See discussions, stats, and author profiles for this publication at: Growing the Pie? The Effect of ResponsibilityCenter Management on Tuition RevenueWorking Paper · January 2017DOI: 10.13140/RG.2.1.2738.6489READS43 authors, including:Dennis KramerBradley R. CursUniversity of FloridaUniversity of Missouri8 PUBLICATIONS 2 CITATIONS23 PUBLICATIONS 128 CITATIONSSEE PROFILEAll in-text references underlined in blue are linked to publications on ResearchGate,letting you access and read them immediately.SEE PROFILEAvailable from: Dennis KramerRetrieved on: 25 June 2016
Growing the Pie? The Effect of Responsibility Center Management on Tuition RevenueOzan JaquetteAssistant ProfessorUniversity of Arizona, College of Education1430 E. Second Street, Room 327ATucson, AZ [email protected] A. Kramer IIAssistant ProfessorUniversity of Florida, College of Education293 Norman HallGainesville, FL [email protected] R. CursAssociate ProfessorUniversity of Missouri, College of Education202 Hill HallColumbia MO, [email protected] Review: Journal of Higher EducationThis version: June 21, 2016
AbstractResponsibility center management (RCM) budgeting systems devolve budgetresponsibility while creating funding formulas that create incentives for academic units togenerate revenues and decrease costs. A growing number of public universities adopted RCMover the past decade. The desire to grow tuition revenue is often cited as an important rationalefor adoption. However, prior research has not assessed the effect of RCM on institution-leveltuition revenue. Traditional regression methods which calculate "average treatment effects" areinappropriate because RCM policies differ substantially across universities. Therefore, this paperemploys a synthetic control method (SCM) approach. SCM approximates the counterfactual foran RCM adopter by creating a synthetic control institution composed of a weighted average ofnon- adopters. Importantly, SCM estimates the effect of RCM separately for each adopter ratherthan estimating the average effect across multiple institutions that adopted RCM. We use SCMto analyze the effect of RCM adoption on tuition revenue at four public research universities thatadopted RCM between 2008 and 2010. We find modest evidence that RCM positively affectedtuition revenue at Iowa State University, Kent State University, and the University of Cincinnati,but no evidence of a positive treatment effect at the University of Florida.
Growing the Pie? The Effect of Responsibility Center Management on Tuition RevenueThe fiscal landscape facing U.S. public universities has changed significantly since theearly 2000s. Uncertain state appropriations have forced public universities to seek alternativerevenues from research funding (Slaughter & Leslie, 1997), auxiliaries (Barringer, forthcoming),and donations and financial investment (Cheslock & Gianneschi, 2008). However, tuitionrevenue is the largest source of revenue growth for most public universities (Desrochers &Hurlburt, 2016). Most scholarship on efforts to increase tuition revenue focuses on initiativesplanned by central administration, for example tuition discounting (Hillman, 2012) and thepursuit of nonresident students (Jaquette & Curs, 2015).However, in addition to these centrally-led efforts, public universities are managingchanging fiscal conditions through what Meek (2010, p. 2) describes as a “centralizeddevolution.” In particular, a growing number of U.S. public universities have replaced“incremental” budget systems – in which the annual budget for an academic unit is largely basedon prior year’s budget – with decentralized budget models that devolve both budgetresponsibility and decision-making authority in ways that motivate lower-level actors to meetlarger organizational goals (Deering & Sa, 2014; Meek, 2010). “Responsibility centermanagement” (RCM) is the most popular decentralized budget model (Curry, Laws, & Strauss,2013; Hearn, Lewis, Kallsen, Holdsworth, & Jones, 2006). RCM devolves budget responsibilityto academic units and develops funding formulas that create incentives for academic units togenerate revenues and decrease costs. Figure 1 demonstrates the growth in the number of publicRCM adopters in the U.S., showing an increase from 10 in 2006–07 to 24 in 2014–15.According to Curry et al. (2013), the growing popularity of RCM is a response to uncertainty
