Skip Navigation

Nearly a decade has passed since the Federal Credit Reform Act required a new budgetary treatment of federal programs that offer direct loans or loan guarantees. Those programs must initially estimate the cost to the government of subsidizing that credit and periodically issue reestimates as financial conditions change. At the request of the Senate Committee on the Budget, this Congressional Budget Office (CBO) paper compiles and analyzes the cumulative reestimates made for credit programs since 1993. It examines trends in the size and direction of reestimates, looks for possible bias, and discusses ways to prepare and report reestimates more accurately.

David Torregrosa of CBO's Microeconomic and Financial Studies Division wrote the paper under the supervision of Marvin Phaup, Arlene Holen, and Roger Hitchner. Mark Hadley of CBO reviewed the analysis at many stages; his assistance is greatly appreciated. Barry Anderson, Coleman Bazelon, Charles Capone, Sunita D'Monte, Pete Fontaine, Kathy Gramp, Deborah Kalcevic, Susanne Mehlman, Dave Moore, Nathan Musick, and Tim VandenBerg of CBO provided helpful comments, as did Bob Kilpatrick and Courtney Timberlake of the Office of Management and Budget, and Kim Burke of Ernst & Young. In addition, various people helped CBO collect and reconcile the data about reestimates, including Maria Leland and Wanda Lopez of the Department of Agriculture; Karen Weiner, formerly of the Department of Education; Maude Williams and Brenda Davis of the Federal Housing Administration; Winifred Chang of the Office of Management and Budget; Fred Rubin and Joyce Northwood of the Small Business Administration; and Rose Roberts of the Department of Veterans Affairs. Avi Ebenstein, formerly of CBO, provided research assistance.

Chris Spoor edited the paper, and Christine Bogusz proofread it. Rae Roy prepared the manuscript for publication with assistance from Linda Lewis-Harris and Angela Z. McCollough. Annette Kalicki prepared the electronic versions for CBO's World Wide Web site (www.cbo.gov).
 

Dan L. Crippen
Director
September 2000
 
 

SUMMARY

INTRODUCTION

INACCURACIES AND POTENTIAL BIAS IN ESTIMATING CREDIT SUBSIDIES

• The Mechanics of Subsidy Reestimates
• Incomplete Reporting of Reestimates
• Mixed Incentives for Accuracy

THE SIZE AND DIRECTION OF REESTIMATES SINCE 1993

• Discretionary Versus Mandatory Programs
• Reestimates by Cohort
• Comparing Reestimates with Actual Losses
• Overall Trends in Reestimates
• Conclusions of CBO's Analysis
• Limitations of the Analysis

WAYS TO IMPROVE THE ACCURACY OF SUBSIDY ESTIMATES

• Improving the Presentation of Reestimates in Budget Documents
• Improving the Quality of Management Information
• Increasing the Use of Private-Sector Information

APPENDIXES

A - OMB's New Estimating Procedures
B - Correcting and Confirming Reported Credit Subsidy Reestimates
C - Summary of Credit Amounts and Subsidies by Fiscal Year
D - Cumulative Reestimates for Direct Loans and Loan Guarantees
E - Reestimates for Cohorts That Are Close to Maturity
 

TABLES
S-1.	Cumulative Subsidy Reestimates, Fiscal Years 1993-1999
1.	Cumulative Subsidy Reestimates, by Department or Agency, Fiscal Years 1993-1999
2.	Cumulative Subsidy Reestimates for the Department of Agriculture, by Agency, Fiscal Years 1993-1999
3.	Cumulative Subsidy Reestimates for the Farm Service Agency's Agricultural Credit Insurance Fund Programs, Fiscal Years 1993-1999
4.	Annual Subsidy Reestimates for FHA's Mutual Mortgage Insurance Program, by Cohort, Fiscal Years 1993-1999
5.	Annual Subsidy Reestimates for SBA's Section 7(A) General Business Loans, by Cohort, Fiscal Years 1995-1998
6.	Cumulative Subsidy Reestimates for the Department of Agriculture's Unsubsidized Guaranteed Farm Operating Loans, by Cohort
7.	Annual Subsidy Reestimates for the Department of Agriculture's Unsubsidized Guaranteed Farm Operating Loans, by Cohort, Fiscal Years 1992-1999
8.	Cumulative Subsidy Reestimates for All Credit Programs Except FCC Auctions, by Cohort
9.	Cumulative Subsidy Reestimates for All Discretionary Credit Programs, by Cohort
10.	Anticipated Reestimates for Fiscal Year 2000, by Department or Agency
B-1.	Cumulative Reestimates for FHA's General and Special Risk Insurance Program as Reported in the Federal Credit Supplement, by Cohort
B-2.	Annual Reestimates for FHA's General and Special Risk Insurance Program as Reported in the Budget Appendix
B-3.	Corrected Annual Reestimates for FHA's General and Special Risk Insurance Program, by Cohort, Fiscal Years 1993-1998
B-4.	Comparison of Cumulative Reestimates of Credit Subsidies for Loans Disbursed Between Fiscal Years 1992 and 1998
C-1.	Summary of Federal Direct Loans and Loan Guarantees by Fiscal Year, 1992-1999
D-1.	Cumulative Subsidy Reestimates for Direct Loans and Loan Guarantees, by Department or Agency, Fiscal Years 1993-1999
E-1.	Cumulative Subsidy Reestimates for the Department of Agriculture's Direct Farm Operating Loans, by Cohort
E-2.	Annual Subsidy Reestimates for the Department of Agriculture's Direct Farm Operating Loans, by Cohort, Fiscal Years 1992-1999
E-3.	Cumulative Subsidy Reestimates for the Department of Agriculture's Subsidized Guaranteed Farm Operating Loans, by Cohort
E-4.	Annual Subsidy Reestimates for the Department of Agriculture's Subsidized Guaranteed Farm Operating Loans, by Cohort, Fiscal Years 1992-1999
BOXES
1.	Are Reestimates More Accurate Than Initial Estimates?
2.	The FCC's Credit Program for Buyers of Spectrum Licenses

 

SUMMARY

The Federal Credit Reform Act of 1990 requires that when a federal agency makes a loan directly or guarantees a loan made by another party, the agency estimates how much of that credit will end up being subsidized by the government. That estimate is recorded in the federal budget when the credit is offered. But projecting losses and costs from federal credit assistance over the uncertain, but often lengthy, life of such loans is difficult, and errors are inevitable. Thus, since fiscal year 1993, agencies have updated their initial subsidy estimates each year through reestimates, which reflect changes in financial conditions and in the performance of individual credit programs to date.

