Fifo i Lifo

LIFO VERSUS FIFO:

UPDATING WHAT WE HAVE LEARNED

Nicole Thorne Jenkins Doctoral Student in Accounting

Morton Pincus

Associate Professor of Accounting

College of Business Administration The University of Iowa

108 PBAB Iowa City, IA 52242-1000 U.S.A.

319/335-0915 FAX 319/335-1956

[email protected]

September 1998 (version 1.2)

1

LIFO VERSUS FIFO: UPDATING WHAT WE HAVE LEARNED

1.0 INTRODUCTION

The statutory mandate in U.S. tax law that firms using the last-in first-out (LIFO) inventory costing

method for tax purposes must also use LIFO for financial reporting purposes makes inventory accounting

an especially interesting research and teaching topic. The constraint on managerial discretion imposed by

tax--book conformity highlights the tension that can exist between tax minimization, on the one hand, and

achieving financial reporting objectives, on the other hand. Trade-offs between tax minimization and other

objectives are a major theme in the Scholes and Wolfson [1992] framework for examining taxation in the

context of business strategy. Moreover, a researcher typically can quantify the cash flow impact a firm

derives by using (or forgoes by not using) LIFO or FIFO, whereas quantifying the cash flow effects of

other financial accounting choices is more problematic.1 This is because the cash flow effects of other

accounting choices typically are indirect (e.g., through contracting costs).

It thus is not surprising that research in the LIFO/FIFO area has a long history. Two reviews of

LIFO research have previously been published in the Journal of Accounting Literature. The first

appeared in the initial issue of the Journal and examined LIFO-related research as part of a more general

review of capital market assessments of alternative accounting methods [Ricks, 1982a]. The second

review, published just six year later, focused exclusively on LIFO. It surveyed three main research

streams: the effect of LIFO adoptions on security prices; the determinants of the inventory accounting

choice; and the appropriateness of capital market participants’ interpretations of LIFO- and FIFO-based

earnings [Lindahl, Emby, and Ashton, 1988].2 Notwithstanding the large number of LIFO studies, a

mixed (albeit rich) set of results characterizes much of the research into questions such as the impact of

LIFO adoption on stock prices and the effect of LIFO and FIFO on reported earnings and firm valuation.

As editor of The Accounting Review, Abdel-khalik introduced a “Forum on LIFO Choice of Inventory

1 Since the early 1970s the S.E.C. has required firms using LIFO to also disclose the excess of current cost over the carrying value of their LIFO inventory. The AICPA’s AuSEC (1984) has recommended disclosure of the “LIFO Reserve” for companies subject to S.E.C. regulations since the mid-1980s. 2 Previous reviews of LIFO/FIFO research also include portions of Lev and Ohlson [1982], on market-based research, and Gonedes and Dopuch [1974], on alternative means of assessing the effects or desirability of accounting methods, including capital market research and behavioral laboratory experiments. A recent paper by Thibodeau-Morin and Patton [1998] surveys the entire body of LIFO/FIFO research to identify major research questions and methodological approaches. In contrast, we provide an update of the literature by detailing, analyzing, and synthesizing major research developments occurring since Lindahl et al. [1988]. Our review and that of Thibodeau-Morin and Patton were written independently.

2

Valuation” by expressing a generally felt frustration about research in the area. He said researchers

“continue to be relatively uninformed about” a number of LIFO-related issues “and know little about the

real reasons that many firms do not switch to LIFO when it appears that they would benefit by positive tax

savings” [1992, 319]. He was referring to the “LIFO puzzle,” which continues to be invoked in some

working papers and textbooks.

However, significant strides have recently been made by researchers such that our understanding

of LIFO’s impact on stock prices and of the LIFO puzzle is very different than it was just a few years ago.

Research findings now strongly suggest that (a) there is not really much of a LIFO puzzle, (b) the capital

market reaction to LIFO adoptions is more complex than the analysis of the traditional tax benefits versus

functional fixation hypotheses suggests, and may not yield a meaningful assessment of the impact of LIFO

on share prices, and (c) the implications of LIFO use for the security price--earnings relation are less

straight-forward than what one might expect.

In this review, we survey the research that has helped to update our beliefs about the LIFO/FIFO

choice and its impact. We organize the review into the sections that represent major research thrusts (i.e.,

questions) over the past 10-15 years. The research thrusts are capital market reactions to LIFO

adoptions, theoretical and empirical explorations of non-tax explanations of the LIFO puzzle, the Kang

[1993] model and subsequent capital market research on LIFO adoptions, share price effects associated

with LIFO’s entry into the tax law, the impact of inventory methods on the price-earnings relation and firm

value, and income smoothing. We discuss the research as it evolved in roughly chronological order within

research areas as well as across areas. We begin several sections by summarizing research previously

examined in more detail in earlier LIFO reviews. Note: the italicized parts of the text contain the

more technical or detailed aspects of the papers we discuss. These can be skipped without a loss of

continuity.

