Announcement 2004-75

The Tax Reform Act of 1986 (100 Stat. 2085) (1986-3 C.B. Vol. 1), created a new tax entity, the Real Estate Mortgage Investment Conduit (REMIC), that was designed to be the exclusive vehicle for the issuance of multi-class mortgage-backed securities. A REMIC may issue one or more classes of regular interests and must issue a single class of residual interest. Section 860B(a) of the Internal Revenue Code (Code) requires that a regular interest be treated as a debt instrument whether or not the interest would qualify as a debt instrument under general tax principles. The holders of the residual interest are required to take into account their proportionate share of the REMIC’s taxable income or net loss.

Prior to 1988, the holder of a REMIC regular interest was required to be entitled to a specified principal amount plus interest at a fixed or variable rate. The Technical and Miscellaneous Revenue Act of 1988 (102 Stat. 3342) (1988 C.B. 1) permits the holder of a REMIC regular interest to receive interest that consists of a specified portion of the interest payments on qualified mortgages if the portion does not vary during the period the regular interest is outstanding. Section 860G(a)(1)(B)(ii). The expanded definition of REMIC regular interest has allowed for the issuance of interest-only REMIC regular interests (REMIC IOs).

A REMIC IO generally provides for a nominal (or zero) specified principal amount and stated interest consisting of a specified portion of the interest payments on mortgages held by the REMIC.^[2] Section 860B(a) provides that a REMIC regular interest is taxed as a debt instrument. Nevertheless, a REMIC IO differs from a traditional debt instrument in that the aggregate of the amounts received by the holder of a REMIC IO may be less than the amount for which the instrument was issued. This may occur if the underlying mortgages are prepaid at an unexpectedly rapid rate. In that case, the amounts of interest paid on these mortgages will be less than expected, and the amounts payable to the holder of the REMIC IO will be correspondingly reduced. As a result, REMIC IOs present novel and difficult questions in the application of tax rules that were designed primarily to account for instruments that qualify as debt under traditional tax principles.

Section 1275(d) authorizes regulations to modify the tax treatment prescribed by sections 163(e) and 1271 through 1275 (relating to original issue discount (OID)) if the statutory tax treatment does not carry out the purposes of those sections. The IRS and Treasury are considering whether to issue regulations, including regulations under the authority of section 1275(d), with respect to the tax treatment of REMIC IOs for issuers and initial- and secondary-market purchasers. This advance notice of proposed rulemaking sets out additional background information, including summary descriptions of possible approaches to the problems described below, and requests public comment.

As noted, the terms of a REMIC IO generally provide both for stated interest consisting of a specified portion of the interest payments on mortgages held by the REMIC and also may provide for a nominal amount of specified principal. The tax rules currently applicable to a REMIC IO depend on whether the stated interest is treated as consisting entirely of interest or as being, in part, a return of the proceeds for which the instrument was issued.

Some taxpayers believe that, if the stated interest is respected as interest, it generally is qualified stated interest (QSI) and so is not part of the stated redemption price at maturity (SRPM). As a result, because the specified principal due on the REMIC IO is, at most, nominal, a holder generally will have paid more than the amount payable when the REMIC IO matures, and thus there will be bond premium. On the other hand, if the interest payments are recast as, in part, a return of the proceeds for which the REMIC IO was issued, the portions so recast are included in the SRPM, and the instrument is issued with OID.

Glick v. United States, 96 F. Supp. 2d 850 (S.D. Ind. 2000), weighed these competing analyses of a REMIC IO. The instrument at issue in the case had been issued for a little over $12 million. The terms of the instrument provided both for specified principal of $362,000, which was based on principal payments on the underlying mortgages, and for much larger expected amounts of stated interest, which were linked to, and contingent upon, interest payments on the underlying mortgages.

Given the terms of the mortgages and the rate at which the mortgagors were, in the aggregate, expected to prepay their mortgages, the prospectus estimated total future cash flows under the REMIC IO of over $14 million. Basing its computation on the specified principal amount, the prospectus identified the resulting estimated interest rate on the REMIC IO as being 1006.7 percent. On the other hand, the prospectus further disclosed that, if a yield computation were to be based on the taxpayer’s purchase price of over $12 million, the anticipated yield to maturity was just under 8 percent.

Because of falling interest rates, the mortgages underlying the instrument were prepaid at an extremely fast rate, and the taxpayer recovered less than two thirds of the original investment.

