«Financing lifelong learning Hessel Oosterbeeka, Harry Anthony Patrinosb,∗ a Universiteit van Amsterdam School of Economics b World Bank Abstract ...»
In such natural or quasi-experiments, researchers exploit circumstances in which comparable observations are treated diﬀerently. In the ideal situation, subjects cannot aﬀect their treatment status. The challenge is to ﬁnd a source of exogenous variation that aﬀects treatment and has at the same time no direct eﬀect on the outcomes of interest. Attention to evidence-based policies requires consideration about how the counterfactual was constructed. There is only evidence in favor or against policy interventions if it results from research with a carefully constructed control group.
3. How to Stimulate Lifelong Learning?
This section presents various forms of funding governments use. Brief descriptions of instruments are provided along with a discussion of how these forms are connected to the reasons for government intervention. Credible research ﬁndings regarding the eﬀects of actually implemented funding schemes are analyzed.
30 H. Oosterbeek and H. A. Patrinos
3.1 Cost-Sharing Schemes
To address the ineﬃciency of capital market failure, governments implement or support ﬁnancial aid schemes in which students can borrow against conditions that are more favorable than what the private market oﬀers. We consider the following ﬁnancial aid schemes: (1) mortgage-type loans, (2) income contingent loans, (3) graduate tax and (4) human capital contracts. We summarize evidence on the eﬀect of ﬁnancial incentives on achievement of students in higher education.
Mortgage-type loans. The most popular student ﬁnance scheme is traditional mortgage-type loans. This scheme is likely to be oﬀered only to families who have assets to serve as collateral-that is, precisely those who need ﬁnancial aid the least.
In a mortgage-type loans model, students are given loans, which they are required to repay in the form of ﬁxed installments. The interest rate is typically below the rate that private banks would charge.
At ﬁrst sight, the mortgage loans model seems to address the capital market failure. The exact details of such models may diﬀer. A textbook example of a mortgage model is the current system operating in the Netherlands. In this scheme, the loan only has to be repaid over a particular period if the borrower’s income exceeds a certain threshold. Also, in this model there is a rule that ﬁxes the number of years in which the loan has to be repaid to 15 years. If the repayment is not completed within that period, it is remitted. This restricted repayment duration is however at odds with the idea that schooling is an investment in future opportunities. Comparing age-earnings proﬁles by level of schooling, it is evident that the returns to schooling materialize at later ages. It would be optimal to synchronize repayment and obtaining the returns. A short repayment period is not appropriate.
Traditional student loans have been collected by the state, by private banks, and by universities. Collection has been poor or costly where the taxing power of the state has not been used as a last resort to collect the loans. In some cases, as in the Philippines, poor collection rates have caused such schemes to operate at a loss.
Income contingent loans. During the past decades economists have advocated a system with income contingent loans (Barr, 1993; Chapman, 2006). Students can take up a loan, which they have to repay in the form of a percentage of their earnings after graduation. Compared with the mortgage-type loans, repayment is more evenly spread over graduates’ careers. That is, the costs of the investment in schooling are repaid when the returns materialize.
A separate issue relates to the interest rate that students pay. From the point of view of the lender, it is desirable that the interest rate includes a premium for defaulters. This is feasible if an individual’s probability of default is as uncertain to the borrower as it is to the lender. That is, however, not very likely; borrowers know more about their own characteristics and, therefore, about their own risk than do lenders. As a result, bad risks will drive good risks out of the market. The relevance of this mechanism increases with the heterogeneity of the student population. Since enrollment in post-secondary education has increased in recent decades, the student Financing lifelong learning 31 population has become diverse. The important message is that the government has to be careful when determining the default premium included in the interest rate.
Moreover, adverse selection may be a problem. If the lender is unable to shift the risk of default to borrowers, it is reasonable that the lender should screen students for eligibility. Academic records are likely to be a good indicator of the student’s future default probability.
The more successful adoptions of income contingent loan schemes took place in Australia, New Zealand, South Africa and the United Kingdom (Chapman, 2006).
Two critical aspects of the schemes in these countries relative to others are noted. The ﬁrst is that these countries all have a taxation system in place that could be used to collect repayments eﬀectively. The second is that in these countries the vast majority of universities are public sector institutions, so that the collection (tax) authority and the education providers operate under the same system and the same terms.
Systems and structures most resembling the successful developed countries are not often available and many developing countries have diﬃculties associated with the establishment of the policy’s integrity, credibility and collection (Chapman, 2006).
Given this policy context, it may be desirable that developing countries proceed with the imposition of up-front fees and scholarships instead of income contingent loans.
An important issue concerning the introduction of income contingent loans is to what extent it aﬀects access to higher education, especially for students from poor families. The only country for which this has been documented extensively is Australia. Chapman and Ryan (2002) compare participation rates for 18-year-olds by family wealth before the introduction of the scheme, after the introduction of the scheme, and after some substantial changes have taken place. Over time participation rates for all family wealth categories - including those from the lowest quartile
- have increased. This is presented as evidence that the system did not result in decreases in the participation of prospective students from relatively poor families (Chapman, 2006). While factually correct, this statement is somewhat misleading.