2.0 HOW DOES THE CAPITAL MARKET RESPOND TO LIFO ADOPTIONS (RESEACH

PRIOR TO KANG [1993])?

LIFO is the tax minimizing inventory method choice under the following conditions. A firm must

(1) be a taxpayer in one of the few countries allowing the use of LIFO for tax purposes, (2) face a positive

marginal income tax rate, (3) expect increasing inventory input costs, (4) expect non-declining inventory

levels, and (5) expect the present value of tax benefits to exceed the present value of LIFO adoption (i.e.,

3

implementation and sustaining) costs. Unfortunately, researchers cannot observe managers’ beliefs about

items (2)-(5). Instead, most (pre-1993 capital market) research into LIFO implicitly assumes that high (or

increasing) rates of price inflation in the economy are sufficient to expect that firms will adopt LIFO and

save (i.e., defer) taxes on inventory profits. Moreover, when firms actually adopt LIFO, investors are

expected to bid up these firms’ share prices to reflect the present value of the expected tax benefits, even

though the earnings reported by these firms will fall. As Lindahl et al. [1988] detail, however, the empirical

evidence of early capital market assessments of LIFO adoptions is ambiguous.

For example, Sunder [1973] observes an increase in LIFO adopters’ cumulative excess security

returns in the year of adoption, and Biddle and Lindahl [1982] find a positive association between excess

adoption year stock returns and firm-specific estimates of LIFO tax savings after controlling for

unexpected earnings. However, Biddle and Lindahl’s results are sensitive to the market model estimation

period and also indicate a negative security return-tax benefits relation in certain subperiods. Abdel-khalik

and McKeown [1978], Brown [1980], and Ricks [1982b] examine security returns over intervals that

likely include firms’ LIFO adoption disclosures. These studies compare excess returns for adoption firms

with those for non-adopting firms during the same interval, and each study finds that the market reacts

negatively to LIFO adoptions.

In an effort to clear up the conflicting results in the literature, Aharony and Bar-Yosef [1987]

examine a question similar to those asked in previous studies but use a different approach. They use the

criterion of second-degree stochastic dominance (SSD) to investigate the impact of LIFO adoption on

shareholders. The model differs from those used in previous event studies in that it does not require strong

assumptions about the underlying market model, nor does it require the estimation of market model

parameters.

While CAPM and market model approaches focus on shareholder wealth effects, stochastic

dominance emphasizes shareholder welfare (or expected utility). Under the SSD criterion, if A and

B are cumulative returns distributions and A intersects B from below, and if the cumulative

difference between B and A is non-negative for all possible outcomes, then an investor will prefer A

to B. No assumptions need be made about the distribution function of possible outcomes or the

investor’s utility function (beyond risk aversion and that it be strictly concave). In contrast, CAPM

requires that the investor have a quadratic utility function or that the distribution of possible

4

outcomes be normal; the latter requirement is also an attribute of the market model. Aharony and

Bar-Yosef also use the SSDR rule, which incorporates a risk-free interest rate into the SSD-

criterion, since SSDR has a greater ability to detect dominance than SSD (i.e., it has fewer non-

dominance cases but deletes firms having return distribution means below the risk-free rate).

Aharony and Bar-Yosef use the 233 matched pair sample of LIFO adopters and non-adopters in

Ricks [1982b]. They find that 26-week cumulative return distributions increase after a LIFO adoption

relative to before adoption for both the LIFO and control groups. They conclude that the change in

stockholder wealth is due to factors other than changing inventory accounting methods, and that the LIFO

adoption does not affect shareholders’ expected utility. Like most LIFO adoption studies, their analysis is

restricted to a sample of 1974 firms.3 Because firms self-select inventory accounting methods, similar

results for LIFO and non-LIFO firms in a given year does not necessarily mean that inventory accounting

methods do not affect security returns. Hence, they are unable to clear up the inconclusive results from the

earlier studies.

Biddle and Ricks [1988] joined forces in the hope of discovering what explains the reliably

negative excess returns observed near the earnings disclosure dates for the 1974 adopting firms. They

examine the association between excess returns and earnings forecast errors. They hypothesize that the

negative excess returns observed in Ricks [1982b] are associated with, and possibly due to, analysts’

systematic overestimates of the earnings (and thus tax) effects of the 1974 LIFO adoptions. They find

evidence that Standard & Poor’s analysts’ earnings forecast errors for the sample are negative, on

average, and significantly correlated with both the excess returns observed near the annual earnings

announcement dates and the earnings effects of the LIFO adoptions. Their results suggest that the

negative returns associated with the 1974 LIFO adoption and earnings disclosures are due to a systematic

overestimation of earnings by analysts, and thus by investors. This could explain why the results for 1974

LIFO adopters appear to be anomalous.4

Hand [1995] uses a large sample of fiscal 1974-75 LIFO adoptions (n = 754) to reexamine three