about state and federal funding, growing tuition reliance, and the increasing administrativecomplexity of research and degree-granting programs.While several administrators and consultants have lauded the benefits of RCM (e.g., GrosLouis & Thompson, 2002; Hanover Research Council, 2008; Strauss & Curry, 2002), empiricalresearch has yet to rigorously assess the effects of RCM adoption. Given the substantial timeand financial costs of adoption (Priest, Becker, Hossler, & St. John, 2002), it is important forprior and potential adopters to understand the consequences of RCM. Extant empirical researchtypically utilizes a qualitative, single-institution case study methodology primarily focused onthe implementation of RCM (e.g., Gros Louis & Thompson, 2002; Hearn et al., 2006; Lang,2002). Missing from this literature are systematic analyses that assess the causal effects of RCMon specific outcomes. In particular, the desire to increase tuition revenue by motivatingentrepreneurial behavior was a commonly cited rationale for adoption (e.g., Kent StateUniversity, 2007; University of Florida, 2009), but prior research has not assessed the effect ofRCM on net tuition revenue.Traditional regression methods that estimate average treatment effects are inappropriatefor studying the effects of RCM for two reasons. First, regression methods generally require alarge number of “treated units,” but relatively few public universities have adopted RCM.Second, regression methods calculate an average treatment effect under the assumption that thetreatment is similar across units, but RCM policies differ substantially across universities anduniversities modify their RCM policies over time. The synthetic control method (SCM) wasdeveloped by Abadie and Gardeazabal (2003) to make causal inferences about idiosyncratic orrare treatments. SCM is a quasi-experimental, quantitative methodology that incorporates designelements from qualitative comparative case studies to create a synthetic control unit that
approximates the counterfactual for the treatment unit. Importantly, SCM estimates the effect ofa treatment on one treated unit rather than estimating the average effect of a treatment formultiple units that receive a “similar” treatment.This study uses SCM to analyze the effect of RCM adoption on tuition revenue at fourpublic research universities—Iowa State University, the University of Cincinnati, Kent StateUniversity, and the University of Florida—that adopted RCM between 2008–09 and 2009–10.The model results indicate RCM positively affected tuition revenue at Iowa State University, theUniversity of Cincinnati, and Kent State University. However, the strength of the results suggestthat we should be cautious about making strong causal claims about the effect of RCM at theseinstitutions. Additionally, the models provide strong evidence that RCM adoption did notsubstantively affect net tuition revenue at the University of Florida.This study makes three contributions to the higher education research. First, it provides arigorous, quantitative evaluation of the effect of RCM on tuition revenue at four institutions.Second, our analyses create an analytic framework for future research on RCM that canincorporate more institutions and assess additional outcomes. Third, because SCM can beapplied to many higher education research topics, this manuscript describes SCM methodologyin detail to help higher education researchers implement the method responsibly.The following section reviews empirical literature on RCM. Second, we describe themajor components of RCM—including how these components differ across institutions—inorder to develop a conceptual explanation for how RCM could affect tuition revenue. Third, wediscuss the methods, and fourth, we present the results. The manuscript concludes by discussingfuture research on RCM and future applications of the synthetic control method.
