Home > Industry Analysis > Research & Analysis > FDIC Banking Review




FDIC Banking Review


Vol. 8 No. 2 - Article I - Published: February, 1996 - Full Article

Reps and Warranties
by Penelope Moreland-Gunn, Peter J. Elmer and Timothy J. Curry*

Representations and warranties (R&Ws) are legally binding statements primarily made by sellers assuring buyers that certain minimum asset quality requirements are met. These minimum quality guarantees are commonly given in all types of asset sales, ranging from consumer products to whole loans. The guarantees may extend for only a few months or for the life of the asset.

R&Ws on loan sales are especially interesting because banks and thrifts have increased sales dramatically in the 1990s and each sale tends to create new R&W liabilities that remain long after the sale is completed. Moreover, these liabilities are almost never tracked, recognized in accounting statements, or reserved for with specific reserves. As such, they are a largely unacknowledged source of risk that remains long after the related assets have disappeared from a seller's books.

R&W risk can be significant for institutions that actively purchase and sell loans, because R&W risk exposure increases each time a purchase or sale is made. The act of selling loans increases the seller's liability to honor R&W obligations it issues as part of the sale. Buyers also increase risk due to the possibility that sellers may not honor the R&Ws they issue. Understanding and controlling these risks are prerequisites for effective risk management of any institution that actively purchases or sells loans.

The early difficulties of the Resolution Trust Corporation (RTC) in trying to sell assets without R&Ws are discussed in the first section of this paper. This is followed by a discussion of the "lemons principle" that suggests that the offering of R&Ws is likely to be beneficial. R&Ws commonly made by market participants are then discussed in detail. This is followed by an analysis of the RTC's claims experience with R&Ws, and a comparison of the estimated costs and benefits of granting R&Ws. Conclusions and caveats are presented in the final section.
.

A $200 Billion Example

R&Ws, which have been granted for as long as sale contracts have existed, are used in all types of sales ranging from toasters to corporations. Sellers are often willing to give R&Ws because the risk of loss appears small as long as they are confident that they are well-managed and their product is of high quality. Buyers demand R&Ws because the potential problems may be difficult or costly to research, especially if they are numerous and data are not readily available to assess their risk. Buyers may be suspicious of products with no R&Ws due to a fear that a seller's unwillingness to offer them signals a higher risk of problems.

Figure 1 documents the explosion in sales of single-family mortgages that occurred during the 1980s. Because R&Ws are given in virtually all sales, including those to the Federal Home Loan Mortgage Corporation (FHLMC) and the Federal National Mortgage Association (FNMA), it is safe to assume that R&W liabilities grew in lockstep with sales activity. The success of the single-family market in the 1980s spread to other loan markets in the 1990s, as trading increased dramatically for all types of loans ranging from commercial mortgages to nonperforming consumer loans. R&Ws have necessarily also increased, especially for products that encounter more varied risks than single-family mortgages, such as commercial loans.

The experience of the RTC with R&Ws exemplifies the problems encountered by many market participants. The RTC was formed in August of 1989 with no sales experience and no policy for giving R&Ws. Initially, the RTC made several efforts to limit R&W liabilities created from the sales of loans and servicing rights. One approach granted R&Ws only from the receiverships created from the dissolved savings-and-loan associations (S&Ls), thereby limiting R&W liabilities to the receiverships. Other approaches included the granting of only very limited R&Ws and replacing R&Ws with rights to put back assets to the RTC in the event of problems.


Under the initial R&W policy, the RTC experienced great difficulty in selling its assets. The process was slow, prices were low, and buyers often resisted doing business with the RTC based on the terms it tried to dictate to the market. Buyers were highly suspicious of the credit quality of receiverships. In some cases, buyers required a portion of the sale proceeds to be placed in escrow to ensure that R&Ws issued by insolvent S&Ls could be honored. The limiting of R&Ws caused buyers to demand extensive reviews of loan and servicing files ("due diligence"), which were costly, time-consuming, and cumbersome to administer.

