House Bill Would Artificially Inflate Cost of Federal Credit Programs
Revised January 24, 2012
The House Budget Committee approved legislation on January 24 that would change the federal accounting of direct loans and loan guarantees in ways that would overstate the federal costs of those programs. As a result, the legislation also would overstate the size of federal deficits.
The Federal Credit Reform Act of 1990 changed the budgetary accounting of federal credit programs. Previously, the budget displayed the costs of credit programs on a yearly basis – that is, based on the federal costs from loans or guarantees in any particular year, offset by loan repayments in that year. Since the 1990 law, the budget now displays the expected total net costs of loans or guarantees up front – when the government issues the loans and loan guarantees – rather than year by year over the course of their lifetimes.
The legislation passed by the House Budget Committee – H.R. 3581, introduced by Rep. Scott Garrett (R-NJ) and endorsed by House Budget Committee chair Paul Ryan (R-WI) – would significantly change the rules in place since the 1990 law. It would require the Congressional Budget Office (CBO) and the Office of Management and Budget (OMB) to add an extra amount to the budgetary cost that they show for loan and guarantee programs, based on the additional amount that private lenders would charge if they, rather than the federal government, issued the loans and loan guarantees. By overstating the federal costs of credit programs, the proposal would overstate federal deficits and force policymakers to offset these phantom costs with phantom offsets to avoid overstating the debt as well.
This proposal is not based on any claim that current estimates of the federal outlays and receipts associated with federal credit programs understate the actual federal costs of these programs. Quite the contrary, by requiring CBO and OMB to add an extra amount to their estimated cost of federal credit programs, the legislation would artificially inflate the programs’ estimated budgetary cost. Consequently, the budget treatment of federal credit programs under H.R. 3581 would conflict with the basic purposes of the budget and with the way that the budget records all other budgetary activities.
Credit Accounting Under Current Law
The federal budget generally records revenues and spending on a cash basis. That is, the budgetary cost recorded for a program in a fiscal year is the actual cash expended on that program in that year, and the federal budget deficit for a year is the difference between total cash expenditures for all programs in that year and the total amount of cash collected as revenues in that year. By 1990, however, there was widespread agreement that showing the effect of government credit programs on a cash basis was problematic because it did not facilitate a meaningful comparison between the cost of credit programs and the cost of other programs, or between the cost of direct loans and the cost of loan guarantees.
The problem was not that incorrect amounts of cash disbursements and receipts were being recorded for credit programs. The problem, rather, was that for those programs, showing cash transactions when they occurred did not provide policymakers considering whether to cut, maintain, or increase those programs with meaningful information about the cost of their decisions over time.
Loans and Loan Guarantees Formerly Recorded on Cash Basis
Before the Credit Reform Act, a $100 direct loan was shown in the budget as costing $100 in the year the direct loan was made. The cash the government subsequently received in repayments of principal and interest was recorded in subsequent years as those payments were received. As a result, a $100 loan in the coming fiscal year appeared to have the same budgetary effect as a $100 grant in the same year, even though the loan had a significantly smaller true impact on the budget than the grant, since a substantial portion of the loan would be repaid in subsequent years.
In contrast, a federal guarantee of a $100 loan appeared under the pre-1990 budget rules to produce income for the government in the year that the guarantee was issued. Federal loan guarantee programs generally require borrowers to pay an up-front premium to the government to obtain a federal loan guarantee. That premium (for instance, $5 on a $100 loan) formerly was recorded as income in the budget in the fiscal year the guarantee commitment was made, while the federal disbursements to cover the guarantee if the borrower later defaulted were recorded as spending in future years (if and when a default occurred). Thus, even if the chance of default was high, the loan guarantee looked like a savings for the government, rather than a cost, in the year the guarantee was issued.
Credit Reform Act Records Full Costs of Loans When They Are Made
To make the budgetary effects of loans and loan guarantees comparable with each other — and with other federal spending programs — the Credit Reform Act of 1990 established rules for recording the full lifetime cost of loans and loan guarantees in the year that they are made. Essentially, the cost recorded for making a direct loan is the cash disbursement of the loan, minus the present value of the estimated repayments of interest and principal that will be received over the life of the loan. This estimate takes into account the terms of the loan (including the interest rate and repayment schedule), as well as the risk that the borrower will default on the loan before it is paid off. If the interest rate the borrower is charged is low or the borrower is likely to default, the cost of the loan to the government will be higher than the cost for a loan that charges a higher interest rate to a more credit-worthy borrower.