With seven complete years of subsidy reestimates to draw on, the Congressional Budget Office (CBO) has compiled and assessed the record of federal credit reestimates. Some tentative but important findings emerge:

• The reestimates reported in the President's budget between 1993 and 1999 are in such disorder that they cannot be used for analysis. The data that CBO used in this report were corrected with extensive help from the credit agencies and the Office of Management and Budget (OMB).



• Although various incentives may exist for agencies to underestimate credit subsidies, the corrected reestimates do not reveal any pattern of bias in initial subsidy estimates.



• A few modest changes in current practice could improve the accuracy of subsidy estimates and reestimates.

Deficiencies in Reported Reestimates

Agencies' reestimates account for various factors--such as interest rates, rates of default and other technical assumptions about the performance of loans, and the volume of lending--that may have changed since a credit subsidy was first estimated. Those reestimates are reported in various volumes of the President's budget. However, inconsistencies in the numbers that appear in different volumes and incomplete reporting of default reestimates make monitoring and compiling reestimates difficult.

The Federal Credit Supplement to the President's budget is the most detailed source of credit reestimates and the only publication that distinguishes reestimates for interest rates from default reestimates by the year the credit was issued. But the reestimates reported there cannot easily be reconciled with the ones in the Appendix to the budget. In fact, fewer than one-third of federal credit programs' reestimates appear to have been correctly reported in the Federal Credit Supplement for fiscal year 2000. Inconsistencies arise in part because estimating, accounting, and reporting are distinct functions that may be handled by different people in an agency. Mistakes can also occur at any stage in the process of producing, recording, and reporting reestimates. For most programs, errors in the reported figures can be corrected, but only with additional, unpublished data from the agencies and OMB.

Errors in reporting and accounting are not limited to the Federal Credit Supplement. In one case, the Appendix incorrectly reported a loan modification (a change in the estimated cash flow of a loan program that results from new legislation or unanticipated agency actions) as a reestimate. Moreover, an agency mistakenly reported (but later corrected) a $300 million reestimate in the Treasury Department's Monthly Treasury Statement, which tracks the government's cash flows.

Conflicting Incentives to Estimate Credit Subsidies Accurately

Because reestimates are so difficult to track, agencies may have less incentive to make their initial estimates accurate. In addition, credit reform may give advocates of federal credit programs a reason to understate the expected subsidy of a program; under current rules, initial subsidy estimates are more important costs politically than reestimates. For discretionary credit programs, an initial subsidy estimate requires a Congressional appropriation, whereas later reestimates that raise the cost of a program draw on permanent and indefinite spending authority. Thus, proponents of discretionary programs at an agency might underestimate initial subsidies in order to lend more and then make up the difference with a "costless" reestimate. For mandatory credit programs, which do not require appropriations, low initial estimates could reduce a program's visibility and attractiveness as a target for budgetary savings.

Other considerations, however, may offset any inclination toward bias in estimating credit subsidies. Agencies have incentives to maintain their reputation for integrity. And if their estimates do show bias, OMB's oversight may detect and eliminate it. Further, any gains to an agency from underestimating initial subsidies might be lessened because large upward reestimates can attract increased oversight from the Congress.

Trends in Reestimates Since 1993

By the end of fiscal year 1999, reestimates had added more than $8.7 billion to agencies' initial estimates of credit subsidies (see Summary Table 1). Most of the upward revision, however, was for a single program: loans offered to successful bidders in the Federal Communications Commission's auctions of licenses to use the electromagnetic spectrum. The total upward reestimate for all other credit programs was $3.0 billion. That reestimate equals about 8 percent of the initial credit subsidies for all federal loans and loan guarantees extended since 1991.
 

Reestimates for defaults were on the whole downward for discretionary programs (a total of -$1.2 billion) and upward for mandatory programs (almost $2.2 billion, excluding the loans for spectrum licenses). The net upward reestimate for mandatory programs mainly reflected large increases for the Federal Housing Administration's Mutual Mortgage Insurance program.

That pattern of downward default reestimates for discretionary programs and upward ones for mandatory programs is inconsistent with the notion that agencies have greater incentives to understate initial subsidy estimates for discretionary programs than for mandatory ones. However, those results are for reestimates to date, without the benefit of final losses, and could change in the future. For example, economic conditions over the remaining life of the outstanding loans may not be as favorable as during the past nine years. The entire period of credit reform has been marked by continued economic expansion, so low initial subsidy estimates may be hidden by the strong economy.

Ways to Improve Accuracy

Federal agencies could reduce their errors in preparing, reporting, and accounting for estimates and reestimates in several ways. One approach might be to increase the visibility and consistency of estimates and reestimates in the budget through the following steps:

• The Federal Credit Supplement could be expanded to show both initial subsidy estimates and total reestimates by program, agency, and department, with separate entries for interest reestimates and technical or default reestimates. Reestimates for loan volume could also be reported for mandatory programs.



• OMB could adopt an editing check to compare the cumulative credit subsidy reestimates reported for each fiscal year's loans in the Federal Credit Supplement with the sum of the annual reestimates for those loans reported in the Appendix.



• The Analytical Perspectives volume of the budget could report all reestimates to date by program, including interest on the reestimates.

Agencies could also improve their management information systems to enhance their ability to project federal credit costs. In addition, increasing their reliance on the private sector could allow agencies to estimate and manage credit costs more accurately. Today, federal agencies cannot match the performance of private financial institutions in managing loan portfolios. But they could sell loans to the private sector, which would provide a market test of their subsidy estimates and eliminate the need for reestimates. Another approach would be to hire bond-rating agencies to assess and monitor the quality of credit and estimate the probabilities of default for participants in major federal credit programs.
 

INTRODUCTION

In adopting the Federal Credit Reform Act of 1990, the Congress sought to improve budgeting for federal credit programs by making the budgetary costs of direct loans and loan guarantees comparable with those of other programs. Federal credit programs generally cost the federal government money because they require it to pay out more than it gets in return. The aim of credit reform is to report the government's entire expected loss from direct and guaranteed loans when the loans are made.

