Download Tax Loss Carryforwards and Corporate Tax Incentives: An Analysis of Effective Tax Rates and more Summaries Law in PDF only on Docsity!
This PDF is a selection from an out-of-print volume from the National
Bureau of Economic Research
Volume Title: The Effects of Taxation on Capital Accumulation
Volume Author/Editor: Martin Feldstein, ed.
Volume Publisher: University of Chicago Press
Volume ISBN: 0-226-24088-
Volume URL: http://www.nber.org/books/feld87-
Publication Date: 1987
Chapter Title: Tax Loss Carryforwards and Corporate Tax Incentives
Chapter Author: Alan J. Auerbach, James M. Poterba
Chapter URL: http://www.nber.org/chapters/c
Chapter pages in book: (p. 305 - 342)
10 Tax-Loss Carryforwards and
Corporate Tax Incentives
Alan J. Auerbach and James M. Poterba
The corporate income tax in the United States provides only limited tax relief to firms that report tax losses. Firms that have paid positive taxes during the three years prior to the loss year may "carry back" their losses and receive a tax refund, provided it does not exceed their total taxes in those three years. For some firms, however, current losses exceed potential carrybacks. This may happen when a firm experiences losses in several consecutive years, or when it incurs an especially large loss in a single year. Firms that exhaust their potential carrybacks must carry losses forward, using them to offset future taxable earnings. For these firms, the marginal tax rate on current earnings, as well as the value of tax deductions, depends critically upon when, and if, they regain their taxable status. Firms that anticipate persistent loss car- ryforwards will in effect face very low marginal tax rates. Imperfect loss-offset provisions may substantially alter the incentive effects of the corporate income tax. Two features of the tax, the in- centive to undertake new investment and the incentive to use debt as opposed to equity finance, are particularly sensitive because loss-car- ryforward firms may be unable to claim the benefits of depreciation or of interest tax shields.
Alan J. Auerbach is professor of economics at the University of Pennsylvania and a research associate at the National Bureau of Economic Research. James M. Poterba is associate professor of economics at the Massachusetts Institute of Technology and a research associate at the National Bureau of Economic Research. This research is part of the NBER Program on Taxation. We are indebted to Lars Bespolka, Sandi Fine, William Gentry, and Julie Harrold for help in gathering data, and to Kevin Hassett for excellent research assistance. David Bradford, Paul Healy, and Lawrence Summers made valuable comments on an earlier draft. We thank the NBER, NSF, and the Sloan Foundation for financial support.
305
307 Tax-Loss Carryforwards and Corporate Tax Incentives
tax on marginal profits. The stock of tax-loss carryforwards in the United Kingdom was nearly three times as large as the annual revenue yield of the corporation tax. This paper presents new evidence on the importance of tax-loss carryforwards in the United States. It uses a new data set gathered from corporate annual reports and 10-K forms to investigate the in- cidence of loss carryforwards, and then examines how effective tax rates on different assets are affected by loss-offset constraints. The most important finding is that tax-loss carryforwards are highly per- sistent and significantly affect investment incentives for some firms. Nearly 15% of the firms in our sample had tax-loss carryforwards in 1984, and the fraction is much higher in some industries. Analyzing the effect of the corporate income tax on tax-loss firms is therefore essential to understanding investment and financing incentives in these industries. We estimate the persistence of loss carryforwards and use the results to calculate effective tax rates on new investments in struc- tures and industrial equipment for both currently taxable and loss- carryforward firms. We find that the presence of a tax-loss carryforward has a dramatic effect on a firm's incentive to invest in equipment but relatively little impact on the incentive to invest in structures. The paper is divided into five sections. The first outlines the tax rules governing loss carryforwards and carrybacks. It also explains the dif- ficulties that arise in using standard data sources to measure tax-loss carryforwards, and describes our new data set. The second section presents our basic findings on the importance of firms with tax-loss carryforwards and examines the persistence of loss carryforwards for the firms that experience them. The third section outlines how loss- offset constraints alter the effective tax rates on various assets and describes our numerical procedures. The fourth section presents our calculations of the effective tax rates on plant and equipment invest- ment for both currently taxable and loss-carryforward firms. A con- cluding section discusses the implications of our results for understand- ing the allocative effects of the corporate income tax and suggests a number of directions for future work.
