Category Archives: Higher education

Should we use the OECD’s analysis of the private financial benefits of tertiary education?

I’m quoted this morning in The Australian‘s report on graduate earnings across the OECD, which is in the latest issue of Education at a Glance.

The reported numbers seemed low compared to work Grattan and others have done for higher education, and I have had a bit more time since to work out why.

An issue I noted in the Oz is that the analysis included people with diplomas. In 2012, diploma holders were 28 per cent of everyone with a diploma or higher qualification. Their lower average earnings will bring down the overall average.

Another issue is that the OECD’s data source may be understating graduate income. They used a source I had never heard of for analysing educational returns, the ABS Disablity, Ageing and Carers survey. It was a general population survey so the issue is not that it is a sample of graduates with a disability. However, looking at the way the unit record data is made available to researchers it seems income is only available in ranges, the top one of which is $1,730 a week or more. We hit this problem in the 2011 census as well, with their top range of $2,000 a week or more. 11 per per cent of diploma holders, 21 per cent of bachelor degree holders, and 33 per cent of postgraduate degree holders reported incomes of $2,000 a week or more. As some of these would have incomes well over $2,000 a week, the average is artificially held down by the income category cap.

The OECD numbers are net present value, which means that income expected to be received in the future is counted as of less value than income received now. There is plausible time value of money theory for discounting the future – for example, an 18 year old prospective student would probably rather receive $1,000 now than $1,100 when they finish their 3 year degree, even though there is a favourable implied interest rate on offer.

But in our Graduate Winners report that was not the way we presented the data, which we left undiscounted in the key sections. This was partly because the undiscounted number is easier to understand, and partly because despite the plausibility of time value of money theory in various contexts I was not sure it was so persuasive in this one. In my view, one reason people pursue higher education is so that they will have a good job and a high income in 30 years time. How much theoretical sense does it make to heavily discount the value of achieving a major objective?

Some interesting data on male hourly earnings by years of experience from the latest HILDA report highlights this issue. For the first five or so years, male graduates don’t earn much more per hour than men with vocational education. But after that time a wide earnings gap develops – in the later years that are most discounted by the OECD methodology.

male hourly earnings

The discounting also affects another issue, which is that they assume students don’t work while studying, and the consequent assumed forgone earnings appear with a low discount and are deducted from gross earnings. But in Australia most students work while studying, so the forgone earnings cost is exaggerated, while future income benefits are under-valued.

There is no perfect method of doing educational returns analysis, and every data source in Australia has limitations. But overall I think the OECD numbers are less useful than existing Australian research on the financial benefits of education.

University research expenditure growth finally moderates

In a Grattan report released last year, I argued that research spending was increasing rapidly, with profits on teaching supporting that growth.

The latest ABS statistics on research spending, released every two years, suggest that this trend might be moderating. Although I have not had time to update the complicated analysis we did to estimate the contributing of teaching revenues to research, the long boom in research spending has stalled. After double-digit real growth rates between 2002 and 2012, from 2012 to 2014 there was only 1% real growth, to $10.1 billion (see chart below).

research spending

As overall university expenditure has not stopped growing in this time period, research expenditure as a share of all university expenditure dropped from the record 41% in 2012 to 39% in 2014 (see chart below).

research as a share of university expenditure

Closer examination of the ABS figures by expenditure type shows that one of the reasons for the moderation in research growth is that capital expenditure dropped by $200 million between 2012 and 2014, but 2012 was an unusually high year for capital expenditure. The ABS records capital expenditure in the year in which it occurs, which means that big projects can skew the results for particular years. That $200 million drop accounts for about 0.7% of the drop in the research share of university expenditure.

Still, it does appear that there may be at least a pause in the trend towards research consuming an ever-larger share of university resources. A small decline in research only staff and a small increase in teaching only staff would also support this conclusion. Whether this moderation is just a pause through a period of Commonwealth Budget constraint and a less lucrative international market, or a more lasting slight re-ordering of university priorities, remains to be seen.

Should people use their superannuation to pay off HELP debt?

For some time former Liberal adviser John Adams and Senator Chris Back have been promoting the idea of using superannuation to pay off HELP debt. Adams put his case on Catallaxy last year, and Senator Back is in the AFR again today on the subject. At The Conversation, Geoff Sharrock has a different take on the same idea, proposing that annual superannuation payments by employers be diverted into meeting that year’s HELP repayments.