The RTC experimented with the sale of whole institutions and giving buyers rights to put back assets to the RTC after the sale and after they had been given a chance to research the loans. However, bidder interest showed little improvement and assets were often returned to the RTC. Also, the buyers of S&Ls and banks began teaming with Wall Street experts to buy whole S&Ls, then "flipping" (immediately reselling) the loans to the Wall Street partners, who in turn either securitized or otherwise resold the loans with R&Ws for a profit.

The RTC researched the costs and benefits of adopting more common sales practices, that is, performing due diligence prior to a sale and giving R&Ws. In May 1990, the RTC's Board of Directors approved the first policy permitting the use of R&Ws in loan and servicing rights sales. This policy expanded the R&Ws given and allowed the RTC in its corporate capacity to give R&Ws instead of only giving them from receiverships. However, the policy established several key restrictions, such as limiting R&Ws to a maximum of five years and indemnifying investors only in the event of an actual loss. As will be discussed later, these restrictions are more than is customary in the market because many situations require long-term indemnification and/or the repurchase of assets. Therefore, investors remained concerned about the risk of buying assets from the RTC.

In August 1990, the RTC's Board of Directors responded to market pressures by expanding its R&Ws to those customary or normal in the marketplace. In particular, R&W coverage was extended to cover the life of the loan and the RTC was permitted to resolve problems by repurchasing assets and/or substituting replacement loans. These revisions conformed RTC policy with market practices, thereby clearing the path for high-volume loan sales that competed directly with other loan sales in the market. It also helped to open the door to the development of securitization, a process that aimed to sell loans for higher prices than those achieved by selling loans directly to buyers as whole loans. 1 Indeed, R&Ws have since facilitated the sale of over $200 billion of loans and loan servicing for a vast array of loan types, many of which had never been sold in volume.

Although the expansion of R&Ws cleared the path for high-volume loan sales, it also increased the RTC's exposure to the risk of losses on R&W claims. Therefore, a R&W claims administration department was created to oversee related issues and to establish reserves to cover R&W claims costs. As will be discussed later, the experience of the RTC provides a rare example of the likelihood and magnitude of R&W claims and costs.

The Lemons Principle

Analysts unfamiliar with the whole-loan market occasionally claim that market efficiency will cause the financial benefit of higher prices on the sales of loans with R&Ws to equal the cost of granting R&Ws. This theory suggests that buyer competition will cause the price of loans with R&Ws to exceed the price of loans without R&Ws by an amount equal to the sellers' expected cost of granting the R&Ws. However, a classic principle of economics, known as the "lemons principle," suggests an alternative view, that the value of granting R&Ws should be greater than their cost.

The lemons principle is a response by Akerloff (1970) to the observation that the value of a new automobile declines significantly as soon as it is driven off the showroom floor. Akerloff explained this as a result of the fact that information relating to the quality of the automobile is not the same for the buyer as it is for the seller. That is, information is asymmetric in the sense that the seller is more likely than the buyer to be aware of the auto's problems and limitations, that is, better able to identify the auto as a "lemon." In the absence of complete information, buyers assume the worst and price the auto as if it were a lemon, even though it may in fact be in perfect running condition.

Information on the quality of a loan is as difficult (costly) to assess as information on the quality of an automobile. While recent payment history and loan characteristics are typically available from the servicer on, for example, a magnetic tape, it is always possible that the tape information is incorrect. More importantly, the tape will normally not show information about a host of potential problems such as underwriting equirements, policies used to obtain appraisals, sources of credit histories, mistakes made in originating or servicing the loans, and legal problems that either have or have not been identified. Fraud in any of these areas is always a possibility as well.

Buyers are acutely aware of the risks arising from the lack of information in a pool of loans. They control the risks by either: 1) requesting that the seller give R&Ws, 2) requesting that the seller perform extensive due diligence and make it available to bidders prior to the sale, or 3) bearing the cost of performing their own due diligence.

The experience of the RTC was that many whole-loan buyers would not even consider buying loans that did not have R&Ws customarily given in the market. Some buyers would consider loans with limited R&Ws, but only after performing extensive due diligence. Many buyers did not wish to bear the cost of their own due diligence and they were fearful that it would not address all risks. The final result was that the offering of loans with either no or limited R&Ws reduced the number of potential buyers significantly and complicated the sale with due diligence-related issues. Buyers that took the added risk and participated in the sales consequently discounted prices heavily based on the presumption of problems.