10.1 The Definition and Measurement of Tax-Loss Carryforwards Loss-offset constraints restrict a firm's ability to obtain tax refunds when it generates negative taxable profits. A firm that realizes a tax loss may carry the loss back against tax payments in the previous three years, provided it does not claim current refunds in excess of total tax payments in those years. Firms that have exhausted their carrybacks
308 Alan J. Auerbach and James M. Poterba
may carry unused losses forward for a maximum of fifteen years, after which the losses expire and can no longer be used to reduce tax liability. Prior to 1981, loss carryforwards expired in five years. For firms with loss carryforwards, an additional dollar of taxable income has no effect on current tax liability. The marginal tax burden on an additional dollar of taxable earnings depends upon when the firm becomes taxable again in the future. It is important to distinguish between firms with loss carryforwards and "firms that pay no taxes."^2 A firm with a tax-loss carryforward in a given year pays no tax, but it may receive a refund if it can carry part of the loss back against previous tax payments. A marginal change in the firm's taxable earnings, however, will have no effect on its current tax liability. Its current marginal tax rate is zero, although if it expects to exhaust its loss carryforwards in the near future it will face an effective marginal tax rate that differs from the statutory tax rate only by the price of an interest-free loan for the duration of its remaining tax-loss period. Not all firms with negative current tax payments have loss carryfor- wards, however. Some firms that are carrying back current losses may not have exhausted their carryback potential. For these firms, the mar- ginal tax rate on additional income is the statutory tax rate, because an additional dollar of earned income will reduce the amount of their carryback. These firms face the statutory marginal rate even though their current tax payments are negative. Loss carryforwards are not the only factor that may cause a firm's marginal tax rate to differ from the statutory rate. Cordes and Sheffrin (1983) explain how constraints on the use of tax credits and the cor- porate minimum tax also affect the distribution of marginal corporate tax rates.^3 Unfortunately, publicly available information is not detailed enough to enable us to measure the marginal tax rates facing individual corporations. To calculate a firm's carryback potential, we would need information on its current tax credits, its credit and loss carryforwards, and even its previous tax payments. These data can only be obtained from a firm's past and present tax returns, which are confidential.^4 One type of tax data which can be gathered from published sources is the identity of firms with tax-loss carryforwards. Corporate annual reports and 10-K filings typically contain some information on carryforwards, so we focus on this source of variation in marginal corporate tax rates. Data limitations prevent us from assessing the significance of firms with tax-credit carryforwards. Most of the firms that we identified as having tax-loss carryforwards also reported credit carryforwards. There may be other firms, however, with credit carryforwards but no loss carryforwards; Cordes and Sheffrin (1983) suggest that these credit- carryforward firms account for a substantial fraction of the firms facing
310 Alan J. Auerbach and James M. Poterba
1981 and 1984. The total market value of the firms in our sample is roughly three-quarters of the total market value of the nonfinancial corporate sector. There are several potential biases in our data sample that should be recognized at the outset. First, COMPUSTAT does not include all of the corporations which file tax returns; there were over three million such firms in 1982! The firms on COMPUSTAT are large, publicly traded firms. If losses tend to be more prevalent among smaller or start- up firms, then we may understate the number of firms with tax-loss carryforwards. Second, the data set follows COMPUSTAT in including only firms that were active in 1984. Some corporations that encountered tax-loss carryforwards in earlier years may either have been taken over or gone bankrupt, and the end-of-sample sampling rule imparts a clear selection bias. This may cause us to understate the number of loss- carryforward firms