Rethinking Policy Diffusion: The Interstate Spread of “Finance Innovations”

Abstract

Recently a number of studies have focused on states' adoptions of postsecondary-specific policies. Cutting across much of this research is the presence and influence of interstate diffusion of policy adoptions, a phenomenon for which support is scant. This paper seeks to address this through broadening the categorization of policies beyond the discrete form traditionally used to one that encompasses a larger conception of "finance policy." Our sample uses 131 finance innovations for 47 states over a 29 year period, finding that upon broadening our definition, we can detect the process of diffusion. However, the findings are striking, showing that while states do learn from one another, the process is dynamic and shifts across time.

Introduction

Historically, state policymakers have attempted to keep tuition low through direct appropriations to public colleges and universities. Since the nineteenth century, the states were the predominant funders of public colleges and universities, a commitment accelerated by the federal land grant acts, the development of teachers colleges, community colleges, the expansion of public research universities and further federal action such as the GI Bill (Johnson 1987). However, as early as the 1940s, policymakers became concerned with higher education’s escalating costs to students.¹ In 1965 the federal government passed the Higher Education Act (HEA) which included the seeds of the federal, need-based financial aid program that would later evolve into the Pell Grant Program. In the first reauthorization in 1972, the Act also included the State Student Incentive Grant program [later renamed the Leveraging Educational Assistance Partnership (LEAP) program], which provided federal matching funds for state-run, need-based grants. The LEAP program resulted in a flurry of state policymakers developing such grants, with all states adopting a program by the end of the decade (Heller 2002).

During the 1980s, college costs and prices began to rise at an even accelerated rate, exceeding annual increases in governmental appropriations, inflation, personal income, and state and federal financial aid. In this environment, the burden of paying for college increasingly moved from states to students while the discourse among policymakers shifted to higher education as a private, rather than a public, good (Lyall and Sell 2006). It was against this changing backdrop that state postsecondary officials began experimenting with new “market-based” higher education policies as well as other, alternative means for financing colleges and universities (Titus 2006; Paulsen and Smart 2001; Wellman 2006; Hemsley-Brown 2011). Identifying the factors associated with states adopting one or more of this class of policy innovations is the focus of this study.

To examine this line of inquiry, postsecondary education researchers have drawn from the comparative state political science and public policy literatures to study states’ adoptions of postsecondary policies, often centering on questions of geographical policy diffusion (e.g., McLendon et al. 2005, 2006, 2007; Doyle 2006; Hearn et al. 2008). Geographical policy diffusion argues that the policies ultimately enacted in states tend to mimic the policies of their neighbors and, as one state adopts a policy, it increases the likelihood that officials in proximate states will follow suit. While support for geographical diffusion has been found in other domains, studies testing hypotheses in the context of postsecondary policies have found little support for this phenomenon (Sponsler 2010).

In studying the development of state postsecondary finance innovations, researchers have chosen to be discrete in defining very specific policies (e.g., performance funding, merit aid). While these distinctions have tangible differences in states’ policy landscapes and their effects on colleges and universities, this deep parsing of larger policy categories may obscure geographical policy diffusion’s influence.² In this study, we wish to test the influence of diffusion after taking a broader view of postsecondary policy types. Specifically, we look at the spread of finance innovations as a category, under the rationale that they are frequently driven by similar motivations and goals. Though most of these policies have been studied individually, we have yet to find a study which attempted to analyze the diffusion of these finance innovations as a whole.³

Our argument for combining these policies under the term “finance innovations” stems from two sources: First, as we detail later, these policies fall under the umbrella of what some articulate as “market-based approaches” to financing higher education (Teixeira et al. 2006). Second, combining innovations in this manner may better mirror what individuals in states respond to when they look across their borders; in the case of postsecondary education, policy emulation might not follow the discrete technocratic policies, but rather broad categories that seek similar outcomes. That is, while the distinctions are important in their mechanics, we hypothesize that these differences may not be apparent, or even meaningful, for public officials to whom problems and solutions do not always precede in a linear fashion. From this perspective problems may be matched with any relevant options from a broad category of solutions, or even multiple solutions at once (Kingdon 1984). In the current national landscape, studying postsecondary policy diffusion may yield insights as to policy creation in general and the influence of higher educations’ unique mix of actors. As the federal government, large foundations, non-profit associations, and intermediary organizations attempt to “scale best practices” across state lines, looking at the origins of existing policies will point towards enabling conditions and barriers to adoption.