Discussions with financial advisors suggested that limiting R&Ws can cause price reductions of two percent to three percent for relatively clean portfolios of single-family mortgages. Higher discounts are likely for loans that are not as widely traded as single-family mortgages and for loans with documentation, underwriting, or servicing problems. The lemons principle suggests that these price discounts should exceed the cost of granting R&Ws and the payment of related claims. Actual RTC R&W claim costs shown below are consistent with this view. However, a comparison of R&W costs and benefits is reserved for later in this study, following a more detailed discussion of the nature of common loan-related R&Ws.

Common Representations and Warranties

R&Ws fall into two categories: 1) institutional, which deal with the legal authority of a firm to sell loans, and 2) loan-related, which deal with risks arising directly from the loans being sold. Most R&Ws are loan-related because the most likely source of problems are loan characteristics such as documentation, underwriting standards, and delinquency status. R&Ws cover a wide range of issues, from innocuous guarantees given in almost every sale to specialized guarantees used only for unique circumstances.

Institutional R&Ws are small in number and can often be used for many types of loan sales. As shown in section A of Table 1, these R&Ws focus on the nature of the institutional seller and its right to sell the assets. While they are found in many types of sale agreements, claims on institutional R&Ws are rare because sellers normally have full authority to execute sales and their obligations are valid and binding. Nevertheless, buyers consider them indispensable because a violation can result in extensive losses and/or litigation. For example, a buyer can seek to return all assets, and sue for return of its funds, if a seller does not have legal authority to sell a package of loans.

Loan R&Ws are used in many types of loan sales, but many must either be tailored, or written precisely, to meet the needs of each type of loan. Section B of Table 1 lists a number of R&Ws common to sales of most types of mortgages. Items B-1 and B-2 are seemingly obvious guarantees that may be given by most sellers with little risk of subsequent claims. In fact, they are found in almost all contracts and result in claims only in unusual situations. For example, they became an issue in the RTC's sale of "participation" loans, that is, loans owned by two or more participants. Documentation on joint ownership for these loans was not always complete or available. Ownership guarantees greatly enhanced the ability to sell the loans because they relieved all parties of the need to find all documentation and resolve all ownership issues prior to sale.

Servicing-related issues, such as items B-3 and B-6 are most likely to result in claims. Because a host of loan characteristics change each month, incorrect information can be easily transmitted to a buyer. The loan balance or coupon could be incorrectly recorded on the seller's computer system or not received prior to transfer. The rate on an adjustable-rate loan may be tied to the wrong index or otherwise reset incorrectly. Mortgage insurance coverage may have lapsed due to failure to make timely payments. Real-estate taxes may not have been paid.

Table 1
Sample of Common Loan Reps and Warranties

A. Institutional
1. Seller has taken action to authorize the execution, delivery, and performance of the sale agreement.
2. All obligations of the seller are legal, valid, and binding.
3. Obligations will not conflict with any provisions of any law.
4. No action, suit or proceeding is pending against the seller that would materially affect the ability of the seller to carry out the transaction.

B. Found in Many Types of Mortgage Sales
1. Seller is sole owner and holder of the mortgage loan.
2. Seller has full right and authority to sell mortgage loan.
3. Mortgage loan schedule information is correct in all material respects.
4. The related note, mortgage and other agreements executed in connection therewith are genuine and each is a legal, valid and binding obligation of the maker enforceable except by limitations of bankruptcy, insolvency, reorganization or other similar laws.
5. Seller is transferring the loan free and clear of any and all liens, pledges, charges or security interests of any nature encumbering the loan.
6. All taxes, governmental assessments, insurance and utilities that came due prior to the closing date have been paid or an escrow account has been established.
7. The related mortgaged property is free of mechanics' and materialmen's liens.
8. There are no proceedings pending related to the property, and property is in good repair and free and clear of any damage.