in 1981 through 1983, although the bias is likely to be small given the relatively low rate of both bankruptcy and takeover for firms on the COMPUSTAT tape. A third source of bias arises be- cause not all firms with losses may report them. Firms are required to report loss carryforwards only if they are "material"; since some firms with small carryforwards may not appear as carryforward firms on COMPUSTAT, we may understate their importance. A final problem with loss-carryforward data gathered from annual reports and 10-K filings is the divergence between the divisions of the firm included on its consolidated tax return and those included on the financial statements. For example, as Dworin (1985) explains, some firms do not include their finance subsidiaries in their financial state- ments although for tax purposes these subsidiaries are consolidated with the parent corporation.^6 We may therefore classify a parent firm as having a tax-loss carryforward even though the total taxpaying entity has no carryforward. This problem is impossible to overcome when using data published in annual reports.^7 It is difficult to evaluate the impact of these biases. One simple check involves comparison of our aggregate estimate of loss carry- forwards with aggregate IRS data. A lower bound for loss carryfor- wards by nonfinancial corporations can be calculated as the current deficit reported by firms with current losses, less total losses carried back. This lower bound was $57.1 billion in 1981 and $75.2 billion in 1982, roughly five times as large as our aggregate estimates. While the source of this discrepancy is unclear, we believe it is due primarily to small firms not included in our data and some large firms which do not report loss carryforwards on their accounting statements. This is likely to contaminate our estimates of the incidence of tax losses much more than our estimates of transition probabilities and effective tax rates.
311 Tax-Loss Carryforwards and Corporate Tax Incentives
10.2 The Importance and Persistence of Loss Carryforwards This section uses our annual-report data set for the post-1981 period, as well as accounting-purpose loss-carryforward data available on COMPUSTAT for a longer sample period, to explore the economic significance of tax-loss carryforwards. We ask how many firms have carryforwards, and then examine the persistence of these losses.
10.2.1 The Importance of Loss-Carryforward Firms Table 10.1 presents summary evidence on the importance of firms with tax-loss carryforwards in the years since 1981. It considers the total population of nonfinancial firms, as well as some particular in- dustries. The table shows that about 15% of all firms are in the loss- carryforward regime. We also weighted firms by their 1984 net book assets, and found that 5.9% of all assets were held by loss-carryforward firms. Table 10.1 also shows that there is substantial concentration of car- ryforward firms in some industries. In the oil industry (SIC codes 1311 and 2911), for example, nearly a quarter of the firms (accounting for 2% of the common stock) had loss carryforwards in 1984. In 1982 and 1983, 40% of the firms (accounting for about 10% of the value of out- standing common stock) in motor vehicles and car bodies reported tax loss carryforwards. In the steel industry, the findings suggest a third of the firms have losses, and, they account for half of the industry's outstanding equity value. Finally, for airlines we also find a high in-
Table 10.
Industry All Nonfinancial corporations Oil (SIC 1311 & 2911) Autos (SIC 3711) Steel (SIC 3310) Airlines (SIC 4511)
Tax-Loss Carryforward Firms, 1981-
Sample Firms (N)
1,
69
7
25
20
% of Firms with
1981
1982
Loss Carryforwards
1983
1984
Note: All calculations are based on the authors' data set, which is described in the text.
313 Tax-Loss Carryforwards and Corporate Tax Incentives
Table 10.3 The Tax-Loss
Firm
- Burlington Northern
- Chrysler Corporation
- U.S. Steel
- General Dynamics
- Syntex Corporation
- Bethlehem Steel
- Penn Central
- LTV Corporation
- IC Industries
- Asarco Inc.
- Inland Steel
- Phelps Dodge Corp.
- Storer Communication
- Clark Equipment
- Datapoint Corporation
- Alleghany Corporation
- American Motors
- Turner Broadcasting
- Best Products
- International Harvester
Carryforward Top Twenty, 1983 Equity Value ($ million) 3,677. 3,365. 3,178. 3,064. 1,799. 1,318. 1,164. 1,017.
