If you even casually follow the news, you've probably heard that Americans owe a record $1.3 trillion in student loans. These loans are now second only to mortgages as the largest source of household debt. Policymakers are scrambling to respond to what is widely perceived as a debt crisis akin to the mortgage meltdown: seven million borrowers are in default, while millions more are behind on their payments.

Readers – and politicians – often assume that higher debt must lead to increased risk of default. But the fact is, default is highest among those with modest student debts. Of those borrowing under $5,000, a whopping 34 percent end up in default. For those borrowing more than $100,000, the default rate is just 18 percent.

What explains this counterintuitive pattern? In a word: earnings. And this should influence the way we think about student debt and government policy that has gone astray.

The Link Between Earnings and Default

The most indebted student borrowers are likely to be those who attended graduate school or who earned undergraduate degrees at expensive, elite institutions. These borrowers typically spent many years in higher education and so racked up many years of debt. But they also build up a lot of human capital, which typically pays off in higher salaries once they enter the job market.

Small borrowers, by contrast, tend to be those who went to for-profit technical schools or community colleges for a year or less. They finished (or dropped out) quickly and so borrowed little. But they also build up little human capital along the way and so do relatively poorly in the labor market.

A recent research paper by Adam Looney of the U. S. Treasury and Constantine Yannelis of Stanford University delved into this link between earnings and student borrowing. They had access to new data on a subject for which figures have been frustratingly incomplete. This new data matches records on federal student borrowing with the borrowers' earnings from tax records (with identifying details removed to preserve privacy). And it includes information about the borrowers and the amounts borrowed, the colleges the borrowers attended, their repayment records and their earnings both before and after college.

Looney and Yannelis found that half of the increase in loans between 2003 and 2013 was driven by a surge in borrowing to attend for-profit and community colleges, where enrollment exploded as workers fled a weak labor market. These borrowers gained little from their investment: the median income for those exiting these schools in 2010 was a paltry $22,000. The explanation lies in who attended and why.

As part of the anti-recession stimulus package, the federal government encouraged unemployed and low-skilled workers to head to college for training. Increases in federal need-based Pell Grants and the new American Opportunity Tax Credit (which was worth up to $2,500 a year for households with annual incomes of up to $160,000) helped workers to take this option. But the huge influx of students taxed the resources and capacity of community colleges.

Wait a moment. Shouldn't an increase in demand make any enterprise better off? In private markets, a surge in demand produces additional sales revenue that can be plowed back into expanding capacity. At public colleges, however, tuition covers only a fraction of the cost of educating students. Public colleges thus rely heavily on state aid to supplement tuition revenue. Prices (tuition) at public colleges would resemble prices at private colleges were it not for this substantial government subsidy.

During the Great Recession, however, states cut their appropriations to public colleges as part of their efforts to offset falling tax revenues. Cuts in state funding to community colleges created a cascade of consequences.

Public colleges, including community colleges, made up part of the loss by raising tuition. In response to this increase in prices, students at community colleges borrowed more. Traditionally, community college students have borrowed little or nothing; this new borrowing therefore reflected a sea change in how these students finance their schooling.

The drop in state funding also increased student borrowing at for-profit colleges. How? Community colleges, strapped for funds, could not accommodate the increased demand. For-profit colleges enrolled many of the students that community colleges could not.

For-profit colleges have long provided training for the same sort of low-skilled, though ambitious, workers who attend community colleges, but they have always been a small part of this market. This changed significantly during the Great Recession. In 2000, for-profits enrolled just 4 percent of undergraduates in the United States; by 2010, they were enrolling 11 percent.

As for-profit enrollment rose, so, too, did borrowing by their students. For-profit colleges charge high tuition and (unlike private, non-profit colleges) provide little aid (which are really discounts). As a result, for-profit students – unlike students at community colleges – have always taken on relatively heavy debt. By the same token, defaults among borrowers at for-profit colleges have always been high, especially during economic downturns. The Great Recession continued this pattern, with one key change: instead of representing a tiny minority of college students, for-profit students were now a bigger chunk of undergraduate enrollment and undergraduate borrowing.

When the surge in for-profit borrowers left school and were obliged to start repayment, defaults rose rapidly. For those leaving for-profit colleges in 2010, the five-year default rate was a painful 28 percent. The default rate was also high (31 percent) among community-college borrowers, but there were far fewer of them. While there are many more community-college students than for-profit students, it is still relatively rare for community-college students to borrow.

