SENSITIVITY OF STUDENT ENROLLMENT TO TUITION IN HIGHER EDUCATION THE CASE OF THE CALIFORNIA STATE UNIVERSITY SYSTEM Sue Youn Kim, Universiy of Souhern California, CA 90089, 323-442-2000, sueyounk@usc.edu ABSTRACT Changes in uiion and he corresponding impac hey possess on enrollmen have been a maer of some concern for differen insiuions of higher educaion over he las weny years. Sudies of he relaionship beween uiion and enrollmen are relevan for boh public and privae schools, bu schools funded by he sae may find he resuls of such research even more inriguing. The purpose of his paper is o measure he impac of uiion increases a he 23 California Sae Universiy campuses on enrollmen. Resuls of he research did no find large uiion hikes o have a subsanial impac on enrollmen. INTRODUCTION The sudy of uiion changes and he corresponding impac on enrollmen a differen insiuions of higher educaion has drawn aenion over he las weny years. Such research is beneficial, no only o sudens wih heir financial planning, bu also o he insiuions hemselves, who mus deermine wheher rising uiion levels affec he flow of funds. This would arac even more aenion when, a imes of economic recession and budge crunches, he absence of sufficien sae and federal funds may force insiuions of higher educaion o consider resoring o higher uiion. While he findings concern boh public and privae schools, schools funded by he sae may find his sudy even more inriguing. The purpose of his paper is o measure he impac of uiion increases a he 23 California Sae Universiy campuses on enrollmen. Residenial suden uiion and enrollmen daa for all 23 campuses, as well as couny-level daa peraining o income and employmen, was colleced for he years 2000 hrough 2009. The analysis and findings of he enrollmen variable, which is specified as a funcion of uiion, payroll, and he unemploymen rae, will be presened in his paper. Resuls of he research did no find large uiion hikes o have a subsanial negaive impac on enrollmen. Review of he Lieraure Given he frequen and somewha large increases in he levels of uiion during he las weny years, a major concern has been wheher access o public higher educaion has been negaively impaced. A he privae level, Vasigh and Hamzaee [2003] found ha uiion consideraions seem o have very lile impac on sudens enrollmen decisions. Mulivariae analyses examining he relaionship beween uiion and enrollmen commonly fall under wo caegories: cross-secional and ime-series sudies. Cross-secional sudies examine individual suden behavior in he face of various pos-secondary opions while ime-series sudies analyze changes over ime in aggregae suden enrollmen [Heller, 1996]. The Unied Saes Governmen Accounabiliy Office [2007] issued a sudy on he enrollmen levels on schools of various ypes hroughou he counry. Their findings indicae ha uiion levels coninue o rise, bu he enrollmen paerns vary depending on he ype of insiuion and educaional expendiures.