To take account of the time value of money, the interest and principal payments received over the course of the loan are discounted at the Treasury’s cost of borrowing. The time value of money reflects the fact that $100 today is worth more than $100 ten years from now. This can easily be illustrated by the fact that if you receive $100 this year, you could invest that $100 in ten-year U.S. Treasury notes. If the interest rate is 3.2 percent and you re-invest your interest earnings in Treasury notes, you will end up with $137 after ten years. So $100 now is clearly worth more than $100 in ten years.
Under the Credit Reform Act and credit budget rules, a similar approach also is taken for loan guarantees. The budget records the up-front cost of the loan guarantee as the difference between (1) any premium that the private borrower whose loan is backed by the federal guarantee pays to the government when the loan-guarantee commitment is made; and (2) the present value of the government’s estimated cost of covering defaults in the future minus any proceeds from selling any assets the government receives when a default occurs.
The key here is that the cost recorded in the budget reflects up front the estimated cash flows related to the loan or loan guarantee over the course of the loan. The only difference from the way in which the cost of other federal programs is shown in the budget is that the future-year cash flows associated with a direct loan or loan guarantee, as adjusted to take account of the time value of money, are booked in the year that the loan is made.
It should be emphasized that the estimated costs of loans and loan guarantees do take account of so-called default risk — the likelihood that some direct loans will not be paid back in full or that a borrower will default on a loan that the federal government has guaranteed. The estimated costs of credit programs under these procedures represent the best estimates of OMB, CBO, or executive branch agency staff of the cash flows likely to occur as a result of a loan or loan guarantee, after taking into account the terms of the loan and the probability that some borrowers will default.
Proposal Would Add a Further Amount to Reflect Private-Sector Risk Aversion
Even as the Credit Reform Act was being debated, some argued that calculating the cost of credit programs by this method understated the “true” cost of credit programs in a broader societal sense, because the method is based on the cost of these programs to the federal government rather than what similar loans or loan guarantees would cost in the private market. The government’s cost of making a loan is less than that of a private lender because it can borrow more inexpensively.
Since 1990, this argument has been refined, particularly in work by Deborah Lucas and Marvin Phaup. Lucas and Phaup argue there is an additional “cost” of credit programs that is not reflected in estimates of the cash flows in and out of the Treasury resulting from loans or loan guarantees. They argue that the loan costs would be higher if the private sector made the loans, due to the variability of the cash flows associated with loans and the fact that private individuals are risk averse — and that the federal budget should show what the loans and loan guarantees would cost if made in the private sector, rather than what it costs the government to make them.
The variability of the cash flows related to loans and loan guarantees stems from the fact that it is impossible to know with certainty exactly how much will be repaid on a given loan (or class of loans), because it is impossible to predict exactly how many borrowers will default and what collateral the government might acquire after a default. As a result, the actual collections flowing from any direct loan or class of direct loans and the actual guarantee payments required to indemnify a lender in the case of defaults on federally guaranteed loans may be higher or lower than originally estimated.
That this variability exists does not mean that the original estimates of the cash flows in and out of the Treasury due to a credit program were faulty and didn’t accurately reflect the risk of default. The variability simply reflects the inherent uncertainty of the cash flows. To understand this, consider an estimate of the number of times that heads will come up if a coin is flipped 100 times. We know the best estimate is 50. But if this exercise were repeated 100 times, the result would rarely be exactly 50 heads out of any 100 flips. The average — or expected value — over those 100 sets of 100 flips would be 50 heads, but most of the time there would be more or fewer than 50 heads.
Lucas and Phaup do not contend that the current estimates of the cost of credit programs misrepresent the cash flows related to loans and guarantees; they do not claim that CBO and OMB underestimate the true expected value of the cash flows. Their argument is different: that regardless of whether the estimates of the cash flows are the best ones possible — indeed, even if they perfectly represent the expected value of the cash flows — the method of calculating the cost of credit programs under the Credit Reform Act does not reflect the full “cost” for a different reason.
Lucas and Phaup base their argument on the variability of the actual cash flows and how individuals respond to risk in financial arrangements. They note that research has found that private individuals are risk averse, which is shown by the response of people who are offered the opportunity to place a bet on the results of 100 flips of a coin where they will receive $1 for each time that heads comes up. We know the expected number of heads in 100 coin tosses is 50, so a person who is not risk averse — and feels as much benefit from winning a dollar as pain from losing a dollar — should be willing to bet $50 on 100 coin tosses. Yet experience shows that people generally are willing to bet only if the price of the bet is less than the expected $50 value. This suggests that the benefit perceived to come from winning more than $50 in return for the $50 bet is less than the perceived pain from losing money on the bet (i.e., getting back less than $50). This asymmetric response reflects people’s unwillingness to accept a financial arrangement with variable financial outcomes at a price that only represents the expected value (or best estimate) of the outcome. This unwillingness is termed “risk aversion.”
