Recent Research on Income-Contingent Student Loans

(This is a joint post with Darragh Flannery of UL and Kevin Denny of UCD).

Income-contingent loans (ICLs) for students were one of the options proposed by the Cassells Report on Future Higher Education Funding when it was published last year (see here). The topic has been back in the news again in recent weeks because of the dissemination of a paper[1] by Shaen Corbet and Charles Larkin, which claims to show that an ICL could not work in Ireland.

Two of us (Doris and Flannery) have done research directly in the area of ICLs – indeed Doris’s research ended up being used by the Cassells Expert Group to provide illustrations of how an ICL might work in Ireland. We both found, using different data sets and different ICL parameters (income thresholds, repayment rates etc.) that the discounted value of loan repayments would be about 75% of the loan values, even when accounting for graduate emigration.[2] Under these repayment rates, there would be no problem operating an ICL in Ireland.

The third poster (Denny) has written papers on the determinants of participation in higher education (HE), the returns to education and related topics and so has a strong research interest in the effects of funding on access to HE.

We were all surprised by the reports of Larkin and Corbet’s results and so went off to read the paper. This had added interest as the research appears to be influencing policy makers.  Given this context and with apologies for the length of the post, we have decided to make our assessment of it public.

While the paper discusses a large number of topics, we focus on their simulation of an ICL system, since this seems to be the basis for their conclusion that an ICL would be ‘imprudent’. Their empirical analysis models Irish graduate earnings, and then examines how repayments respond over time to changing four model parameters: the ‘default’ rate, the repayment cap (an upper limit on annual student loan repayments), the interest rate and the size of the loan. Based on some assumptions about which parameter values are most likely, the authors conclude that it would be over 20 years before the ICL system would return a profit and that this renders such a scheme imprudent.

Apart from the fact that is not clear why the authors think an ICL system should return a profit, we have many criticisms of this empirical analysis. First, the authors use almost no real data. They have one data point – mean earnings of recent graduates, as published in the 2015 HEA first destinations report – and from that starting point, they assume an arbitrary rate of earnings growth that, being linear, does not reflect real-life data (lifecycle earnings are concave in age typically).

A more important point is that you cannot say anything about non-repayment (‘default’) rates in ICL models without having information about the full graduate earnings distribution. It is because some graduates have low earnings that they fail to repay their loans. To estimate non-repayment rates, you must have a dataset that covers a representative sample of the graduate population so you can estimate how earnings will evolve over the life cycle across the whole distribution. You cannot talk about non-repayment rates with information only on the mean.

To assess the viability of an ICL scheme, the Corbet and Larkin paper varies the ‘default’ rate as though it were exogenous. It is not. The default rate should be the outcome of the model rather than an input, since it is determined by the lifecycle earnings profiles and the parameters of the ICL scheme. The approach they have taken is unheard of in the existing literature.

Non-repayment in an ICL system arises if earnings are too low. This is a feature of the system, not a flaw; the system is designed to subsidize those who turn out not to benefit from their higher education. The only way default can occur is if graduates emigrate and refuse to repay. Measures to encourage payments by emigrants should be designed into the system from the beginning, as our research shows that non-repayment by emigrants can make a substantial difference to the cost of the system.

Apart from our criticisms of the paper’s failure to estimate non-repayment rates, we also have many more minor quibbles with the empirical exercise. For example, the authors assume a cap on annual repayments by the graduates (this is one of the parameters they vary), which has never been proposed anywhere and they assume 6% non-employment and 10% emigration, both figures that are probably too low.

In other places, basic errors are made. For example, the authors claim that the HEA First Destinations report indicates that up to 19% of graduates may “have irregular contact with officials further reducing their probability of repayment”. Rather than reflecting an enormous black economy for graduate employment, closer inspection of Figure 4.1 in the report indicates that the 19% reflects the proportion of new graduates whose institution failed to include details of the region of employment with their returns to the HEA. These are not potential defaulters.

The paper makes several other claims about how ICLs would operate in Ireland. In each case, there is no analysis in the paper that backs up their claims, and the experience of other countries contradicts them.

One notable claim is that the Irish system is “neither small enough nor large enough to make an ICL system work”. Despite a section entitled ‘The Scale Problem’, there is nothing in the paper that substantiates this claim. There is a vague assertion that a larger size allows for risks to be spread over a wider pool of loan holders – presumably, this refers to the risk of low earnings. But New Zealand, which is very similar in size to Ireland, has been successfully running an ICL system since 1991. It is unclear why a country smaller than Ireland would enjoy some advantage in running an ICL.

A further theme running through the paper is that an ICL scheme might be workable for the universities, but not for the Institutes of Technology (IoTs). This is hard to understand. At several points, the authors make statements that suggest that they believe that students borrow from and repay their own institutions. This is not the case, so we can see no implications of an ICL for individual higher education institutions (HEIs). It is the overall non-repayment rate that matters for the viability of an ICL scheme.