Tuiion in privae insiuions increased he mos in dollars, bu increased he mos in percenage in public insiuions. Despie rising uiion levels, more sudens are enrolling in colleges han ever before. Heller [2001] used financial aid as a criical facor for deermining suden enrollmen levels. In he summary of he exising research, he paper clarifies ha educaion is considered a normal good. Tha is o say, as price rises, individuals are less likely o consume more of i, all oher hings equal, price in his case being he uiion and he good iself being he educaion offered a a higher insiuion. The findings however, sugges ha alhough uiion has been rising, public higher educaion in California is sill affordable relaive o oher saes. As such, California is performing well in erms of moving sudens o enroll and obain bachelor s degrees. The empirical findings in Heller [1997] indicaed ha higher college prices would reduce he probabiliy of enrollmen. The following key observaions were made in his review: 1. In response o every $100 increase in uiion, a drop of 0.50 o 1 percen in enrollmens migh be seen across all ypes of insiuions. 2. Enrollmens are more sensiive o gran awards han o work sudy or loans. 3. In comparison o middle and high-income families, low-income sudens were more sensiive o uiion and financial aid. 4. Similarly, black sudens were more sensiive o uiion and aid han were whie sudens. 5. Finally, communiy college sudens show more sensiiviy o uiion and aid in comparison o sudens of four-year public colleges and universiies. Using daa from he inegraed Possecondary educaion sysem, Hemel and Marcoe (2008) examine rends in uiion a public universiies and approximae effecs on enrollmen. They found no evidence o sugges ha especially large increases from one year o he nex have a disproporionaely large negaive effec on enrollmen. Similarly, Shim and Milon [2006] conduced a sudy direcly on a public insiuion, aking wage premiums, financial aid, and unemploymen levels ino consideraion. The resuls of heir research indicae ha uiion does no have an effec on he growh of enrollmen. The uiion of compeing insiuions, however, was shown o have a posiive, and significan, effec on college enrollmen growh. Daa, Daa Sources, and Variables In he curren paper, hree main variables are used in consrucing he model o measure uiion impac on enrollmen. Enrollmen and uiion daa from he years 2000 o 2009 were rerieved from he California Sae Universiy online archives. Of he weny-hree campuses in he CSU sysem, only six are quarer based, wih he res being semeser based. Addiionally, here are missing observaions in he records, which lead o significan gaps in he daa. Thus, o compensae for he differences and he missing inpus, he daa are reorganized o include only Spring and Fall erm daa o use in he regression analysis. Given ha previous sudies incorporaed variables ha represen some form of income for sudens, he second variable ha is used in his sudy is annual average payroll. The couny locaions of all weny hree campuses were idenified, and he corresponding annual payroll daa was hen rerieved from a California financial archive. To esimae, i sands o reason ha payroll from he previous year would be
a facor in deermining curren enrollmen levels. Therefore, payroll daa from he years 1999-2008 were exraced. The daa was hen divided by he oal couny populaion o generae daa on a per capia basis. Similarly, counywide unemploymen levels were rerieved from he Labor Marke Informaion source. As unemploymen raes are volaile from monh o monh, monhly saisics are used in place of yearly saisics. Addiionally, i is assumed ha sudens would normally ry o predic where unemploymen raes are heading by observing a hisorical rend. As such, he average unemploymen rae of he las hree monhs before he beginning of a erm is used, insead of adoping he rae for he monh ha immediaely precedes a erm. Framework Model The calculaions in his research include undergraduae sudens a he weny hree California Sae Universiy campuses (referenced as CSUj). Based on he demand heory and he previous sudies reviewed here, enrollmen a colleges and universiies can be expressed as: ( T, U Y ) E = f, (1) which in urn can be expressed as he linear funcion where: j ( T ) + β ( U ) ( Y ) E β 0 + β1 2 + β 3 = (2) E, = suden enrollmen a campus period j j = he weny hree California Sae Universiy campuses (1,2,,23) = he ime period measured in Fall and Spring quarers/semesers