C. Single-Family Mortgage Sales
1. Mortgage loan is secured by a mortgage on a 1-4 family residential real property or a condominium unit or a unit in a planned unit development or is a co-op loan.
2. Each mortgage loan was underwritten generally in accordance with FNMA or Federal Home Loan Mortgage Corporation (FHLMC) standards in effect at the time of origination except that 1) the original principal balance may exceed FNMA/FHLMC limits.
3. Each mortgage loan with a loan-to-value ratio (LTV) greater than 80 percent is covered by a primary mortgage, FHA, or VA insurance policy. Such policy is in full force and effect and insures the excess over a 75 percent LTV, and such policy shall remain in effect until the unpaid principal balance is reduced below 80 percent LTV.
4. Loan is serviced in accordance with the terms of the note.

D. Commercial Mortgage Sales
1. Borrower possesses all valid licenses, permits and other authorizations necessary to conduct business.
2. If required, seller has inspected or caused to be inspected property within the past 12 months.
3. Credit file includes environmental assessment that does not disclose any disqualifying conditions. If a disqualifying condition exists, remediated cost should be equal to or less than $10,000.
4. Seller represents that, as of the sale closing date, property does not contain hazardous materials such that the mortgage loan would be ineligible for purchase by FNMA (assuming that the loan were otherwise eligible for purchase). In the event of a breach of this representation, the seller will cure such breach or repurchase the affected mortgage loan.

E. Vehicle Loan Sales
1. Seller has not and does not make any representations, warrants or covenants regarding financial condition or status of obligor.
2. Seller has not transferred its interest in and to the vehicles to any other person. Seller has the right to sell its interest in and to the vehicles in accordance with the terms of the contract and free of liens.
3. Purchaser waives any claims or cause of action it might have against the seller for any loss, damage or expense caused by or to any vehicle.
4. Purchaser is a sophisticated buyer of contracts and loans similar to loans in package.

The wide variety of servicing problems implies that there is no single rule or procedure for dealing with them. Some can be easily identified and resolved soon after the sale is completed, either as a result of servicing the loans or a post-sale audit or reconciliation. Claims generated at this stage tend to relate to performing loans for which delinquency and default-related losses are not an issue. Claims generated at later stages are more likely to arise after a loan defaults and a careful check is made of the loan file. Incorrect real-estate tax payments and title problems are often caught at this point because back taxes must be paid and a title check made in order to complete the foreclosure process. Because delinquency and foreclosure may not occur until many years following a sale, R&W claims may not be made for a similar period. Thus, it is not surprising that buyers often require R&Ws for the life of the loan.

Section C of Table 1 contains a sample of R&Ws specific to single-family mortgage sales. This loan market is made unique by the widespread influence of the FHLMC and the FNMA. Firms that sell directly to the FHLMC and the FNMA must agree to voluminous R&W origination and servicing requirements stated in lengthy volumes published by the two agencies. Aside from these requirements providing the agencies with R&W protection, their wide publication and acceptance enable other mortgage market participants to adopt similar R&W requirements easily. As shown in item C-2, a simple R&W extends FHLMC and FNMA underwriting requirements to mortgage sales that do not involve either agency. Item C-3 is also a FHLMC and FNMA requirement even though it does not directly reference them by name.

Commercial-mortgage R&Ws differ significantly from those of single-family mortgages because the loan and underwriting characteristics are different and there is far less standardization of the origination and servicing functions. Also, the FHLMC and the FNMA are not permitted by their charters to purchase commercial mortgages, so standardized R&W guidelines and benchmarks are not as common as for single-family mortgages.

The unique aspects of commercial mortgages are evident in the four R&Ws listed in Section D of Table 1. Because commercial-mortgage borrowers are usually corporations, they may be required to have and maintain various licenses and authorizations required to conduct business. Servicers are often required to inspect periodically commercial properties and monitor the borrower's financial health by reviewing financial records. Environmental problems are a major issue, as suggested by items D-3 and D-4. Item D-4 is especially important because the cost of environmental clean-up can be very high and the requirement places this cost on the seller, even if the problem is unknown to anyone at the time of sale. In some cases the environmental clean-up costs exceed the value of the loan and may involve years of litigation to resolve.