Credit reform requires officials to identify what percentage of a credit transaction will be subsidized by the government and to record the cost of that subsidy separately in the budget from the cash flows of the transaction.⁽¹⁾ For direct loans, the subsidy cost is the present value (in today's dollars) of the government's expected future losses from net interest costs, late payments, and nonpayments. For loan guarantees, the subsidy is the present value of the government's expected payments in excess of the fees collected for the guarantees.⁽²⁾ An important provision of credit reform is that for discretionary credit programs, the Congress must provide budget authority for credit subsidies before the government obligates loan funds or commits to guarantee a loan. For new mandatory credit programs and for modifications to existing ones, any added costs must be offset by decreases in other mandatory spending or increases in revenues, as specified in the pay-as-you-go provisions of the Budget Enforcement Act.

From the beginning of credit reform, many budget analysts expressed concern about the ability and willingness of agencies to estimate subsidies accurately. To ensure that the estimates of subsidy costs in the budget are consistent with subsequent cash flows, and to inform policymakers about the performance of federal loans, the Credit Reform Act requires agencies to update those estimates periodically for outstanding direct loans and loan guarantees. Those revisions, or reestimates, appear in the President's annual budget submission to the Congress and other budget documents, such as the Administration's Mid-Session Review.

Now that credit reform has been in place for nearly a decade, analysts can begin assessing that policy, especially the early concerns about the accuracy of subsidy estimates. The General Accounting Office has found that many agencies have had trouble producing reliable subsidy estimates.⁽³⁾ This analysis focuses on reestimates of subsidy costs--particularly on inaccuracies in the reported figures and incentives that agencies face to underestimate subsidy costs, the patterns visible in reestimates to date, and some ways in which to improve the accuracy of estimates.
 

INACCURACIES AND POTENTIAL BIAS IN ESTIMATING CREDIT SUBSIDIES

The Congress needs information about the accuracy of subsidy estimates for credit programs to control government costs. But incomplete and inaccurate reporting of reestimates in the budget makes it difficult to monitor costs and hold agencies accountable for their estimating errors. Moreover, because of the way in which reestimates are financed under credit reform, agencies may have less incentive to make their initial estimates accurate.

The Mechanics of Subsidy Reestimates

For each federal credit program, subsidy reestimates are reported in the budget as increases or decreases in outlays at the time of the reestimate. Upward reestimates raise the cost of credit subsidies and thus lower the budget surplus (or increase the deficit). Downward reestimates reduce credit subsidies and thus increase the budget surplus (or lower the deficit) when the reestimate is made.⁽⁴⁾ In general, agencies must calculate reestimates annually for each credit cohort--all of the loans or guarantees that a program obligates in a given fiscal year. Ideally, reestimates are based on new information, some of which comes from repayment experience as the loans age.

All credit reestimates have at least two components: revisions for errors in interest rates and reestimates for changes in default rates or other technical assumptions. Interest rate reestimates correct the projected subsidy for the difference between the discount rate (the interest rate used to calculate the present value of a future stream of receipts and spending) that was assumed in the initial estimate and the actual rate prevailing when the credit is disbursed. In most cases, interest reestimates are first made when 90 percent of a loan cohort has been disbursed. Final interest reestimates are made when disbursement is complete, which can be several years later. The Office of Management and Budget (OMB) requires agencies to use its forecast of interest rates in formulating initial subsidy estimates. Thus, agencies have no control over the initial interest rate assumption or the errors that may result from it.⁽⁵⁾

Technical or default reestimates correct for borrowers' actual performance in repaying loans. Those revisions adjust the subsidy estimate for unanticipated changes in delinquencies, default rates, recoveries, prepayments, and receipts from fees.⁽⁶⁾ Some credit programs require additional adjustments for changes in the interest rates that borrowers pay. (Student loans, for example, are made at variable interest rates, subject to a cap.)⁽⁷⁾ For most programs, technical or default reestimates are prepared at the end of each fiscal year; however, OMB may permit agencies to make them less frequently.

Unlike the assumptions about interest rates, agencies control their assumptions about default and other technical factors. Hence, any estimating bias on an agency's part is likely to show up in default reestimates. However, upward default reestimates could also reflect a genuinely unexpected deterioration in the credit performance of borrowers.

Incomplete Reporting of Reestimates

Inconsistencies in the reestimates reported in various parts of the President's budget and incomplete reporting of default reestimates make it difficult--if not impossible--to monitor reestimates. Moreover, no complete listing exists of annual technical or default reestimates for each cohort of each credit program. That missing information may be critical to assessing initial subsidy estimates and program performance because cumulative reestimates, even by cohort, cannot reveal the volatility of yearly reestimates.

A complete set of reestimates can be compiled from various volumes of the budget. Even so, the reported data from those sources differ in format and contain significant discrepancies. The Federal Credit Supplement to the President's budget is the most detailed source of credit subsidy reestimates: tables provide the cumulative reestimates for credit cohorts by agency, program, and bureau. For each cohort, the tables generally show the percentage subsidy rate originally estimated, the current reestimated subsidy rate, the distribution of that reestimate between interest rate and technical or default changes, and the total dollar amount of the reestimate. However, for some large programs--such as student loans and the Federal Housing Administration's (FHA's) loan guarantees--the Supplement does not separate interest rate and technical or default reestimates because the programs' agencies combine those into a single total. (For a more complete description of problems with the reported data, see Appendix B.)

The Appendix volume of the budget also reports reestimates by program, but those figures are not broken out by cohorts, and technical, default, and interest rate reestimates are consolidated into a single number.⁽⁸⁾ However, the Appendix does list programs' upward and downward reestimates separately.⁽⁹⁾ A third volume, Analytical Perspectives, reports only the net annual reestimate for each credit program.