from 2000-2009 (more specifically 20009-20091, where 20009 is he Fall semeser/quarer for he year 2000, and 20091 being Spring quarer/semeser for he year 2009) T, = uiion per fall or spring quarer/semeser a campus j and period U, = he civilian unemploymen rae of he couny wih school j a period j Y, = he per capia annual couny payroll of he locaion of he corresponding campus j a period. j B 0 = he inercep Alernaively, esimaions are also conduced based on he widely used log-linear funcional form of equaion (2), in which he coefficiens represen he elasiciy esimaes: log[ E ] = β 0 ' + β1 log[ T ] + β 2 log[ U ] + β3 log[ Y ] (3) In order o find he bes resuls, his paper esimaes equaions (2) and (3) using ordinary leas squares (OLS) wih he addiion of an error erm in each. Using hypohesis esing, he noion ha uiion did no significanly impac enrollmen became he null hypohesis, while he alernaive was he noion ha i did impac enrollmen significanly. To capure possible fixed effecs from ime-specific and/or crosssecion-specific facors, eigheen 0-1 ime dummy variables, and weny hree 0-1 campus dummy
variables were creaed. The ime dummies represen he eigheen quarers during he nine year span, while he campus dummies represen he weny hree campuses in he California Sae Universiy sysem. Despie he fac ha he Channel Islands campus did no open unil 2002, he regression was noneheless conduced wih blank enries for he firs wo years under CalSae Channel Islands. Analysis A muliple regression analysis was conduced o es he impac of uiion on enrollmen levels. For he regression, hisorical daa on enrollmen and uiion from he years 2000 o 2009 was rerieved from he main California Sae Universiy websie. While seveneen campuses operae under he semeser forma and six under he quarer sysem, he daa per year for each school was divided ino hree academic seasons (Fall, Winer, and Spring). Of hese seasons, only he daa for Fall and Spring were used. Addiionally, he specific couny locaion of each CSU campus was recorded and he corresponding payroll and unemploymen daa was also rerieved. The able below presens he esimaed coefficiens under equaion (2), as well as he significance levels for he hree main variables using ime series daa for he period of 20009-20091 (ha is, he Fall semeser/quarer of 2000 hrough he Spring semeser/quarer of 2009) and across he 23 CSU campuses. TABLE 1 Regression Resuls of Equaion (2) M1 M2 M3 M4 TUI UNEMP PAYROLL -1.358 (0.206) 234.819 (0.040) 196.822 (0.006) -2.938 (0.051) -733.437 (0.001) 297.033 (0.000) -0.006-1.677 (0.984) (0.120) 14.084 (0.846) N/A 268.645 (0.000) N/A Where R^2 0.937 0.224 0.933 0.935 F 129.001 5.625 212.996 131.782 N 409 409 409 409 M1: Model 1, he full model wih all hree main variables and dummy variables M2: Model 2, he model excluding he campus dummy variables M3: Model 3, he model excluding he ime dummy variables M4: Model 4, he model using only uiion (no ime or campus dummies uilized) I should be noed ha wih he excepion of Model 4, one ime 0-1 dummy and/or one campus 0-1 dummy variable was excluded from he regression in order o provide more accuracy. Addiionally, while here should be a sample size of 413 for each model, he daa for he firs four enries for CSU Channel Islands is missing. Therefore, four enries were deleed from he regression, hus resuling in a smaller han expeced sample size. The resuls above indicae ha uiion, in comparison o unemploymen rae and payroll, is an insignifican facor in deermining enrollmen levels. The coefficiens in each model are indeed negaive,
indicaing a negaive relaionship, bu he reliabiliy of uiion as a facor is quesionable as a resul of he significance levels. Even Model 4 (in which uiion is he only variable in he model) suggess ha uiion is only marginally significan as a predicive variable. In all four models, payroll possesses he highes level of significance, indicaing ha i impacs enrollmen more heavily han any oher facor. The posiive coefficiens also show a posiive relaionship; as payroll levels increase, he likelihood of enrollmen a a California Sae Universiy campus also increases. The same process was repeaed under equaion (3), he logarihmic model. The condiions for he firs se of resuls also hold for he logarihmic models. The daa for CSU Channel Islands sill has no records for he firs wo years, resuling in a gap and a smaller sample size. Time dummy and campus dummy variables were also removed o