An assumption maintained throughout the paper is that the introduction of an ICL would necessitate the establishment of a Special Purpose Vehicle (SPV) so that the debt could be taken ‘off balance-sheet’. In support of the claim, the authors note several details of the Fiscal Stability Treaty that limit expenditure and require debt reduction. Several points are worth noting here.

First, the Cassells Report itself included a discussion of this point that was provided by the Department of Finance (see footnote 56). This discussion concluded that the introduction of an ICL would restrict flexibility in achieving our fiscal targets (as indeed would other spending commitments), but would not be precluded by the Fiscal Treaty. We bow to the Department’s expertise on this issue.

Secondly, there is nothing about student loans that requires them to be taken ‘off balance sheet’ and indeed they are not in Australia or New Zealand, and have not been in the UK until recently. The process of taking them off balance sheet by selling the loan book has begun in the UK, but this is in order to reduce the calculated value of the national debt rather than to reduce costs. In fact, most financial commentators regard this is a mistake (see here): since private investors have less capacity to absorb earnings risk than governments this leads to heavy discounts when the loan book is sold.

The alternative proposal given in the paper for funding HE is the introduction of a new tax – an ‘education levy’ – which would replicate the existing Universal Social Charge (USC); in other words, for most income tax payers, an additional 5% of all income would be payable to fund HE. If such additional revenue were available, it could just as easily be used to cover any non-repayment of ICLs, or indeed any other method of funding HE. There is no discussion of the implications for efficiency of such a tax increase, nor of its political feasibility.

The point is that, for a given increase in HE funding, a system partially financed by graduates themselves must be less costly to the taxpayer than a system funded entirely by the taxpayer. To the extent that the system is ‘leaky’ – i.e. to the extent that there is non-repayment or ‘default’ – the savings to the taxpayer are reduced, but it is hard to see how the system can be more costly.

Thus, if the Fiscal Treaty constrains our ability to raise HE funding via an ICL with any shortfall funded by government borrowing, it must constrain our ability to raise that funding entirely from government borrowing even more. And if political preferences constrain our ability to fund non-repayments in an ICL system by increased taxation, they must constrain our ability to fund a system with zero repayments (i.e. ‘free fees’) even more.

The focus in the Corbet and Larkin paper on the conditions necessary for selling off the student loan book misses this point, and allows the authors to introduce some fairly apocalyptic conclusions, including that an ICL would create ‘an Anglo in slow motion’. There is nothing in the paper, nor in the experience of other countries, that could justify such a conclusion.

We have previously communicated our concerns about the contents of the paper to the authors; Doris sent an email containing many of the above comments on April 10 and Doris and Flannery reported our concerns to a meeting of the Joint Oireachtas Committee on Education and Skills on May 2.

At the Oireachtas Committee meeting, Drs. Larkin and Corbet presented an updated analysis that seemed to have taken at least one of our criticisms on board – that default rates need to be estimated rather than assumed. Their account of their revision (which can be found here, along with the statement to the Committee by Doris and Flannery) was unclear, but seems to suggest that they pulled together various statistics from various published sources to approximate a representation of the Irish graduate earnings distribution. This would be no substitute for real data, and cannot be statistically valid.

At the same Committee meeting, the authors mentioned that the research was still work in progress, suggesting that it may be open to further revision. We make two comments about this.

First, we would suggest that in revising the paper, the authors familiarize themselves more thoroughly with the extensive research literature on higher education financing,[3] the social and private returns to education as well as research on higher education access. Having spent much of our careers working in the economics of education, our judgement is that the current version of the paper does not reflect a sufficient understanding of this topic.

Secondly, we note that an apparently unfinished piece of research is being used to influence public policy on a very important topic. Apart from appearing at the Oireachtas Committee, the authors have also been meeting politicians in Leinster House (see here) and publishing opinion pieces (see here) and the paper has been cited by politicians and educators in support of their opposition to ICLs (see here and here).

Because of the flaws in the current version, and the fact that the research seems to be ongoing, we believe it is inappropriate that the research should be used to influence public policy. Accordingly, we suggest that the authors should consider withdrawing the current version of their paper.

We do not contend that ICLs are workable under any and all circumstances, but simply wish to highlight that the debate surrounding it should be as informed as possible.




[1] There are several versions of the paper available. Here, we’ll refer to the version most recently posted on SSRN.

[2] For an overview of our research in this area, see Darragh Flannery, Aedín Doris & Bruce Chapman, 2017, ‘Student Financing of Higher Education’ in John Cullinan and Darragh Flannery (eds.) Economic Insights on Higher Education in Ireland: Evidence from a Public System, Palgrave Macmillan.