Because individuals are risk averse, Lucas and Phaup argue, the government should be risk averse as well, on their behalf. As a result, Lucas and Phaup contend, the cost of credit programs should be shown in the federal budget as exceeding the best estimate of their actual cost to the Treasury (that is, as exceeding the best estimate of the cash flows that will result from the loans and loan guarantees). As they put it, “[I]ncluding a risk premium in subsidy cost produces a cost estimate that, on average, exceeds outlays for realized losses.” Because the government should be risk averse, they believe, it should be considered as losing more if collections are lower than estimated than it will gain if collections are higher than estimated. They argue that this risk aversion should be converted into a dollar figure and added to the cost of credit programs shown in the federal budget, as well as to the cost of legislation related to credit programs.
Lucas and Phaup would have the government estimate this extra “cost” that they believe should be assigned to direct loans and loan guarantee program by estimating what private markets would charge to issue or guarantee the same set of loans. They would estimate, for example, how much the private sector would pay to acquire the government’s portfolio of direct student loans. Presumably, risk-averse private investors would value the portfolio at a lesser amount than the government is expected to collect in loan repayments (after accounting for expected defaults and for the time-value of money).
To make this budget adjustment, H.R. 3581 defines two separate costs: (a) the government’s actual cash cost in operating credit programs, as calculated under the existing Credit Reform Act rules; and (b) the additional amount associated with risk aversion on the part of private investors (the notion that private individuals generally value an unexpected gain of $100 less than the cost of a $100 unexpected loss). The bill would require that these two amounts be added together and the sum be treated as the cost of a credit program in the federal budget, thereby raising the budgetary cost of these programs and legislation related to them.
Why the Proposal Is Flawed
This legislation suffers from several serious flaws.
Government May Be Less Risk Averse than Individuals
The Treasury generally does need not to be risk averse just because individuals are. And where the government is or ought to be risk averse, it need not be as risk averse as private individuals.
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. They will need their savings when they retire, when their children are in college, or when they encounter a sudden severe illness or disaster, and cannot simply “ride out” a down financial market during those periods by borrowing instead of cashing in their assets. Put simply, individuals may be forced to “sell low” if they need cash when times are bad.
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. (Consider the current low interest rates the Treasury is paying, which are near or below zero in real terms.) The government is thereby able to spread risk across decades or even generations, while individuals generally cannot.
Risk Aversion Is Not a Budgetary Cost
A more fundamental problem with adding a risk-aversion adjustment to the cost shown in the budget for loan and loan guarantee programs is that risk aversion is not, in fact, a budgetary cost. The risk-aversion adjustment that Lucas and Phaup propose and H.R. 3581 would require would reflect amounts that the Treasury would never actually pay to anyone; the amounts would not be dollars that the government spends. The obvious question then is: why should the budget record risk-aversion as a cost when the Treasury never pays that cost?
Answering this question requires thinking about what the budget is supposed to do. For over 200 years, the answer has been that 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 (and to the extent that deficits have exceeded surpluses, to cover the difference by borrowing and to record the amount of that borrowing as debt). To meet this purpose, the budget must measure accurately the amount spent on programs and the amount collected in taxes and fees, and the resulting deficits and debt — what budget analysts call the nation’s fiscal position.
Adding a risk-aversion adjustment to the spending side of the budget would add an extra “cost” that the government does not actually incur — and that doesn’t need to be covered by additional taxes or borrowing. It would consequently cause the budget to mismeasure deficits and debt and no longer serve the basic purpose of accurately presenting the nation’s fiscal position. With respect to nation’s fiscal position, a risk-aversion adjustment is a phantom cost.
Some policymakers, upon hearing about this legislation, may assume the legislation’s existence signifies that current procedures understate the costs the government incurs in operating credit programs and thus understate the programs’ effect on deficits and debt. But that is not the case, as Lucas and Phaup acknowledge. The current system produces the best estimates that CBO and OMB can provide of the actual outflows from, and inflows to, the Treasury as a result of the credit programs and their impact on deficits and debt. H.R. 3581 would add an additionalcost that the government is not expected actually to incur and base this extra amount on the additional cost that private lenders would charge if federal loans and loan guarantee programs were privatized.
