

1986 – 1997

–

48

528

2.61 (p = 0.00)

7.52 (p = 0.00)

0.437

0.507

−0.0003 0.56

−0.00001−1.62

−0.01

−0.0002 −0.52

0.0001

−0.0001 −2.01

0.12

−0.30

Coef.

Public tuition ratio

Table 3. Public and private tuition

–

0.178

0.283

–

–

–

1.00

–

1986 – 1994

452

–

3616

2.51 (p = 0.00)

2.63 (p = 0.00)

–

–

–

0.0000010

–

−8.76

–

−22.28

−0.34

t-stat

−0.23

Coef.

Percent

0.230

0.328

–

–

–

1.45

–

–

−1.39

−16.34

t-stat

1986 – 1994

452

–

3616

3.16 (p = 0.00)

3.33 (p = 0.00)

–

–

–

0.0132

–

–

−0.04

−0.37

Coef.

Level

Private tuition per student

State competition in higher education: A race to the top, or a race to the bottom? 69

70

M.A. Bailey et al.

would likely see a lower public tuition rate in the following year. A state with a 10% lower public tuition ratio in a given year would see a rise in the ratio by 1.5% the following year. Finally, we observed convergence in both faculty salaries and research spending (Table 4). But while there are signs of a RTB in research spending, there are no such signs for salaries. For research spending, however, states that spend more on research per capita in a given year decrease their spending more rapidly than relatively low spending states increase their spending. Other things equal, a state spending 10% more than its neighbors on research in a given year would decrease its spending by 5.5 percent in the subsequent year. A state spending 10% less would raise spending by 2.1%. Faculty salaries are an interesting exception to the patterns observed so far for public schools, as there is no sign of a RTB. This indicates that while states are sensitive to spending more across a variety of categories, this is not the case for salaries. The salary variable should be taken with some caution, however, as salaries are highly dependent on the rank and experience of its faculty, something that we are unable to control for. It is possible, for example, that an ambitious state making many competitive hires at the junior level would appear to be a low salary state, while another state making few competitive hires may appear to be a high salary state if it has many senior faculty members. We are mainly interested in estimating the impacts of interstate competition on educational policies, with the other independent variables serving mainly as statistical controls. Coefficients for the control variables by and large show rather little of interest. But a few results are worth pointing out. First, Democratic strength generally does not matter, although Democratic strength in southern states is associated with lower spending per student (see Table 1), while Democratic strength in northern states is associated with higher faculty salaries. Second, median income had only minor effects, despite our initial hypothesis that higher incomes would lead to higher state generosity. States with higher median incomes have only slightly higher public appropriations per capita: a 10% increase in the state median income would lead, other things equal, to a 0.4% increase in public appropriations per capita. Third, a more surprising result comes from the elderly, measured as the percentage of the population over the age of 65. Both public appropriations per student and public appropriations per capita (Table 1) respond positively and in a statistically significant fashion (i.e., above 0.95 significance) to an increase in the population of the elderly. A 10% increase in the percentage of the population over age 65 would contribute to a 0.5% increase in public appropriations per student and a 0.34 percent increase in public appropriations per capita. Fourth, a state with a higher deficit per capita is likely to spend slightly more, rather than less, in public education appropriations per student and per capita. In addition, such states have slightly lower tuition rate increases at their public schools. Lastly, we find that the ratio of students to the remainder of the population has a significant effect on spending. Often, the direction is downward; for example, as the relative population of students increases, public spending per student declines. This is true for both public spending and public appropriations. This makes sense, as there is less for each student when there are many students, or because there are

48 1986 – 1997

Number of states

Range of years

1986 – 1997

48

528

3.18 (p = 0.00)

528

3.58 (p = 0.00)

F statistic for no state effects

N

0.397 6.47 (p = 0.00)

0.473

0.461

0.528

1.95

−1.51

8.16 (p = 0.00)

R2

0.27

0.95

1.23

−0.49

0.64

−4.98

−1.67

t-stat

Level

−0.002

0.74

0.14

−0.01

0.01

−0.40

−0.09

Coef.

F statistic for no fixed effects

Adj.

R2

0.003550

1.19

−1.75

−0.000044

State budget deficit

−0.23

−0.000561

Median income

Student/population

−0.04

−0.000019

Democratic strength – South 0.83

−0.08

−0.000024

Democratic strength

0.001389

−3.84

−0.34

“Compete"

Elderly

−7.03

−0.21

t-stat

“Converge"

Coef.