The Government argued that the instrument was issued at a discount and that the taxpayer’s loss on the instrument was capital and would be recognized only in the year the instrument was retired. The taxpayer, on the other hand, claimed that the instrument was acquired at a premium and that ordinary deductions were allowable under section 171 during the entire period that the taxpayer held the instrument. Explaining that it had resolved the question by “[e]xamining the economic reality of the transaction,” 96 F. Supp. 2d at 867, the court issued summary judgment for the Government.

REMIC regular interests are among the debt instruments for which the accrual of OID is calculated taking prepayments into account. This is accomplished by using a method commonly known as the prepayment assumption catch-up (PAC) method, which is provided in section 1272(a)(6). Under this method, it is necessary to estimate first the rate at which any outstanding principal on the underlying mortgages will be prepaid and, then, the yield to maturity of the instrument. These estimates remain constant in all PAC method computations throughout the life of the instrument.

In each accrual period, the daily accruals of OID are equal to the ratable portion of the excess (if any) of the sum of (1) the present value of the remaining payments under the debt instrument as of the close of the period (end-of-period present value) and (2) the payments during the accrual period that are included in the SRPM (accrual-period SRPM receipts), over the adjusted issue price of the debt instrument at the beginning of the period.^[3]

The end-of-period present value is calculated using the two estimates referred to above. First, the amount and time of the remaining payments are determined on the basis of both the specified principal actually outstanding at the end of the accrual period (taking into account any prepayments occurring before the close of the accrual period) and the previously estimated, static assumption about the rate at which any outstanding principal will be prepaid. Second, the present value of these remaining payments is determined by discounting them at the previously estimated original yield to maturity.

A holder of an OID debt instrument includes in gross income the sum of the daily portions of the OID for each day during the taxable year on which it holds the debt instrument. An issuer’s interest deduction for OID accruals is computed in a similar fashion.

In the case of a traditional debt instrument that is issued with OID or a REMIC regular interest that is issued for less than its specified principal amount, prepayments increase the instrument’s yield to maturity. Failure to anticipate prepayments would result in uneconomic deferred accrual of OID inclusions, and the holder would recognize capital gains when the instrument is finally sold or retired. To prevent such uneconomic deferral of OID inclusions, the PAC method, in each period, recognizes more OID than would be recognized if no anticipated prepayments were taken into account. However, the PAC method may result in uneconomic acceleration of OID accruals in certain circumstances.

When section 1272(a)(6) became law, an instrument subject to it generally provided for payments of a fixed amount of specified principal, plus payments of QSI, which were based on the amount of principal still outstanding. If the issue price of the instrument was less than the specified principal, that difference resulted in a fixed amount of OID, which had to be accrued over the life of the instrument.

For such an instrument, if actual prepayments occur at a slower rate than the original estimate, OID will be accrued more rapidly under the PAC method than the actual prepayment rate would justify. If prepayments are particularly slow, the OID remaining to be received at the end of a period may be greater than the excess of the original OID on the instrument over the amount of the OID that had been accrued in prior periods. As a result, the amount of OID for the current accrual period under the formula in the PAC method may be a negative number (Negative OID).^[4] This occurs if the adjusted issue price at the beginning of an accrual period (which reflects prior OID accruals) exceeds the sum of (1) the end-of-period present value and (2) the accrual-period SRPM receipts.

Because the amount of OID to be received over the life of the instrument is fixed, and thus the OID that had been previously accrued will be received eventually, the premature accruals may be addressed by a period of nonaccrual of OID. An alternative approach would be to reverse the premature accruals by recognizing Negative OID in the current period and then to accrue the OID again later.

In enacting the PAC formula, Congress expressed its intent that the rules implementing the PAC method would use a period of nonaccrual to correct possible premature accruals and would not accrue and recognize Negative OID.

2 H.R. Conf. Rep. No. 841, 99^th Cong., 2d Sess. II-239, 1986-3 (Vol. 4) C.B. 239. The IRS and Treasury understand that taxpayers generally comply with this intent not only for ordinary REMIC regular interests but also for REMIC IOs.

The quoted expression of Congressional intent occurred before the 1988 amendment permitting REMIC IOs. In the case of a REMIC regular interest that resembles a traditional debt instrument (such as the regular interests that existed before the 1988 amendment), a Negative OID computation is evidence that unexpectedly slow prepayments may have caused OID to accrue more rapidly than, in hindsight, it should have. In such a situation, disallowing Negative OID causes a timing issue. To the extent that OID has been overaccrued, the accrual period is extended until the computation for the extended accrual period produces a positive result. This future positive result of the computation has to occur eventually as principal on the debt instrument is repaid.