Those differences in the format and structure of reported reestimates would not be a serious handicap if the data were consistent between sources. But for several programs, the Federal Credit Supplement reports significantly different reestimates than the Appendix does. Moreover, the reestimates in the Analytical Perspectives sometimes fail to agree with those in either the Supplement or the Appendix. Discrepancies arise in large part because of reporting errors. A frequent mistake in the Supplement is for reestimate rates and dollar amounts to be inconsistent: for example, upward rate reestimates should result in upward dollar reestimates, but that is often not the case for data reported in the Supplement. In addition, annual reestimates are occasionally reported as cumulative reestimates.⁽¹⁰⁾

Reporting errors are not limited to the Federal Credit Supplement. In the case of FHA's Mutual Mortgage Insurance program, neither the Supplement nor the Appendix gives the correct cumulative reestimate: both wrongly report $1.2 billion in savings for that program in 1996 as a downward reestimate rather than as a loan modification.⁽¹¹⁾ Further, the Small Business Administration mistakenly reported (but later corrected) a $300 million reestimate for the disaster direct loan program in the Treasury Department's Monthly Treasury Statement, which tallies the government's receipts and payments.

Mixed Incentives for Accuracy

Agencies face conflicting incentives in the process of estimating subsidies. The strength of those incentives differs for discretionary and mandatory credit programs--largely because of the ways in which the two types of programs are funded or created. On the whole, agencies' impetus to underestimate subsidy costs is considered strong for existing discretionary programs, weak for existing mandatory programs, and probably nonexistent for new credit programs.

The Federal Credit Reform Act provided for reestimates because legislators correctly anticipated that initial estimates would contain errors. In fact, it is impossible to precisely predict subsidy costs because it is impossible to precisely predict future economic growth, personal and other income, inflation, interest rates, and other economic factors that have major effects on credit costs.⁽¹²⁾ In addition, some federal credit programs, especially student loans, give borrowers a choice of repayment options, whose costs to the government vary; projecting which option borrowers will select is difficult. Consequently, the Credit Reform Act provides permanent and indefinite budget authority for reestimates. Although that provision is intended to avoid penalizing agencies for revising their initial subsidies, it may also dampen their incentives for accuracy.

Indeed, an agency might have reason to understate the expected cost of credit. Initial subsidy estimates for discretionary programs must compete with other spending programs for appropriations, and those estimates count against the statutory caps on discretionary spending. Reestimates, by contrast, need not compete with other spending priorities and do not count against the caps because of their permanent and indefinite spending authority. (Just as significant, if an agency overestimates a subsidy initially, later downward reestimates do not free up money for other uses.) That incentive to understate subsidy costs would affect only discretionary credit programs, because mandatory programs are not constrained by appropriations.

Agencies may also have a more fundamental incentive to underestimate subsidies for both mandatory and discretionary credit programs: to the extent that agencies are advocates for their programs, they may be inclined to understate costs to reduce a program's visibility and attractiveness as a target for budgetary savings. Since the Credit Reform Act increased the visibility of credit costs, the Congress has frequently taken steps to reduce subsidies for both mandatory and discretionary programs. For example, it lowered the cost of the student loan program by limiting the participation of schools with high default rates and reducing the interest rate subsidy paid to private lenders.⁽¹³⁾ The Congress also reduced the percentage of each Small Business Administration (SBA) loan that is guaranteed and raised the fees that SBA charges.⁽¹⁴⁾ If heightened cost-consciousness causes bias to creep into agencies' subsidy estimates for mandatory as well as discretionary programs, mandatory programs will not serve as a useful basis of comparison to identify bias in estimates for discretionary programs.

When agencies are in doubt--which is always the case with subsidy estimates--they may be inclined to err on the low side. If so, reestimates will tend to be upward. However, bias may not be immediately evident from the size and direction of reestimates (see Box 1).
 

Although the various factors described above suggest a reason for agencies to systematically underestimate credit subsidies, countervailing pressures exist. Agencies may wish to aim for stable estimates and to maintain a reputation for professional integrity. In addition, the expected gains from underestimating subsidies would be diminished by the prospect of detection by OMB, which generally tries to root out bias. Moreover, large reestimates--whether up or down--can attract Congressional oversight. For example, the House Banking Committtee has voiced concern about upward subsidy reestimates for FHA's single-family loan program.⁽¹⁵⁾ The Congress later passed legislation to increase the efficiency of FHA's disposal of property, and the committee directed FHA to improve its administrative oversight of loan originations. Also, upward reestimates for older cohorts may force agencies to raise the estimated subsidy cost of new cohorts. Stable cost estimates are less subject to criticism by either the Congress or a program's constituency.

In general, agencies have little to gain by providing low estimates for new mandatory credit programs or for legislative changes to existing ones, because agencies' estimates are not used in the Congressional budget process. Instead, the Congress usually relies on the Congressional Budget Office's (CBO's) estimates of the cost of legislation to create credit programs, such as the direct student loan program, or to modify existing programs. In addition, the Congress generally uses CBO's subsidy estimates for mandatory credit programs.⁽¹⁶⁾ In line with its mandate to provide objective, nonpartisan information, CBO tries to produce accurate and unbiased subsidy estimates.
 

THE SIZE AND DIRECTION OF REESTIMATES SINCE 1993

During the 1990s, reestimates added more than $8.7 billion to initial credit subsidy estimates (see Table 1). In other words, according to evidence to date, agencies as a whole initially underestimated the cost of federal credit. That underestimate--as measured by the subsequent reestimates--amounted to 24 percent of the initial $36 billion subsidy for credit issued over the period. (For a summary of total credit amounts and subsidies by fiscal year, see Appendix C.) However, most of the upward reestimate was for a single program: loans offered to successful bidders in the Federal Communications Commission's (FCC's) auctions of licenses to use the radio spectrum. The total upward reestimate for all other credit programs was just $3.0 billion, or about 8 percent of the initial subsidy estimate.
 

Default reestimates lowered initial subsidy estimates for discretionary credit programs by $1.2 billion. That result is inconsistent with the view that, overall, agencies are underestimating subsidies for existing programs that are constrained by the availability of appropriations. Mandatory programs (excluding the FCC's auctions) reported upward reestimates of $2.2 billion for default losses. Most of those upward reestimates were for FHA's Mutual Mortgage Insurance program, which is the largest federal credit program, with outstanding loan guarantees of $411 billion in 1999.⁽¹⁷⁾

The difference between reestimates for discretionary and mandatory programs is not the only split that might reveal systematic under- or overestimating. Reestimates for specific departments, agencies, programs, and cohorts might also disclose such bias, if it exists. Accordingly, this analysis examined reestimates by every available breakdown. However, it found no evidence of bias in any of them.