3 Pincus and Wasley [1994] find LIFO adoptions are the most common accounting change over the 1969-88 period, and, by far, most LIFO adoptions occur during 1974. 4 Elliott and Philbrick [1991] examine a sample of voluntary and mandatory accounting changes made over the 1976-84 period to investigate accuracy and dispersion of I/B/E/S analyst earnings forecasts in the year of change. LIFO adoptions dominate their sample of voluntary accounting changes (132 of 285) and have the largest average negative effects on earnings. While most of their results are aggregated across all voluntary changes, Elliott and Philbrick report no evidence of a significant bias in analyst forecasts in

5

seemingly anomalous results in Biddle and Ricks [1988]. First, Hand documents that the result in Biddle

and Ricks of mean negative excess returns accompanying LIFO adoption announcements is restricted to a

negative reaction associated with NYSE/ASE listed firms; no such negative effect is present for firms

traded over the counter (OTC). However, he is unable to explain why these differing reactions occur and

thus identifies a new puzzling result. Second, he finds that the evidence in Biddle and Ricks of a negative

reaction on the annual earnings announcement date for firms that had previously disclosed they would

adopt LIFO is explained by there being uncertainty about whether analysts’ earnings forecasts were based

on a LIFO or a FIFO basis. Third, Hand [1995] shows that after controlling for forecast age, there is no

bias present in Standard and Poor’s analysts’ forecasts of the earnings effects of 1974 LIFO adoptions.

This is consistent with analysts updating forecasts in a sophisticated manner.

3.0 ARE THERE ALTERNATIVE EXPLANATIONS FOR THE LIFO PUZZLE?

The inability of capital market studies to consistently document reliably positive security returns

accompanying LIFO adoptions was disconcerting to many researchers. It raised the possibility of firms

choosing to remain on FIFO even though they would benefit by adopting LIFO. They would do this out

of fear that their stock price would be adversely affected by the lower earnings they would report under

LIFO. Managers would thus be rewarded for sacrificing substantial tax benefits in favor of reporting

higher earnings. Further, the evidence that most companies did not use LIFO (e.g., see Pincus [1993,

Table 1]; Bowen et al. [1995, Table2]) suggested that the tax benefits of using LIFO were not as large as

some might believe or that other factors outweighed the potential tax benefits. These inconclusive and

often anomalous findings gave rise to the perception of a “LIFO puzzle” (e.g., Biddle [1980]; Morse and

Richardson [1983]), and encouraged researchers to seek non-tax explanations of inventory method

choices. We first discuss theoretical models aimed at understanding the LIFO puzzle, and then we review

non-market-based archival studies having the same purpose.

3.1 THEORETICAL MODELS OF THE LIFO PUZZLE

Amershi and Sunder [1987] demonstrate that an efficient capital market may not be sufficient to

discipline managers whose beliefs about investors’ decision rules and behavior are incorrect. They argue

that managers’ apparent failure to maximize cash flows in their LIFO/FIFO decisions is consistent with

their theory.

years of LIFO adoptions.

6

They assume (a) a well-functioning stock market. Shares of a number of not necessarily

identical firms are traded by a large number of investors, and the market is perfectly competitive,

valuing shares at the future cash flows market agents expect. (b) Managerial compensation takes

the form of cash and is perfectly aligned with the market value of the firm’s stock. (c) Managers

and shareholders are risk neutral, can have non-homogeneous beliefs, and are rational since each

maximizes his own expected cash flows. (d) Investor decision rules that generate market prices are

not common knowledge among managers. This last assumption is essential to their results.

In their model, while some managers (“C-type”) assume investors value firms based on cash flows,

other managers (“I-type”) mistakenly model investors’ valuation decisions as being based on reported

earnings. Hence, while all parties behave rationally, given their beliefs, Amershi and Sunder do not assume

everyone necessarily believes that all other parties use the identical decision criteria they use. Also,

managers and investors (in aggregate) receive the same information signal about the state of nature (y1 or

y2); thus, the capital market is strong-form efficient. Moreover, investors know whether an I-type or C-

type manager runs the firm.

Amershi and Sunder construct an example in which the market value maximizing decisions

of C-type and I-type managers differ. The I-type manager makes the market value maximizing

decision based on his model of maximizing expected accounting income, given y1. Investors

observe y1 and the I-type manager’s decision, and they arrive at the market price for the firm’s

stock (using their model of maximizing expected cash flows). The resulting stock price is the same

price the I-type manager expects to observe if investors follow the decision rule (maximizing

expected accounting income) he incorrectly assumes they follow. However, investors have set the

price by maximizing expected cash flows in light of the I-type manager’s decision and y1. The I-

type manager’s decision does not maximize expected cash flows; given y1, the C-type manager

would make the opposite decision than the one the I-type manager has made. But, the resulting

stock price is the same price that would result if the C-type manager had made the non-market

value maximizing decision. Accordingly, the share price provides no clue to I-types that their

model of investor behavior is incorrect. Rather, I-types’ mistaken beliefs are reinforced since the

observed market price fulfills their expectations. The authors introduce noise into the market

pricing mechanism to establish that their example reflects a more stable and generic phenomenon

7

and to show that managers do not necessarily learn from stock price changes of other firms.