State Postsecondary Finance Innovations

Over the past thirty years state policy increasingly utilized market-based approaches to the financing and governance of the public sector (Rabovsky 2012). In state higher education finance policy, market-based approaches are those that: (1) attempt to incentivize personal or institutional behavior towards certain larger outcomes deemed important by state policymakers (as opposed to regulations and other top down directive policy approaches); and/or (2) shift the burden of paying for college from the state to the individual. The specific innovations under this umbrella are: tuition decentralization, voucher programs, 529 savings plans, pre-paid tuition programs, performance funding, and broad-based merit aid programs (Titus 2006; Hauptman 2006; McLendon et al. 2009; Protopsaltis 2008; Wellman 2006; Hemsley-Brown 2011).

One of the primary arguments supporting the decentralization of tuition authority was that markets should drive prices and, once freed from artificial price controls, institutions would be able to compete on quality and students would “vote with their feet” (Wellman 2006). Spurred by new economic and political realities, in some states tuition authority has become decentralized from central state offices, allowing institutions—as opposed to postsecondary governing structures or legislatures—to set their own tuition rates which predictably led to increased tuition (Lyall and Sell 2006). Realizing that enrollments continued to grow even in the face of increasing costs, state legislatures have continued to disinvest in higher education, support other public priorities, and allow institutions to increase tuition (Breneman and Finney 1997; Hovey 1999). Between 1987 and 2006, 18 states decentralized tuition setting authority.

Perhaps one of the most innovative approaches to reorganizing the financial relationship between the state, students, and institutions was the creation of the Colorado Opportunity Fund, a program that functions as a voucher program for higher education. Using money that otherwise would be appropriated directly to institutions, Colorado established a stipend available to all lawfully present Colorado residents to use for offsetting their total costs at the public (and eligible private) higher education institution of their choice. While Colorado remains the only state to experiment with vouchers in the postsecondary sector, they have a long history as a policy innovation in the K-12 education arena (Rouse and Barrow 2009). In both cases, vouchers are an extreme case of enacting legislation which pushes institutions to compete for students (Western Interstate Commission on Higher Education 2009; Fox 2006; Protopsaltis 2008).

College savings and prepaid tuition plans are another means through which policies aim to alter student and family behavior. Similar to Roth IRAs and other retirement plans, 529 plans are college savings accounts where principal is invested and not subject to capital gains tax, with many states allowing citizens to deduct contributions from state income tax. Conversely, prepaid tuition plans allow purchasers to lock in a certain tuition rate by purchasing future credits at today’s prices, theoretically providing a hedge against tuition inflation. Often coupled together as options in policy discussions, both plans were designed to increase college affordability and personal savings and financial planning. The approaches may be seen as natural market-based approaches to college affordability as they do not regulate college costs, use incentives as opposed to new laws, are voluntary and require individual investment, and almost uniformly allow students to continue to chose the institution in which they ultimately enroll.⁴ Between 1986 and 1999, thirty-one states adopted some form of a 529 plan and 21 states adopted prepaid tuition plans (Doyle et al. 2010; Hurley 2006; Olivas 2003).⁵ Importantly, adoption of 529 plans continued after 1999, though not all at once. In 2000 five states adopted a plan, followed by four in 2001, and six in 2002.

Also subsumed in this era of market-based financial policies, was a shift in the oversight and accountability environment for public higher education. The movement was from a concentration on regulatory compliance and rudimentary reporting of inputs and expenditures to measuring performance and accounting for outcomes and results (Burke 2005; McGuinness 2005; McLendon 2003; Volkwein 2007; Volkwein and Tandberg 2008), a postsecondary change that mirrored a larger shift in public education policy referred to as the “new accountability” movement (Gorbunov 2004; McLendon et al. 2006; Zumeta 1998; Toutkoushian and Danielson 2002). A primary manifestation of this movement was state-level performance funding policies, which are programs that link institutions’ funding levels to a set of outcomes (e.g., student retention, graduation rates, student scores on licensure exams, job placement rates, faculty productivity, and campus diversity). As opposed to other policies that attempt to alter individuals’ behaviors, these programs seek to incentivize institutions to meet state educational priorities (Burke 2002).⁶