Literature ReviewWe reviewed empirical research on the effects of RCM, emphasizing scholarship onrevenue and enrollment outcomes. Most scholarship on RCM is qualitative and focuses onimplementation at a single institution (e.g. Lang, 2002; Courant & Knepp, 2002; Gros Louis &Thompson, 2002). Typically, these case studies describe RCM policy details, implementationschallenges, campus reception, and the benefits and drawbacks of RCM across a broad range ofoutcomes. For example, Gros Louis and Thompson (2002) analyzed RCM implementation at theUniversity of Indiana, drawing largely from interviews of campus stakeholders about theirperceptions of RCM. The authors stated that RCM provided benefits such as increased studentsatisfaction, created incentives to generate income, and encouraged bottom-up decision makingand long-run planning. However, potential drawbacks included reduced collaboration betweenschools, pressure on schools to offer more courses, and encouragement of grade inflation.Only a few studies have examined the effect of RCM on tuition revenue or enrollment.The most rigorous analysis is Hearn et al. (2006), a mixed-methods analysis of the 1997–98RCM implementation at the University of Minnesota. For academic units, the correlationbetween full-time equivalent enrollment and operating revenue increased following RCMimplementation, suggesting that the transition to RCM increased the payoff for enrollmentgrowth. Related, colleges that grew enrollment experienced larger budget increases thanacademic units with stagnant enrollment. Interview data suggested that RCM increased theentrepreneurial and decision-making capacities of individual academic units. However, theanalyses did not show whether RCM affected university-level revenue generation.Other publications make claims about the effect of RCM on institutional revenues basedon potentially problematic empirical methodologies (e.g., Hanover Research Council, 2008;
Lang, 1999). For example, the NACUBO monograph by Strauss and Curry (2002) sought toanswer the question, “How is responsibility center management working?” (p. viii). However,the methodological basis for their claims often relied on descriptive data before and after RCMimplementation. This approach fails to account for change over time in factors—aside fromRCM—that affect university budgets. As a specific example, the authors used the University ofDenver to support their claim that RCM increases institutional revenue-generating capabilities:Indicators of financial health were pointing and moving in the wrong direction . . . but theimpacts were not immediately visited upon the schools responsible. Denver changed therules, forcing the coupling of choice with consequences, by implementing RCM.Denver’s financial (and academic) health indicators have been pointing strongly upwardsever since. (p. 32)Similarly, Gros Louis and Thompson (2002) used descriptive statistics before and afterRCM implementation as evidence that RCM positively affected student satisfaction. The authorsstated that “such improvements may have taken place without RCM, but RCM surely was acontributing factor” (p. 95).Given the growing number of RCM adopters and the cost of implementation, the currentstate of scholarship on RCM is lamentable. On one hand, pro-RCM publications, often writtenby consultants and targeted at administrators (e.g., Curry et al., 2013; Hanover Research Council,2008), use anecdotes and descriptive data as evidence of the positive effect of RCM. On theother hand, anti-RCM essays often criticize the “corporate” ideology underpinning RCM, butlack empirical evidence to support their gloomy assessments (e.g., Adams, 1997; Dubeck, 1997;Murray, 2000). Empirical case studies have provided valuable insights by analyzing particularRCM policy details (e.g., Courant & Knepp, 2002; Hearn et al., 2006; Lang, 2002). However,
these studies analyze one institution and many outcomes. Therefore, it is difficult to compareresults across studies to identify the systematic effect of RCM on intended outcomes. Noempirical research has established a causal link between RCM adoption and tuition revenuegeneration, which is often cited as a rationale for adoption (Curry et al., 2013). Therefore, thegoal of this study was to conduct the first systematic quantitative analysis of the effect of RCMadoption on tuition revenue.Policy Background and Conceptual FrameworkThis section briefly describes incremental budgeting, as most RCM adopters havetransitioned from incremental budgeting to RCM. Next, we discuss the major components ofRCM and how these components differ across institutions, and conclude by describing thepotential mechanisms linking RCM to tuition revenue.Historically, most public universities have utilized some form of centralized budgeting, inwhich central administration owns all unrestricted revenues and decides how to allocate theserevenues to academic units (Curry et al., 2013). Incremental budgeting is the dominantcentralized budgeting model. Here, the annual budget allocated by central administration to eachacademic unit is largely based on the amount allocated last year and is relatively unresponsive toyear-to-year changes in unit performance. The benefits of incremental budgeting arepredictability and simplicity. However, Curry et al. (2013) highlighted several criticisms ofincremental budgeting. First, the disconnect between output production and budget allocationcompels academic units to view annual budgets as an “entitlement,” thereby discouragingentrepreneurial behavior. Second, incremental budgeting requires central administrators toacquire substantial knowledge of the complexities of each academic unit because central