The final sample of R&Ws applies to a group of loans unrelated to mortgages, namely, vehicle loans. Because these loans have relatively low balances and are collateralized by automobiles, lenders place much less reliance on the security gained from foreclosure and the subsequent sale of the underlying collateral. Extended or costly litigation between the borrower and the lender is unusual. These attributes combine to greatly simplify the types of R&Ws given. For example, one requirement might be that the seller guarantees its ownership or right to sell the loans (item E-2) in order to prevent fraud. Given the changed nature of product risk, sellers may be as likely as buyers to request R&Ws. Specifically, sellers may ask buyers to give R&Ws to the effect that they are sophisticated and capable of understanding the risks of the loans being sold (items E-3 and E-4).

Claims Trends

A consequence of the RTC's unprecedented sales volume is that it provides the large-scale R&W claims experience needed to gauge the likelihood of problems that normally have only a low likelihood of occurrence. The large program size also caused the creation of a claims administration department and a database of claims losses. These data provide one of the largest records of R&W claims costs and trends ever created.

Many interesting characteristics of the data can be observed by examining the claims experience of a sample of deals. Tables 2a and 2b provide such a sample for fixed-rate and adjustable-rate mortgage sales that occurred relatively early in the RTC's history (1991). 2 The sales chosen were five securitizations, three containing single-family mortgages (1991-11, -12, and -14) and two with multi-family mortgages (1991-M1 and -M5). Early securitizations provide convenient examples because their collateral characteristics are well-documented and claims experience is seasoned. A review of other sales around the same time found that the sample chosen was reasonably representative.

 Table 2a  
Rep and Warranty Claims Statistics For a Sample of RTC Sales
Repurchased Total Initial Sale Data Cured Claims Claims Approved Claims Mortgage Sale UPB Number Average Average Average Sale Date $Mil Loans Type Number $000s Number $000s Number $000s (Security) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Single-Family 1991-11 10/91 $888 16,826 FX 28 $2 1 $ 32 29 $ 3 1991-12 10/91 527 2,330 ARM 9 2 0 0 9 2 1991-14 11/91 545 7,126 ARM 176 2 0 0 176 2 Multi-Family 1991-M1 8/91 373 218 ARM 0 0 6 5,960 6 5,960 1991-M5 11/91 386 548 FX 0 0 1 1,536 1 1,536
Note: Data reflect all RTC R&W claims filed as of December 31, 1994, for all R&Ws contracted by the RTC at the corporate level. The UPB in column 2 is unpaid principal balance of collateral placed in trust as of the starting date of the security. The average claims figures in columns 6, 8, and 10 are computed by dividing total claims costs by the number of claims shown in columns 5, 7, and 9, respectively (rounded to $000s).

The initial sale data in columns 1-4 of Table 2a provide a snapshot of the sample. The loans were sold into security trusts between August and November 1991. These start dates imply three to four years of R&W claims experience prior to the 1994 database cutoff used for this study. The securities were mid-sized offerings in the $300-900 million range that contained considerable collateral variation.

Columns 5-10 of Table 2a break total approved claims into two components, cured and repurchased. A limitation of the data is that the costs of cured claims are recorded as the actual cost of curing the claims, whereas the cost of repurchases is recorded as the entire unpaid principal balance (UPB) of the repurchased loan. 3 The cost of curing a repurchase is not tracked because repurchased loans may be resold in a number of ways that create problems for tracking losses. For example, a repurchased delinquent loan might later have the delinquency cured and be included in a package of several thousand loans, none of which are priced separately when the securitization is completed. The important point is that the recorded cost of repurchases is many times larger than the cost of cures, and the two numbers should not be compared directly as a measure of claims costs.

In spite of data limitations, several points can be noted from the approved claims data in Table 2a. First, on the single-family mortgage side, the average claim for all three sales shown was only $2,000. This is because R&W problems for single-family mortgages tend to relate to relatively solvable and inexpensive problems such as tax penalties and incorrect payment adjustments. Second, on the multi-family mortgage side, all claims are repurchases. 4 Multi-family claims are more likely to relate to environmental or other problems that are not as easily solved as single-family problems. Repurchase may be required because the problems are either too costly or troublesome to resolve. Multi-family problems are aggravated further by large loan balances ($5,960,000 average balance for the six claims in 1991-M1 and $1,536,000 for the one claim in 1991-M5).