One of the most frequently studied state higher education policy adoptions is the widespread initiation of merit-based student aid programs (Doyle 2006; Ness 2008; Cohen-Vogel et al. 2008). Similar to several other higher education finance innovations (performance funding, 529 college savings plans, and prepaid tuition programs), merit aid programs seek to encourage particular behaviors, in this case student academic performance and enrollment in an in-state institution. At the same time they address affordability (for those who qualify) and allow for (limited) college choice. In the early 1990s, the state of Georgia initiated the first large-scale, statewide merit aid program with 12 other states adopting similar programs (Heller and Marin 2004).⁷

Conceptual Framework

Our conceptualization of the factors influencing state postsecondary policy adoption draws from the political science literature which argues that these innovations stem from both within state characteristics and, in part, the behavior of other states. The first (intrastate forces) comes from the “new institutionalism” perspective which, as March and Olsen (1984) succinctly assert, “is simply an argument that the organization of political life makes a difference” (p. 747). While this thinking acknowledges the importance of state economic and demographic factors, politics is an important driver as well. Recently the new institutionalism perspective migrated to the state higher education policy and finance literature. It has been used, often in combination with other perspectives, to explain state political actors’ higher education policy decisions (Doyle et al. 2010; Hearn and Griswold 1994; McLendon et al. 2007; Nicholson-Crotty and Meier 2003). This has helped scholars move away from viewing postsecondary policy creation as solely influenced by economic and education related factors. The second influence of this conceptual framework comes from the policy diffusion literature, and involves interstate forces. This perspective argues that state officials and policymakers emulate, learn and compete with each other, resulting in policy innovations diffusing from one state to another (Berry and Berry 1990, 1992; Walker 1969).

These adoption studies differ from research that describe policies (e.g., Burke 2002; Breneman and Finney 1997; Hovey 1999; Hurley 2006; Olivas 2003) and those that focus on their impacts (e.g., Richardson 2005; Volkwein and Tandberg 2008; Lehman 1990; Dynarski 2004; Henry et al. 2004). Despite the recent turn to examining single policies, this paper is directly influenced by two seminal works in the postsecondary policy adoption literature.⁸ Hearn and Griswold (1994) examined states’ adoptions across the domains of “academic policies, policies relating to teacher education, and financing policies” all which were broad categories that contained individual innovations. McLendon et al. (2005) also explored several broad categories, including “Financing Innovation.” Using these wider definitions, both studies found interesting patterns related to the adoption of these policies. Most notably, McLendon et al. (2005) remains one of the few studies to find evidence of interstate diffusion.

In this study we aim to build upon these pioneering works through examining policy adoption in the ten years following McLendon et al.’s (2005) study and using techniques that may aid us in understanding the mechanisms at work at interstate diffusion and why it so rarely is found to influence adoption of postsecondary policies. Theoretically we differ as we hypothesize that what is important are the multiple and continuous actions toward the postsecondary financing, which places this study in literature conceptualizing the policy process as continuous (e.g., Anderson 2011; McLaughlin 1987). From this perspective problems are never “solved” and are therefore continue to reemerge. In addition to diffusion, we envision a variety of additional hypotheses regarding state adoption of postsecondary finance innovations. In specifying the mechanisms at work, we outline several broad categories and our rationale for including specific indicators.

Diffusion

Within the United States’ federated system of government, the 50 American states may function “laboratories of democracy,” in which new policies are enacted, tested, and possibly emulated. In the study of policy adoption, there are several theoretical arguments for what may contribute to policies’ spread across states. The broad form of this is the idea of policy “learning” and “emulation” in which effective policies are transmitted across states and (theoretically) ineffective policies are not. In his early analysis, Walker (1969) suggested that policymakers may simply seek “shortcuts” to aid them in their chaotic and busy roles; diffusion thus results from policymakers’ attempts to address the confusion inherent in their duties. In the second age of policy diffusion studies, much of the reasoning focused on competition between states. This conception is perhaps best represented by Berry and Berry’s 1990 study of lottery adoption, in which states suffer an economic disadvantage if a neighbor has a lottery and they do not. While within the political science and public policy literatures some support exists for the idea of interstate diffusion, the results in regard to postsecondary policy diffusion are less certain and, at best, mixed (e.g., McLendon et al. 2005; Doyle 2006).