Discretionary Versus Mandatory Programs

Among discretionary credit programs, the Department of Agriculture (USDA) and FHA's General and Special Risk Insurance program are the major sources of upward technical or default reestimates, and the amounts are small. Within USDA, the size and sign of reestimates vary considerably among different agencies and different types of credit (see Table 2). Thus, even for departments whose overall data might be consistent with bias, program data do not suggest a pattern of bias. Specifically, USDA's direct loan programs (with one exception) reported downward technical or default reestimates, whereas the revisions for loan guarantee programs were all upward. Further, the default reestimates were not consistently upward even within a single agency. The Farm Service Agency's Agricultural Credit Insurance Fund, for example, reported upward cumulative default reestimates for four programs and downward ones for five programs (see Table 3).
 

Unlike the overall downward default reestimate for discretionary programs, default reestimates for mandatory programs as a whole raised costs by nearly $2.2 billion (excluding the $5.7 billion reestimate for the FCC's auctions). That distinction between types of credit programs, however, may be less sharp than it appears. The $3.2 billion upward technical/default reestimate for Mutual Mortgage Insurance more than accounts for the entire positive reestimate for mandatory programs.⁽¹⁸⁾ Other mandatory programs had upward or downward reestimates roughly comparable with those reported by discretionary programs. For example, the technical or default reestimate for the Department of Veterans Affairs (VA) program that guarantees loan sales raised that program's cost by more than $500 million, but the technical or default reestimates for the other two mandatory VA housing programs lowered costs by more than $1.5 billion.⁽¹⁹⁾

In examining mandatory credit programs, exclusion of the installment loan program created for the FCC's auctions of spectrum licenses may be appropriate because that program is no longer extending credit, and its original subsidy estimate did not affect the decision to conduct the auctions. That estimate was made after the fact and consequently had no impact on policy or spending. Moreover, the FCC no longer provides installment loans or other forms of credit to bidders. (For more details on that program, see Box 2.)
 

Besides default reestimates, another difference between discretionary and mandatory programs is the possibility of unexpected changes in the volume of credit. Loan volume is not a source of reestimates for discretionary programs because their total volume of lending is constrained by the size of the appropriation for subsidies. But mandatory programs saw a significant upward reestimate ($719 million) between 1993 and 1999 for unexpected credit volumes. That revision is noteworthy because volume reestimates were not anticipated by the Credit Reform Act. Consequently, there are no budgetary or financial reporting requirements for volume reestimates. CBO separated the volume reestimates from total reestimates for mandatory programs using unpublished agency data.

Although changes in program volume do not affect subsidy rates and are largely beyond an agency's control, they do affect total costs for VA's housing programs and the Department of Education's student loans. Volumes for both VA direct loans and loan guarantees were underestimated; as a result, costs rose by more than $330 million and were reported as reestimates. Higher-than-expected volume in the guaranteed student loan program increased subsidy costs by $418 million, whereas lower-than-expected volume in the direct student loan program reduced spending by $30 million. (For more details about how credit subsidy reestimates differed for direct loan and loan guarantee programs, see Appendix D.)

Reestimates by Cohort

The reestimates reported in budget documents consolidate figures for various cohorts in a way that could mask estimating bias.⁽²⁰⁾ In other words, subsidies for some but not all cohorts might be understated initially and then corrected with reestimates that would not seem excessive when spread over all cohorts. However, examining annual reestimates by cohort for the Mutual Mortgage Insurance program and SBA's Section 7(a) general business loan program fails to show any consistent pattern of upward reestimates.⁽²¹⁾

Federal Housing Administration. Mutual Mortgage Insurance, the biggest federal credit program, has had the largest upward revisions (apart from the spectrum auction program). Through 1999, those reestimates totaled almost $2.8 billion, or 0.5 percent of total loan guarantees and 23 percent of the original subsidy estimate (see Table 4). All cohorts except that of 1992 had upward total reestimates. Moreover, the annual reestimates for individual cohorts appear to follow a cyclical pattern: consecutive downward reestimates followed by a run of upward revisions.
 

The program's largest upward reestimate for any year was reported in the 1999 Mid-Session Review following poor performance of the program's portfolio in 1998. In particular, losses on adjustable-rate mortgages in California were much larger than expected, and housing prices nationwide rose less than anticipated. Thus, the $3.2 billion upward reestimate in mid-1999 appears to be justified by new information rather than to reflect systematic underestimates earlier. Nevertheless, that reestimate was quite large: about the same size as the program's total negative subsidy for 1999 ($3.1 billion). However, no reestimate for the program is anticipated in 2000.

Small Business Administration. SBA's Section 7(a) program, which is discretionary, guarantees about $10 billion in new business loans annually. Downward reestimates for that program have been getting successively larger and extending to more cohorts (see Table 5). Unlike the generally upward reestimates for cohorts of the Mutual Mortgage Insurance program, all SBA cohorts showed downward reestimates in the most recent available data. In all, reestimates for the program totaled -$965 million, or more than 1 percentage point of the program's original subsidy rate. SBA has already recorded a reestimate of about -$300 million for its various business loan programs in 2000.
 

Comparing Reestimates with Actual Losses

Another way to determine whether initial subsidy estimates are systematically too high or too low is to compare them with programs' final losses. Unfortunately, since credit reform took effect, no major federal credit program has had a loan cohort that has been completely retired; all have outstanding unpaid balances. Three smaller programs of USDA's Farm Service Agency--Direct Farm Operating Loans, Unsubsidized Guaranteed Farm Operating Loans, and Subsidized Guaranteed Farm Operating Loans--have statutory loan terms of seven years, so actual losses (and presumably reestimates) for their early cohorts may be approaching final values.⁽²²⁾ But refinancings and bankruptcies have extended effective maturities in those programs, so the early cohorts still have some time to go before reaching final maturity.⁽²³⁾

The Unsubsidized Guaranteed Farm Operating Loan program is the Farm Service Agency's largest credit program, with more than $8 billion in loan commitments made between 1992 and 1999 (see Table 6). It does not subsidize borrowers' interest payments; instead, borrowers pay the same rate of interest on their loans as the government pays on its debt. (That feature distinguishes the program from the Subsidized Guaranteed Farm Operating Loan program, in which the government pays some of the borrowers' interest costs.) The unsubsidized loan program has had a net upward technical/default reestimate totaling $71 million, or about 1 percent of loan principal. All of the early cohorts (1992 through 1995) show upward reestimates for default losses, which are equivalent to 1 percent to 2 percent of loans but 137 percent to 318 percent of the original subsidy cost.
 