You would have to be a bit desperate use superannuation, on which you can reasonably hope to earn a 5-10% a year rate of return on average, to pay off a debt with 2-3% interest. But the proposals are about cash flow, not long-term financial advantage. Any HELP debtor who earns the repayment threshold has to repay more than $2,000 a year, and someone on $100,000 a year will have to repay at least $8,000. While people on these incomes are not poor by general community standards, they could reasonably regard their current needs as more important than additional wealth or consumption in the future.

There is good evidence that people manipulate their income to stay below the HELP repayment threshold (see for example figure 22 on page 40 of our recent report), even though many of them probably will repay eventually. So there would probably be some demand for trading in super for repaying HELP. The Sharrock plan is likely to have more voter appeal than the Adams plan, which will only have the desired effect for debtors who have enough accumulated superannuation to clear their entire debt (as HELP repayments are only based on current income, not outstanding balance*). Most new bachelor-degree graduates will have too little in their super accounts to clear the $20,000 to $40,000 they will typically owe after completing an undergraduate degree, but as Sharrock shows the 9.5% of income annual compulsory superannuation investment is always higher than the compulsory HELP repayment.

But should the government allow superannuation to be used in this way? The two big issues are whether superannuation should be diverted from its core retirement savings purposes, and whether it will save the government money as well as being more convenient for debtors.

Adams is aware of the criticisms, particularly around the long-term impact on retirement income. He says people could be required to make up the contributions later. As with HELP repayment on family income, this would be complex to administer and enforce. It needs a counter-factual amount that would have been saved, and a plan for how it is going to be reached. Sharrock just observes that in his plan most graduates would still have decades of work ahead of them. In part, how big an issue this is will also depend on the level of HELP debt.

Perhaps the bigger danger to retirement savings is the precedent using superannuation for HELP would set. While people who have not been to university don’t have HELP repayments, they do on average have lower incomes, so it is hard to say that HELP debtors have any unique cash flow issues (apart, perhaps, from those created by lifestyle expectations). Adams provides an intellectual differentiation between using HELP for other purposes, such as buying a house, and using it to repay HELP. But inevitably the details would be lost as other people tried to free their superannuation savings for more urgent uses.

The other big question is the effect on the government’s finances. There is value to the government in earlier repayment of HELP debt through lower interest subsidies, and possibly in reduced doubtful debt (although people who think they might not repay in full will be less likely to use their super for early repayment). But the government will lose the taxation it would receive on superannuation funding earnings. Super for HELP also seems open to rorting, by salary sacrificing into superannuation and then using the money to repay HELP debt. It would be a back door way of restoring the bonus for repaying ‘early’ that will otherwise be abolished next year. Many years in the future, there may also be additional issues with people whose super savings are too low for retirement who end up having to rely on government payments.

While I do want HELP debt to be repaid more quickly, on balance I don’t think diverting money from superannuation is the way to do it.
—–
*Unless HELP repayment is completed during the year.

Should HELP repayments be based on family income?

The Fairfax papers this morning have stories on a ‘push’ to repay HELP debtor from family income, in which I am quoted extensively. This idea is considered in our latest Grattan report on lowering the HELP threshold for individual debtors, but we did not propose it.

The appeal in the idea is that our analysis suggests that many HELP debtors who are not currently repaying, and likely a much larger proportion of debtors at risk of never repaying, live in reasonably affluent households. The reason they are not earning more than the $54,000 threshold is that they don’t need to, because another family member makes enough money that the household can maintain reasonable living standards without two full-time earners. It’s a little hard to read, but the slide below from the report shows the disposable (ie after tax and with non-taxable benefits) income of the households affected by our proposed $42,000 threshold. That’s not the whole situation – it excludes households where all HELP debtors earn less than $42,000, or over $54,000. But it illustrates how personal HELP debtor income is not a good guide to overall personal living standards.

family income

While repaying HELP from family income would make a big difference, it was not recommended for a range of reasons:

* How would we determine whether a couple was a couple for HELP repayment purposes? There are some clear potential markers, such as marriage or kids. But only about half the 20-something graduates who were living together as a couple in the 2011 census were legally married. What if you were legally married but had split?

* It would make life more difficult for employers, who currently deduct most HELP repayments. Now they can use their own payroll; in future they would have to know spouse or partner income as well. More people would make incorrect repayments during the year, and could be hit with major additional repayments at tax return time.

* Who would be legally liable to pay? Presumably it would have to be the debtor, with an assumption that partners would often choose to cover it. But that could mean people getting bills that exceed their income. If the principal income earner refused to pay, it would mean that HELP caused default, which it is not supposed to do. If the principal income earner was forced to pay, it would be an unusual case of someone becoming liable for debts they never took out.

