The Washington PostDemocracy Dies in Darkness

Everything you could possibly need to know about student loan budgets (with gifs!)


On Monday, we wrote a post explaining that "No, the federal government does not profit off student loans" (as the original headline put it). A number of people have pushed back at that conclusion, pointing out that it depends on how you do the arithmetic.

So here's a fuller explanation of what is in dispute in the debate over whether federal student loan programs are profitable for the government or not. And the details show why people can follow the same set of facts to such different analytical conclusions. Of course people disagree all the time about budget priorities. But in budget debates, it also matters quite a lot how you add up what we spend on what -- and student loans are a prime example.

Liberals crunch numbers like this, but conservatives crunch numbers like this

Among the more heated of budget accounting debates has to do with how to deal with loans, and in particular student loans, that the government makes. Like with anything else in the budget, you want to compare what goes out (in this case, the initial money loaned to students) to what comes in (the payment checks sent by previous loan recipients).

For other budget categories, that's easy. It's simple enough to compare, say, the amount we're collecting in cigarette taxes in 2013 to the amount we're spending on preschool in 2013, since you're looking at actual numbers in one year.

But it doesn't make any sense to compare the loans that go out in 2013 to the payments that come in 2013. Those payments are on loans made years, sometimes decades in the past, with different interest rates and histories than the loans being made in 2013. So rather than comparing those two numbers, budgets generally compare the amount going out in the form of an initial loan with how much they expect to come back from students repaying that specific loan. How exactly to do that, however, is very much disputed.

No, seriously, pay attention

This probably sounds super-boring. But it's important. In the student loan case, it's often the difference between federal programs being counted as turning a profit and them being counted as losing money. That matters. If they turn a profit, they're just a business the government runs because the private sector, for a number of reasons, can't do the job properly, and the question becomes whether to reduce that profit by cutting interest rates or to use it for other purposes. If they're operating at a loss, they're a form of government aid to students like Pell Grants, and it's worth asking whether they're the best form that aid could take, and whether raising interest rates might make sense.

That's why the approach that the CBO currently uses, due to the Federal Credit Reporting Act of 1990 (FCRA), is often characterized as liberal, and the alternative, "fair-value reporting," is characterized as conservative, or at least "neoliberal". The FCRA method shows student loans as profitable (as seen in the CAP chart above), which suggests that the government is taking money that could be in the hands of needy students. The fair-value method show, at the very least, lower profits, and in many cases that student loans are a subsidy. That suggests that the programs should perhaps be less generous, that the money would be better spent elsewhere (perhaps on Pell Grants).

That's why NYU Law's David Kamin wrote, in a piece arguing against fair-value accounting, that it "puts a thumb on the scale against government action." If that's true, it's a really important problem with the approach. But if the FCRA method puts a thumb on the scale for government action, that's an equal and opposite problem. So who's right?

How the government calculates loan cost now

The cost of the student loan program in 2013 is currently calculated by taking the net present value (NPV) of every loan issued in 2013. That means taking all the payments you expect to be made on the loan throughout its life, and "discounting" them back to 2013 dollars.

You discount because money that you can only access in the future is considerably less valuable than money you can access right now. I'd be pretty mad if The Washington Post decided to hold onto my paycheck and give it to me in 2043. A dollar in the future is worth a lot less to me than $1 now does, both because a dollar in 30 years will be worth less than a dollar now due to inflation, and because that's 30 years when I could have used the money for other stuff. I could have invested it and earned 30 years of interest, for example, or take a vacation now instead of in the distant future. In short, the government understands the principles of YOLO.

And caring more about my well-being now isn't necessarily irrational, since I know less about what future me will want than about what present me wants. Derek Parfit has a good argument along these lines. Grist's David Roberts has a very good explanation of this stuff if you're interested, complete with pictures of otters to make it easier to swallow.

So first, when determining the cost of these programs, the CBO estimates what cash flows it expects from the loans in question. Those cash flows take into account "the probability of default each year, the recovery rate on defaulted loans, the planned repayment (amortization) schedule of a loan, estimated voluntary prepayments, and the fee schedule." Basically, everything that could possibly affect what money's going out and what money's coming in. So the issue isn't that the CBO isn't taking the risk of default into account - that's already baked into the picture.

Then, once that's all calculated, you discount. Currently, the government always discounts using the same rate: the interest rate on Treasury bonds.

What that leaves out

That, suffice it to say, is not how discounting works in the private sector. Private firms, as the CBO has explained in arguing against the way Congress currently forces it to evaluate loan programs, generally operate by "discounting those flows using the rates of return on private loans (or securities) with similar risks and maturities."

Given that those rates of return are generally higher than the rates of return on Treasury bonds, that means a higher discount rate. And a higher discount rate means predicting less money coming into the government in the future. Generally, private sector practices would predict considerably less revenue for loans than the government predicts when it issues loans.