Although those large upward reestimates for early cohorts suggest bias, reestimates have been much smaller for later cohorts. Moreover, default reestimates for the first three cohorts were revised downward in 1998 and 1999 (see Table 7). Thus, the overall pattern of reestimates does not provide strong support for the notion that initial subsidy estimates were systematically understated.
 

Analyzing reestimates for the other two short-term loan programs of the Farm Service Agency also reveals no evidence of bias. However, for both programs, early reestimates appear to move the estimated credit subsidy farther away from the final value rather than closer to it. (For more details on reestimates for those programs, see Appendix E.)

Overall Trends in Reestimates

Agencies can be expected to improve their initial estimates over time as they gain more experience with the estimating process and as information about the performance of loans and the characteristics of borrowers increases. More accurate estimates should mean smaller reestimates, divided more evenly between positive and negative values. However, reestimates for recent cohorts suggest that although agencies may be making smaller revisions to subsidies for discretionary programs, they are making more numerous and larger upward revisions to subsidies for mandatory programs than in the early years of credit reform.

Looking at all credit programs (except the FCC's auctions) combined, recent cohorts are more likely to have upward reestimates, which offset the downward reestimates for earlier cohorts. The first two cohorts since credit reform have large downward technical or default reestimates: -$857 million for the 1992 cohort and -$509 million for the 1993 group (see Table 8). The next five cohorts all have positive reestimates for defaults.
 

Looked at by cohort, discretionary credit programs gradually shift from consistently downward technical or default reestimates to a mix of upward and downward changes (see Table 9). The size of those reestimates is also declining, which suggests that the accuracy of initial estimates may be improving.⁽²⁴⁾ The large swings in unidentified ("unknown") reestimates, however, make any interpretation tentative.
 

Based on the Office of Management and Budget's projections, -$4.4 billion in new downward reestimates for all credit programs are expected in 2000 (see Table 10). The largest expected reestimates for mandatory programs are for student loans--a reestimate of -$2.4 billion for direct student loans and a revision of nearly $800 million for loan guarantees, mainly because of higher projected interest rates. The largest discretionary reestimate is for the Small Business Administration, which has already recorded reestimates of more than -$800 million in this fiscal year.⁽²⁵⁾ USDA remains one of the few departments expected to report upward reestimates.
 

Conclusions of CBO's Analysis

The pattern of reestimates under credit reform provides no visible evidence to support the idea that departments and agencies are abusing the permanent and indefinite spending authority provided for reestimates. Moreover, most reestimates have not substantially raised overall spending. (The multibillion-dollar reestimates for the FCC's auctions of spectrum licenses and FHA's mortgage insurance programs, which together total $9 billion over seven years, are the exceptions.)

Mandatory programs have seen net upward reestimates for defaults. But again, no strong indication of bias exists. The Mutual Mortgage Insurance program, which has a negative subsidy, accounts for the bulk of those upward reestimates. The revisions for mandatory programs cast doubt on claims that similar reestimates for discretionary programs are evidence of gimmickry. In both types of programs, upward reestimates could result from the same random shocks or analytic imprecision.

Three other findings of CBO's analysis stand out. First, taking all credit programs as a whole, reestimates for earlier cohorts (especially in mandatory programs) are downward, whereas those for later cohorts are upward. That trend may reflect changes in the general economy since the early 1990s, when credit reform took effect. Initial estimates for those early cohorts may have been influenced by the recession and slow recovery at the beginning of the decade, and their downward reestimates may be explained by the subsequent booming economy. Second, early default reestimates are unlikely to contain much new information and are more likely to reflect random factors, as indicated by the fact that the direction of those reestimates often changes from year to year. Third, the significant upward reestimates made for some housing programs in mid-1999 show how rapidly expected subsidy costs can change.

Limitations of the Analysis

The above conclusions are necessarily tentative, and the situation may look different five years hence, for several reasons. First, no final actual losses are yet available to compare with the initial or revised subsidy estimates. Big surprises could lie ahead.

Second, the entire period of credit reform has been one of continued economic expansion. Default rates for corporate bonds show a dramatic connection between economic expansion and defaults. For example, cumulative default rates for bonds rated BBB, the lowest investment-grade rating by Standard & Poor's, are thought to be several times higher during recessions than during economic expansions.⁽²⁶⁾ Thus, subsidy estimates that may have been biased downward could be hidden by the strong economy.

Third, CBO has little confidence in the reestimate data for some of the programs it analyzed because of inconsistencies that could not be resolved. For example, CBO could not reconcile the cumulative reestimates reported in the Federal Credit Supplement with those reported in the Appendix for the Agency for International Development and three USDA agencies. In the largest unexplained discrepancy, the Appendix reports a reestimate of $50 million for USDA's P.L. 480 direct loan program, but the Supplement reports a reestimate of -$37 million. The unresolved differences for other programs are smaller. (For more details, see Appendix B.) No matter how those remaining discrepancies were resolved, however, the basic results reported here about the direction of reestimates for discretionary and mandatory programs would not change.
 

WAYS TO IMPROVE THE ACCURACY OF SUBSIDY ESTIMATES

CBO's analysis and other reports illustrate the need for more accuracy in estimating credit subsidies. In auditing the government's 1999 financial statements, the General Accounting Office (GAO) cited agencies' inability to estimate the cost of some important federal credit programs as one of the government's "major problems."⁽²⁷⁾ That problem could be ameliorated in at least three ways. First, increasing the visibility of estimates and reestimates in agencies' budget submissions to OMB and in the President's budget request to the Congress could encourage more oversight. Second, agencies could improve the quality of their management information systems more quickly than they are now. Third, agencies could derive better estimates of the cost of federal credit programs by increasing their reliance on private-sector credit evaluation.