560,
Tax-Loss Carryforward
1,600. 1,200.
2,097.
unknown
Note: Firms are ranked by outstanding equity value at the end of 1983. The Alleghany Corporation reported the presence of tax-purpose loss carryforwards, but it did not report their amount.
tions and Turner Broadcasting, for example) appear because substantial investment programs generated depreciation allowances significantly greater than taxable earnings from current operations.
10.2.2 The Persistence of Tax-Loss Carryforwards The extent to which the restricted loss-offset provisions in the cor- poration tax affect investment and financing incentives depends upon the duration of nontaxable spells. If firms with loss carryforwards can expect to recover their taxable status within a year or two, then the absence of loss-offset provisions will have relatively little effect on incentives. If firms with carryforwards tend to be constrained for many years, however, then they may face incentive effects which are sub- stantially different from those of taxable firms. We adopt two different approaches to analyzing the persistence of tax-loss carryforwards. First, we use our data for the last four years to fit simple Markov models for transitions into and out of the loss- carryforward state. This provides the basis for our analysis of effective tax rates in later sections, but it is limited by the fact that our data span a period of only four years. Moreover, since these years include
314 Alan J. Auerbach and James M. Poterba
a very deep recession, transition probabilities from the recent period may be unrepresentative of those confronting firms over a longer ho- rizon. To obtain information on long-term persistence of loss carry- forwards, we therefore perform the same calculations using a second data source, the partially contaminated accounting loss-carryforward data from COMPUSTAT, for the period 1968-84. These data are also used to construct empirical distributions of the number of firms with losses that persisted for one year, two years, three years, etc. Although the differences between tax and book loss-carryforwards make these tabulations an imperfect source of information on persistence, they do permit us to compare the recent experience with that in prior years. Table 10.4 reports summary statistics, based on our post-1981 data sample, for transitions into and out of loss-carryforward status. The top panel shows probabilities based on the first-order Markov as- sumption, i.e., treating a firm's current status as containing all relevant information about its transition prospects. These estimates show that for the 1983-84 period, the probability of a firm without a loss car- ryforward in period t experiencing one in period t + 1 is .026. For a
Table 10.4 Tax-Status Transition Probabilities
Previous State No loss carryforward Loss carryforward
Previous State
First-Order Markov Model Probability of Moving to State of No Loss Carryforward . . Second-Order Markov Model State in Period / + 1: No Loss Carryforward
Loss Carryforward . .
Loss Carryforward
No loss carryforward (/ - 1) No loss carryforward (i) .977. No loss carryforward (t - 1) Loss carryforward (t) .099. Loss carryforward (t — 1) No loss carryforward (t) .680. Loss carryforward (t - 1) Loss carryforward (?) .083. Note: All calculations are based on the authors' data set, described in the text, which yields 2,849 firm years of data. The estimates are for transitions observed in 1983 and
316 Alan J. Auerbach and James M. Poterba
The most significant drawback of our post-1981 data is that we cannot examine the long-term persistence of tax-loss carryforwards. We can address this issue using the data on accounting tax-loss carryforwards drawn from the COMPUSTAT tape, however. To evaluate the potential biases associated with these data, we compared their second-order Markov transition probabilities for the 1983-84 period with those ob- tained from our annual reports data. The probability that a firm with two previous years of loss carryforward would remain in the loss- carryforward state was .928 in the COMPUSTAT data, compared with .917 in the annual reports data. The probability of remaining carryfor- ward-free after two years of being currently taxable was .966 rather than .977. The COMPUSTAT data therefore probably overstate the persistence of tax losses because the chances of experiencing a tax loss in a given year, for firms that have as well as those that have not experienced them in the past, are higher in these data. This is consistent with our finding that financial-purpose loss carryforwards, because they include asset write-offs and other losses, may appear more significant than the comparable tax-purpose losses. Nonetheless, the close agree- ment between the COMPUSTAT and annual report-based data suggest that valuable information can be obtained by studying COMPUSTAT transition probabilities over time.