In hindsight, this was almost inevitable. The very poor employment prospects for low-skilled workers make borrowing very risky for them. Indeed, their low earnings make it difficult to support even small amounts of debt.

How Much is Too Much?

Borrowing is not risky for all students. Graduates of four-year colleges, especially colleges with selective admission standards, tend to earn good salaries and to pay back their loans. At selective schools, the default rate is only about 6 percent. Graduate students also have little problem repaying their debt and have a similarly low default rate.

Note, too, that student loan debt is lower than is widely perceived. Consider students who first enrolled in college in 2003-4. Six years later (the length of time it takes a typical undergraduate to earn a bachelor's degree), 44 percent had borrowed nothing and another 25 percent had borrowed $10,000 or less. That is, 69 percent of undergraduates borrowed $10,000 or less.

Another 29 percent borrowed between $10,001 and $50,000, while just 2 percent borrowed $50,001 or more. Today's entering college students appear to be on a similar path. Thus, while attention is focused on extreme cases, only a very small share of undergraduate borrowers are liable for the six-figure debts that dominate the headlines.

Remember, too, that a four-year college education remains one of the best investments that a young person can make. Median earnings among all young workers (those aged 22 to 27) with a 4-year college degree in 2015 was $43,000, compared with $25,000 for those with just a high school degree. Take that difference and multiply the figure across four decades of work, and it is quickly apparent that $30,000 in student loans – the typical debt for a student graduating from a four-year public college – is a sensible investment.

Could Free Community College Solve Our Student-Debt Woes?

If we are determined to reduce distress and default among student borrowers, the target should not be the graduates of elite four-year institutions but rather those who dropped out of non-selective programs. Going forward, one way to reduce their debt distress is to keep them from borrowing in the first place.

Community colleges are the traditional entry point for low-income students whose parents did not go to college, and historically they have charged very low prices – a few hundred dollars a semester was not atypical even two decades ago. With a need-based Pell Grant from the federal government, a community college student could get by without borrowing. Students who dropped out could thus return to the labor force without debt. In essence, society took on much of the risk of the college bet by bearing the direct costs. Bringing prices at community colleges back to their historic standard – near zero – would be one way to reduce borrowing and, thereby, debt distress.

However, making community college free would not end the debt problems of those who attend for-profit institutions. As noted above, for-profit colleges draw students from the same pool as community colleges. Yet most of them would still need to borrow large sums to attend. Reducing the price of community college would presumably draw some students away from their expensive private counterparts. But it's unlikely to shut down the whole sector. So, whether or not we make community college free again, we'll need a well-functioning system for borrowing and repaying student loans.

A Flexible System of Student Loans

One simple way to reduce defaults is to lengthen the time frame of loan repayment. In the United States, the standard period is 10 years. Other countries let students pay back their loans over a far longer horizon. In Sweden, students pay their loans back over 25 years. For a loan with an interest rate of 4.3 percent (the current rate), stretching amortization to 25 years would halve monthly payments.

Lengthening the horizon of loan repayment makes economic sense. A college education is an investment that pays off over many decades. With most loans, the length of repayment corresponds to the useful life of the collateral. We take out 30-year mortgages on houses, but just five-year loans on cars. Cities and states sell long-term bonds to fund highways and other infrastructure that will provide services for many decades.

A complementary way to make debt manageable is to link payments to income. In an income-based repayment system, payments would rise and fall with earnings, thereby reducing pressure on borrowers. Australia, Chile, New Zealand, Thailand and Britain have all gone this route with student loans. In Britain, for instance, borrowers pay 9 percent of their annual income that exceeds £21,000 (about $30,000); any remaining balance is forgiven after 30 years.

The federal government does have income-based repayment options for student borrowers, such as the Pay As You Earn program. But PAYE is not the default repayment plan. The automatic option for borrowers is a 10-year mortgage-style fixed payment. Borrowers must proactively apply to the income-based programs before being admitted. Eligibility must be renewed annually.

The Consumer Financial Protection Bureau has documented the difficulties that borrowers have in navigating this process. With PAYE, and all the other income-based repayment programs, every change to earnings requires a new application to adjust the loan payment. Even if earnings don't change, staying in an income-based plan requires an annual round of complicated financial paperwork. Those who most need a helping hand are probably the least able to navigate this bureaucracy. Indeed, the number of borrowers in these flexible repayment plans is much lower than the number in distress and default, which is evidence that the current system isn't working.