While I would not say we should never, ever, consider a different basis for calculating repayment income, for this report I thought there were too many practical and philosophical issues for it to make the list of recommendations.

Polling on the initial HELP threshold

A headline in The Australian this morning reads “Student loan: Poll backs current repayment threshold”.

In a American-style voting system where the largest single minority wins, I guess that’s right. The current threshold is $54,126, and 27% of respondents chose the nearest $10,000 increment of $50,000 (slide below).

MetaPoll

But another way of looking at it is that half of all the sample, and about two-thirds of those who offered a view, came up with a number below the current $54,126. The 2014 Budget proposed reducing the threshold to $50,638. Nearly a quarter of respondents offered a number below the $42,000 we recommended in our report last week.

According to The Australian‘s article respondents were not told the current threshold. If they had been, I expect that there would have been stronger status quo bias. Still, for an issue that has not been prominent until recently I think these poll results are reasonably good, from a reformist perspective.

—-
The poll was conducted by MetaPoll. No precise details on the methodology were published, although younger people were less in favour of a lower threshold than older people.

Student debt repayment times: England compared to Australia

The Australian‘s higher education gossip column yesterday reported some (anonymous) objections to our plans to change HELP’s repayment system.

Some of these are based on the belief that HELP debtors are in a special category who should be treated much more generously than their contemporaries who did not do quite so well at school. For reasons outlined earlier in the week, I do not think this is a convincing point of view.

But there are some more interesting points about repayment times:

A reader contacted us to say it would “perpetrate some asymmetries in the system”. How? “Not all graduate jobs begin with the same salary and not all jobs have the same prospects for wage increases over the working life of a graduate. These two parameters (starting salary and wage trajectory) have a dramatic effect on repayments.” Indeed and both would much more dramatically impact women than men, no doubt.Our reader astutely pointed us to the UK’s loan repayment calculator which from which we can see an investment banker might pay of their debt in 12 years 5 months while someone in sales would take 18 years 11 months to pay off the same debt. Graduates on really low wages (teachers, child care workers, social workers) would take more than 30 years and still have 45 of their debt overhanging because of the power of interest. With a 30 year debt forgiveness model, doubtful debt in the UK looks set to skyrocket…. Our reader says it would be helpful if the Grattan Institute or some other interested body produced a similar modelling tool so we could look at the long-term effects of such proposals.

A loan repayment calculator website that helps students understand their repayment obligations is an excellent idea. However, the analogy with England doesn’t work for either the current HELP system or what we propose in our report. There are some important differences:

Underlying borrowing. In England, most undergraduate courses are a flat £9,000 a year. Given different occupational earnings, that inevitably produces very wide variation in projected repayment periods. In Australia, student contributions are roughly linked to likely future earnings. Admittedly, the amounts are based on a back-of-the-envelope calculation made 20 years ago, and need an update. But work we’ve done before suggests that it tends to cluster by-discipline repayment times much more around the median than would a system based on flat fees, costs, or a fixed proportion of costs.

Interest rates. In England, there is a 3% real interest rate, here it is zero except for the minority of students charged loan fees. But Australia’s system means that compounding interest is a minor issue for debtors compared to England. There, slow repayment can substantially increases the real value of the debt; here it won’t.

Repayment system. In England, debtors repay 9% of their income above the threshold. Here, at each threshold you repay a percentage of all your income that increases depending on how much you earn, from 4% to 8% now. Because their threshold is lower than Australia’s (coincidentally, about the $42,000 we suggest for here) more lower-income debtors make some repayment. But except for English debtors earning below the Australian threshold, this leads to lower annual repayment than here, as seen in the slide below from our report:

England repay

Point of write off. The English write off debts after 30 years if the debtor is still alive. Here it is only on death. Australia will get more repayment from women returning to full-time paid work after they finish their main childcare period.

So high original debts, slow repayments, and real interest on the debt all join to produce long repayment times in England, with the consequent higher risk of doubtful debt exacerbated by premature write-off of debt.

Under the new thresholds we propose in our report (including lower upper thresholds), Australian repayment times would be reduced. That would be the only effect for the vast majority of higher education HELP debtors, who will eventually repay even under the current system. Their total costs stay the same in real terms, and slightly decrease in nominal terms as they will incur less indexation of their debt.

