Would the prospect of having to borrow to go to university dissuade someone from enrolling in an academic program that was attractive to them and that held the promise of earnings that were enough to repay the loan? That is, does debt aversion exist and does it affect high stakes decisions? What is generally meant by debt aversion is essentially that being in debt carries a psychic cost, apart from any of the explicit costs and risks associated with the loan.
Several recent studies look for debt aversion in the context of student loans.
The NYU Law School Experiment
Over time, rising law school tuition has led to average borrowing levels among graduating lawyers in the range of $70,000 to $80,000. The law school at New York University (NYU) is well-known as a school that trains public-interest lawyers. Acknowledging the wide and rising gap in earnings between public-interest jobs and private sector jobs, a number of law schools including the NYU law school, have instituted loan repayment assistance programs (LRAP). An LRAP pays off the loans of graduates who work in public-interest jobs The LRAP at NYU makes the loan payments for graduates who are working in public-interest jobs, essentially forgiving the loan for graduates who work in such jobs for ten years after graduating. In 1997, the NYU law school announced a variant of the LRAP called the Innovative Financial Aid Study (IFAS); for those eligible, IFAS would pay two-thirds of the tuition of students who took up public-interest jobs after graduation. If they still needed to borrow, their loans would be eligible for LRAP. Graduates who chose not work in public-interest jobs had to repay their own loans (under LRAP) or, under IFAS, have their tuition subsidy converted to a loan for which they were responsible.
Rather than selecting particular students to get the IFAS tuition subsidies, NYU decided to make IFAS the “treatment” in a randomized trial. To be eligible for IFAS, applicants to the NYU law school first had to agree to be part of the experiment. Among the volunteers, those randomly selected to be in the treatment group would be offered the IFAS tuition subsidy while those in the control group would be offered LRAP. The two programs were designed to have exactly the same net present value; that is, they were financially equivalent, taking into account all interest on the loans. The only difference was that the LRAP students had to take out larger upfront loans than the IFAS students, but they did so having been promised that NYU would pay off the loans if they worked in public-interest jobs. Only if borrowing carried a psychic cost, apart from any financial or risk-related considerations, would the enrollment rates of the IFAS applicants differ from the enrollment rates of the LRAP applicants. The result? Field (2009), who analyzed the data arising from the experiment, reports (Field, 2009, p.11)) that among applicants for the 1999 class, 42 percent of the applicants offered IFAS enrolled compared to only 32 percent of the LRAP applicants; among the 2000 applicants, the gap between the IFAS and LRAP applicants was even larger, 20 percentage points in favour of IFAS. These effects are as close as researchers have gotten to documenting educational debt aversion in high-stakes decisions.
The World Bank Survey
Caetano, Partinos and Palacios (2011) used the answers to questions on a World Bank survey to measure the existence of debt aversion. All survey respondents were asked to choose between two methods of financing their postsecondary education. The two methods were described differently but were in fact financially equivalent. The first method required a fixed monthly payment, except when income was low, in which case the payment would be a percentage of income. The second method required the repayment of a percentage of the respondent’s income except when income was high, in which case a fixed monthly payment would be required. With the appropriate choices of payment, these two methods are simply two different ways of presenting the same payment conditions. Caetano et al. assert that the first method will be understood as a loan and expect that debt aversion will imply a preference for that method.
The respondents were randomly divided into two halves. Both halves were presented with the same two methods. The first half was presented with the first method explicitly labelled as a “loan” and the second method explicitly labelled as a “human capital contract”. The second half was presented with the same two options but without any explicit labels.
The results were clear (Caetano et al., 2011, p.22). The treatment group, whose choices were explicitly labelled “loan” and “human capital contract”, were about 13 percent points more likely to choose the human capital contract over the loan, even though both options were financially equivalent. Control group respondents, without the explicit labelling, were less likely to choose the loan option, but only by about two percentage points.
Two Canadian Experiments
Two related Canadian laboratory experiments attempted to measure the existence and importance of educational loans (Eckel, Johnson and Montmarquette, 2007; Johnson and Montmarquette, 2010). Participants were asked to make a large number of binary choice between cash, on the one hand, and various forms of loans and grant, on the other. In making these choices, they knew that at the end of the experiment, one of their choices would be randomly chosen to be honoured. For example, suppose one of the choices was between a $1000 loan and a $500 grant. The participants would have an incentive to choose the option they really preferred because they knew that their choice might actually be implemented, if it were the one chosen for them at the end of the experiment.