provide accuracy. The following char displays he resuls of he regressions. TABLE 2 Regression Resuls of Equaion (3) M1 M2 M3 M4 TUI 0.002 (0.994) 1.071 (0.053) 0.003 (0.967) 0.007 (0.966) UNEMP 0.081 (0.134) 0.313 (0.090) 0.100 (0.775) N/A PAYROLL 0.483 (0.119) 0.979 (0.000) 0.811 (0.000) N/A R^2 0.976 0.124 0.975 0.000 F 350.326 2.761 590.605 0.002 N 409 409 409 409 The resuls of all four models sugges ha uiion is posiively relaed o enrollmen levels. This runs couner o he belief ha rising uiion hurs enrollmen. However, his asserion is quesionable due o he unusually large significance levels in all four models. The only variable ha possesses any reliabiliy is payroll, and while i has a large significance level in M1, i is near zero in M2 and M3. The low R square value in M2 also suggess ha M2 is no a very represenaive model. The resuls of his sudy indicae ha uiion consideraion seems o have very lile effec on sudens decisions. Despie he uiion increase adoped in he pas 10 years, here has acually been an upward rend in enrollmen a all weny hree campuses. Possible jusificaions ha his could be aribued o are several facors such as improvemen in CSU repuaion, a significan decline in couny unemploymen raes, and relaively lower uiion raes in he CSU schools as compared o oher insiuions of higher educaion. Hence, he upward rend can be explained hrough righward shifs in he shor-run demand funcions. Conclusion and Summary The resuls of his sudy indicae ha CSU uiion is no he mos significan deerminan of is enrollmen levels. Specifically, he framework analysis presened in his repor shows ha here is virually no relaionship beween uiion and enrollmen. In realiy, enrollmens in he pas 10 years a CSU have increased despie higher uiion raes. Several reasons o jusify his conclusion include he fac ha he uiion raes of CSU schools are sill significanly lower han ha of oher California sae schools, or he locaion and operaing formas of he schools were mos compaible wih he sudens work lives. This research deails a sudy conduced using he collecive daa from weny hree
California Sae Universiy schools, which aimed a measuring he impac of uiion increases and oher facors on enrollmen. The sudy was based on hisorical daa o deermine he facors imporan in he college decision of enrollees and non-enrollees. The daa were used for he regression esimaion as presened in he earlier par of he paper. The mos imporan finding of his sudy was ha uiion consideraions seem o have very lile impac on sudens enrollmen decisions in he CSU sysem. In he saisical analysis of he uiion raes for all en years, i is eviden ha uiion raes do no play an imporan role in he enrollmen decisions of sudens. Furhermore, for undergraduae enrollees, he cos of aendance was he leas imporan reason for enrolling a any CSU. Payroll levels and counywide unemploymen raes were he mos imporan reasons among he undergraduaes for enrolling a CSU schools. The sensiiviy analysis also reveals ha he enrolled undergraduaes were likely o enroll a CSU even if he uiion raes were o rise higher. A key limiaion of his work lies wih he fac ha he variables used in he daa may have a mulicollineariy issue, paricularly wih he rae of change models. Decisions could have been jusified by oher facors perceived o be more aracive a rival insiuions han a any CSU school. Fuure sudies may resolve his limiaion by drawing upon furher avenues of research. References [1] Hemel, S.W., & Marcoe, D.E.. Rising uiion and enrollmen in public higher educaion. Insiue for he Sudy of Labor, 2008. [2] Heller, D. Tuiion Prices, Financial Aid, and Access o Public Higher Educaion: A Sae-Level Analysis. Annual Meeing of he American Educaional Research Associaion, 1996. [3] Heller, D. Suden Price Response in Higher Educaion: An Updae o Leslie and Brinkman, Journal of Higher Educaion. 1997, 68 (6), 624-659. [4] Heller, D.E. Effecs of uiion prices and financial aid on enrollmen in higher educaion. EdFund. 2001. [5] Sco, G.A. Tuiion coninues o rise, bu paerns vary by insiuion ype, enrollmen, and educaional expendiures. Unied Saes Governmen Accounabiliy Office, 2007. [6] Shim, J., & Milon, S. Rehinking uiion effecs on enrollmen in public four-year colleges and universiies. The Review of Higher Educaion. 2006, 29(2). [7] Vasigh, B., & Hamazee, R. Tesing sensiiviy of suden enrollmen wih respec o uiion a an insiuion of higher educaion. Inernaional Advances in Economic Research, 2003,10(2).