Based on their model, Amershi and Sunder comment on the results of empirical studies. Cross-

sectional studies, they believe, would include both I-type and C-type managed firms. Since their model

predicts opposite stock price effects when I-type and C-type managers make the same decision, Amershi

and Sunder expect the cross-sectional average stock price change to inventory method decisions to be

positive or negative, depending on the relative proportions of each type of firm in a sample. They are thus

not surprised by the inconsistent empirical results of prior studies.

The Amershi and Sunder model demonstrates the possibility that some managers can fail to make

a cash flow maximizing decision while nonetheless acting rationally. Demonstrating existence is useful, but

the question arises why I-type managers should be expected to persist in their incorrect beliefs. The

authors identify possible alternative mechanisms to the capital market that might serve to inform managers

of their errors. These include direct communication between investors and managers, the markets for

corporate control and managerial labor, and education. These alternative mechanisms are assumed to be

slower in disciplining managers than the capital market.5 Empirically, the speed of adjustment of such

alternative disciplining mechanisms is an open question, as is the extent and turnover of I-type managers.

Bar-Yosef and Sen [1992] develop a model to capture the trade-offs between tax savings and the

interests of managers as an alternative way to explain why managers might forgo the tax benefits of a LIFO

adoption. When managers’ compensation is tied to their firm’s earnings, shareholders’ desire to maximize

cash flow may be contrary to managers’ desire to maximize earnings. There is no signaling of private

information via the inventory method choice in the model.

When tax effects are present in the model and the owner is unaware of the state of demand, the

choice of LIFO versus FIFO depends on certain conditions. Bar-Yosef and Sen find, not unexpectedly,

that lower tax rates increase the preference for FIFO, while increases in tax rates cause the inventory

accounting choice to swing toward LIFO. However, the results are indeterminate with regard to

increasing acquisition prices. Increasing costs increase the importance of the compensation incentive

effect, which favors the choice of FIFO, but also increases the tax benefits, which favors the choice of

5 Amershi and Sunder suggest that the few students who learn about capital market research in accounting courses probably remember that earnings and stock price are positively correlated. However, they think that few such students will likely retain an appreciation of the relation between share prices and accounting methods since it is more ambiguous (see below).

8

LIFO.

Hughes and Schwartz [1988] consider the inventory method choice in a game theoretic setting,

utilizing adverse selection. Their analysis is based on a game played between a firm’s management and

investors in the presence of asymmetric information within the capital market. The authors identify

incentives and disincentives for firms to use FIFO, and they hypothesize that a firm’s choice of FIFO

serves as a credible signal that management has favorable private information about the firm and its

prospects. Hughes and Schwartz assume that firms using FIFO signal they are high quality firms by

forgoing LIFO-based tax reductions and the opportunity to reduce expected bankruptcy costs. Investors

will assume a firm choosing FIFO is a good firm if the cost and benefits are such that firms least able to

forgo the tax benefits opt for LIFO.

Hughes and Schwartz model the expectation of inventory accounting choice as a function of

expected after-tax cash flows under FIFO, minus fixed bankruptcy costs, and plus the tax benefits

from the adoption. Tax benefits from adoption are only expected under LIFO; this component is

zero for FIFO firms. Managers’ inventory accounting method choices are based on changes in

their wage functions. Wage functions are based on the same components as the inventory method

expectation function. All other financial measures are assumed to be exogenously determined.

Two types of firms exist in this model: good firms and bad firms. Based on certain economic

conditions, Hughes and Schwartz conjecture that either a separating or a pooling Nash equilibrium

may exist. The existence of either equilibrium is a result of the beliefs of investors. The authors

describe in detail the various scenarios that may exist under either regime. In particular, they

describe scenarios in which FIFO Pareto dominates LIFO despite significant forgone tax savings.

Fellingham [1988] provides a discussion of the Hughes and Schwartz paper and raises three key

points. First, what makes inventory accounting choice a viable signal of firm quality? Second,

shareholders can control the behavior of managers through compensation contracts, yet the wage function

in the Hughes and Schwartz model does not allow for shareholder influence. Shareholders can be

expected to maximize the available benefit of compensation contracts in an attempt to advance their own

goals. Third, the authors link bankruptcy costs to inventory costs, but such a link has not been empirically

observed. Moreover, the authors do not provide any theoretical framework as to why this relation should

exist. Fellingham points to prior research that documents instances where firms in financial distress

9

abandon LIFO to increase reported earnings and provide an additional cushion for debt covenants (see

Johnson and Dhaliwal [1988]). He concludes there is little evidence supporting the authors’ claim that

good firms choose FIFO.

Another problem with the model is that Hughes and Schwartz fail to cite any empirical or

theoretical work that supports their assumption of only two types of firms. While this is a fairly common

generalization in models, it may pose significant confounding effects in empirical work since firms can

exhibit various degrees of financial distress. In general, separating equilibrium do not exist when there are

more than two types represented.