Proposal Does Not Treat All Programs the Same
Another problem with the proposal is that it would result in inconsistent and discriminatory budgetary treatment of different categories of federal programs. To help Congress and the nation allocate public resources among competing priorities, the budget should record the costs of all government programs in the same way. It is essential that $100 in costs for one program mean the same thing as $100 in costs for another program, so that policymakers can know how much cost a policy will impose on the Treasury as they decide how to allocate resources. When allocating public funds, the budget must reflect costs comparably across all programs.
H.R. 3528 would violate this principle. It would make credit programs appear more expensive to the Treasury than they truly are without making similar adjustment for other programs whose actual costs also are uncertain. Much of the budget involves programs whose costs are only known for certain after the fact — that is, programs for which the best, unbiased estimates of expected costs nevertheless entail uncertainty, and for which actual costs will almost certainly turn out to be either lower or higher than the original estimates. Social Security and Medicare are two examples. Even some programs for which fixed rather than variable dollars are appropriated, such as weapons procurement, involve uncertainty because it is never known whether the items will end up costing more or less than budgeted, and Congress almost always feels it has to cover any additional costs. If a risk-aversion adjustment were added to credit programs, it should be added to all such other costs as well. Not doing so would disadvantage credit programs relative to other forms of government assistance or investment and distort the budget as a tool for allocating public resources.
Phantom Costs Require Phantom Offsets
Since the risk-aversion adjustment that H.R. 3581 would mandate would not reflect the amount the Treasury actually spends, including these phantom costs in the budget would cause the budget to record a total amount of federal spending that exceeds what the Treasury pays out. It also would mean that the deficit figures shown in the budget would be inaccurate, as they would exceed the true difference between actual expenditures and actual revenues — and the figure shown as the amount of debt held by the public also would be inaccurate, as it would be higher than the amount the Treasury actually has borrowed. To avoid these bizarre results, advocates of adding a risk-aversion adjustment tacitly or explicitly advocate accompanying that adjustment with a phantom offset.
Three types of phantom offsets have been suggested by proponents of this change. In proposing these phantom offsets, the proponents explicitly or implicitly acknowledge that the government will not actually incur the additional “cost” they would require to be shown for credit programs, which is why they are compelled to propose phantom offsets to prevent the deficit and debt figures from being out of whack. The obviously unsatisfactory nature of these phantom offsets (which are described below) underscores the point that the budget should measure actual costs and receipts and should not include either phantom costs or phantom offsets.
Lucas and Phaup propose recording a phantom tax receipt equal in size to the phantom additional cost they propose for the credit programs. In other words, the budget would show both more spending than the Treasury actually spends and more tax revenues than it actually collects, in order to keep annual deficits and total debt from being inaccurate. (Note that under this approach, an increase in a credit program would be shown as increasing revenues as well, and hence would run afoul of “no tax” pledges and be unconstitutional under versions of the Balanced Budget Amendment that bar increases in taxes.)
OMB’s recent experience with a mandate to display phantom costs demonstrates how hard it is to make sense of the results. Specifically, a provision of the 2008 Troubled Asset Relief Program (TARP) required that OMB record as a cost a risk-aversion adjustment, on top of TARP’s expected effects on government cash flows, but failed to specify an offset. OMB handled this in two ways. First, to avoid overstating the deficit, it created a phantom offset — lower interest payments on the debt, spread over time. In other words, over the lifetime of the portfolio of assets that Treasury might acquire under TARP, OMB offset the phantom cost by including a phantom downward adjustment in net interest payments (that is, it showed a figure for interest costs lower than the true amount of interest that OMB expected Treasury to pay). This produced the inexplicable result that an increase in up-front spending (for TARP) was shown as reducing interest costs; but at least the deficit and debt totals would be correct over time. Second, OMB proceeded to unwind this phantom scorekeeping in each subsequent year by re-estimating downward each year both the risk-aversion adjustment and the offsetting interest-payment adjustment.
H.R. 3581 adopts yet another phantom offset approach. It requires that the phantom cost not be offset by phantom revenue increases or phantom interest reductions, and it consequently leaves the total amount of federal spending — and the deficit figures — at inflated levels. But it creates a phantom downward “means of financing” offset that, over time, would prevent the debt from being too high even though the annual deficits would consistently be overstated. One result of this approach is that the sum of deficits over time would diverge more than it already does from the amount of debt held by the public.
Is There a Place for a Risk-Aversion Estimate?
Estimates of the extent (if any) to which government credit activities impose a risk-aversion cost on taxpayers should not play a part in budget accounting, because they do not represent an actual government cost and their inclusion in the budget would mismeasure the government’s fiscal position. Nevertheless, the concept that governmental transactions can impose risk on the public is not without merit. To the extent that the government does not spread such risk across generations (or ameliorate it by spreading it to well-off people who are less risk averse), 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.