Introduction of the income contingent loan scheme is Australia was not associated with a reduction in participation of students from poor families when participation rates before and after the introduction are compared. This does not exclude, however, the possibility that participation rates of this group measured after the introduction would have been higher if the system had not been introduced. In other words, Chapman’s (2006) statement takes the before participation rates as counterfactual, thereby ignoring changes that would have occurred in the absence of the policy. The positive evaluation of the Australian model would have been much more convincing if based on a randomized experiment or some other design with a proper comparison group.
Graduate tax. In a system with graduate taxes, students receive a public subsidy while studying and repay this in the form of a special tax after graduation. Graduate taxes are often regarded as an imperfect substitute of the ideal system with income contingent loans. Graduate taxes and income contingent loans are similar in many ways, but graduate taxes have the disadvantage that they also provide funding to students who do not want it or need it, and that graduates may end up paying (much) 32 H. Oosterbeek and H. A. Patrinos more than they received in the form of a grant. This latter feature may provide a disincentive to study to those who would have earned a high income even without the education. The problem is that the graduate tax is levied based on total earnings rather than the earnings increment due to the extra education.
The relative appraisal of graduate taxes changes when prospective students are debt-averse. In that case it makes a diﬀerence that under an income contingent scheme students carry debt whereas under a graduate tax scheme they do not carry debt, even though the actual ”repayment” of the loan/grant are exactly the same.
Theory predicts that debt-averse individuals are more willing to enroll in education when ﬁnancial aid is in the form of a graduate tax then when given in the form of an income contingent loan.
Graduate taxes are not in operation anywhere, and it is thus impossible to evaluate them. Yet, some insight about their potential can be gained from studies that inform us about the relevance of debt aversion. Two recent empirical studies from selective colleges in the United States provide evidence for the presence of debt aversion.
In a recent study, Field (2006) reports on an experiment in which law students at New York University were randomly assigned to one of two ﬁnancial aid conditions. The ﬁrst, standard, condition is a loan repayment assistance program (LRAP) which forgives all graduates who choose careers in the public sector or other low paying ﬁelds of law the majority of educational loans incurred during law school through quarterly prospective funding for up to ten years following graduation. The second, innovative, condition consists of public service scholarships (PSS) that provide grants of two-thirds tuition that convert to a loan in the event the recipient does not pursue a public interest law career. The two schemes are identical in terms of ﬁnancial consequences and diﬀer only in duration of indebtedness. Under the LRAP scheme, people have a debt position from the start. Under the PSS scheme, people only carry a debt burden when they do not pursue a public interest law career. Field’s (2006) results show diﬀerences in outcomes between the two treatments. In classes for which the lottery was announced prior to enrolment, those in the PSS treatment are twice as likely to enroll. Moreover, those in the PSS treatment are also substantially more likely to have a ﬁrst job in public interest law. Field (2006) interprets her ﬁndings as evidence of debt aversion: behavior is consistent with utility being negatively aﬀected by carrying debt loads.
Data from a highly selective university that introduced a ”no-loans” policy under which the loan component of ﬁnancial aid awards was replaced with grants was recently analyzed (Rothstein and Rouse, 2007). This natural experiment identiﬁes the impact of student debt on employment outcomes. Debt causes graduates to choose substantially higher-salary jobs and reduces the probability that students choose lowpaid ”public interest” jobs. There is also some evidence that debt aﬀects students’ academic decisions during college. The authors oﬀer two potential explanations for their ﬁndings: (1) young workers are credit constrained and (2) they are averse to holding debt.
Financing lifelong learning 33 Data from a recently collected survey among higher education students in the Netherlands indicate that debt aversion is signiﬁcantly higher among students whose father has a lower level of education. A regression of debt aversion on level of father’s education gives a coeﬃcient of -0.028, suggesting that especially children from poorer backgrounds may beneﬁt from the introduction of a graduate tax scheme.
Human capital contracts. A human capital contract is a contract in which students agree to pay a percentage of their income for a speciﬁed period after graduation in exchange for funds to ﬁnance their education. Originally proposed by Milton Friedman (Friedman and Kuznets 1945; Friedman 1955), the idea of such contracts has re-emerged in recent years.
An essential characteristic of human capital contracts is that investors determine the percentage of future income that students have to commit, which could vary depending on the type of learning undertaken and the investor’s judgment about the borrower’s likely future income. From an eﬃciency perspective, optimal results are achieved when market forces determine the percentage of income that learners have to commit and externalities are covered by public subsidy. For the outcomes to optimize social welfare, distributional considerations must also be taken into account by targeting public subsidies in order to achieve equity.
Implementation of human capital contracts is constrained by the diﬃculty of obtaining information on learners, the need for a developed tax collection agency, and the problem of adverse selection (Palacios, 2004). There are now some examples of human capital contracts in practice. MyRichUncle was the ﬁrst company oﬀering human capital contracts in the United States. The company switched, however, to oﬀer student loans, and more recently stopped lending altogether. CareerConcept in Germany has until now ﬁnanced hundreds of students and continues growing. Lumni in Chile and Colombia is still very small but continues growing. With this limited scale of operation, and without any attempt to evaluate, little can be said about the eﬀectiveness of this instrument.
Financial incentives for students. A randomized experiment involving two strategies designed to improve outcomes among ﬁrst-year undergraduates at a large Canadian university was evaluated (Angrist et al., 2006). One group was oﬀered peer advising and organized study group services. Another was oﬀered substantial meritscholarships for solid ﬁrst year grades. A third group combined both interventions.