What are the effects of student loans on educational attainment?

Responsible researcher: Viviane Pires Ribeiro

Article title: STUDENT LOAN NUDGES: EXPERIMENTAL EVIDENCE ON BORROWING AND EDUCATIONAL ATTAINMENT

Article authors: Benjamin M. Marx and Lesley J. Turner

Location of intervention: United States

Sample size: 18,800 entries

Sector: Education

Intervention Type : Effects of Student Loans

Variable of Main Interest: Student Loans

Evaluation method: Others - Field experiment

Assessment Context

In the United States, student enrollment has increased by more than 30% since the year 2000. Several studies provide evidence that community college students receive substantial returns in the job market. While some grant programs help low-income students enter and complete their studies, the design of federal student aid programs can make it difficult for students to access these resources. Thus, even as the federal government subsidizes loans for low-income students, it also encourages colleges to discourage borrowing, and there is little empirical evidence on the effect of borrowing on student outcomes to guide policy.

Student loan debt in the United States has grown steadily over the past decade, reaching $1.34 trillion in 2017. Despite the fact that community college students have greater financial need and are less likely to take out loans than students at private and more selective institutions, efforts to reduce borrowing have been especially pronounced within this sector.

Intervention Details

Marx and Turner (2019) study the effect of student loan offers on borrowing and educational attainment with a field experiment at a large community university. The experiment was implemented at “Community College A” (CCA), an anonymous community college, during the 2015 to 2016 academic year. CCA's costs are comparable to the costs faced by college students across the country. For example, in the university district, fees for the 2014-2015 school year were approximately $3,100 versus $3,249 nationwide. As with most community colleges, pricing depends on residency and is a linear function of credits taken. Community College A has a significantly larger student body than the average community college, with 18,800 enrollments compared to a 12-month average of 4,300. Financial aid receipts are similar for CCA students and other community college students. Approximately 45% of CCA students received Pell Grant aid and 25% received federal loans in 2013-2014, compared with 41 and 19% of students, respectively, at community colleges.

Students at CCA have substantially lower completion rates and slightly worse job market outcomes than the average community college student. Only 5% of CCA students completed a credential within 150% of the wait time, compared to 21% of students nationwide. Median earnings among federal aid recipients who were no longer enrolled for ten years after entry are similar for CCA and community colleges nationwide (about $28,000 and $30,253, respectively). Other results follow similar patterns, with CCA students experiencing worse job market outcomes than national averages.

During the 2014-2015 academic year, CCA offered loans to all students with less than $25,000 in federal student loan debt. All students who listed CCA on their Free Application for Federal Student Aid (FAFSA) received financial aid electronically, in a web system. In addition to federal requirements, CCA students had to confirm whether they wanted to apply for a loan and the loan amount through an electronic form. CCA eligibility criteria and application procedures were not changed for the trial, meaning the default loan amount was $0.

Methodology Details

To provide evidence about the causal effect of student loans on educational attainment, Marx and Turner (2019) used experimental variation that mirrors the most common practices at community colleges. Although many federal student loan features, such as maximum amounts and eligibility requirements, apply uniformly to students and institutions, colleges have criteria about whether to include loan “offers” in students' financial aid award letters. Listed loan amounts, commonly referred to as “offers,” do not change students' choice set, but can affect borrowing through choice architecture, that is, the design of the decision-making environment.

The experiment applied random assignment of loan offers to students. On a daily basis beginning in May 2015, the CCA financial aid office provided data on each batch of students for whom an aid letter was generated the following day. The experimental sample included all students eligible for financial aid. Students were assigned to either the treatment group or the control group using stratified randomization by Expected Family Contribution (EFC) bins and all possible combinations of binary variables for freshman versus senior, dependent versus independent, and with debt versus no debt. student loan.

Results

Students were randomly assigned to receive a $0 loan offer or an offer of $3,500 (for “freshmen” who accumulated fewer than 30 credits) or $4,500 (for seniors). Students who received a non-zero loan offer were 40% more likely to borrow than those who received a $0 offer, with each additional borrower taking up a $4,000 loan on average. Non-zero loan offers also generated considerable gains in educational attainment.

Non-zero loan offers were randomly assigned, generating a 40% increase in loan probability. This effect is larger than changes in loans produced by more onerous, intensive interventions and consistent with prior evidence that student loan decision-makers are disproportionately affected by information and administrative costs imposed when making such decisions. The experimental results suggest that the cost of collecting information about the willingness to borrow contributes to the fixed cost of borrowing.

Students induced to ask for loan suggestions received, on average, 3.7 additional credits and improved their grades by 0.6 points in the year of the intervention. In the following academic year, they were 11 percentage points (178%) more likely to transfer to a public four-year institution. The authors estimated that non-zero loan offers increased the short-term achievement of community college students. However, it cannot be concluded that offering a non-zero loan improves each student's well-being, but it was projected that the average responder benefits financially from the loan even with a discount rate as high as 12%. Using a simple theoretical model that allows for important effects, information costs, and default bias, the authors provide evidence about the channels through which nonzero offers affect student behavior. The pattern of responses suggests that at least 78% of loan responses are driven by a reduction in information costs, suggesting that non-zero offers improve the well-being of most students.

Public Policy Lessons

Marx and Turner (2019) emphasize that the results of the study are relevant for colleges, policymakers and future research on the effects of “nudges”. In the United States, more than five million students attend colleges that do not offer loans in financial aid award letters, and nearly one million students attend colleges that do not participate in federal loan programs. The analysis suggests that loan offers to students enrolled at these colleges can generate substantial increases in educational attainment. The authors show that nudges can affect behavior by communicating information in a better way than other methods used to communicate the same information. Students appear to benefit substantially from clear communication of the loan opportunity when they are making loan decisions and therefore being informed of their set of options. At the same time, a better choice within this set also requires knowledge of expected costs and benefits. Future research could examine how to help each student obtain a loan amount that best meets their needs.

References

Marx, B. M., & Turner, L. J. (2019). Student Loan Nudges: Experimental Evidence on Borrowing and Educational Attainment. American Economic Journal: Economic Policy , 11(2), 108-141.