Cost-benefit analysis, however, is not budgeting. A cost-benefit analysis serves a different purpose — to provide information on whether a program or project is worthwhile. To illustrate the difference between budgetary costs and cost-benefit analysis, consider two bridges, each of which would cost $50 million to construct. A bridge from nowhere to nowhere is a waste of money, while a bridge connecting two bustling sister cites might have substantial economic and social benefits. The budget should reflect $50 million in cost for each bridge — not more and not less — but a cost-benefit analysis that helps inform policymakers should take into account all of the pros and cons of the two bridges. In this context, risk aversion on behalf of the taxpayer, to the extent that it may exist, is a legitimate cost to include in the cost side of a cost-benefit analysis.
Note that under the approach H.R.3581 takes, however, other important aspects of cost-benefit analysis would not be reflected as phantom budget costs — not the social costs and benefits of regulation, for example, nor the large risk-mitigation benefits of social insurance programs such as Social Security and Medicare. This extends our earlier point that most government spending programs have uncertain rather than fixed costs; they also have uncertain rather than fixed benefits.
This discussion raises one final point about the concept of a risk-aversion budget adjustment. H.R. 3581 looks only at the cost that a credit program might impose on risk-averse taxpayers, while failing to consider the benefits to risk-averse borrowers such as students, farmers, or homebuyers. If the subsidy cost under a loan program turns out to be higher than the original cost estimate, taxpayers will eventually have to bear higher costs — but borrowers will have received more help. Put differently, being able to borrow from the government creates a benefit of an uncertain amount for the borrower; to the extent the subsidy ultimately proves larger than expected, it may impose a social cost on risk-averse taxpayers, but it also confers a social benefit on risk-averse students, farmers, homebuyers, or other borrowers. The proposed legislation would recognize only the costs, not the benefits. Our view — that risk aversion can be one of a number of appropriate aspects of cost-benefit analysis, but not of budget accounting — would still demand that all risk-aversion aspects of government programs be taken into account in a fair cost-benefit analysis.
 H.R. 3581 is very similar to section 4 of S. 1651, introduced in October 2011 by Sen, Jeff Session (R-AL) and Sen. Olympia Snowe (R-ME).
 Aside from credit programs, there are only a few instances — such as the recording of some Treasury interest costs as they accrue rather when they are later paid — in which the budget records spending on other than a cash basis.
 Before credit reform was enacted in 1990, all the various cash flows of a credit program were shown in the year that they occurred, and Treasury debt increased or decreased (as would interest payments) as cash left the government or flowed back to the government. When the loan finally matured, the sum total of all the cash transactions equaled the amount by which the debt held by the public had increased as a result of the loan’s issuance. Credit reform aggregated these credit transactions into a single subsidy cost shown up front. After a loan matures, the sum of that subsidy and the interest that the Treasury pays on the money it borrowed to finance that subsidy is exactly the same as the amount that would have been recorded in the budget before credit reform was enacted and represents the amount by which the debt held by the public has increased. Thus, credit reform did not change the lifetime budgetary cost of credit programs; it simply shifted the timing of when that cost is recognized. The total cost is now shown up front so Congress can better see what cost it is imposing on the Treasury at the time that it votes to impose that cost.
 Estimates are based on calculations for a class of similar loans or guarantees, not for individual loans or guarantees.
 Deborah Lucas and Marvin Phaup, “Reforming Credit Reform,” Public Budgeting & Finance, Winter 2008, pp. 90-110.
 The federal administrative costs of direct loans and loan guarantees are recorded in the budget on a cash basis rather than on a present-value basis, because administrative funding is determined by current and future appropriation bills. Although Lucas and Phaup point to advantages in treating the administrative costs of credit programs on a present-value basis, neither Lucas and Phaup nor H.R. 3581 offers a way to do so.
 Lucas and Phaup, page 92.
 In the same vein, if the government tried to purchase reinsurance from the private sector to cover the defaults associated with a government portfolio of loan guarantees, a risk-averse private investor would charge more to reinsure that portfolio than a perfect estimate of what the government will actually have to pay on the defaults, after accounting for the time-value of money.
 The authors are indebted to an unpublished article by David Kamin, “Risky Returns: Accounting for Risk in the Federal Budget,” October 2011, for its presentation of the arguments against including nonbudgetary costs of credit programs in the federal budget.
 Similar issues also arise on the revenue side of the budget, since tax collections are also uncertain.