Table 2b 
  Rep and Warranty Claims Statistics For a Sample of RTC Sales (continued)  


                               Total Claims                      Claim Filing Date   Claim Processing Time
                  Approved Claims     Denied/Other Claims
     Mortgage               Average             Average
      Sale        Number    $000       Number   $000          0-1 Yr   1-2 Yr   &gt2 Yrs      0-2 Mo    2-4 Mo    &gt4 Mo
    (Security)     (1)       (2)        (3)      (4)            (5)      (6)       (7)          (8)      (9)       (10)
 Single-Family      
     1991-11        29         $3       133      $15              1       26       135           65       29        68
     1991-12         9          2         8       23              7        4         6           15        2         0
     1991-14       176          2       165       25             44       57       240           44      108       189
 Multi-Family 
     1991-M1         6     $5,960         8      429              4        4         6            1        0        13
     1991-M5         1     $1,536         6    3,090              2        1         4            6        1         0
Note: Data reflect all RTC R&W claims filed as of December 31, 1994, for all R&Ws contracted by the RTC at the corporate level for the five securities listed.

Table 2b begins by comparing approved claims to those either denied or subject to some other resolution status. 5 Relatively large percentages of the claims were denied or otherwise not approved. For example, in three sales (1991-12, 1991-14, and 1991-M1) approximately 50 percent of all claims were denied, while the remaining two sales had approximately 80 percent of all claims denied. Moreover, the average approved claim tended to cost considerably less than the average denied claim. Combining these statistics suggests that substantial savings accrue to reviewing claims carefully.

The remainder of Table 2b deals with the time required to file and process claims. The most interesting points are that large numbers of claims may be made several years after the loans are sold and a review of these claims may take considerable time and/or resources. In all of the cases shown the largest number of claims occurred in the longest filing date category (over two years) while the shortest category (up to one year) contained either the lowest or almost the lowest number of claims. This trend primarily results from the fact that RTC securities provide an indemnity against loss, which implies that claims are more likely to occur several years after the securities are created and after the loans have had time to default. In contrast, the RTC's non-securities sales generally did not indemnify against losses and their claims tended to occur in the first year following sale. This trend can be seen in Figure 2, which shows that R&W claims for sales consummated in 1991 peaked in 1992 and declined thereafter for every class of RTC sales except securities. Therefore, the likelihood of 1991 sales claims occurring in 1995 or later is low for all of the classes of RTC assets shown in Figure 2, with the exception of securities.

Compounding the filing time problems is the fact that substantial numbers of claims required a long period (over four months) to process, that is, to approve, deny, or otherwise conclude. The long processing times reinforce the notion that significant resources and costs may be required to administer claims even apart from the direct cost of curing claims and repurchasing loans.


Table 3 RTC Claims Losses by Product Type

                                                                      Estimated     Estimated   
                                   Claims    Claims        UPB          Losses        Losses      Estimated 
Estimated
                             UPB   Losses    Losses    Repurchased   Repurchases  Reperchases      Total    
 Total
Product                     $Bil    $Mil    as % UPB      $ Mil         $ Mil       as % UPB        $ Mil      as % UPB
                             (1)     (2)       (3)          (4)           (5)          (6)            (7)         (8)
Servicing Rights          $ 139 B  $ 69 M     0.05%       $ 44 M       $ 0-7 M       0-0.01%       $69-76 M   0.05-0.06%
Single-Family
 Securitizations             23      27       0.12          83          0-13         0-0.06         27-40     0.12-0.18
Multi/Commercial/Other      
 Securitizations             20      40       0.20         134          0-33         0-0.17         40-73     0.20-0.37
Whole Loans                  25      12       0.05         178          0-27         0-0.11         12-39     0.05-0.16
Other                        19      17       0.09          64          0-10         0-0.05         17-27     0.09-0.14
Total $ / Average %        $226 B  $165 M     0.07%       $503 M      $ 0-90 M       0-0.04%     $165-255 M   0.07-0.11%
Note: Data reflect all RTC R&W claims paid as of December 31, 1994, for all R&Ws contracted by the RTC at the corporate level. Estimated losses in columns 5 and 6 are calculated as the likely range of losses on loans repurchased by the RTC to cure R&W deficiencies. These ranges assume a minimum loss of zero for all categories of loans. The maximum loss is 15 percent of UPB repurchased for all categories of loans except multi/commercial/other securitizations, and 25 percent of UPB repurchased for multi/commercial/other securitizations.