By contrast, in the case of a REMIC IO, a Negative OID computation may occur because unexpectedly rapid prepayments reduce the amount of OID that will ever be received or paid under the terms of the instrument. Rather than the right amount of OID being accrued too fast, the wrong amount has been accrued. In the case of a REMIC IO, therefore, the prohibition against Negative OID may result in denying the holder current recognition of an overall actual loss that will not be reversed in future periods and may only be realized upon the sale or maturity of the REMIC IO.

There is also a corresponding distortion to the net income or net loss of the REMIC (and thus to the income or net loss of the holder of the residual interest). Even if one or more holders of the REMIC IOs sell their interests and recognize losses that correct their own overaccrual of OID income, nothing corrects the REMIC’s overaccrual of OID deductions until the instrument is finally retired. This asymmetry may result in an understatement of the overall tax base attributable to income from mortgages held in REMICs (the total amount taxable to holders of REMIC regular interests and REMIC residual interests).

Section 1276(b)(3) provides that the accrual of market discount on a debt instrument the principal of which may be paid in installments shall be determined under regulations. Regulations have not yet been issued.

The legislative history of the Tax Reform Act of 1986, however, states that, until regulations are issued, if a debt instrument is issued with OID and the principal of the instrument may be paid in two or more installments, then holders of the instrument may elect to accrue market discount for the instrument either on a constant yield basis or in proportion to the OID accruals on the instrument. Under the latter method, the amount of market discount that accrues during an accrual period is determined by multiplying the total remaining amount of market discount on the instrument as of the beginning of the period by a fraction the numerator of which is the amount of OID for the period and the denominator of which is the total remaining OID at the beginning of the period.^[5] See 2 H.R. Conf. Rep. No. 841, 99th Cong., 2d Sess. II-842 (1986), 1986-3 (Vol. 4) C.B. 842. The IRS and Treasury understand that, under current practice, during any period for which the PAC method produces Negative OID, the numerator of the fraction is treated as zero, and no market discount is accrued. In some cases, this practice may uneconomically defer recognition of market discount.

If the rules in section 1272(a)(6) apply to a debt instrument (without regard to whether the instrument is issued with OID), this legislative history indicates that accruals of market discount on the instrument are to be determined using the same prepayment assumption as that used under section 1272(a)(6) (whether or not the taxpayer elects under section 1276(b)(2) to accrue market discount on a constant-yield basis). See id.

The IRS and Treasury are aware of several possible methods, discussed below, for addressing the foregoing problems.

Because of the range of instruments to which section 1272(a)(6) applies and the breadth of the new accounting methods about which comment is being requested, any new method might not necessarily be limited to REMIC IOs. For example, a new method might apply to interest-only strips from fixed investment mortgage trusts. In addition, a new method might apply to all instruments that provide for disproportionately high interest payments (as defined in §1.860G-1(b)(5)). Under this approach, the new rules would apply to REMIC regular interests whose issue price exceeds 125% of the specified principal amount and to similar non-REMIC interests.

Although the PAC method may sometimes fail to clearly reflect the income of the holder or the issuer of a REMIC IO, the method is not without merit. The method is specifically designed to deal with debt instruments that are subject to prepayments, like traditional REMIC regular interests. Under the PAC method, if loans are actually prepaid faster than expected, the projected future cash flows are adjusted immediately to more accurately reflect income. To a large extent, the problems arising from the application of the PAC method to REMIC IOs arise from the prohibition against taking Negative OID into account.

Because REMIC IOs did not exist when the 1986 legislative history discussing Negative OID was drafted, that discussion related to a Negative OID computation that would indicate that the affected taxpayers had accrued some OID too soon, rather than that they had accrued OID that would never be paid or received. Congress might have articulated a different intent concerning Negative OID if it had addressed the issue once REMIC IOs were permitted.

Accordingly, the IRS and Treasury are considering whether to propose a regulation that would follow the section 1272(a)(6) formula in the current PAC method, except that the regulation would specifically allow holders of regular interests to accrue Negative OID deductions and would require the REMIC (and thus the holder of the REMIC residual interest) to accrue and recognize income from Negative OID.

The considerations supporting recognition of Negative OID by initial purchasers may not apply with equal force to secondary-market purchasers. Secondary market prices are likely to reflect both prepayment history and revised expectations regarding future prepayments, with the result that the Negative OID deduction that might be appropriate for an initial purchaser may exceed any actual economic loss sustained by a particular secondary-market purchaser. The secondary-market purchaser’s depressed purchase price, however, is likely to result in a substantial amount of market discount. See section 1278(a)(2). The rules for accruing Negative OID and market discount will have to be coordinated to produce a net result that is economically sensible.