Improving the Presentation of Reestimates in Budget Documents

Ideally, reestimates should give policymakers a timely, accurate indication of how a loan program is performing compared with expectations. The current budget documents do not do that effectively, however. The reported reestimates contain many errors. Moreover, assessing loan performance requires having reestimates for defaults separate from those for interest rates, but not all agencies provide that. Finally, no published document or unpublished source shows all of the default reestimates for each cohort.

The following specific steps could improve the budget presentation of reestimates.

• OMB should check the cumulative credit subsidy reestimates in the Federal Credit Supplement against those reported in the Appendix and Analytical Perspectives volumes of the budget.



• The Supplement should report initial subsidy estimates and total reestimates to date by program, with separate entries for interest reestimates and technical or default changes. Volume reestimates should also be shown for mandatory programs. Subtotals should be included for agencies, bureaus, and departments.



• Periodically (perhaps every five years), agencies should report how their estimates of various loan characteristics--such as default and recovery rates--for individual cohorts match experience.

Developing and maintaining an accurate set of reestimates, however, would entail increased costs for agencies as well as OMB.

Improving the Quality of Management Information

The basic tool for developing accurate subsidy estimates--a good management information system--is not in place at many federal agencies. Without the capacity to monitor loan performance according to the characteristics of borrowers and loans, agencies cannot predict or effectively manage federal credit costs. As GAO has documented, most agencies have been making progress in improving their information systems, but major deficiencies remain.⁽²⁸⁾ Developing integrated financial databases requires spending large sums of money up front, which must compete with other priorities for scarce budgetary resources.

Increasing the Use of Private-Sector Information

Relying more heavily on the private sector could permit agencies to estimate and manage credit costs more accurately. Federal agencies lack the personnel, motivation, training, and systems to match the performance of private financial institutions.⁽²⁹⁾ Consequently, many agencies contract with consulting firms for the creation and maintenance of databases and software to predict default losses. Others, including the Department of Education, have contracted out the servicing of loans.

Most federal agencies could make further use of private-sector capabilities by increasing the sale of federal loans. Those sales would provide a market test of subsidy estimates and eliminate the need for reestimates.⁽³⁰⁾ Loan sales would also encourage federal credit managers to adopt practices that enhanced the value of loans.

When SBA began selling its loans on the secondary market, for example, its information about those loans improved dramatically because of demands from buyers. The agency now intends to stop servicing loans itself and focus on soliciting and processing loan applications--areas in which it may have an advantage over the private sector. In fact, in its first auction, SBA received nearly $95 million above its anticipated value of $100 million for those loans. However, most federal loan sales are not expected to result in realized gains.⁽³¹⁾

Agencies could also hire private-sector bond-rating agencies or investment banks to assess and monitor credit quality and estimate probabilities of default for participants in major federal credit programs. Private credit-rating agencies are experienced in identifying weaknesses in credit portfolios, and several investment banks have developed credit-scoring and portfolio-management systems. Moreover, their ratings could provide a more accessible source of information than current data do. Downgrades would be an unmistakable and public signal of the worsening quality of a credit portfolio.

Some budget analysts already use bond ratings to project default rates, especially when estimating the cost of new programs. For example, CBO considered the credit ratings of firms in the satellite television business when it estimated the subsidy cost of loan guarantees for rural television service.⁽³²⁾ In addition, the Department of Transportation uses external credit ratings to assess federal credit risk for assistance provided under the Transportation Infrastructure Finance and Innovation Act of 1998 (TIFIA). The Congress requires all TIFIA projects to get an investment-grade credit rating before receiving federal assistance. Reestimates are based on rating agencies' updated assessments of the government's default risk.⁽³³⁾

1. For more details on credit reform accounting, see Marvin Phaup, "Credit Reform, Negative Subsidies, and FHA," Public Budgeting and Finance, vol. 16, no. 1 (1996), pp. 23-36; Congressional Budget Office, An Explanation of the Budgetary Changes Under Credit Reform, CBO Staff Memorandum (April 1991); and Congressional Budget Office, Credit Reform: Comparable Budget Costs for Cash and Credit (December 1989).

2. The total cost of making loans or providing loan guarantees should also include the costs of administering the loans; however, the Credit Reform Act specifically prohibits administrative costs from being included in estimates of subsidy costs. See Congressional Budget Office, Budgeting for Administrative Costs Under Credit Reform (January 1992).

3. General Accounting Office, Credit Reform: Greater Effort Needed to Overcome Persistent Cost Estimation Problems, GAO/AIMD-98-14 (March 1998), and Credit Reform: Key Agencies Had Difficulty Making Reasonable Loan Program Cost Estimates, GAO/AIMD-99-31 (January 1999).

4. Some credit programs take in more money from fees and interest than they pay out and thus have negative subsidies. For such programs, upward reestimates reduce the amount of the negative subsidy and downward reestimates increase it.

5. OMB is changing the treatment of interest rates in the subsidy estimating process; for more details, see Appendix A.

6. To ensure that a program's financing account stays in balance, reestimates also include the interest that would have been earned or paid if the reestimate amount had been included in the original estimate. See Office of Management and Budget, Preparation and Submission of Budget Estimates, Circular A-11 (July 12, 1999), p. 294.

7. The Congressional Budget Office's estimates of the subsidy rate for student loans (both direct and guaranteed loans) differ substantially from those produced by the Department of Education, largely because CBO's estimates allow for interest rate volatility and the possibility that rates will rise. CBO's estimates also reflect the possibility that interest rates will exceed the cap placed on the rates that borrowers must pay--generally 8.25 percent. For details of CBO's estimates for student loans, see Congressional Budget Office, Estimating the Costs of One-Sided Bets: How CBO Analyzes Proposals With Asymmetric Uncertainties, CBO Memorandum (October 1999), and A Framework for Projecting Interest Rate Spreads and Volatilities, CBO Memorandum (January 2000).

8. For the instructions that agencies are given on preparing credit subsidy estimates and reporting them in the budget, see Office of Management and Budget, Preparation and Submission of Budget Estimates, pp. 279-329. For details on financial accounting, see Federal Accounting Standards Advisory Board, Amendments to Accounting Standards for Direct Loans and Loan Guarantees in Statement of Federal Financial Accounting Standards No. 2, Exposure Draft (March 1999).

