using tuition elasticity to forecast enrollment
TRANSCRIPT
Agenda
§ What is elasticity? Why should you care? § Project goals, from each perspective
§ University Provost & Enrollment Management § Chief Budget Officer’s Office
§ The data § The methodology/techniques applied § The general results and their interpretations
Definition: Tuition Elasticity
§ Price elasticity of demand (from economics) applied to higher education § The responsiveness (elasticity) of the quantity
demanded (enrollments) to a change in the price (tuition)
§ Answers the question “If we increase price by 1%, how much will demand change?”
– Elasticity = -1, then 1% decrease in demand. – Elasticity = -0.5, then 0.5% decrease in demand – Elasticity = -2, then 2% decrease in demand
Higher Education Complicating Factors
§ Financial Aid § Student Loans
§ Different pricing models (resident/non-resident) § Program Fees/Differential Tuition
§ Institutional change § Value perception
Substitutes
§ Elasticity varies, in part, based on the available substitutes § “Go to College/Don’t Go” is one option, BUT
once that decision is made there are a LOT of possible substitutes
– Other within-sector institutions – Other sector institutions – Community colleges – Specialty schools
Elasticity, old school
§ Traditionally determined through phone surveys of potential students § Expensive and time consuming § Usually involves bringing in consultants
§ Few prior studies at the institution level that are published § Curs & Singell (2010) § Moore, Studenmund & Slobko, T. (1991) § Ehrenberg & Sherman (1984)
Overall Project Goals
§ Estimate how sensitive different student groups are to changes in tuition pricing § Sticker price § Net price § Relative price (ASU vs. substitutes)
§ Quantify change between years and relate to changes in tuition, 2008-2012 § Validate predictive model against 2013
§ Create estimates that can be used to inform enrollment forecasting models
Project Goals: Provost’s Office Perspective § Understand which subgroups of students are
most influenced by price changes
§ Provide factors (% change assumptions) to inform enrollment models
§ Understand the differential effects of tuition and gift aid to hone financial aid policies
§ Particular attention to elements we can control or use in making financial aid decisions
– Residency – Effective Family Contribution (EFC) range – Academic ability composite score range
Project Goals: Chief Budget Officer’s Office Perspective
§ Inform discussions around whether tuition prices can and/or should be raised further
§ Understand whether students make decisions
based upon the current year’s tuition and fees or prior year numbers
§ Understand the differential effects of tuition and
gift aid to improve accuracy of net revenue estimates (tuition and financial aid budgets)
Methodology and Tools
§ Probit model § Log transformation of continuous variables
§ SPSS for data manipulation and management § JMP for exploratory analysis of which variables
to include § Stata 12 for Probit models, marginal effects and
collinearity diagnostics
Population
§ Fall First Time Freshman Applicants § Excluded groups due to different decision criteria
– International students – Transfers – Online – Athletic Scholarships – Waivers
§ Fall 2008 – Fall 2012 (5 years)
§ Outcome Variable: § Enrolled on 21st day
Variables
§ Anything the institution would know about the student prior to their decision § Information from the application § Information from the FAFSA § Information from transcripts § Information from marketing lists
§ Anything the student knows about the institution § Tuition and fees § Rankings of departments/majors § Initial aid offer
§ In other words, the kitchen sink!
Consistently important variables § PRICE
§ List price including mandatory fees § Relative price for those who went elsewhere (ASU list price minus Substitute list
price) § FINANCES
§ Pell Eligibility § Effective Family Contribution (EFC) § ASU first and only school listed on FAFSA § Aid received § Loans approved
§ ACADEMIC § HS Rank / HS GPA / SAT or ACT Scores in a composite variable (CI) § Honors College Admission § College (but only some of them) § Weeks Out 1st contact with ASU
§ SOCIAL § Gender § Minority Status § High School Mean SAT § Neighborhood Median Income
Model 1
§ Limited variables included: § Residency Status § List Price (tuition + mandatory fees) § Gift Aid § Effective Family Contribution (EFC), bucketed § Composite academic ability (CI), bucketed
Model 1: Resident Incoming Freshman
§ Most subgroups are inelastic ( |elasticity| < 1) § EXCEPTIONS: lower ability students with either higher
levels of need
§ Coefficient for aid is ~1/4 the size of Price § For most groups you need to raise aid 3-4x more than you
raise tuition in order to offset lost enrollments § EXCEPTIONS: Highest ability groups are closer to 1 to 1.
§ Effect sizes are small, even for the neediest groups (< 0.3), with less needy groups being even smaller § Pseudo-R2 goes up (more important) as CI goes down
Model 1: Non-resident Incoming Freshman
§ Very high need and high need students are sensitive to price regardless of ability level ( |elasticity| > 1 )
§ Aid coefficient is only 1/7th the size of price coefficient
§ Effect sizes are small, even for the neediest groups (< 0.2), with less needy groups being even smaller § Pseudo-R2 goes up (more important) as CI goes down
Model 2 § Full relevant variable list
§ Residency Status § List Price (tuition + mandatory fees) § Gift Aid § Effective Family Contribution (EFC), bucketed § Composite academic ability (CI), bucketed
§ Gender § Minority Group § ASU Choice Number (from FAFSA) § High School Mean SAT § Neighborhood Median Income § Admitted to Barrett, the Honors College § Professional College (Engineering, Business, Education) vs. Liberal Arts § Pell Eligibility (yes/no) § Loan Amount § Weeks out at time of 1st contact with ASU § Geographic subgroup (Residents: Maricopa County or not, NonResidents: CA or
other state)
Model 2: Resident Incoming Freshman
§ Generally inelastic (|elasticity| <1) for nearly all combinations of attributes § Approaches elasticity in the higher income and/
or higher CI groups
§ Aid coefficient a fraction of the size of the price coefficient § Loan coefficient the same or larger than gift aid,
suggesting that students don’t differentiate
§ Additional variables vastly improve effect size (R2=0.7, n=36540)
Model 2: Non-resident Incoming Freshman
§ Overall elastic, nearing -2 (for each 1% price increase, 2% enrollment decrease) § Only students under the Max Pell EFC level and
in the lower ability groups were not sensitive
§ Aid and loan coefficients about the same size and often not significant at all
§ Effect sizes also significantly improved (R2=.6, n=31868)
Other findings
§ Students who list ASU as the 1st institution on the FAFSA are 5+ times as likely to enroll
§ Students admitted to Barrett, the Honors College are 3 times as likely to enroll
§ Each additional $100 in NET tuition makes a student ~2% less likely to enroll § However it takes more than $100 in aid to make
up for $100 increase in tuition for most groups
Implications
§ Driving enrollment forecasts directly from the model likely to be inaccurate § R2 on model 1 too low § Model 2 includes elements we can’t consider for
policy setting § Level of financial need is crucial to
understanding reaction to price changes § High need students are less sensitive than low
need students