Accordingly, it may be appropriate either to develop explicit rules to effect this coordination or to limit recognition of Negative OID in the case of secondary-market purchasers. For example, recognition of accrued Negative OID might be limited to the aggregate of amounts that the secondary-market holder previously included in income as accrued OID or accrued market discount. However, in the case of a secondary-market holder who has suffered a real economic loss on a REMIC IO, such a limitation could uneconomically defer recognition of that loss.

Moreover, if a limitation on the allowance of Negative OID is applied to secondary-market purchasers, perhaps a similar limitation for initial purchasers will be needed to avoid disparate treatment of similarly situated holders (for example, initial purchasers and secondary-market purchasers that purchase shortly after original issuance at a price substantially the same as the issue price). However, such a limitation would also perpetuate many of the problems previously described.

Any rule recognizing Negative OID would have to deal with a variety of collateral consequences, such as adjustments to the instrument’s adjusted issue price and the holder’s basis in the instrument to reflect any deduction for Negative OID. Comments are requested concerning both the range of collateral consequences of recognizing Negative OID and the ways in which these consequences should be dealt with.

Another way to more clearly reflect the income of holders of REMIC IOs would be to issue regulations under section 166 (which concerns deductions for bad debts). These rules might both determine when (prior to realization) a holder has sustained an economic loss and also allow a deduction for the loss under section 166.^[6] Section 166(a) provides a deduction for any debt that becomes wholly or partially worthless during the taxable year. Indeed, some holders of REMIC IOs have claimed deductions for partial worthlessness under section 166(a)(2) and §1.166-3. The rules for determining worthlessness and partial worthlessness, however, were developed with reference to debts that become worthless or partially worthless because of the issuer’s anticipated failure ever to make required payments, not because certain contingencies (such as rapid prepayments) have reduced the amounts required to be paid. Thus the existing regulations under section 166 focus on whether a debt instrument is uncollectible and cannot be fully satisfied through foreclosure on collateral. See, for example, §§1.166-2 and 1.166-6. By contrast, the existence of Negative OID for a REMIC IO is evidence that the amounts contractually owed under the terms of the instrument are being reduced, not that the holder cannot collect whatever amounts are so owed.

Comments are invited regarding (1) whether, in the absence of any default by the issuer, the policy underlying the allowance of a deduction for worthlessness and partial worthlessness should be extended to a change in the amount that the issuer is required to pay, and (2) whether any rule allowing a deduction under section 166 can be extended to, or combined with, rules respecting corresponding income inclusions for REMICs and the timing of the inclusions.

The foregoing discussion attempts to provide a method for recognizing interest income and deduction from a REMIC IO by altering an existing method applicable to traditional debt instruments. Although it may be possible to alter an existing method, doing so is difficult because existing methods are designed to apply to debt and a REMIC IO is unlike most debt. Furthermore, as previously indicated, altering an existing method often leads to collateral problems that must be addressed. Therefore, an alternative method created especially for REMIC IOs, and similar instruments, may better reflect the income and deductions for these instruments.

Economically a holder of a REMIC IO (like other investors) has invested cash in an instrument and expects to receive cash flows from that investment. What is distinctive about a REMIC IO is that the amount and duration of the cash flows are unknown at the time of making the investment. Given the economics of the REMIC IO, a method for distinguishing between receipt of income and recovery of the amount originally invested could be based on the projected (but uncertain) cash flows under the instrument and not on the expectation of a fixed return. The following method attempts to achieve that objective.

First, the holder of a REMIC IO would include payments made on the REMIC IO in income as they are received. The holder would then be allowed an offset to any payments included in income for the period. The offset would be equal to an amount that bears the same ratio to the investment as the payments for the period bears to the total expected payments (based on a prepayment speed assumption). The total expected payments would be calculated each period taking into account both an updated prepayment-speed assumption and any payments made on the REMIC IO. For this purpose, the investment is the total investment cost (i.e., the issue price).

At the maturity of the IO, and perhaps at earlier times, a look-back regime may be appropriate to correct any under- or over-accrual of interest. See section 167(g)(2).

For an example of this method, see the appendix.

Comments are requested on two aspects of this IO-specific method in particular. First, can a variation of the method be applied to determine appropriate interest deductions for the REMIC? Second, in the typical REMIC IO, cash-flows start high and then decline to zero. For these instruments, the new method may clearly reflect income. One of the method’s weaknesses, however, is that, unlike OID accrual generally, the method does not accrue OID prior to the receipt of the cash representing the OID. An issue exists as to what regime should apply if the application of existing regulations to tiered structures produces REMIC IOs the cash flows on which are not expected to begin until well after the issue date.