As mentioned previously, this study departs from the traditional approach of studies of postsecondary policy adoption, moving from the study of individual policies towards a larger category. The primary motivation for this approach is to question the mechanism through which these policies diffuse while accounting for the density of activity in this broadened definition. To borrow from another theory of the policymaking process, we hypothesize that it may not be the “solutions” that diffuse, but rather attempts to address the “problems.” In utilizing this Kingdonian rationale, we believe that state policymakers are not responding to specific, codified policies but instead emulate their neighboring states’ attempts to address the financing of postsecondary education. For this case, it may not be that policymakers feel pressure to adopt a prepaid tuition policy because their neighbor recently adopted it, rather neighbors’ activities compel them to enact any policy that addresses college cost.

Postsecondary Environment

Principal among the factors one would believe to influence the adoption of postsecondary finance reforms are those surrounding the cost of public higher education within states. To address this, we include an indicator for tuition and another for need based aid. As the rationale behind many finance innovations involves rising tuition, we specify this variable as the 3-year moving average of the percent change in tuition at the state flagship institution.⁹ Thus, we suspect that states with rapidly increasing tuition will be more apt to adopt finance innovations. In addition to tuition,we include a states’ need-based grant coverage of Pell recipients. We argue that if the current state need-based grant program sufficiently offsets the cost of attending college for low and middle income students, the need and demand for new finance policies will be lower (Ness 2006). Activity in state need-based aid may also proxy an orientation to more traditional forms of higher education finance policy and make them less likely to engage in novel reforms.

In addition to tuition and aid, we also propose that policymakers use other indicators as a barometer for the health of postsecondary education in their state. Though there are many potential indicators explored in other research, the interrelated nature of many candidates makes an exhaustive list less appropriate in this study. The indicator we selected is out-migration, or the percentage of students from a state who pursue postsecondary education beyond its borders. We believe that states with high patterns of out-migration will be more likely to enact new finance policies in an effort to stem the movement of native postsecondary students out of the state (Doyle 2006).

Political Environment

States also vary on their political indicators, with the existing adoption literature finding overwhelming support for the influence of states’ political environments in the formation of higher education policies. First we consider unified party control of government. This hypothesis suggests that states in which the same party controls both houses of the legislature and the executive office are more likely to adopt a policy than those where control is split and greater roadblocks exist to passage (Huber et al. 2001). Following the conception of McLendon et al.’s (2007) study of non-directional governance reform, we believe that when state government is unified they will usher in new finance reforms.

Beyond the political control of state government, we include other aspects that characterize a state’s political landscape - notably those of analytic capacity and power. States vary in their level of legislative professionalism, a measure that attempts to capture the latent analytic capacity of a state’s legislative body. We propose that states with higher levels of legislative professionalism will better understand the landscape of policy tools available to address postsecondary finance and thus be more likely to adopt these policies (Barrilleaux et al. 2002; Squire 1993, 2007).

We also seek to highlight differences in states’ executive offices, hypothesizing that states where a greater degree of budgetary power is invested in the gubernatorial branch will be at a decreased risk of adopting a finance innovation. Here we propose that higher education finance innovations would codify into law a portion of state finances, thus limiting the ability of the governor to use his or her discretion to divert resources away from higher education and redistribute to other policy areas. Prior research has confirmed that variation in the institutional powers of governors impacts the likelihood of policy innovation in the states (Beyle 2004; Dometrius 1987). The majority of the innovations either require additional state resources (e.g., broad based merit aid), lock the state into a form of financial commitment (e.g., performance funding) or limit the state’s fiscal powers over higher education (e.g., tuition decentralization). There is little reason why a governor with significant budgetary powers would want to cede this power and thereby limit his or her ability to alter budgetary priorities in the future.

The relative strength of a state’s higher education interest group lobby may also be hypothesized to relate to the adoption of finance policies. Referred to as the higher education interest group ratio, it indicates the strength of the higher education lobby relative to the larger interest group universe in a given state. In large part, this measure is driven not only by institutions, but the more visible and prestigious institutions. While on one hand, one may suspect that the regulation of performance funding policies would oppose these measures—these institutions are likely those that will benefit from these changes. Also comprising this are public institutions and various advocacy groups that would stand to benefit from policies that benefit individuals.

Some research finds that gubernatorial tenure can influence policy outcomes, as newer governors seek to establish their own policy portfolio and expend whatever political capital they earned through their elections (e.g., Dometrius 1987; McLendon et al. 2007; Schlesinger 1965). Therefore we hypothesize that the longer a governor is in office in a given state the less likely that state will be to adopt a finance innovation.