administration assumes responsibility for the budgets of academic units. Third, the opaquenessof incremental budgeting makes budget allocation susceptible to political power dynamics.RCM, the most popular decentralized budget model (Curry et al., 2013), is based on theidea that increasing university-level resources depends on creating incentives for organizationalsubunits to become more focused on growing revenues and minimizing costs. Commonly citedgoals of RCM include increasing responsibility of lower-level actors by devolving ownership ofrevenues and costs; increasing lower-level planning capacity through (a) better information and(b) by clarifying the links between activities and associated revenues and costs; growing revenueby encouraging financially viable entrepreneurial activities; increasing efficiency of nonacademic units; and increasing transparency about budget allocations (Curry et al., 2013; Hearnet al., 2006; Strauss & Curry, 2002).Major Components of RCMTable 1 describes the RCM systems of the four universities analyzed in this paper—IowaState University, the University of Cincinnati, Kent State University, and the University ofFlorida—with an emphasis on the revenue components of RCM. The initial step in RCMimplementation is deciding what role each academic and non-academic unit will play in theRCM system. First, “responsibility centers”— the fundamental entity in RCM—areorganizational subunits that assume responsibility for generating sufficient revenues to covertheir costs. For most RCM adopters, including the four universities in this study, academiccolleges (e.g., the College of Science) become responsibility centers. Second, RCM “supportunits” are non-academic units (e.g., student services, library, IT, facilities) that provide servicesto academic units and are folded into the RCM economy; the services they provide generaterevenue for the non-academic unit and incur costs for academic units. Third, some universities
(e.g., Iowa State University) fund particular “administrative units” outside the RCM economy.Fourth, revenue-generating auxiliaries such as dining and athletics that do not serve academicunits are often excluded from RCM (e.g., University of Cincinnati).RCM allocation formulas determine the revenue to each unit and create incentives forresponsibility centers to achieve goals valued by central administration. RCM creates incentivesfor academic units to increase enrollment through revenue formulas that allocate tuition revenueto academic units on the basis of credit hour production and the number of majors. For instance,undergraduate tuition revenue at Iowa State University is split by allocating 75% of tuitionrevenue to the instructional unit and 25% to the academic major. RCM funding formulas for theallocation of graduation tuition revenue often place higher weight on majors whereas formulasfor undergraduate tuition revenue often place higher weight on credit hours (e.g., Iowa State andKent State). The rationale for the different treatment of undergraduate and graduate tuitionrevenue is that individual academic units can directly increase the number of graduate studentsbut have less control over total undergraduates.State appropriations revenue is often treated the same as tuition revenue. However, atIowa State University, state funds are combined with tuition revenue only after paying for thePresident’s Office and for the RCM investment/subvention fund (described below). Similar tothe allocation of tuition revenue, the rationale for allocating state appropriations to academicunits on the basis of credit and major generation is to incentivize academic units to increaseenrollment, which in turn should lead to increased university-level tuition revenue.Though not the focus of this study, RCM purports to incentivize cost containment byrequiring that academic units pay for both their direct personnel costs and the costs of servicesprovided by support units (Curry et al., 2013). RCM cost funding formulas dictate how much
academic units pay support units for services. At most RCM universities, cost formulas vary bysupport category (e.g., facilities, IT), thereby creating incentives to be more efficient aboutspecific activities. Finally, most RCM systems divert revenues to a fund that subsidizes unitsrunning a deficit and/or investing in areas of targeted growth.Potential Effect of RCM on Tuition RevenueHaving described the major components of RCM, we motivate the potential effects ofRCM adoption on tuition revenue. The hypothesis that RCM positively affects net tuitionrevenue is motivated by the “new public management” approach to the provision of publicservices (Meek, 2010). Drawing from principal agent theory (Hansmann, 1980), new publicmanagement argues that public bureaucracies are inefficient because the incentives of lowerlevel actors are not aligned with the goals of the principal (e.g., policymakers). Proponents ofnew public management argue that public bureaucracies can become more efficient by creatingmarket mechanisms that align the goals of agents with those of the principal (Walsh, 1995).From this perspective, RCM devolves budget