Percent

Public research per student

0.142

0.219

– –

1982 – 1998

48

768

1.49 (p = 0.02)

2.91 (p = 0.00)

–

–

0.53 −1.00

0.000330

−0.35

1.72

0.22

−3.12

t-stat

−0.000384

−0.000050

0.000146

0.02

−0.25

Level Coef.

0.128

0.206

t-stat

–

–

−1.15

0.59

−0.29

1.65

0.32

−2.62

1982 – 1998

48

768

1.45 (p = 0.03)

2.74 (p = 0.00)

–

–

−194.82

16.19

−1.85

6.16

0.03

−0.25

Real salaries Percent Coef.

Table 4. Specific spending by public universities

State competition in higher education: A race to the top, or a race to the bottom? 71

72

M.A. Bailey et al.

more students to cover fixed costs. As the student ratio of the population increases, on the other hand, public appropriations per capita increase, as each taxpayer has to support relatively more students. In sum, the results across a variety of dependent variables and specifications indicate several strong patterns. There is always convergence, with low spenders moving up and high spenders moving down. For public schools, there is asymmetric convergence, with high spenders moving down faster than low spenders move up; this pattern is more consistent with a RTB than a RTT. For private schools, spending has the same pattern as for public schools, but tuition has a different pattern, one in which high tuition schools tend to move down (or increase more slowly) than low tuition schools tend to move up. This is consistent with a RTT on private schools’ tuition.

5. Conclusion One key to understanding higher education in the United States is understanding how external environments affect higher education policies. Does competition among states and private schools push them to spend more on education and charge less for it? Or, does it push them to spend less and charge more? Does the effect of competition vary across public and private governance structures? The results here indicate that competition generally dampens public spending and, if anything, boosts tuition charges. There are broad convergent tendencies, but the convergence comes more from the generous states cutting or slowing spending than from the less generous states raising spending. This pattern of responses to neighbors is more consistent with the view that states treat higher education spending like redistribution; in short, states appear more concerned with spending too much or charging too little than with the opposite. The results are different for private schools, especially on tuition. Private colleges and universities that charge more than their comparable neighbors lower or slow the increases in tuition more starkly than schools with low tuition increase their tuition. This indicates that there is price competition among private schools and that they are sensitive to becoming non-competitive in tuition rates. However, since private schools cannot rely on the potentially unlimited resources of state taxpayers, their sensitivity on tuition is matched with a tendency for the high spending schools to decrease spending more rapidly than low spending schools increase spending. Taken together, the results indicate that competition and governance structure exert important effects on higher education in the United States. Competition matters because neighboring states and schools influence policies. Governance structure matters because public and private schools respond differently to such competition.

State competition in higher education: A race to the top, or a race to the bottom?

73

Appendix: Data descriptions and sources Variables

Description

Source

Enrollment variables9

Complete data was available for academic years 1986– 1987 through 1995–1996. Preliminary data was available for 1996–1997 and 1997–1998; data was missing for fewer than 5 public schools and these were extrapolated from previous data. Part-time students were converted into “full-time student equivalents” (FTEs) by dividing part-time enrollment figures by three.10

IPEDS; ICPSR (1986–1988)

Finance variables

Data was available for academic years 1986–1987 through 1995–1996.

IPEDS; ICPSR

Faculty salary data

The average faculty salary data comes from a variety of Department of Education sources from 1982–1998. Figures for years which were missing data (1983– 1984, 1986–1987, 1988-1989) was interpolated.

DES; IPEDS

Democratic strength

This variable is percent Democratic in the state House + percent Democratic in the state Senate + 1 if the governor was a Democrat

BOS

State cost of living

Index set to 1995 Texas = 1.

Berry, (2000)

Median income

Median income for four person families by state, measured in thousands

Census

Demographic variables

Both the percentage of state populations aged 65 years or older and the percentage of state populations aged 19 years or younger were calculated by the authors from the raw data.

Census

Student/pop

Number of students per thousand population.

IPEDS; ICPSR (1986–1988); Census

State budget deficit

Deficit per 1,000 people in state

State finances

et.al.