Lastly we include an indicator for whether or not a state has a consolidated governing board. Traditionally the thinking around state governance agencies has fallen under the category of postsecondary environment, yet we believe these structures in part define the boundaries of a state’s “politics of postsecondary education.” Often referred to as a “fourth branch of government,” these consolidated governing boards wield a greater degree of power and influence when compared to less centralized governance structures. On one hand, some suggest these boards may act like academic cartels (McLendon et al. 2006; Zumeta 1996) in which finance policies may empower individual students or create incentives among its members to certain externally desirable outputs. On the other, these structures may view the adoption of finance innovations as an attempt to circumvent their power and authority over the state’s postsecondary system writ large. Under both of these views, we believe that states with consolidated governing boards are less likely to adopt a finance innovation (Doyle et al. 2010; McLendon et al. 2006; Deaton 2006).

Economic Factors

Prior research has found that economic conditions may increase the likelihood of policy adoption with median family income positively associated with the adoption of a number of postsecondary finance innovations (McLendon et al. 2005; Doyle 2006; Deaton 2006). Therefore our study controls for median family income and hypothesizes that it will be positively associated with the adoption of these policies. Middle class and wealthy families clearly stand to benefit from these policies, as they are more likely to attend college, to make use of the college savings and 529 plans and to benefit from the merit aid programs (Hurley 2006; Cornwell et al. 2009).

Data and Methods

The data for the dependent variable come from the aforementioned studies and were updated using reliable secondary sources and state legislative documents.¹⁰ Our definition of “innovation” draws from Walker’s 1969 articulation that defines innovations as when a state adopts something novel to itself and our model includes 47 states over the time period 1979–2008. Alaska and Hawaii were removed from the analysis because diffusion was specified as the total number of finance innovation adoptions by neighboring states. As is typical in comparative state studies, Nebraska was removed from the analysis due to its unique quality of having a unicameral legislature. We do include the finance innovations it adopted in the diffusion measure for its neighboring states. Our data begins in 1979 when Tennessee adopted its long-standing performance funding policy. The final adopter in our analysis is Florida’s 2006 tuition decentralization policy, though our time frame continues till 2008, the final year for which we have data for the dependent variable. In the years of the analysis, 131 policy innovations occurred with each state adopting at least two finance innovations. A limitation of our conceptualization is that states can and do adopt multiple finance innovations in a single year yet our model observes this phenomenon as single events, which brings the observed number of events to 116.¹¹ A list of these policies and the years that states adopted them may be found in Appendix 2.

Figure 1 presents a “heat map” for the total number of adoptions in the U.S. states, circa 2008. In this map, darker areas represent larger numbers of total adoptions and lighter areas fewer. As indicated by the dark shades, three states (Kentucky, Florida, and South Carolina) adopted five out of six finance innovations. Seven states Arizona, Delaware, Iowa, Montana, New Hampshire, Rhode Island, and Vermont only adopted a 529 savings plan and appear lighter on the map.¹²

Fig. 1

State adoptions of total finance innovations, 2008 (darker areas indicate a greater number of adoptions)

Independent Variables

Table 1 provides the means and standard deviations for the independent variables in the model. These data come from a variety of reliable secondary sources and are described below.¹³

Table 1 Summary statistics

The higher education interest ratio variable is constructed by dividing the total number of state public higher education institutions and registered non-college or -university public higher education interest groups by the total number of interest groups in the state, minus any registered colleges and universities or other registered higher education interests groups that may lobby for more money for public higher education (Tandberg and Ness 2011). The interest group data has been retrieved from state websites and government archives, from the Council on Governmental Ethics Laws (CGEL) Blue Book, and data provided by David Lowery. Data on the number of public institutions were retrieved from the National Center for Education Statistics’ Digest of Education Statistics.

Legislative professionalism is measured via the Squire Index (2007), which is a latent variable that seeks to account for the robustness of legislative staff resources available for policy development and deliberations. The measure is operationalized by accounting for legislative salary, how long they remain in session, and the number of staff assigned to individual legislators.