Costs vs. Benefits

An advantage of the RTC's experience is that its claims and sales data may be combined to compare the financial costs and benefits of granting R&Ws. That is, the data may be used to perform a cost/benefit analysis of the economic efficiency of granting R&Ws.

The cumulative claims loss experience for the RTC's R&Ws is shown in Table 3. These data include claims paid for all R&W agreements made by the RTC in its corporate capacity as of December 31, 1994. As such, they include the seasoned claims experience arising from sales completed in 1990 and 1991, as well as only recent claims experience from sales completed in 1994.

Column 1 of Table 3 displays the RTC's R&Ws by type of sale. Of the total $226 billion UPB covered by R&Ws, the largest component, $139 billion, or approximately 62 percent of the total, was given on sales of servicing rights. 6 The remaining 38 percent were from four similar-sized categories of loan sales: single-family securitizations, multi-family/commercial securitizations, whole loans, and "other" loan sales.

Claims losses for each sale type are shown in columns 2-8 of Table 3. The UPB of all loans covered by RTC corporate R&Ws is shown in column 1 in order to provide a benchmark for gauging the magnitude of the losses. The UPB is a useful benchmark because it reflects the maximum potential R&W loss in most circumstances. Losses are normally below the UPB because lenders may opt not to foreclose on properties with projected losses greater than the UPB. While it is possible for legal or environmental problems to cause losses to exceed the UPB, these situations are rare.

The most easily tracked costs are for cures, which equaled $165 million, or 0.07 percent (seven basis points) of the $226 billion UPB benchmark. The claims loss rate in column 3 was highest for multi-family/commercial securitizations (20 basis points) due to a higher likelihood of complicating factors such as environmental problems. The remaining claims loss rates ranged from five basis points for whole loans and servicing rights to 12 basis points for single-family securitizations. Whole-loan loss rates were the lowest because most RTC whole-loan sales were limited to either unusually clean packages of loans or loans with extensive due diligence.

As noted earlier, the cost of R&W claims resulting in repurchases was not tracked by the RTC. The fact that many repurchases cause either no or only small losses complicates matters. For example, a repurchase of high-quality adjustable-rate mortgages may be required because they were sold into a security containing only fixed-rate mortgages.

The problems of calculating losses on repurchases are dealt with by estimating potential losses as ranges (see columns 5 and 6) based on the UPB of loans repurchased, as shown in column 4. The range for each product has a lower limit of zero and an upper limit of 15 percent of the UPB of loans repurchased for all types of loans except multi-family and commercial loans. A 25 percent upper bound is used for multi-family and commercial loans due to their higher risk. The 15 percent and 25 percent upper bounds are chosen based on the authors' best estimates of higher-than-expected or "conservative" estimates of the average loss on repurchases. 7 Given the range estimates in column 5, column 6 estimates that repurchases add an upper limit of between one basis point and 17 basis points to the estimated R&W losses of the various product classes. The average upper limit loss across all products is four basis points.

The costs of granting R&Ws can now be directly compared to the benefits for a key asset class, single-family mortgages. As noted, financial advisers and early RTC experience estimated the benefit of granting R&Ws on single-family mortgages in the range of two percent to three percent (200 to 300 basis points) of the UPB sold. That is, single-family mortgages sold with R&Ws can be expected to sell for 200 to 300 basis points more than if they are sold without R&Ws.