Why are they different? Because private firms generally have to pay investors for what's called "market risk." That's the risk that remains in an investment even if it's well-diversified, even if you're managing the risk of specific investments well. The risk that a recession will send the value of all investments plummeting, for instance, is a market risk. If taxpayers made student loans on their own, they'd demand compensation for that kind of risk, which would push the market price of the loan up.

So the government's current accounting methods use a price below the market price for loans. And it's only for loans. Let's say the government wants to buy your house using eminent domain. They don't get to offer a below-market rate. They have to offer the market price for the house.

That's why many observers want to switch to fair-value accounting, the way the private sector does business, from FCRA, the way the government does now. The Congressional Budget Office, for one, supports the change, as did the bipartisan Pew-Peterson Commission on Budget Reform and the nonpartisan Financial Economists Roundtable.

The Ryan budget backs the change, as did Nobel laureate Peter Diamond and former budget director Peter Orszag in their book, "Saving Social Security". Suffice it to say, Orszag and Ryan don't usually agree on budget matters. It's enough for Kamin to write that there's a "growing consensus among economists, budget experts, and policymakers" behind fair-value.

Is the government special?

A lot of people object to using private sector methods because the government really is special, especially on student loans. For one thing, you can't discharge student loans in bankruptcy. That means that defaulted-on loans actually get paid back at remarkable rates. In fact, the Department of Education finds that defaulted-on borrowers actually pay back more than the principal and interest owed at the time of default. They still lose money on defaults, because of the cost of collection and discounting, but not nearly as much as, say, a bank would lose on a mortgage default.

The cost of collection is also much lower for the government than for private entities. It can hold back tax refunds to pay off loans. It can garnish wages without a court order. If you make it to your 60s with outstanding debt, or get on disability, it can take your Social Security. To quote The New Inquiry's Malcolm Harris, saying the government is better at collections than private lenders is "like if the mob claimed to be better at security for your small business than the guard you hired." It plays by different rules entirely.

But here's the thing: all of these advantages should be reflected in the expected cash flow predicted by the federal government. "That cash flow has any advantages the government has baked into it," New America's Jason Delisle explains. All that changes between FCRA accounting and fair-value is the discount rate. Now, if those cash flows estimates are wrong, that's a problem — and we'll get to that later on — but any inherent advantages the government has over private lenders should already be reflected, and won't bear on which accounting method is better.

Is treating it as special okay?

But, critics of fair-value argue, there are ways in which the government is not like a private business in this regard that aren't reflected in the cash flow, and that need to be reflected in the discount rate. The government doesn't have to be risk averse on behalf of its citizens, just because they are, the argument goes. They therefore don't have to account for risks that would trouble private investors. "The fundamental premise of the argument in favor of so-called fair-value scorekeeping is that, because private markets are (appropriately) loss averse, the government should be, too," Richard Kogan of the Center for Budget and Policy Priorities said. "We disagree."

"Individuals are risk averse in part because their financial assets are likely to be needed at specific times, even when the value of those assets has declined," write CBPP's Kogan, Jim Horney, and Paul Van de Water. "The general fund of the Treasury, in contrast, is rarely or never in that position because, as history shows, when times are bad it can borrow very inexpensively." That means the government can spread risk over generations, which individuals often can't.

But spreading costs more widely definitionally doesn't get rid of market risk, which exists even when portfolios are fully diversified and risks are spread as widely as possible. Even if the Treasury can always borrow at low rates, it still has to pay those loans off eventually. The cost of the borrowing doesn't go away. "Even if the federal government can spread risks widely," the CBO writes, "it cannot eliminate the component of risk that is associated with fluctuations in the aggregate economy—market risk—and which investors require compensation to bear."

Moreover, the fact that individuals are risk-averse should bear on the government. It uses our money. "The federal government is just a passthrough for your money and my money," Delisle explains. "To the extent that they don't use a value that you and I think are legitimate, now you've got a real problem."

A budget is a budget is a budget is a budget

The more serious argument against fair-value accounting is that while market risk is a cost, it's not a budgetary cost, since it doesn't necessarily reflect money that the government is going to have to pay out. As CBPP writes, "the federal budget is supposed to record the amount that the Treasury disburses on spending programs and the amount it receives in revenues, and to show the difference as a surplus or deficit."

They argue that just because the additional riskiness of some loans poses a cost to taxpayers upon whose behalf the government is making them doesn't mean that that cost should necessarily be reflected in the actual budget. That's why the Center for American Progress and former CBO director Robert Reischauer oppose using fair-value for budgeting.

But here's the thing. Even the most student critics of fair-value accounting admit that the costs it identifies are real social costs, that need to be taken into account in any cost-benefit analysis of a policy. As the CBPP authors write, "The concept of a risk aversion cost can and should play a part in the cost-benefit analysis that policymakers should undertake in deciding whether a government program constitutes wise public policy." Kamin agrees, writing, "Policy tradeoffs should be made based on a cost-benefit analysis that fully incorporates social costs and benefits."