[3] An accessible introduction for the interested reader can be found at

9 replies on “Recent Research on Income-Contingent Student Loans”

Q to the authors…did you speak to the other authors prior to this publication? Critiquing policy making being done on working papers via, in effect, a rebuttal to same in an effort to..influence policy making ..seems rather circular?
Anyhow, fwiw in my view a finding that an ICL system would result in mass warts would make no difference,as the decision seems to be ‘in’.


Student Debt Rising Worldwide [good graphics including Ireland]

Postsecondary education, regarded as essential in a competitive global job market, is credited for increasing prosperity for individuals and their communities. Greater numbers of students pursue higher education, but rising costs of college force many to rely on loans. Tuition for attending a public college varies – free in 40 countries like Brazil or Germany and thousands of dollars in the United States and the United Kingdom. The average student loan debt is $37,000 in the United States and $55,000 in the United Kingdom; even students in Sweden, where tuition is free, leave with an average debt of about $20,000. The struggle to fund college is particularly tough for students from rural areas and developing nations. Governments must weigh the arguments of citizens who favor higher rates of taxation to fund education programs and loan forgiveness versus those who support private lending and tough repayment enforcement. Demographer Joseph Chamie warns that the rising costs and growing debt could slow economies as many young adults delay decisions to marry, purchase homes or start families while others forgo higher education altogether. – YaleGlobal [h/t]

Nice work. Another point of concern for me is that the work of Larkin and Corbet appears to have been sponsored by the THEA (the umbrella body for the Institutes of Technology). Perhaps their conclusions, particularly in reference to different funding options for different parts of the higher education sector, should be interpreted in this light. It is a shame, but perhaps not surprising, that most coverage in the media didn’t allude to this issue.

No more than Aedin being involved in the modelling underlying the ICL (See the appendices in Cassels report) does the funding here suggest we should discount their models. Nor accept.
There is more than a hint here today of this site slipping back into playing the player not the ball. Pity, that.

Putting on my two editorial hats : whats needed here is a lot of openness. Both sides of the debate should put out into the open their models – whether on spreadsheets or whatever – and let people input what are seemed to them to be sensible parameters. Indeed, I would think that as we are in essence dealing with simulations then the densities of the outcomes as they depend on the inputs would be useful to see.

As a point of clarification: I did not receive a penny for the work I did on the Cassells report. I was doing research on the topic with Bruce Chapman anyway, and I offered to share my results with the group working on it.

Not meant as a personal attack Brian, but I do think there is more of a difference here. The contribution of Doris to the Cassels Report is in the context of an independent group commissioned to examine the funding options. The THEA, as is absolutely their right, publicly set out their stall against ICLs. If the Larkin and Corbet work is sponsored by the THEA (as appears to be the case) and their conclusions in this paper support that position and/or suggest free education in the IoTs and ICLs in the universities (as they do), then I think it is an important caveat for interpretation of their work. That doesn’t make them wrong. Or right. But it should be factor in the reader drawing those conclusions.

Student loans are a bit like Schodinger’s Grants. For many, they are a possibly-to-be-written-off loan, depending on the politics of the future, the relevant wave function collapsing after observation of a general election result some time in the future.

You market them as ‘not really debt’ – but a higher rate of tax above an income threshold.

As David points out they introduce uncertainty, except for the institutions that the loan-funded fees are paid to. They definitely collect, while later on, the bill is argued over between the students and the rest of the taxpayers.

“Blessed be The Rate-of-Growth. May it continue without ceasing.” [a political incantation]

– Well it will not so continue, its hours are numbered and of this I am certain. But what of economic modellers? I understand that success eludes them half of the time.

I shall eschew the need to enquire into the economic justification that WEIRD* societies need to provide more than 36% of their emergent adult population with a Level 8 qualification. Or the more absurd notion that Levels 9 + 10 are essential ingredients to sustain the abovementioned Rate-of-Growth in positive values.

You see; debt exhibits a Rate-of-Growth which is truly magical, nay celestial: it grows absent a single real, physical input. Jack-and-Beanstalk sort of stuff.

And our unwise counsellors wish to foist this debt object onto unsuspecting volunteer undergraduates – to be carried forward and paid down, or else! So what is this ‘or else’ then? Imprisonment? Indentiture until disabled by physical and mental impariment? Bankruptcy? Because, if that abovementioned Rate-of-Growth should (over the longer run) prove insufficient to generate the steady increments in graduates’ incomes …..

Forget this research stuff and its non-sensical modelling. Its time for Ye to re-acquaint Yerselves with reality. Unfortunately were you do do so, you’d be redundant. Such is life.

Income Contingent Loans are economically, socially and politically – mad, bad and dangerous. A significant financial hazard with an associated high level of human risk.

Please try to avoid inflicting social experiments on a febrile public. When engineers design aircraft or bridges they do not assume they will fail: they design them to be fail-safe. Failure is not an option. Think about this.

*Western, Educated, Industrialized, Rich and Democratic. [Henrich, J. et al. 2010. No. 139; www://]

Comments are closed.