Table 10.5 presents the pattern of transition probabilities from the COMPUSTAT sample. It reports our estimates of the four basic tran- sition rates for each year between 1968 and 1984, as well as the prob- abilities for the full sample period and two subsamples. Two conclu- sions emerge. First, the probability that a firm with loss carryforwards in the two previous years will experience another year of tax loss increased substantially in 1981. We denote this probability as pLLL, where the subscripts refer to the tax status in periods t—2, t—1, and t, respectively. The subscript takes the value L for loss carryforward, and Tfor currently taxable. The probability pLLL, which never exceeded .90 in the years prior to 1980 and which was frequently below .80, averages .928 since 1981. The probability of remaining in the loss po- sition rises between 1981 and 1983, then declines in 1984, reflecting in part changing business-cycle conditions. There is also a discontinuity in 1981 in the probability that a firm which has experienced a taxable year followed by a loss year will remain in the loss state, pTLL in our notation. From a pre-1981 average of .787, pTLL changes to a post- value of .909. The table also shows a substantial post-1981 increase in the proba- bility that a taxable firm will experience a loss carryforward. From. before 1981, pTLL increased by nearly 40% to .034. There is a smaller increase in the chance that a firm which has experienced a loss- carryforward year followed by a taxable year will reenter loss status.
317 Tax-Loss Carryforwards and Corporate Tax Incentives
Table 10.
Year
1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 Means 1968- 1968- 1981-
Tax-Status Transition Compustat Sample
PTTL . . . . . . . . . . . . . . . . .
. . .
Probabilities Estimated from
PTLL . . . . . . . . . . . . . . . . .
. . .
PLTL . . . . . . . . . . . . . . . . .
. . .
PLLL . . . . . . . . . . . . . . . . .
. . .
Note: Each column reports the transition probabilities calculated from the COMPUSTAT data set of financial purpose tax-loss carryforwards.
These movements in the Markov transition probabilities correspond to changes in the steady-state distribution of firms with respect to tax status. The pre-1981 probabilities imply that in the steady state 10.9% of all firms have tax-loss carryforwards. The comparable steady-state value for the post-1981 probabilities is 33.5%, a striking increase.^9 This undoubtedly overstates the long-run effect of the 1981 tax reform, since it is difficult to disentangle the effects of the 1981 tax reform from the post-1981 recession. Our estimates of second-order Markov transition rates are incom- plete because they shed no light on the behavior of firms that have experienced losses for many periods. One way to study this long-term persistence is by calculating the probability that a firm with a loss in a particular year will experience losses for one more year, two more years, etc. Table 10.6 presents calculations of these long-term transition rates from the COMPUSTAT data sample for the period 1974-83.^10 The table shows that a significant fraction of firms that experience tax losses in a given year will continue to have such losses for at least four more years. The probability of this much persistence has also risen
319 Tax-Loss Carryforwards and Corporate Tax Incentives
over time, from .32 in 1974 to .50 in 1980, the last year for which it is possible to calculate the four-year-later transition rate. The estimates presented in this section are at best a rough charac- terization of the transition probabilities confronting firms. In the next two sections, we calculate effective tax rates for hypothetical firms whose movements into and out of the tax-loss state are given by our estimates. This analysis, which is primarily illustrative, demonstrates the potentially important effect of loss-offset provisions on effective tax rates.