Key IPEDS: U.S. Department of Education. “Integrated Postsecondary Education Data System (IPEDS): Higher Education Finance Data” Various years. http://nces.ed.gov/Ipeds/data.html; July 2001. ICPSR: U.S. Department of Education. “Integrated Postsecondary Education Data System (IPEDS): Higher Education Finance Data, 1986-87 and 1987-88.” Accessed via Inter-university Consortium for Political and Social Research; http://www.icpsr.umich.edu; July 2001. DES: U.S. Department of Education. Digest of Education Statistics. Washington, D.C., various years. BOS: The Book of States. Lexington, KY: Council of State Governments, various editions. Berry et al.: Data was obtained from calculations by Berry et al. (2000). 9

With both the enrollment and finance variables, milityry academies have been removed from the sample. In Analyzing private schools, the authors removed schools with fewer than 200 undergraduate students, as well as scholls in Alaska and Hawaii. 10 The FTE conversion method was created by Lowry (Lowry 2001,11).

74

M.A. Bailey et al.

Census: Department of Commerce, Economics and Statistics Administration, Bureau of the Census. Data accessed at http://www.census.gov; July 2001. State Finances: State and Local Government Finances from the Census Bureau, Data accessed at http://www.census.gov/govs/www/estimate.html; June 2001.

References Anselin, L. (1988) Spatial Econometrics: Methods and Models. Kluwer Academic Publishers, Boston, MA Bailey, M.A., Rom, M.C. (2001) A Wider Race? Interstate Competition across Health and Welfare Programs. Unpublished manuscript, Georgetown University, Washington, DC Balla, S.J. (2001) Interstate Professional Associations and the Diffusion of Policy Innovations. American Politics Research 29: 221–245 Baltagi, B. (1995) Econometric Analysis of Panel Data. John Wiley & Sons, New York, NY Berry, F.S., Berry, W.D. (1999) Innovation and Diffusion Models in Policy Research. In: Sabatier, P. (ed.) Theories of the Policy Process. Westview Press, Boulder, CO Berry, F.S., Berry, W.D. (1990) State Lottery Adoptions as Policy Innovations: An Event History Analysis. American Political Science Review 84: 395–416 Berry, W.D., Fording, R.C., Hanson, R.L. (2000) An Annual Cost of Living Index for the American States, 1960–1995. Journal of Politics 62: 2 Besley, T., Case, A. (1995) Incumbent Behavior: Vote-Seeking, Tax-Setting and Yardstick Competition. American Economic Review 85: 25–45 Brown, R.D. (1995) Party Cleavages and Welfare Effort in the American States. American Political Science Review 89: 23–33 Caillouet, L.S. (2001) Agency to Ask for Millions to Boost College-Grad Rates. Arkansas Democrat Gazette. Thursday, July 19. 1A, 3A Donahue, J.D (1997) Disunited States. Basic Books, New York, NY Dye, T. (1990) Competitive Federalism: Competition Among Governments. D. C. Health, Lexington, MA Eisinger, P.K. (1989) The Rise of the Entrepreneurial State: State and Local Economic Development Policy in the United States. University of Wisconsin Press, Madison, WI Erikson, R.S., Wright, G.C., McIver, J.P. (1989) Political Parties, Public Opinion, and State Policy in the United States. American Political Science Review 83: 729–750 Erikson, R.S., Wright, G.C., McIver, J.P. (1993) Statehouse Democracy: Public Opinion and Policy in the American States. Cambridge University Press, New York, NY Figlio, D.N., Kolpin, V.W., Reid, W.R. (1999) Do States Play Welfare Games? Journal of Urban Economics 46: 437–454 Fox, S.E. (1996) The Influence of Political Conditions on Foreign Firm Location Decisions in the American States, 1974–1989. Political Research Quarterly 49: 51–75 Gray, V. (1973) Innovation in the States: A Diffusion Study. American Political Science Review 67: 1174–1185 Hearn, J.C., Griswold, C.P. (1994) State-Level Centralization and Policy Innovation in U.S. Postsecondary Education. Educational Evaluation and Policy Analysis 16: 161–190 Hearn, J.C., Griswold, C.P., Marine, G. (1996) Region, Resources and Reason: A Contextual Analysis of State Tuition and Student Aid Policies. Research in Higher Education 37: 241–278 Heller, D.E. 2002. The Policy Shift in State Financial Aid Programs. In: Smart, J. and Tierney, W. (eds.) Higher Education: Handbook of Theory and Research. Volume XVII. Algora Publishing, NewYork, NY Hsiao, C. (1986) Analysis of Panel Data. Cambridge University Press, New York, NY Huber, J.D., Shipan,C.R., Pfahler, M. (2001) Legislatures and Statutory Control of the Bureaucracy. American Journal of Political Science 45: 330–345 Hughes, K.L., Bailey, T.R., Mechur, M.J. (2001) School-to-Work: Making a Difference in Education. Columbia University, New York, NY

State competition in higher education: A race to the top, or a race to the bottom?