Budget powers of the governor is a scale of 0–7 and includes data from 1976 to 2004 across all 50 states. The items included are whether state agencies make requests directly to the governor or to the legislature; whether the executive budget document is the working copy for legislation or if the legislature can introduce budget bills of its own, or whether the legislature or the executive introduces another document later in the process; whether the governor can reorganize departments without legislative approval; whether revenue estimates are made by the governor, the legislature, or another agency, or if the process is shared; whether revenue revisions are made by the governor, the legislature, or another agency, or if the process is shared; whether the governor has the line item veto; and whether the legislature can override the line item veto by a simple majority. Each of these has a value of 0 or 1 (Tandberg and Ness 2011). The sources for the data are Council of State Governments’ The Book of the States, the National Association of State Budget Officers’ Budget Processes of the States, and The National Conference of State Legislatures data.

The governance variable is a dummy variable, where consolidated governing boards receive a value of 1 and other types of state boards (i.e., state regulatory coordinating boards, state advisory coordinating boards, and state planning boards) receive a value of 0. Data were gathered from the Education Commission of the States’ (ECS) website, ECS’s State Postsecondary Education Structures Handbook, and State Postsecondary Education Profiles Handbook.

Unified political control of government is a dummy variable indicating instances where the legislature and the governorship are controlled by the same party. The governor’s tenure measure the number of months the governor has been in office. These data were accessed via Klarner’s personal website, available at: www.indstate.edu/polisci/klarnerpolitics.htm (Klarner et al. 2013). The diffusion measure was derived using a computer program that counts the timing and spread of the individual policies that comprise finance innovations. These measures were then aggregated for our final indicator.

Given the longitudinal nature of our study, we wish to descriptively focus on four key variables that exhibit change over time and are particularly pertinent to finance innovations and state postsecondary education writ large. These variables are the state need-based grant coverage of Pell recipients, 3-year avg. % change in state tuition, median income, and out-migration.

The state need-based grant coverage of Pell recipients is expressed as a percentage and the data were collected from the Tom Mortenson’s Postsecondary Opportunity (www.postsecondary.org). The 3-year avg. % change in state tuition data were collected from the Digest of Education Statistics. Median income data were collected from the Bureau of Economic Analysis (BEA) and out-migration, or the percentage of students from a state who pursue postsecondary education beyond its borders, data were collected from www.higheredinfo.org which developed the measure using IPEDS data. Figures 2 and 3 attempt to show the movement of these variables over time. In both figures the grey dots represent our states at the beginning of the study, the black dots the constellation of states in 1995, and the abbreviations the values in 2008.

Fig. 2

Tuition and need aid

Fig. 3

Income and out-migration

In Figure 2, one notes that both the middle and end of our study were times of increasing tuition—especially compared to the late 1970s.¹⁴ Over the same time periods, one notices a contraction in the coverage of state need-based aid. These shifts, as we hypothesized in our conceptual framework, capture the movement from the traditional positions of states at the time of large-scale innovation. Figure 3, shows the increases in median income and student mobility. It is worth noting that our measure of income correlates with many measures of educational attainment found in other studies.¹⁵ Because the importance of income to our conceptual framework, we elected to include it rather than this class of variables.

Methods

For over two decades the field of political science has used the broad category of “Event History Analysis” to study state policy formation, an approach that was later adopted in the by researchers focusing on postsecondary education issues (e.g., McLendon et al. 2005; Doyle 2006; McLendon et al. 2006).¹⁶

Because we broadened the definition to “finance innovation,” it is possible all states may experience multiple adoptions over the time period of our study. We acknowledge that there are multiple estimators one could use to model this phenomenon, but due to the nature of our data, our question, and theoretical framework we use a Cox Proportional Hazards Model for repeating events.¹⁷ The Cox Proportional Hazards model is well suited to social science problems as it enables researchers to examine the effect of various covariates without specifying a parametric form for the baseline hazard rate. More importantly, the “repeating nature” of finance innovations cannot be ignored as we argue that interstate diffusion rests on both the mimicry of the broad category and the continual and repetitive nature of these “reforms.”

While studies focusing on adoptions of single policies remove a state from the analysis after the year of adoption as they are assumed to no longer be at “risk,” we include all 47 states throughout the duration of our time frame.