decision-making authority and responsibility bycreating funding formulas that increase the incentive for academic units to achieve revenue andcost goals valued by central administration. In particular, funding formulas that reward credithours and majors encourage academic units to develop classes and majors demanded by students.Proponents of RCM argue that, in aggregate, the efforts by individual academic units to increasemajors and credit hours will positively affect institution-level tuition revenue (Curry et al., 2013).RCM could also have a null effect on tuition revenue. First, RCM funding formulas forundergraduate tuition revenue incentivize majors and credit hours. However, individualacademic units typically do not affect the total number of undergraduates because undergraduaterecruitment is typically the domain of centralized support units. Therefore, a criticism of RCM
is that academic units compete for undergraduates already enrolled as opposed to competing fornew students that would increase institution-level tuition revenue (Priest et al., 2002). Second,students typically do not pay additional tuition for credit hours in excess of some threshold (e.g.,12 credits per semester), such that convincing a student to enroll in 15 credits rather than 12credits does not garner additional tuition revenue. Third, RCM formulas create incentives at theacademic unit-level (e.g., a college of education), but not necessarily at the academicdepartment-level which is ultimately responsible for teaching.Overall, the effect of RCM on tuition revenue is unclear and must be assessedempirically. Furthermore, because RCM policies differ substantially across institutions, theeffect of RCM on tuition revenue likely differs across adopters. Therefore, analyses of the effectof RCM on tuition revenue should be conducted separately for each adopter.Empirical MethodologyThe primary challenge of estimating policy effects is the creation of the counterfactual,the outcome that would have occurred in the absence of policy adoption (Khandker, Koolwal, &Samad, 2010). Analyses of policy adoption effects often utilize the difference-in-differenceestimator, which estimates a counterfactual on the basis of the behavior of a set of comparisonunits by assuming that change over time for these units represents what would have happened tothe adopting units in the absence of adoption. However, difference-in-difference is inappropriatefor analyses of RCM because this strategy estimates average treatment effects acrossuniversities, but RCM policies differ dramatically across universities. Thus, difference-indifference estimates of the effect of RCM are likely biased toward zero because the estimates arean average of the effect of effective RCM models and the effect of ineffective RCM models.
This study utilized the synthetic control methodology (SCM) to estimate the effect ofRCM adoption. SCM was introduced by Abadie and Gardeazabal (2003) to analyze the effectsof terrorist conflict in the Basque Country on economic production. Subsequent articles byAbadie and colleagues further developed the method and provide guidance for itsimplementation (Abadie, Diamond, & Hainmueller, 2010, 2011, 2015). In essence, SCMcompares the change in outcome for the treated unit after treatment to a “synthetic control” unit,which approximates the counterfactual for the treated unit. SCM creates the synthetic controlunit (e.g., a university that adopted RCM) from a weighted average of non-treated units (e.g.,universities that did not adopt RCM). During the pre-treatment period, the outcome values of thesynthetic control should mirror those of the treated unit. In post-treatment periods, the treatmenteffect is defined as the difference in the outcome between the treated unit and the syntheticcontrol unit.Researchers have used SCM to analyze rare and/or idiosyncratic treatments, such as theeffect of earthquakes (Barone & Mocetti, 2014), the effect of winning a major academic award(ieg., the John Bates Clark Medal in economics) (Chan, Frey, Gallus, & Torgler, 2014), theeffect of affirmative action bans (Hinrichs, 2012), and the effect of the federal auto industrybailout (Fremeth, Holburn, & Richter, forthcoming). SCM is attractive for analyzing the effectof RCM because it estimates the effect of a treatment for one treated unit rather than estimatingan average treatment effect across many treated units. Below, we describe the data and RCMadopters analyzed and then discuss the implementation of SCM in detail.DataWe created an institution-level panel dataset using data from the IntegratedPostsecondary Education Data System (IPEDS) and state- and county-level population data from