A proportion of additional repayment will be from people who might not otherwise repay. As the report and associated coverage note, most of the personal debt holders in these cases will be female, but the actual effects are likely to be much more gender neutral. In our analysis, it is long-term part-time work by second-income earners that comes out as a major doubtful debt risk, and having some of those debtors repay 3% of their income in HELP repayment will reduce this. However, the repayment will be much less than 3% of household income, and the effects will in practice be shared by usually male partners and children. Using HILDA for analysis, the singles affected are 58% female, which is almost exactly their share of the student population.

Should you only repay HELP debt if you get a financial benefit?

Not surprisingly for an Opposition in an election year, Labor is opposing the recommendation of my latest report to lower the initial HELP threshold from $54,000 to $42,000.

These are the reasons they give:

The income contingent loans scheme (HECS-HELP), introduced by Labor, is based on the principle that you repay your contribution only if you benefit from your education in the way of higher salaries.

It is not a loan scheme as is understood in financial circles, but a social insurance program.

These arguments have their origins in the 1988 Wran report, which recommended the creation of HECS, but they don’t sit easily together. A social insurance program implies some relationship to other social insurance programs, while higher salaries implies relationship to some other counter-factual income, above what the graduate could have earned if they did not continue with their education. Given Australia’s means-tested welfare system for working age people, which is designed to make welfare financially unattractive compared to work, the two arguments are in inherent tension.

In practice, neither benchmark has been actively used to set the threshold. The current threshold is around $20,000 above the minimum wage and common social security programs. But at $54,000 in 2014-15, the threshold is below all persons average weekly earnings (the original Wran suggestion) of about $59,000. The threshold was last re-based in 2004-05 (we have the historical thresholds in an appendix to the report), but since then has increased by 17 per cent in real terms, as although not set according to average weekly earnings, it is indexed to movements in average weekly earnings.

Even if we accepted Labor’s financial benefit principle, it is not at all clear how it should be implemented. What is the counter-factual for financial benefit? Should it be age-adjusted? Should we have a different threshold for diploma borrowers? Should it be based on household rather than personal income? So far as I can see, while several people other than Labor have offered the financial benefit idea since the report was released, none have attempted to think through how it should be set.

Our report argues that ‘we should not turn HELP features that are based on political judgments of their time, or on assumptions that no longer apply, into principles that cannot be overturned’. Like most Grattan reports, it is at least an implicit argument against status quo bias, against all the rationalisations that attach themselves to existing policy arrangements, and an attempt to think about what a policy should be trying to achieve, and whether that can be done in a better way.

The average weekly earnings idea was a by-product of the late 1980s politics involved in ending free (to the student) higher education. But it sets HELP debtors up in quite a privileged position compared to other people being protected from financial hardship by the government, and we should ask why. The most obvious answer is that we need to protect people from financial risk to attract them to higher education. That’s an empirical issue, but nobody has shown that the current threshold is the right one from that perspective. England and New Zealand now, and Australia in the past, have all had lower thresholds without significant demand-side issues. Until recently, the barrier to higher education participation has always been the supply of places, not demand for them.

In the absence of a strong principled or empirical argument for the current repayment system, threshold reform is an element of moving HELP towards focusing subsidies on debtors with long-term low household income, ie social insurance. This keeps the justifiable element of Labor’s argument, while leaving open the idea of reducing the initial repayment threshold.

Low ATAR offers result in few low ATAR completions

The SMH has lots of NSW university offers data to contribute to the annual ATAR controversy. It’s the usual story of students being offered places with very low ATARs.

There are always lots of threads to this controversy. Are low-ATAR admissions a sign of declining academic standards? Do we place too much emphasis on ATAR anyway? Should universities be more open about their admission practices? Do low-ATAR enrolments risk putting incompetent professionals out into the workforce?

While there are real issues here, some numbers are useful for putting things into perspective. I’m using 2014 numbers because I have enrolment data for that year but not yet 2015.

The first three columns in the chart are from tertiary admission centre data. The first thing to note is that in 2014 about 60% of very low ATAR – 50 or below – applications did not result in any offer. Of the applications that did result in an offer – the issue in the news this week – about half resulted in the prospective student rejecting it. The offer might have been for a course that was not their first preference, or possibly they only applied to keep options open, without having a firm intention to go to university.

diminishing ATARs

The next bar on the chart is enrolments of students who reached the HELP census date, usually around the end of March. It shows that about a third of below 50 acceptances leave before the census date, presumably having decided that higher education was not for them, at least not at this time. The low-ATAR issue is starting to look at lot smaller than it did at the offers stage. Final enrolments of 50 or below ATARs from 2013 school leavers in 2014 were only 3% of all that cohort (although total low-ATAR enrolment is higher than this, due to students who finished school in other years).