75

Johnstone, D.B. (1992) Tuition Fees. In: Harmon, G. (ed.) Encyclopedia of Higher Education. Pergamon Press, New York, NY, Volume 2: 1501–1509 King, G., Keohane, R., Verba, S. (1994) Designing Social Inquiry. Princeton University Press, Princeton, NJ Lowry, R.C. (2000) Governmental Structure, Trustee Selection and Public University Prices and Spending: Multiple Means to Similar Ends. American Journal of Political Science 45: 845–861 Lowry, R.C. (2001) Governance, Markets, and University Priorities: Evidence on Undergraduate Education and Research. Paper presented at the conference “It’s Better to Rely on Well-Designed Institutions than Well-Behaved People.” UCLA, May 2001 McLendon, M.K., Heller, D.E., Young, S.P. (2001) State Postsecondary Policy Innovation: Politics, Competition and the Interstate Migration of Policy Ideas. Paper presented at the Midwest Political Science Association annual meeting. Chicago, IL Mintrom, M. (1997a) Policy Entrepreneurs and the Diffusion of Innovation. American Journal of Political Science 41: 738–770 Mintrom, M. (1997b) The State-Local Nexus in Policy Innovation Diffusion: The Case of School Choice. Publius 27: 41–59 Mintrom, M., Vergari, S. (1996) Charter School Laws Across the United States, 1996 Edition. Institute for Public Policy and Social Research, East Lansing, MI Mintrom, M., Vergari, S. (1997) Charter Schools as a State Policy Innovation: Assessing Recent Developments. State and Local Government Review 29: 43–49 Mooney, C.Z. (2001) Modeling Regional Effects on State Policy Diffusion. Political Research Quarterly 54: 103–124 Peterson, P.E. (1995) The Price of Federalism. The Brookings Institution, Washington, DC Peterson, P.E., Rom, M.C. (1990) Welfare Magnets: A New Case for a National Welfare Standard. The Brookings Institution, Washington, DC Rocke, A.J. (2000) Nationalizing Science. MIT Press, Boston Rogers, E.M. (1995) Diffusion of Innovations. 4th edition. Free Press, New York Rom, M.C., Peterson, P.E. Scheve, K.F. (1998) Interstate Competition and Welfare Policy. Publius 28: 17–38 Rom, M.C. (1989) The Family Support Act of 1988: Federalism, Developmental Policy, and Welfare Reform. Publius Saavedra, L.A. (1998) A Model of Welfare Competition with Evidence from AFDC. Institute of Government and Public Affairs Working Paper 63 (October). University of Illinois at Urbana-Champaign Schram, S.F., Beer, S.H. (1999) Welfare Reform: A Race to the Bottom? Woodrow Wilson Center Press, Washington, DC Schram, S.F., Soss, J. (1998) Making Something out of Nothing: Welfare Reform and a New Race to the Bottom. Publius 28: 67–88 Smith, P. (1991) An Empirical Investigation of Interstate AFDC Benefit Competition. Public Choice 68: 217–233 Tiebout, C.M. (1956) A Pure Theory of Local Expenditures. Journal of Political Economy 6: 416–424 Tweedie, J. (1994) Resources Rather Than Needs: A State-Centered Model of Welfare Policymaking. American Journal of Political Science 38: 651–672 Volden, C. (2002) The Politics of Competitive Federalism: A Race to the Bottom in Welfare Benefits? American Journal of Political Science 46: 352–363 Volkwein, J.F. (1987) State Regulation and Campus Autonomy. In: Smart, J. (ed.) Higher Education: Handbook of Theory and Research. Volume III. Agathon Press, New York, NY Volkwein, J.F., Malik, S.M. (1997) State Regulation and Administrative Flexibility at Public Universities Walker, J.L. (1969) The Diffusion of Innovations Among the American States. American Political Science Review 63: 880–899 We owe special thanks to Sam Barbett and Charlene Hoffman at NCES for providing a number of data series and answering many questions.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.