9. Because reestimates are made for each cohort, a program can have upward and downward reestimates in the same year. Upward reestimates are reported in the individual program account, whereas downward reestimates are shown as outlays of the program's nonbudgetary financing account and are paid to a general fund receipt account.

10. Another type of error is excluding interest on the reestimates (see footnote 6).

11. Loan modifications result from government actions (such as new legislation) that change the estimated cost of an outstanding loan or guarantee. The permanent and indefinite budget authority that governs a reestimate does not extend to a loan modification.

12. For discussion of the consequences of uncertain economic forecasts, see Henry J. Aaron, "Presidential Address--Seeing Through the Fog: Policymaking with Uncertain Forecasts," Journal of Policy Analysis and Management, vol. 19, no. 2 (2000), pp. 193-206.

13. For details, see Margot A. Schenet, Federal Student Loans: Program Data and Default Statistics, CRS Report for Congress RL30048 (Congressional Research Service, December 22, 1999); Katharine Fraser, "Banks' Victory in Adjusting Education Loan Subsidies Would Last Only 3 Years," American Banker, December 13, 1999, p. 2; and Congressional Budget Office, The Profitability of Federally Guaranteed Student Loans, attachment to a letter to Senator Domenici (March 30, 1998).

14. In addition, SBA has taken steps to improve collections on loans in default. For details, see Budget of the United States Government, Fiscal Year 2000: Analytical Perspectives, pp. 191-193; and the statement of Judy A. England-Joseph, Director of Housing and Community Development Issues, Resources, Community, and Economic Development Division, General Accounting Office, before the House Committee on Small Business, published as General Accounting Office, Small Business Administration: Credit Subsidy Estimates for the Section 7(a) and 504 Business Loan Programs, GAO/T-RCED-97-197 (July 16, 1997).

15. See House Committee on Banking and Financial Services, FY2000 Budget Views: Committee Overview (available at www.house.gov/banking/fy2000bv.htm).

16. OMB's subsidy estimates are more likely to determine how many loans can be made with a given amount of discretionary budget authority, an area in which the Administration's implementation is key. And OMB's reestimates determine the cash flows--and budget outlays--for both discretionary and mandatory programs.

17. Although the Mutual Mortgage Insurance program receives an appropriated negative subsidy and is subject to a cap on lending, the program is treated as mandatory for two reasons. First, the cap is not binding; it is bumped up whenever the level of lending activity approaches it. Second, the receipts generated by new loan guarantees are deposited into the liquidating account to help cover losses associated with guarantees issued prior to credit reform. Before 2000, part of the receipts were used to offset the discretionary administrative expenses of running the program. However, that practice has changed; now all of the receipts are deposited into the liquidating account, and all funds required to administer the program must be appropriated.

18. FHA does not distinguish between technical/default and interest reestimates. However, the entire reestimate for Mutual Mortgage Insurance in the 1999 Mid-Session Review, which reflected unexpectedly poor performance of the program's loan portfolio, was assumed to result from technical or default assumptions.

19. For more information, see General Accounting Office, Internal Controls: VA Lacked Accountability Over Its Direct Loan and Loan Sale Activities, GAO/AIMD-99-24 (March 1999).

20. The General Accounting Office has analyzed the volatility of subsidy rates over time by type of reestimate for cohorts of selected programs; see General Accounting Office, Credit Reform: Greater Effort Needed to Overcome Persistent Cost Estimation Problems, GAO/AIMD-98-14 (March 1998).

21. To track reestimates for individual cohorts over time, CBO used unpublished data from several agencies. The numbers reported here are annual totals that include all types of reestimates, but after the first few years they should primarily reflect defaults. No reestimates for interest rate changes are made after loan cohorts have been fully disbursed.

22. The Farm Service Agency gave CBO a complete set of reestimates, by cohort, for those programs and identified the types of reestimates. That information included annual reestimates for all cohorts for 1998 and 1999, which were prepared too late to be included in the Federal Credit Supplement for 2000, and reestimates for the 1999 cohort of each program.

23. Bankruptcies can involve lengthy court proceedings and result in the rescheduling of payments to debtors and write-offs of debt.

24. In general, spikes in reestimates occur when new estimating models are used. Thus, the swings in reestimates may also be declining because analysts are less likely now than they were earlier to completely change their models.

25. After a $326 million revision for the disaster assistance program in September 1999, SBA recorded a -$516 million revision in February 2000. That decline partly resulted from the use of a new estimating model with lower default rates.

26. That conclusion comes from an analysis of cumulative five-year default rates by Nathan Musick of CBO.

27. In its financial audit, GAO singled out the Department of Agriculture as the primary source of estimating problems. See Department of the Treasury, Financial Report of the United States Government, 1999, p. 20.

28. General Accounting Office, Internal Controls: VA Lacked Accountability Over Its Direct Loan and Loan Sale Activities, GAO/AIMD-99-24 (March 1999).

29. Peter O. Davis and Jozelyn R. Davis, "Realigning Incentives: Improving the Management of the U.S. Government's Trillion Dollar Loan Portfolio," May 25, 1999 (available at www.homestead.com/jpdavis/files/RealigningIncentives5-25-99.pdf). Peter Davis is a manager in Ernst & Young's risk-management and regulatory practice; Jozelyn Davis is a graduate student in economics at Columbia University and a summer associate in Lehman Brothers' risk-management and credit divisions.

30. Thomas H. Stanton, "Using Loan Asset Sales to Improve the Management of Federal Credit Portfolios," The Financier, vol. 5, no. 1 (Spring 1998), pp. 5-16.

31. Because SBA already had authority to sell those loans, the sale will result in a subsidy reestimate rather than a loan modification. For details of SBA's transactions, see Scott Barancik, "Small Business Administration Auctioning $8.5 Billion of Loans," American Banker, August 16, 1999, pp. 1, 3.

32. Statement of Dan L. Crippen, Director, Congressional Budget Office, before the Subcommittee on Telecommunications, Trade, and Consumer Protection, House Committee on Commerce, March 16, 2000.

33. Additional details of the TIFIA project and the Department of Transportation's Quick Score model are available at the TIFIA Web site (http://tifia.fhwa.dot.gov).

---