the U.S. Census. Although RCM creates incentives for academic units to change behavior, weanalyzed the effect of RCM on institution-level outcomes because academic unit-level data wereunavailable.Dependent variable. The dependent variable was tuition revenue. This choice wasmotivated by several factors. First, public universities cite tuition revenue growth as an importantrationale for RCM adoption (Curry et al., 2013). Although RCM revenue formulas generallyreward credit hours and majors, these formulas are premised on the assumption that if academicunits grow credit hours and majors, then institution-level tuition revenue will increase (Curry etal., 2013). Second, SCM typically requires a long pre-treatment period, and tuition revenue canbe measured consistently from 1986–87 to 2013–14. For several alternative outcomes (e.g.,annual credit hour production), IPEDS did not begin collecting data until the early 2000s.Ideally, the outcome would be tuition revenue net of all scholarship allowances(NACUBO Accounting Principles Council, 1997), which is the difference between the stickerprice and the amount paid on behalf of the student. IPEDS did not begin measuring net tuitionrevenue for public universities until Government Accounting Standards Board (GASB) 34/35standards were phased in starting in 2001–02. To create a measure that was consistent acrossaccounting standards, we defined tuition revenue as gross tuition revenue net of unfundedinstitutional aid. For pre-GASB 34/35 years, this measure was defined as gross tuition minusexpenditure on unfunded institutional aid (Delta Cost Project, 2016). For GASB 34/35 years,this measure was defined as net tuition revenue plus scholarship allowances applied to tuitionminus expenditure on unfunded institutional aid.RCM adopters. Data on which universities adopted RCM and the timing of adoptionwere based on Curry et al. (2013), supplemented with information from websites, reports, and
press releases of individual institutions. We analyzed the effect of RCM adoption at Iowa StateUniversity, the University of Cincinnati, Kent State University, and the University of Floridabecause these four universities adopted RCM between 2008–09 and 2009–10, thereby allowing asufficient number of post-treatment years to estimate the treatment effect and a sufficiently longpre-treatment period for SCM.The Synthetic Control MethodologySCM notation and overview. Here, we describe the implementation of SCM in detail.Following Abadie et al. (2010), assume there are ! 1 institutions (e.g., universities), with 1referring to the treated unit (e.g., the university that adopts RCM) and 2, , ! 1 referring tountreated units in the “donor pool,” which is the sample of untreated units drawn from to createthe synthetic control unit. The number of observed pre-treatment periods is )* and the numberof post-treatment periods is ) , such that the total periods observed is ) )* ) . Let ,-. referto the outcome for unit at time /.For each post-treatment period, / )* , we would like to estimate the causal effect, 1 . 1 23 4 , , 1 2 , as:6 . , . , . 1 ,where , . for periods / )* represents the counterfactual, defined as the post-treatment outcomefor the treated unit had the treated unit not received the treatment. Since counterfactuals do notexist, SCM creates a “synthetic control” unit, a weighted combination of the ! 2, ! 1 unitsfrom the donor pool, to approximate the counterfactual for the treated unit. The intuition behindSCM is that a combination of untreated units can yield a better approximation of thecounterfactual than any single untreated unit. Creating a sound synthetic control unit depends on
(a) choosing an appropriate donor pool and (b) selecting the weight, :- , applied to each unit inthe donor pool.Choosing the donor pool. The first step in SCM is choosing the donor pool. Becausethe synthetic control is meant to approximate the outcome values of the treated unit in theabsence of treatment, the donor pool should be restricted to untreated units for which the value ofthe outcome variable is driven by “the same structural process” (Abadie et al., 2015, p. 497) asthe treated unit. Whereas traditional regression methods assume a probabilistic sample, choosingthe SCM donor pool requires judgment about which untreated units should be included orexcluded from contributing to the counterfactual (Abadie et al., 2015). However, these decisionsmust be rationalized with respect to the assumptions of the method.SCM relies on four assumptions (Fremeth et al., forthcoming), three of which are relatedto choice of donor pool. First, the donor pool should contain a sufficient number of untreatedinstitutions that are similar to the treated unit with respect to pre-treatment outcome values andpre-treatment determinants of the outcome. Second, SCM assumes that the treated unit does notexperience post-treatment shocks to the dependent variable that were not experienced by thedonor pool, and vice versa. To satisfy these two assumptions, researchers should excludeuntreated units that experienced idiosyncratic shocks in the years leading up to the treatment orin the years following the treatment (Abadie et al., 2015). Third, SCM assumes that untreatedunits are unaffected by the treatment.Because this study analyzes the effect of RCM adoption at four public researchuniversities, our potential donor pool consisted of the population of non-military, undergraduateserving public research universities (defined as research-extensive or research-intensive by the
2000 Carnegie Classification) that had not adopted RCM prior to 2013–14, the end of theanalysis period (N 140).We investigated each institution in this potential donor pool and excluded institutions thatexperienced idiosyncratic shocks to the outcome prior to or after to treatment period. First, weexcluded public research universities from L
Responsibility center management (RCM) budgeting systems devolve budget responsibility while creating funding formulas that create incentives for academic units to generate revenues and decrease costs. A growing number of public universities adopted RCM over the past decade. The desire t