Jung [1989] develops a two-period inventory choice signaling game, similar to the one developed

by Hughes and Schwartz. Jung’s game is played between a firm and a potential competitor within the

firm’s product market. Information asymmetry exists between the firm and its potential competitor, and

the latter is unaware of many of the specific costs incurred by the firm.

In the model, the firm is assumed to be a monopolist. Several variations of the game are

discussed; the simplest form being when there are no potential competitors. The monopolist makes

the inventory accounting choice that maximizes cash flow, which would be LIFO under rising

prices. A pre-entry game is comprised of two components: the competitor’s decision to enter the

market and the firm’s inventory accounting choice. The competitor’s decision to enter the market

is based on the costs it will incur in relation to the costs incurred by the incumbent firm. Three

scenarios may arise. The first two are based on the competitor calculating some profit level. If

profit is positive the competitor enters the market, and if profit is negative it remains out of the

market. In the third scenario the competitor bases its decision of the perceived relative costs

incurred by it and the monopolist firm. If the latter’s costs are less than those to be incurred by the

potential competitor, no entry take places, but if its costs are more than those of the potential

competitor entry will take place. The inventory decision made by the firm in the first two scenarios

is irrelevant because the potential competitor bases its decision on elements outside of the firm’s

control. However, the third scenario is a game of incomplete information.

The potential competitor’s strength is strong, indifferent, or weak depending on whether its

expected profit upon entry, which is based on its belief about the monopolist’s costs, is positive, zero, or

negative. Depending on the potential competitor’s strength and on parameter values in the model, different

10

types of equilibrium emerge. One that potentially explains the LIFO puzzle is when the monopolist faces

high input costs. Adopting LIFO will signal to the market that the firm faces high costs. When the

expected LIFO tax benefits exceed the value of deterring entry into the market, a separating equilibrium

exists because the high cost monopolist will choose LIFO. If the two values are equal, the firm

randomizes its LIFO/FIFO decision. However, when LIFO tax benefits are less than the value of

deterring entry, the firm will forgo the LIFO tax benefits and remain on FIFO, and a pooling equilibrium

will result. It is a pooling equilibrium since a monopolist expecting low input costs will also choose FIFO.

Before the market knows the accounting choice, the firm is valued based on the weighted-average

of its FIFO firm value and its LIFO firm value. If the firm adopts LIFO, its share price is made up of the

expected after-tax cash flow from the first period and the discounted expected after-tax cash flow from

the second period. These cash flows are conditioned on all information available to investors. The

adoption of LIFO reveals increasing costs and a potentially higher probability of competitors entering the

market. This realization causes the market to decrease the value of the LIFO adopting firm. Thus, when

the LIFO choice is revealed to the market, firm value decreases.

The framework set forth by Jung [1989] is tractable and produces well-defined results. However,

the premise that firms make the accounting inventory method choice to deter potential competitors seems

less than compelling. At the very least, large, well-established firms in industries that have high entry

barriers would not seem to be as concerned with potential competitors as are small firms in industries with

low barriers to entry.

Hughes, Schwartz, and Thakor [1994] extend the model previously developed by Hughes and

Schwartz [1988] to examine signaling of firm type via capital structure and inventory accounting method

choices. These signaling vehicles are hypothesized to communicate management’s private information

about the value of their firm to investors; i.e., whether the firm is good or bad. The authors develop

separating equilibrium under certain economic conditions and use capital structure and inventory

accounting choice as independent and joint-signaling tools. The analysis hinges on the assumption that

firms are motivated by the probability of bankruptcy. Hughes et al. use debt as an indication of how close

a firm is to bankruptcy.

Unlike Hughes and Schwartz [1988] in which debt is exogenously determined, this model

allows debt and thus the probability of bankruptcy to be a choice. The analysis also extends the

11

earlier model by allowing for three (rather than two) types of firms: high, intermediate, and low

quality. Quality is a function of future of cash flows, about which managers have private

information. However, strong assumptions are imposed on the intermediate firm. Intermediate

firms can only signal with debt. Thus, a separating equilibrium will exist only between firms of

high and low quality in the current model, as in the Hughes and Schwartz model, and intermediate

firms cannot signal their type via their choice of inventory methods.

The scenario proposed by Hughes et al. is that the firm decides to undertake a project, some

portion of which is financed by issuing debt. At that time the firm also selects its inventory method. When

profit or loss is realized in the future, taxes are paid and the debt is repaid. If the realized cash flow is less

than the amount needed to service the debt, the firm and management suffer bankruptcy costs. Managerial

bankruptcy costs are derived from a loss of reputation, difficulty in finding an equivalent job, etc.

Moreover, the model assumes managers suffer relatively higher bankruptcy costs than their firms do, which

provides an incentive for managers to issue less debt than is optimal from the firm’s point of view.

Any increase in debt increases the probability of bankruptcy more for low quality firms as

compared to high quality firms, and thus capital structure can be used to signal firm quality. However, the

relatively high bankruptcy costs managers face makes signaling quality via debt level a losing proposition

for them.