Unlike initial purchasers, taxpayers who acquire REMIC regular interests subsequent to issue may have to take into account not merely accruals of OID but a combination of OID and market discount or a combination of OID and acquisition premium. As discussed above, the issues concerning OID accruals and the possible recognition of Negative OID require separate consideration with respect to secondary-market acquisitions.

The IRS and Treasury are considering alternative rules for the accrual of market discount attributable to REMIC IOs. One possible rule is to require accruals under a formula similar to the PAC method, including the use of a prepayment assumption and discount rate that remain static. However, instead of the projected prepayment speed and the projected yield to maturity being fixed as of the date on which the REMIC issues all of its regular interests, they would be fixed for a subsequently acquired REMIC IO at the time of the acquisition. Essentially a holder of a REMIC IO would apply the same methodology regardless of whether its acquisition was on the issue date (with the holder calculating OID based on estimates that were fixed on that date) or on a subsequent date (with the holder calculating market discount based on estimates that were fixed on the subsequent acquisition date).

If the amount of market discount is based on the revised issue price, as provided in section 1276(a)(2) and (4), the rules will need to integrate accrual of market discount (which will be specific to each holder) and accrual of OID (which will be the same for all holders). If the amount of market discount is based on remaining SRPM at the time of acquisition, accrual of the market discount will be a substitute for any OID accrual. In either case, a holder with any market discount will need substantial amounts of individualized data from the REMIC servicer. Comments are requested as to the REMIC servicer’s ability to provide the necessary individualized data.

It would be possible to revise the rules for accrual of market discount without adopting a rule recognizing Negative OID. As described above, however, if this recognition is permitted generally and is made available to secondary-market purchasers as well as initial purchasers, additional questions will be presented for secondary-market purchasers. These would include whether the amount of market discount should be redetermined and, if so, what the effect of that determination would be on collateral consequences of market discount such as the deferral of interest deductions under section 1277. One possibility would be to condition the recognition of Negative OID for secondary-market purchasers on an election by the holder to be taxable under the OID rules on both OID and market discount or premium. (See the election under §1.1272-3.)

The IRS and Treasury are aware that there are some REMIC IOs for which the prepayment speed that the servicer projected at the pricing date produces a projected negative yield. Arms-length investors do not voluntarily enter transactions with anticipated negative yields. Rather, such an investor may subjectively anticipate a different prepayment speed, or the investor may be “making a bet” on the occurrence of a prepayment scenario with a rate of return that more than compensates for its low probability of occurring. Mathematically, “discounting” a cash flow at a negative yield produces a present value that is greater than the sum of the future values of the cash flow. Unmodified application of the PAC method would therefore be unreasonable because it would require the holder to include amounts in income that are based on unrealistically high deemed present values of future cash flows. Comments are requested on whether the PAC method should be altered by requiring the use of a discount rate that is no less than an economically reasonable discount rate or whether some other adjustment would be more appropriate.

The IRS and Treasury request comments on the desirability of adopting special rules for taxing REMIC IOs, high-yield REMIC regular interests, and apparent negative-yield instruments, and whether those special rules should also be applied to other similar instruments (including how to identify such similar instruments). Comments and suggestions are also requested regarding possible approaches to what additional special rules may be desirable, including the possible recognition of Negative OID, the formulation of special guidelines for the application of section 166 to REMIC IOs and similar instruments, and the adoption of a new alternative method applicable to REMIC IOs and similar instruments.

Persons providing comments may want to consider, among other things, the following questions. Should recognition of Negative OID be limited to prior inclusions of OID, to prior inclusions of OID and market discount, or to some other amount? If any limit is imposed, should the limit apply to all holders or only to those who do not acquire their interests at original issue? If recognition of Negative OID by initial purchasers is limited to prior OID inclusions, should recognition of Negative OID be permitted for secondary-market purchasers to the extent of prior market discount inclusions as well as OID inclusions? If recognition of Negative OID is unlimited for initial purchasers, should it be limited for secondary-market purchasers? Should recognition of Negative OID for secondary-market purchasers result in a redetermination of a purchaser’s market discount and, if so, should the redetermination affect the application of the interest deferral provisions in section 1277? Alternatively, is the situation addressed adequately by currently recognizing both Negative OID and currently accruing market discount? Should recognition of Negative OID by secondary-market purchasers be conditioned on an election to treat all discount and premium on the instrument as OID?