To address the issue of previous events influencing the probability of ensuing adoptions we use a conditional gap time Cox model that incorporates both “time from entry” and “time from the previous event” into estimation. That is, we believe that the act of adopting a fifth finance innovation is likely to have important differences from adopting the first innovation. To illustrate this process we turn to Tennessee which adopted its Performance Funding policy in 1979. After this adoption, the count for duration for Tennessee was reset to 1 in 1980, while other states’ duration continued to 2. This continued in Tennessee till its adoption of a 529 Savings Plan in 2000, after which t is again set to 1.¹⁸ To help understand this difference, the first equation presents the conventional Cox Proportional Hazards Model and the second the conditional gap time variant:

$$\begin{aligned} { h} (t)&= h_0(t)e^{\beta ' x_j} \end{aligned}$$

(1)

$$\begin{aligned} { h}_k (t)&= h_{0k}(t)e^{\beta ' x_{kj}} \end{aligned}$$

(2)

In both, $h(t)$ is the hazard of adopting a a policy at time t, $h_0$ the baseline hazard, $\beta 'x$ is the matrix of regression parameters and covariates and the subscript j the clustering of standard errors (Box-Steffensmeier and Jones 2004; Hosmer and Lemeshow 1999). The notable distinction between the two is the inclusion of the subscript k in the second equation which indicates the ${k}^{th}$ failure, or the number in the sequence of events. That it indexes both x and $h_0$ is not a trivial matter. While there are a number of methods for estimating repeating events, this varies the hazard rate by the events because we assume that experiencing the fifth event is substantively different than experiencing the first.

To account for the occurrence of tied events the Efron method was used. We selected the Efron method for its ability to allow us to cluster standard errors on states, a correction we believed needed to be done a priori. We also tested for the proportional hazards assumption, said plainly, the assumption that the effect of the covariates are constant through time. To test this assumption we conducted diagnostic tests of the Schoenfield residuals. Our indicator for interstate diffusion was found to violate this assumption and was corrected through an interaction with the natural log of time t.¹⁹

Findings

Table 2 presents the results from the Cox Proportional Hazards model, nesting the variables into groups that reflect our broad categories outlines earlier. For the remainder of the paper we focus on interpreting the full model that finds median income, outmigration, legislative professionalism, a governor’s budgetary powers, and the adoption activity of neighboring states to influence the risk of adopting a finance innovations. We caution readers to not interpret the variable interacted with time from the coefficients listed in the table, instead we direct you to the following graphical displays that provide the combined effects of the interacted terms.²⁰

Table 2 Results of the Cox Proportional Hazards Mode (standard errors in parentheses)

We first focus on our primary hypothesis of geographic diffusion. Initial diagnostics indicated that our measure for diffusion violated the assumption of proportionality. To address this we interacted the variable with the natural log of time, where time is denoted by 1, 2, 3...30. Rather than interpret the coefficients themselves, Fig. 4 displays the combined effect of our measure for contiguous diffusion and its interaction with time. In this figure, the x-axis is the year and the y-axis the corresponding estimation for the combined $\beta$ for that year. Table 3 presents the same data in tabular form.

Fig. 4

Interaction of diffusion and time

From Fig. 4, one sees that in the early years of the analysis our hypothesis and the traditional thinking surrounding diffusion is confirmed. That is, as the number of contiguous states adopting a finance innovation increases, so does the hazard of a state adopting any finance innovation. However, as we move through time this relationship diminishes with the 95 % confidence intervals containing zero from 1991–1997. Then, from 1998 to the end of the analysis the relationship reverses with the hazard of adopting a finance innovation decreasing as the number of neighboring states' innovations increases.

Table 3 Diffusion and time interaction

This finding underscores the complex relationship found in the diffusion of policies. Why would the relationship be positive in the early years but negative in later years? With this in mind we argue that this diffusion finding suggests competition in the early years and learning in later times. The former represents competition through simply responding to neighboring states’ actions in this policy environment. States may not be competing on tangible outcomes per se, but the appearance of being active. The latter may speak to a type of learning that stems from the long term implications of these policies. For many of our policies, research suggests that changes in the financing of postsecondary education can lead to less desirable behaviors at both the individual and institutional levels (e.g., Heller and Marin 2004; Burke 1998; Western Interstate Commission on Higher Education 2009; Cornwell et al. 2009; Dougherty and Reddy 2013). In light of this, we argue that as the second order effects of finance innovations became known, policymakers were in turn less likely to enact such changes because the consequences were understood.

Beyond our primary hypothesis of diffusion our model suggests several other findings that merit discussion. We find a positive relationship for legislative professionalism, supporting our thinking that more professionalized legislatures would have the necessary staff to digest policy activity occurring elsewhere, prompting these states towards adoption. Though not below the conventional $\hbox {p}<.05$ threshold, the results pertaining to governor’s budgetary powers were consistent with our initial thinking where governors with strong powers over the budget would be less likely to effectively cede power through codifying it through law. There is little reason why a governor with significant budgetary powers would want to limit the ability to change priorities in the near future.