Now consider the inventory accounting choice. Adopting LIFO increases a firm’s cash flow

because of tax savings, which increases firm value. Hence, low quality firms have a greater incentive to

switch than high quality firms do. High quality firms, of course, can also benefit from LIFO, but to

distinguish themselves from low quality firms that also adopt LIFO, high quality firms will have to issue

more debt after adoption to signal their quality. But this will increase the probability of bankruptcy, which

falls disproportionately more heavily on managers than on their firms. Hughes et al. show that managers of

high quality firms may find it optimal to choose FIFO instead, thereby signaling that they are a high quality

firm that can afford to forgo LIFO tax benefits, and then lower their firm’s debt level, thereby reducing the

probability of bankruptcy. Hughes et al. suggest that the differences in capital structure implied by their

model is one explanation for why firms within a particular industry do not adopt the same inventory

accounting method.

Signaling is a plausible explanation for the choice of inventory methods. This is because the

12

requirement that there be differential costs cross-sectionally to make the signal credible is present in the

LIFO/FIFO choice due to the direct cash flow effects from taxes. However, one must be skeptical of

signaling arguments in models that do not include a menu of signaling mechanisms from which to choose.

While Hughes et al. allow firms to signal their type in two different ways, other seemingly lower cost

signaling options are not considered. For example, why not adopt LIFO, reap the tax benefits, and then

issue an earnings forecast or increase cash dividends?

3.2 NON-CAPITAL MARKET STUDIES OF LIFO/FIFO CHOICE AND CHANGE

We turn now to non-capital market-based empirical research that was motivated in part, if not

primarily, by the LIFO puzzle. Non-tax explanations of accounting method choice or change (e.g.,

contracting cost theory) were considered in some early studies, and it is important to recognize that tax

and non-tax explanations of inventory choice need not be independent. Rather, the choice may reflect a

simultaneous consideration of operating, investing, and financing issues (e.g., input costs, inventory levels,

and income taxes).

Among the studies reviewed in earlier reviews are Abdel-kahlik [1985] and Healy, Kang, and

Palepu [1987]. They investigate a link between inventory method choice and management compensation,

but their results are inconclusive with regard to changes in parameters of compensation plans following

LIFO adoptions. Niehaus [1989] finds that the level of managerial ownership is a determinant of inventory

method choice, although not as he predicted. LIFO tends to be used by companies having managers who

are not shareholders (or who are very large shareholders). One would expect LIFO use to increase as

managerial ownership rises and the interests of shareholders and managers naturally become more aligned.

Morse and Richardson [1983] and Hunt [1985] find little or no evidence that inventory choice can be

explained by the debt covenant hypothesis. That is, firms facing binding provisions in debt contracts have

an incentive to increase earnings and inventory carrying values by remaining on FIFO, but these authors do

not find much support for this in their tests.

In another early study, Lee and Hsieh [1985], using both univariate and multivariate analyses,

document that inventory turnover, inventory variability, and the frequency of price increases are associated

with inventory method choice. However, differences between LIFO and FIFO firms could be due to

substantive economic differences or they could be induced by alternative accounting methods. Lee and

Hsieh do not attempt to discriminate between these two possibilities by adjusting for the impact that

13

different inventory methods have on reported financial statement amounts. Johnson and Dhaliwal [1988]

examine LIFO abandonments. Relative to non-abandoning firms, firms abandoning LIFO are more highly

leveraged, have less slack in their working capital constraints, have weaker earnings performance, and

exhibit greater earnings volatility. The results support the conjecture that abandonment firms are suffering

financial distress and benefit from abandoning LIFO because of improvements in financial statement

presentation.

Dopuch and Pincus [1988] examine the characteristics specific to firms, drawn from the 1962-

1981 period, that are long-term users of FIFO (n = 102), long-term users of LIFO (n = 29), or FIFO

firms that switch to LIFO (n = 70). A key aspect of their analysis is the computation of variables “as-if”

the alternative inventory accounting method were used. To compute as-if amounts, Dopuch and Pincus

adapt an algorithm developed by Biddle [1980]. They do not base the as-if computations on LIFO

Reserve disclosures since such disclosures are not available for more than half of their sample period and

no comparable as-if disclosures are available for FIFO firms.

Dopuch and Pincus estimate the tax benefits reaped or forgone by firms using LIFO or FIFO.

They find the average estimated tax benefits garnered by LIFO firms dwarf the estimated LIFO tax

benefits forgone by FIFO users. Contrary to what had been previously suggested, long-term FIFO firms

do not forgo substantial tax benefits by remaining on FIFO. Moreover, the small dollar amount of tax

benefits that FIFO firms do pass up by not adopting LIFO may reflect a lower bound on the costs of

adopting LIFO. These results support Biddle [1980] and Morse and Richardson [1983] who argue that

firms switch to LIFO after their estimated tax savings reach a high plateau. The idea is that LIFO

adoptions are costly and firms do not switch until the expected tax benefits exceed the expected adoption

costs. Dopuch and Pincus also compare FIFO firms that remain on FIFO versus those that switch. Firms

that eventually switch to LIFO forgo larger estimated LIFO tax benefits prior to the year of adopting

LIFO than do their FIFO counterparts that do not switch. Further, this difference persists after the year of

change. Overall, the results support the conclusion that both LIFO and FIFO firms minimize taxes subject

to non-tax constraints.