10.3 The Incentive to Invest in the Presence of Tax Losses
Unlike more direct forms of investment subsidy, tax-loss carryfor- wards are not likely to have uniform effects on different firms and asset types. A firm with substantial unused tax benefits may appropriately view itself as temporarily "tax exempt," while a firm with a small carryforward which it expects to utilize during the next year regardless of its current decisions should take no account of it in making invest- ment decisions. The differences across asset types stem from differ- ences in the timing of taxable income. Many assets, such as equipment under current law, may be expected to generate negative taxable income in their initial years. If a firm has unused tax benefits when the project begins, this will decrease the asset's after-tax income. Since the ac- cruing losses must be carried forward until the firm achieves a positive tax liability, some investments may actually be discouraged by the presence of unused tax benefits.^11 This section describes our meth- odology for quantifying these incentive effects. There are a number of approaches to measuring the impact of tax- law asymmetries on investment incentives. Ideally, one would specify a dynamic model of firm value maximization in which risky investment would be affected by, and in turn would affect, the magnitude of unused tax benefits present at different dates in different states of nature. This problem is complicated by the joint endogeneity of investment and the firm's tax status.^12 To make the problem more tractable, if less general, one may restrict the endogeneity of either the firm's investment be- havior or its tax status. The former approach is taken by Majd and Myers (1985, 1986). They value the tax payments associated with risky projects, taking account of the project's impact on the firm's future tax status. Their approach highlights the interaction between the project's risk and the risk of other random changes in the firm's tax status, but it ignores potential changes in corporate behavior which may result from variation in tax status. An alternative approach, the one taken here, assumes that the prob- ability distribution of future tax status is determined by the firm's
320 Alan J. Auerbach and James M. Poterba
history alone. This can be interpreted as treating the marginal invest- ment project as small relative to the rest of the firm, so that the firm's tax status is determined by the stochastic returns on its prior invest- ments. The assumption that the probability distribution of tax status is invariant with respect to marginal decisions is justified if this distri- bution is the direct result of firm optimization decisions. This inter- pretation highlights a shortcoming of this approach, however, in that it is necessarily restricted to partial equilibrium analysis of changes in tax rules or other components of the economic environment. We cannot predict how a change in tax regime would affect the incentive to invest, since it could both change the firm's statutory tax benefits holding its investment decisions fixed and alter the probability distribution of its future tax status.
10.3.1 Effective Tax Rates with Loss-Offset Limits The summary statistic used throughout our analysis is the effective tax rate on a marginal investment project, calculated as the percentage difference between the internal rates of return on expected cash flows before and after tax. We assume that these marginal investments are inherently risk-free, and that the only source of uncertainty is the time profile of future tax payments. We designate the project's before-tax rate of return as p, which is set equal to .06 in all calculations. The asset depreciates at a constant rate, 8, so an investment made at the beginning of period O yields a gross return in period t of (p + 8)(1 — 8)'"' per dollar of initial investment. We assume that the investment tax credit and the first half-year depreciation allowance accrue at date O. Thus, the firm's project-specific accrued tax liability (Bt) at date t is:
- 8)'-» - A d " 0'] t> 0)
[ t = 0 where T is the corporate tax rate, k the investment tax credit, Dt is the nominal date t depreciation allowance, and / is the inflation rate. These expressions describe an equity-financed project; with debt finance, in- terest deductions would also enter the formula for Bt. Under a symmetric tax system with full loss offset, equation (1) would describe actual tax payments. The project's after-tax internal rate of return, r, would be defined implicitly by the expression:
(2) -Bo + 2, (1 + r)-'[(p + 8)(1 - 8)'-> - Bt]
= 2 (1 + r)-'[(p + 8)(1 - 8)'-^1 ] - T(r) = 1 _t=_
322 Alan J. Auerbach and James M. Poterba
ability to subsequently carry back future tax losses. We will use vts to denote the shadow value of additional tax payments in year t + s, viewed from the perspective of year t. With these complications, the present expected value of tax payments, T, becomes
f