Median income and outmigration were also found to influence the adoption of finance innovations. Our hypothesis was that higher levels of median income would be associated with adoption but our results point towards the opposite relationship. However, this notion is at the individual level which is best examined within the policy context of individual states. While research suggests that these policies benefit the middle to upper–middle classes, in this study we are effectively comparing state’s middle classes to one another; in light of this finding, future research may alter the question to ask “are states with less wealthy middle classes more likely to adopt these policies?”

We also find states with higher levels of out-migration are more likely to adopt finance innovations. This is in contrast to previous studies (e.g., Doyle 2006) that found counterintuitive, opposite effects; within our broadened definition we find support for the hypothesis that policies are enacted to retain students. One possible explanation is that the political will to enact a particular policy could differ among the states. That is, while a particular policy may have an easy path to adoption in one state a different finance innovation may better address out-migration elsewhere.

Conclusions

As indicated in the title, our motivation was to “rethink” the diffusion of postsecondary policies through both returning to the broad conceptualizations used in pioneering studies and using an estimator that more closely mirrors the theoretical conceptualization of a continuous policymaking process. By broadening the category of innovation we, in part, uncovered the frequently hypothesized effect of diffusion, though the variation of the effect over time suggests an even more complicated relationship; our model found the conventional, positive diffusion in the early years of the analysis, in the end years we found the opposite, negative relationship for contiguous adoptions. For future studies of policy adoption this “anti-diffusion” finding should compel researchers to redefine the exact mechanisms of diffusion by stating the underlying process they believe to influence the spread of policies. Is it interstate competition? If so, then there should be strong theoretical arguments for how competition manifests itself. If it is learning, one must acknowledge that decision-makers within states may learn from one another in multiple ways and seek to refine a measure that can account for the timing of this learning and how it too might spread. The question recalls Ness’ (2010) discussion of “research utilization” in the postsecondary policy process, in which he speculates the avenues through which research about postsecondary policies are disseminated to policymakers. How much of the “learning” might be attributable to Ness’ more technocratic conceptualization and to what extent are policies' effects transmitted through less formal means?

Beyond contributing to an understanding of interstate policy diffusion, this paper demonstrates the leverage researchers can gain over data in comparative state studies of postsecondary policies. Through studying policies as interconnected, we may better understand the process itself. Conceptually, this is a return to not only the early studies of postsecondary policy adoption, but the work of Walker and Gray in the late 1960s and early 1970s, an approach augmented through methodological advancements. Though we believe this is an advancement, we also wish to acknowledge what we believe to be the most flawed aspect of our modeling choice: the model does not differentiate between single and multiple adoptions in a given year, a phenomenon occurring approximately 13% in the period of our study. Logically it would follow that the outcome of adopting two finance innovations in a given year is substantively different than a single adoption. This echoes a perennial issue of studies of this nature that we, admittedly, do not resolve. While our multi-event study touches on “intensity” of policymaking by means of frequency over time, it does not differentiate between the strength of the policies or wholesale state postsecondary reform. As within many categories there can be policies “with teeth,” and we suggest that future researchers measure comprehensive policy reform. Future investigations may wish to build on this by accounting for the repeating events and their intensity, as well as using other models that account for unobserved heterogeneities.

Despite these shortcomings, broadening the categories of postsecondary policies will help researchers and policymakers alike to better understand this critical arena that shapes much of higher education. At a time when many state legislatures, governors, and postsecondary policymakers have once again begun to address the iron triangle of affordability, efficiency, and accountability, and adopted their own “completion agendas,” the spread of these priorities will undoubtedly shape the environment in which institutions operate. Directly related to policy adoption is the current, potentially shifting sands of policy diffusion in the postsecondary arena. While the current higher education reform movement is characterized by large foundations and intermediary organizations that attempt to provide cross-state technical assistance to state decision-makers—effectively hastening the diffusion of policy innovation. While this may represent a new facilitation for the spread of policies, our study suggests that these efforts that could enable or inhibit these goals.

Appendices

Appendix 1

See Table 4

Table 4 Policy sources table

Appendix 2

See Table 5.

Table 5 State policy adoptions

Appendix 3

See Table 6.

Table 6 Independent variable descriptions and sources