A 200-to-300 basis point financial benefit for single-family mortgages compares very favorably with total R&W costs in columns 7 and 8 of Table 3. The cost category most directly applicable to single-family sales is single-family securitizations. Column 8 shows the cost of single-family securitizations as 12 to 18 basis points. However, because Figure 2 suggests that securitization claims may continue in the future, an adjustment needs to be made to recognize the possibility of future claims. If we project future single-family securitization R&W claims costs to equal all costs incurred through December 31, 1994, then the projected R&W cost implied by past securitization claims equals 24 to 36 basis points. 8 Thus, the cost of single-family mortgage R&Ws is only a small fraction of the 200-to-300 basis point benefit that accrues from granting them in single-family mortgage sales.

The dramatic excess of financial benefits over costs suggests that sellers of single-family mortgages are well-advised to grant R&Ws in spite of the expectation of some claims-related costs. Indeed, the excess may help to explain the RTC's experience that it is customary for sellers to grant R&Ws, and why sellers who do not grant R&Ws are viewed with suspicion by buyers who willingly pay a price premium to avoid the risk of purchasing a "lemon."

But does the cost/benefit analysis of the the RTC's single-family mortgage experience extend to sales of other types of financial assets, such as commercial mortgages or "other" loans? Three points suggest that the RTC's experience applies to many other types of assets. First, the R&W cost statistics in Table 3 are of the same order of magnitude for all types of assets. While the costs vary by product, all column 8 estimates are far below the 200-to-300 basis point range found for single-family mortgages. 9 Second, all types of loans are characterized by information problems similar to those of single-family mortgages. Moreover, the RTC's initial policy of limiting R&Ws resulted in the same liquidity and bidder interest problems regardless of the types of loan-related assets sold. Finally, it is a widespread market practice for sellers to give, and for buyers to request, R&Ws for all types of financial assets.

The weight of evidence points to the conclusion that it is cost-effective to grant R&Ws in many types of loan sales. Given the lemons principle, we expect that the excess of benefits over costs increases as the quality or availability of risk-related information declines.

Conclusions

A number of interesting lessons may be drawn from the RTC's extensive R&W experience. First, R&Ws may represent significant risk because considerable expense may be required to cover claims costs and claims may continue for several years after the loans are sold. The risk may be a problem especially for firms that actively sell loans to the secondary markets because R&W liabilities grow with each sale. Second, administrative costs exacerbate the risk because claims may require additional expenses in the form of file reviews, the repurchase of assets, legal advice, and other problems. Finally, in spite of the likely costs and administrative problems, the experience of the RTC suggests that it is financially efficient for firms to grant R&Ws. R&W claims costs for single-family mortgages sold by the RTC were dramatically less than the financial benefits estimated to accrue from higher market prices on loans sold with R&Ws. The evidence points to a similar conclusion for many other types of assets as well.

Two caveats should be kept in mind regarding the RTC's experience relative to that of other institutions. First, the bulk of the RTC's claims experience occurred in a relatively favorable economic environment between 1992 and 1994. If this had been an extended period of unfavorable economic conditions, then claims-related losses could have been higher, especially for the default-sensitive portion of claims found in RTC securitizations. Similarly, an economic downturn could cause future claims experience, for the default-sensitive portion of the claims, to compare unfavorably from that documented in this paper. Second, the RTC's experience is made distinct by the fact that the S&Ls it took over, and whose assets it sold, were highly distressed. Therefore, the RTC may have been more likely to encounter R&W claims problems than, for example, healthy firms, and this may have affected both the benefits and the costs of granting R&Ws. While the net effect of these factors is difficult to quantify, the RTC's R&W claims experience nevertheless seems to provide a useful guide to projecting potential claims costs and trends, especially for firms that actively sell loans to the secondary markets.

REFERENCES Akerloff, George. "The Market for Lemons: Quality Uncertainty and the Market Mechanism." Quarterly Journal of Economics (August 1970): 488-500.

Resolution Trust Corporation. Sketchbook of RTC Securities. Washington, DC: Resolution Trust Corporation, December 1994.

Last Updated 8/2/1999 Questions, Suggestions & Requests