I have added a projection of their final completions, based on the Department’s 2005 cohort analysis. If that is a good guide to the future, less than 20% of below 50 ATAR school leavers who receive an offer will eventually get a degree.

My perspective on this is primarily a consumer protection one. Prospective students are not being informed of the risks they are taking. Universities say that they are looking at what predicts success other than ATAR, but only rarely do they release any evidence of this that can be checked by independent analysts. The regulator’s vague statements are less than confidence inspiring. The numbers of people taking major risks are not that large in the context of total enrolments. But we should be doing more to ensure that they are making decisions that are in their own long-term interests.

ATAR and higher education admission

In this morning’s papers, there is controversy over the declining proportion of Victoria universities publishing ‘clearly-in’ ATARs – the ATAR above which all applicants are admitted. In Victoria, the share of courses with a clearly-in ATAR has declined from 40% last year to 25% this year.

Some universities say that ATARs, and especially ATAR cut-offs, can be misleading. ATARs are not necessarily the basis of admission or the only basis. As the chart below shows, in the lower ATARs a significant minority of recent school leavers are admitted to university on some basis other than their secondary education. More will be admitted based on secondary education but without relying on ATAR.

ATAR school leavers

The overall admission process is larger and more complex than for recent year 12 students. For 2014 commencing undergraduate students at public universities, people who had completed school the previous year were around two-thirds of those with ATARs in the enrolment data. The remainder were people who had finished school in earlier years, and people admitted on other grounds, the largest of which were prior vocational education, prior higher education, and mature age entry.

As can be seen in the chart below, this pushes down the share of people admitted based on secondary education, especially in the lower ATAR groups. Only just over a quarter of people with ATARs below 50 were admitted based on their secondary education. The biggest single category was vocational education (31% of admissions). About half of people with ATARs who were admitted based on vocational education had ATARs below 60, confirming it as a pathway into higher education for people who did not feel ready for higher education, or could not get into higher education with their school results alone.

ATAR admission color switch

I’ve consistently opposed minimum ATARs as too blunt a tool for regulating admissions. But the universities have conflicts of interest in this area. They want to fill all their places and avoid any reputational issues from admitting low-ATAR students. Given what we know about the link between low ATAR and non-completion, going to university may not be the best option for low-ATAR students (and indeed most don’t go to university).

And while the published ATAR cut-offs can be misleading, information on the range of ATARs that previously resulted in offers or enrolments would be helpful. They would give prospective students a guide to whether they are likely to be admitted, and to the academic strength of their potential fellow students.

Labor supports reducing HELP costs

In a surprise move, Labor has agreed to back abolition of the 10% discount for paying student contributions upfront and the 5% ‘bonus’ for voluntary repayments of HELP debt.

For reasons we don’t fully understand, the proportion of people paying upfront has been in decline for a long time, as the chart below shows. Halving the discount in 2012 accelerated a long-term trend but was not a major turning point. The government will hope that the latest change will not dramatically affect the numbers, as borrowing has costs and risks for them.

Previous research has suggested that many of the upfront payments are from parents wanting their kids to be debt free, so perhaps the final abolition won’t make much of a further difference. Indeed, discount abolition may already be partly evident in the data, as anecdotal evidence suggests many people believe it has already been abolished, following on from the original publicity in 2013 when Labor first proposed it (to pay for Gonski; they later reversed their support saying the Coalition was not meeting Gonski commitments, before going back to their original position today).

HECS-HELP
Source: Department of Education Student liability statistics, various years

Any further decline may be due in part due to expanded eligibility for HELP loans. Currently New Zealanders and permanent residents are entitled to a subsidised place have to pay student contributions upfront, but soon some long-term NZ residents will be able to take out HELP loans.

Abolition of the 5% bonus for voluntary repayments is a sensible move. Previous ATO data suggested that it was mostly used by people towards the end of their repayments anyway, when the government would be better off just waiting a little longer and getting the full face value of the loan.

Removing the bonus will also weaken an incentive to borrow under FEE-HELP. Some students can borrow unnecessarily and then repay ‘early’ (ie pay the full amount much later than they could have) and use the bonus to pay less than the face value of the loan. I did this a few years ago to draw attention to this rort. I’m glad it is finally going to be abolished.

Hopefully this signals that Labor is taking HELP’s costs seriously. There is plenty more to do.