Formula Changes and College Enrollment: A Simulated Instrument Analysis of Pell Grants

• In Event: Financial Aid

Abstract

Background

The Pell Grant is the main source of federal grant aid for undergraduates to pay for college in the United States, with a maximum award of $7,395 and $27 billion in grants disbursed to 6.2 million recipients in Award Year 2023-24. The Pell Grant is targeted to lower-income students, with generosity decreasing roughly in proportion to students’ household income, phasing out completely around a family AGI level of $80,000. Despite this substantial investment, credible causal evidence on the program's effects remains mixed and limited, with most studies focusing on discontinuities at eligibility thresholds.

Research Question

This study uses data from the US Department of Education on the universe of 11 million FAFSA filers between Award Years 2008 and 2023 to estimate the effect of Pell Grant formula generosity for income groups throughout the population of federal aid recipients, taking advantage of differential year-to-year changes in generosity across income groups. Access to the full aided student population allows me to take advantage of quirks in formula changes to answer the following research question for the full aided student population: How does access to additional Pell grant dollars affect (1) borrowing and (2) re-enrollment for at least 90 days, at least half time in any college in the next year?

Methods

To address endogeneity concerns in estimating Pell Grant effects, I employ a simulated instrument approach within a two-way fixed effects framework. This method exploits plausibly exogenous year-to-year variations in the Pell Grant formula parameters that differentially affect students across income groups. I construct the simulated instrument by calculating Pell eligibility for a fixed population (2016 FAFSA filers) under each award year's formula rules, then averaging these simulated values within groups defined by parents' income, household size, and state. To calculate these estimates I use the following instrumental variables setup: 

1. 
   

   P_igt = π Z_gt + θ_g + λ_t + ν_gt

   
   


2. 
   

   Y_igt = β_1 P̂_igt + α_g + δ_t + ε_gt

   
   

where:

Y_igt: Dummy for individual i in Parents' Income group reenrolling in following award year t

P_igt: is actual Pell eligibility for student i

P̂_igt: predicted Pell eligibility from first stage

Z_gt: fixed simulated eligibility for group g in year t

θ_g, α_g: Parents' Income Group fixed effects

λ_t, δ_t: Award Year of application fixed effects

π, β_1: coefficients indicating the effect of simulated eligibility and Pell on outcomes

ν_gt, ε_gt: error terms

Results

For every $100 increase in loan eligibility, I see a $123-$168 decrease in loan dollars disbursed to a student. Also, for each $100 in additional Pell a student has access to, I see a very small reduction (0.07 ppts) in re-enrollment the following year. 

Conclusion

The evidence here largely confirms previous findings in previous Pell studies of minimal effects on enrollment and persistence, even as there is quite dramatic substitution out of loans into grants when Pell is more generous. This suggests that null enrollment effects result from students changing how they pay for college, not whether they go. 

Author

• Carmello Libassi, Teachers College, Columbia University
  Presenting Author