Note, too, that while PAYE theoretically limits payments to 10 percent of income, outlays can actually consume a much larger share of a borrower's earnings in a given year. That's because, with all the income-based plans in the United States, payments are calculated as a percentage of the previous year's income. But income can change a lot over the course of a year, especially for the low-skilled marginally employed. For those patching together several part-time jobs, hours and earnings can bounce around weekly. The payment that would have been affordable last year may well be unaffordable this year.

To effectively buffer earnings shocks as they arrive, payments need to adjust dynamically with earnings. And this is not beyond the capacity of law and finance.

Social Security is a good model. Workers do little paperwork to make Social Security contributions: they complete an initial W-4 form and employers handle the rest. Social Security contributions then automatically rise and fall with earnings. Loan payments could be handled the same way. I've written a brief on this idea for the Hamilton Project.

There are now several proposals circulating in Washington that would get more troubled borrowers into income-based repayment plans. Some would keep the standard 10-year repayment plan but automatically shift borrowers into income-based plans if they fell behind on payments. Others make income-based payment universal, as it is in England and Australia.

What About Cutting Interest Rates?

In Washington, it is frequently argued that high interest rates are at the root of the student loan mess, increasing defaults and discouraging college attendance. But there are good reasons to doubt that they have much impact on attendance, yet they are an expensive way to reduce defaults.

Econ 101 predicts that a lower interest rate would increase college enrollment because it would reduce the lifetime cost of college. A rational decision maker would take this into account, summing the projected lifetime benefits of college and subtracting the projected lifetime costs – including interest due on loans after leaving college. But building on the work of psychologists, behavioral economists have shown that this rational model fails predictably when people have to accept burdens in the present to gain benefits in the future. This applies to a variety of important life decisions ranging from deciding whether to save for retirement, to exercise for health or to go to college. Strong evidence suggests that, in these situations, tangible incentives that are felt at the moment of decision making are most effective in changing behavior.

Interest-rate subsidies fail the tangibility condition: students are handed the same amount of cash now whether the loan's interest rate is 2 percent, 4 percent or 10 percent. By the same token, they fail the "in the moment" test, since the benefits of reduced interest rates (lower monthly payments) arrive only after students have left college.

Lower interest rates might reduce defaults, but the effect is likely to be quite small. The payments students make are determined mostly by principal rather than the interest rate. Most borrowers are enrolled in a 10-year mortgage-style payment program, in which payments are fixed when the students leave college. For a loan of $20,000, cutting the interest rate by 2.5 percentage points would reduce the monthly payment by only about $25 (from $230 to $205).

Not only does reducing interest rates have little impact on struggling borrowers, it is hugely expensive. When rates are cut, everyone gets the benefits, including those with high earnings who have no difficulty repaying their loans. An interest subsidy is therefore a poorly targeted, expensive and ineffective tool for reducing loan defaults.

If we want to hand borrowers a windfall, with little effect on college attendance or default rates, lowering interest rates on student loans would do the job well. This is a perfectly reasonable policy goal, but we should be clear that this is what we are doing – transferring wealth to borrowers.

If our goal is to increase college attendance, grants or lower tuition would do the job more effectively than lower interest rates. If our goal is to reduce defaults, an income-based plan in which payments flex automatically would also accomplish more at lower cost than changing interest rates.

The Big Picture

Governments across the world provide student loans, allowing students to borrow against the expected lifetime income gains created by a college education. While borrowing has risen in the United States, so too has the return from schooling. The typical student holds debt that is well below the lifetime benefits of a college education. To put it another way, the typical student borrower is not "under water," unlike many homeowners during the mortgage crisis.

The real problem is the mismatch in the timing of the arrival of the benefits of college and its costs, with payments due when earnings are lowest and most variable. The solution is an income-based-repayment structure for student loans, with payments automatically flexing with earnings over a longer horizon than the current 10-year standard.

As noted above, income-based repayment options are already available in the United States, but the administrative barriers to accessing them are formidable – and especially daunting to the students at greatest risk of default. Further, the existing options do not adjust loan payments quickly enough to respond to the high-frequency shocks that characterize young people's earnings, especially during a recession.

There's no doubt that Americans view exploding student debt as a problem. And there's little doubt that many of them would be willing to pay something in order to relieve the debtors' pain. It would be unfortunate if policymakers wasted this window of opportunity by making changes that did little for those who most need help.