The dispute, then, is over the rather technical point of whether this cost should be in the budget or in subsequent analysis. No one disagrees these are real costs that need to be taken into account. For policy analysis purposes, fair-value accounting includes both what everyone agrees are budgetary costs, plus costs that everyone also thinks are important. It's hard to argue, then, that it doesn't give fuller information than the current accounting approach. "My understanding is that it's pretty much a slam dunk for the fair value approach," Matthew Chingos, an education expert at the Brookings Institution, concludes.

Does the metric you use really matter?

Maybe all this is for naught, though. As Malcolm Harris pointed out on Monday, the latest CBO numbers show that even under a fair-value estimate, the loans originated in fiscal year 2013 will turn a profit. A much smaller profit, to be sure — the difference between the two types of estimates is about $31 billion — but a profit nonetheless. The subsidized Stafford loan still costs the government money, amounting to a 11.5 percent subsidy, but the subsidy rate on unsubsidized Stafford loans and PLUS loans is negative for that year. That doesn't mean that the government's actually turning a profit, not least because it doesn't take into account administrative costs, but it's notable.

It also contradicts previous years' numbers. In 2007, the CBO found that Stafford loans started in 2006 were subsidized at an average of 8.3 percent. In 2010, it estimated that directly provided loans would provide an average subsidy of 12 percent a year from 2010 to 2020, and that the since-discontinued guaranteed-loan program would offer a 20 percent subsidy. But now the average subsidy is negative.

There are two things going on here. One is that these programs vary from year to year. Returns on all kinds of assets are depressed at the moment, owing to the recession, which means that even discounting using market rates will produce high paybacks in the future. That suggests that when interest rates get to a more normal place, these programs are going to start showing up as subsidies again.

Another is that the cash flow estimates could be wrong. "I'm at this student loan conference now, and was talking to lenders in the hallway, and asked, "What do you think of the expected default rate?" Delisle tells me. "They say, 'they're outrageously low.' My suspicion is that they're outrageously low because they want to record a good number." If the default rate is, indeed, outrageously low, then even a fair-value estimate would show too much money coming in in the future, and show a profitable student loan program where none exists.

It's hard to say whether this is the case, since there's no research independent of the Department of Education on the default rate, and DOE has a clear interest in making it look low. But it matters. "When you want to figure out the cost of the program, you're basing your models on repayment behavior, you're trying to guess what the repayment behavior will be on loans originated today," Chingos says. "So even fair value stuff you have to take with a grain of salt."

One thing that isn't affecting this much at all is the transition to all direct loans that happened in Obamacare. Recall that in 2010 even the direct loan program showed a big subsidy. It follows that in years with similar interest rates to 2010, the current policy regime will show up as a big subsidy. But there is another policy change that affects this picture…

Why income-based repayment (IBR) will change everything

However, one reason to believe that repayment rates being forecast now are almost certainly too low is that a ton of student loans are set to be forgiven in coming decades. That's because of a bevy of programs that allow graduates to pay back loans based on their income rather than their loan amount.

"Pay as you earn," for instance, lets borrowers pay out 10 percent of their disposable income, if that amount is less than their normal student loan burden, for 20 years, and then have their loan forgiven. People in public service organizations get their loans forgiven after 10 years.

That's only for people with loans started after 2007, but other folks can do income-based repayment, which charges 15 percent of disposable income, and forgives after 25 years (10 for public service). There's also the income-contingent repayment program and the income-sensitive repayment program, which operate similarly for other categories of borrowers.

What all this means is that the government is going to be taking in less in monthly payments going forward, and then forgiving a whole lot of debt. There are a few cases where people will pay more interest under these programs, as a result of the longer repayment period, but it's almost certainly going to make loans more costly going forward.

"It dramatically increases the subsidy rate on the loan. Even under the official rules, OMB and DOE will tell you that there's a 15 percent subsidy rate on these loans if they use income-based repayment," Delisle says (update: the exact subsidy rates for Stafford loans using income-based repayment were around 17 percent in 2012 and are projected at 10 percent this year, not even using fair value). The true, fair-value subsidy rate is significantly larger than that.


We've come a long way at this point from "how does arithmetic work." But the short version is that (a) fair-value accounting is a better way of tracking the true costs of these programs and (b) even if the CBO numbers show a profit in 2013 for student loan programs under fair-value, that doesn't mean there's actually a profit.

The effects of income-based repayment, the possibility that default rates are higher than CBO predicts, and administrative costs all means that it might well have taken a loss this year, and will most likely take losses going forward. IBR, especially, represents a massive subsidy to students that didn't exist before.

The takeaway, then, is that student loan programs aren't just businesses the government happens to run. They're subsidies that spare students the true market cost of loans. Maybe those subsidies are a good idea. Maybe paring them back in favor of increased grants would be a better plan. Maybe IBR is a good subsidy, but interest rates should be higher.

The point is that the conclusion that student loan programs are subsidies doesn't necessarily tell you what policy is best. It just means that they're not a free lunch. Understanding that is crucial for making good policy in this area.