The term in brackets is the expected present value of a one-dollar tax accrual in period t. Equation (5) defines T{r), which can in turn be substituted into (2) to compute effective tax rates based on expected tax payments.^14
10.3.2 Computing the Time Distribution of Tax Payments To implement these effective tax-rate calculations, we need the prob- ability distribution of tax payment dates for each accrued tax liability. We compute these distributions from the second-order transition prob- abilities in tables 10.5 and 10.6. These calculations are facilitated if we introduce new variables corresponding to the probability that a firm is in each of the four possible states, TT, TL, LT, and LL, in a given period. We use q\j to represent these probabilities. For a firm which is known to have a tax loss in the period before, and period of, a new investment, q°LL = 1 and q°LT = q\L = q\j = 0. In general, the prob- ability that a firm will be taxable in period one is (6) TX\ = qiT + qxrr =(q°LLPLLT + QTLPTLT) + (irrPTTT +
The second part of the equation shows how the year-one probabilities can be built up recursively from the starting conditions, the qfj, and the transition probabilities that were discussed in the last section. Sim- ilar calculations permit us to derive the probabilities of finding the firm in other tax states in period one. The probability that the firm will carry its taxes from the investment year forward exactly one period is TT£J = qriPrLT + QILPLLT- Parallel calculations show that the unconditional probability of carrying taxes forward for two years or more is TT^X = qriPTLL + QLLPLLL, and the probability of carrying a loss forward for exactly two years is TT^LT = PLLT^LL- Probabilities corresponding to longer carryforwards can also be calculated recursively. While these calculations have considered the distribution of tax ac- cruals from period zero, it is straightforward to apply this approach to compute the distribution of tax payments corresponding to accruals later in the project's life. The initial conditions are just the {q$ cor- responding to the firm's probabilities of being in each tax state at the beginning of period t. These can be calculated recursively from the
323 Tax-Loss Carryforwards and Corporate Tax Incentives
$ and the transition probabilities as in equation (6). As we iterate forward, however, the firm's tax status in year zero becomes less im- portant as a predictor of its period / status, and the TT vector converges to a steady-state value. In practice, we truncate our calculated IT vector after twenty elements and let the twenty-first element capture all of the remaining probability.^15 We incorporate loss carryforwards by assuming that all deferred tax payments may be carried forward N years, where N is the statutory maximum for carrying losses forward.^16 Incorporating carrybacks is more complicated, since the opportunity to carry losses back has the effect of making every tax payment potentially valuable in facilitating the accelerated deduction of future tax losses. This imparts a shadow value to tax payments; we calculate this shadow value in two stages. First, we compute a distribution of expected tax payments under the assumption that there are no carrybacks. Then, we account for car- rybacks by reducing each dollar of estimated tax payments by a shadow value which depends upon the firm's current tax status and the esti- mated transition probabilities. The calculation of the carryback shadow value is described in greater detail in the Appendix.
10.3.3 Qualifications All of the analysis in this section presumes that the effective tax rates which apply to a firm's investment choices are a function of its own tax status. This need not be the case. Leasing arrangements are one example of a channel through which the effective tax rates of the firms using and owning an asset can be separated. These institutions have been particularly popular in some of the industries with a signif- icant incidence of tax losses, such as airlines. It is important to realize, however, that although leasing can reduce the present value of tax payments for a constrained firm, its impact on the firm's incentive to invest in new capital is less clear. A firm that has a loss carryforward would be better off if it could utilize this tax benefit right away, since the tax benefit loses value over time and may expire. Given that the firm cannot use this tax benefit, however, it may be encouraged to invest more, since additional taxable income generated by new in- vestment will enable it to offset part of the loss carryforward. A second limitation inherent to our analysis is its exclusive focus on tax policies. For some of the large firms that have tax-loss carryfor- wards, taxation is just one of the many ways in which the government and the firm interact. Examples of other policies that clearly affect the performance of the firms and the welfare of their shareholders include direct loan guarantees, regulation (especially for airlines and railroads), tariff policy, and in some cases government purchasing policy. Ana-