Additionally, Dopuch and Pincus investigate how long-term LIFO and FIFO users (and also

LIFO adopters and non-adopters) differ in terms of their economic characteristics. They use accounting

ratios identified in the literature and by intuition to proxy for differences in firm-specific characteristics.

14

Their initial cross-sectional comparisons of long-term FIFO and LIFO users indicate that several

operating and investing characteristics vary systematically. Specifically, LIFO firms, on average,

are larger and more capital intensive, have lower variability of inventories, lower inventory/total

assets, higher inventory turnover ratios, and face more rapidly increasing input prices. These

results are generally consistent with Lee and Hsieh [1985], whose sample included more LIFO

firms.

When Dopuch and Pincus calculate as-if variables, i.e., when they adjust for differences in

inventory methods, they find that long-term LIFO firms are larger, more capital intensive, and face more

rapidly increasing input prices than long-term FIFO firms. The differences in inventory variables

documented in their initial analysis are no longer significant. Those differences appear to be artifacts of the

inventory accounting methods used.

Dopuch and Pincus also report that firms switching to LIFO have certain distinctive characteristics

relative to FIFO firms that do not change methods. For example, change firms are larger and have lower

variability of inventory levels than long term FIFO firms. Relative to long-term LIFO firms, however,

change firms are smaller and have higher inventory volatility. Estimated LIFO tax benefits are largest for

long term LIFO users, followed by change firms and then FIFO firms. The overall results for change

firms, coupled with the larger estimated LIFO tax benefits change firms forgo in the years prior to their

adoption of LIFO, suggest capital market agents may be able to identify firms that adopt LIFO prior to the

year they switch.

The Dopuch and Pincus results suggest that the tax explanation remains the most viable one

underlying managers’ choice of inventory costing methods. However, they also find compelling evidence

that LIFO firms are larger than FIFO users (see also Morse and Richardson [1983]; Lindahl [1989]),

which is consistent with there being large fixed costs associated with LIFO adoptions that larger firms can

more easily bear. However, this empirical regularity linking firm size and inventory method choice,

combined with the finding that LIFO users typically are more capital intensive (also see Hagerman and

Zmijewski [1979]), is consistent with another explanation. Because GAAP precludes charging earnings

for the cost of equity capital, capital intensive firms are more likely, ceteris paribus, to report higher

earnings than non-capital intensive firms do. If capital intensive firms are more vulnerable to government

expropriation of wealth due to their sheer size -- i.e., if they are subject to political costs (see Watts and

15

Zimmerman [1986, 1990]) -- then they have an incentive to choose LIFO to report lower levels of

earnings and assets. Whether size and capital intensity proxy for political costs or for economies of scale

in information systems (or in-house tax expertise) remain unresolved questions.

Hilke [1986] uses a survey instrument to investigate why small firms in the retail industry do not

adopt LIFO at the same rate as large firms. The survey lists seven possible factors postulated to be

important considerations when making the inventory accounting choice. The factors are: financial

reporting conformity requirements; costs of changing to LIFO; incompatibility between LIFO and a firm’s

current inventory system; difficulty in understanding LIFO requirements; write-downs; increased IRS

scrutiny of LIFO users; and IRS recapture provisions. Additionally, a tax savings and “other” category

are included. Respondents rated each item for importance.

The National Association of Wholesaler-Distributors administered the survey in 1983.

Approximately 24% of the respondents use LIFO, while, in fact, only 4% of wholesalers used LIFO

during 1983. Large companies represented only 7% of wholesalers in the country, but they

represent 36% of all survey respondents. This overrepresentation of LIFO users and large

companies may limit the generalizability of the survey results.

Relative to LIFO users, non-LIFO users give greater importance to all factors identified in the

survey except write-downs. There is also evidence of a significant difference between large and small

non-LIFO users regarding the importance of the cost of adopting LIFO. The respondents estimate the

present value of the average cost of converting to and maintaining a LIFO system as 2.91% of an average

year’s profits. However, for small companies the conversion cost estimates amount to more than 30% of

profits. Adoption costs thus appear to be a deterrent to LIFO adoption by small firms. Not discussed in

the paper is the related tax savings obtained by companies adopting LIFO. Tax savings could possibly

offset the perceived costs of LIFO adoption.6

Cushing and LeClere [1992] examine long-time FIFO and LIFO users to identify economic

variables that are particular to those groups. Their sample includes 175 FIFO and 48 LIFO firms that use

FIFO or LIFO exclusively during 1975-84; thus they exclude firms changing methods. Both a univariate

6 Hilke suggests that government price indexes should be made available to assist firms in calculating price changes. He believes this would significantly lower the fixed cost of implementing and maintaining LIFO, especially for the smallest firms. Purtill [1996] reports expanded use of external price indexes became increasingly permissible under I.R.S. rules throughout the 1980s.