The exact choices offered to participants varied across the two experiments but were similar in that the grants and loan, if chosen, would be available only if the respondent enrolled in a full-time postsecondary program within two years. Respondent could decide not to accept the loan if they later decided they did not want to borrow.
One of the goals of the experiment was to measure the price sensitivity of the respondents. The relevant prices could be manipulated by assuming that the “price” of the aid was the amount of cash that would have to be given up if the aid were chosen. For example, the price of the $1000 grant mentioned above would be 50 cents per dollar of grant because, to get the $1000 grant, the respondent would have to given up the $500 in cash. The calculation for loan was a bit more complicated because the subsidy involved had to be calculated. In the end, however, each choice involving a cash option had an implicit price associated with it and demand curves for different types of aid could be constructed by mapping the proportion accepting the financial against the calculated price. One of the major findings in both experiments was that “once the price of educational subsidy is accounted for, demand for student financial aid was not much affected by the type of aid” (Johnson and Montmarquette, 2010, p. 39) That is, holding price constant, loans and grant were generally viewed by respondents as equivalent, suggesting that debt aversion was not very important.
The second goal was to see if debt averse respondents could be identified. In the second experiment, Johnson and Montmarquette defined as “debt averse” the relatively small group of respondents — just over 12 percent (152 of 1248) of the participants — who always chose a grant over cash and never chose a loan over cash. Such respondents were “…insensitive to price and complete sensitive to subsidy type” (Johnson and Montmarquette, 2010, p. 40). When Johnson and Montmarquette incorporate the rich set of covariates that they collected from respondents into the analysis, they find that some traditional “at risk” groups — first generation students, members of First Nations — were unexpectedly less likely to be debt averse than others. That is, debt aversion was present but was not concentrated among the groups often thought be debt averse.
To summarize, debt aversion was found in the two experiments (Field, 2009; Caetano et al, 2011) that presented participants with choices that were financially equivalent. It was also found in an experiment (Johnson and Montmarquette, 2010) where it was defined as present when respondent always chose grants but never chose loans when offered a choice between cash and the two forms of financial aid.
Implications for the Design of Student Aid Programs
What implications for student financial reform follow from the existence of debt aversion?
One idea is to translate the large subsidies that are attached to government loans into upfront grants. A cost-neutral version of such a plan is suggested by Gandhi (2008) and would presumably increase the access and persistence of debt averse students. The removal of the loan subsidies, however, might well create other problems for students once the repayment period starts. Without an in-school interest subsidy and with market rates of interest accumulating, many more students would fall behind on loan repayment. Given the difficulty of discharging student loans in bankruptcy — impossible in the US and difficult in Canada — some of those who borrowed (even with the new grants in place) would face severe hardship.
The finding by Caetano et al. that a key part of debt aversion in their study was the explicit labeling of loans suggests the possibility of presenting loans with language that does not explicitly use the words “loan” or ‘debt”. That is, the psychic cost of educational debt might be easily reduced by a simple change in language. For example, the Australian income-contingent loan program was initially called the Higher Education Contribution Scheme (HECS) and the use of the words “loan” and “debt” were not common in program descriptions.
Before rushing to change student financial aid programs, the Canadian finding that debt aversion is not concentrated among groups with lower than average enrolment and graduation rates should be seriously considered. If debt aversion is randomly distributed across all students, is it less a problem?
Caetano, G., , H. Patrinos, & M. Palacios. (2011). Measuring Aversion to Debt: An Experiment Among Student Loan Candidates. Policy Research Working Paper 5737. World Bank. Available at: http://elibrary.worldbank.org/content/workingpaper/10.1596/1813-9450-5737
Eckel, C., C. Johnson, C. Montmarquette, & C. Rojas (2007). Debt aversion and the demand for loans for postsecondary education. Public Finance Review 35(2), 233-262.
Field, E. (2009). Educational debt burden and career choice: Evidence from a financial aid experiment at NYU Law School. American Economic Journal: Applied Economics 1(1), 1-21.
Gandhi, S.J. (2008). Understanding Students from a Behavioral Economics Perspective: How Accelerating Student Loan Subsidies Generates More Bang for the Buck. Kansas Journal of Law and Public Policy 17(2), 130-167.
Johnson, C. & C. Montmarquette. (2011). Loan Aversion Among Canadian High School Students. Scientific Series. CIRANO. Available at: http://www.cirano.qc.ca/pdf/publication/2011s-67.pdf