16

analysis and a multivariate logit in which the LIFO/FIFO choice is the dependent variable are conducted.

Seven of eight variables are statistically significant in the univariate analysis. Specifically, compared to

LIFO firms, FIFO firms, on average, have lower estimated tax savings, larger tax loss carryforwards,

greater variability in inventory levels, greater inventory obsolescence, higher leverage, lower current ratios,

and are smaller.

Their first multivariate model focuses on hypothesized non-tax factors affecting the

LIFO/FIFO choice. The results indicate significance for three variables: LIFO firms are larger,

have higher current ratios, and lower levels of leverage than FIFO firms. This model correctly

predicts the choice of inventory method for 79% of the sample. However, since 78.5% of the

sample is FIFO, the model’s prediction accuracy is no better than chance. The second model

includes only tax factors. Four tax-related variables are significant. These indicate that LIFO

firms have higher tax savings and lower inventory variability and inventory obsolescence vis-à-vis

FIFO firms, but also lower inventory materiality. The sign of the inventory materiality proxy is

opposite to expectations. The second model has prediction accuracy of 82%, and its psuedo R2 of

17.0% is only slightly above that of the first model’s 14.3%. The authors argue that these results

indicate that tax variables explain only slightly more of why firms choose particular inventory

methods than do non-tax factors. However, they perform no formal tests of the significance of the

enhanced explanatory power or predictive accuracy. Model 3 includes both tax and non-tax

variables. The model has a psuedo R2 of 25.4% and prediction accuracy of 85%. Except for

leverage, all variables found significant in the previous two models are significant here as well.

Thus, both tax and non-tax factors are important determinants of the LIFO/FIFO decision when

considered simultaneously.

Cushing and LeClere also survey the firms in their sample and ask financial officers to

identify the reasons why their company uses the inventory method it does. LIFO respondents

identify tax savings as the most important reason for using LIFO. In addition, they say LIFO

provides a better matching of revenues and expenses, and also that they were motivated by

industry conformity to use LIFO. LIFO respondents also claim that political pressures due to high

levels of income are of little concern to their firms. FIFO respondents state that FIFO better

reflects the physical flow of their inventories and provides a more accurate valuation of inventory

17

on their balance sheets. All of the items identified in the research as to why a company would

remain on FIFO are selected by at least two respondents. Additional explanations offered include

LIFO is too complex for the staff, FIFO is required by government contracts, FIFO is more

desirable due to fast turnover of products, or it is used because the firm has substantial income

from a foreign subsidiary domiciled in a country in which LIFO is not allowed. Of the 12 reasons

identified by FIFO respondents, eight relate to a lack of significant tax benefits for their firm under

LIFO.

Skinner [1993] incorporates a firm’s investment opportunity set (IOS) into the analysis of the

determinants of accounting choice as a way to allow for the simultaneous impact (i.e., endogeneity) of

multiple factors. In the context of the LIFO/FIFO decision, the IOS overlaps with, but in principle is

broader than, the tax and inflation dimension. Skinner [1993] finds inflation and thus income taxes are

important in the LIFO/FIFO choice after controlling for the effects of the IOS.

Skinner attempts to explain inventory, depreciation, and goodwill amortization period

accounting choices for a sample of unregulated industrial firms for 1987. IOS variables include

assets-in-place (gross PP&E/market value of equity, R&D/sales, and Tobin’s q) and the riskiness of

firms’ assets (beta*market value of equity/total firm value). He controls for firm size and prior

profitability and, regarding inventory choice, investigates the significance of financial leverage,

accounting-based bonus plans, and the level (low, medium, high) of inflation.

He estimates coefficients from multinomial logits for samples of 256 to 357 firms. The

significant results follow. (a) When only IOS variables are in the model, LIFO is linked with firms

having assets-in-place. (b) When only non-IOS variables are considered, LIFO is associated with

larger firms, good prior earnings performance, and high levels of price inflation. (c) When all

variables are in the model, firm size and prior profitability continue to be highly significant, the

inflation variable is marginally significant, and the IOS variables are no longer significant. Skinner

argues, a priori, that tax and IOS factors probably are correlated since firms’ production processes

likely affect the variability of their inventory levels and determine the inputs required. Thus, for

example, firms with relatively good growth opportunities, because they also tend to be riskier, are

less likely to choose LIFO.

Bowen, DuCharme, and Shores [1995] investigate the role and importance of implicit contracts in

18

the choice of accounting methods. A firm’s ability to negotiate with stakeholders (consumers, suppliers,

employees, and short-term creditors) depends, in part, on its reputation for satisfying the implicit claims

against it. Bull [1987] and Tesler [1980] demonstrate implicit contracts are self-enforcing because of the

negative effect that failing to fulfill them has on a firm’s reputation.

Stakeholders have an incentive to monitor firms with which they have implicit contracts, and one

way to do that is to use the firms’ financial reports. Bowen et al. assume that managers believe that at

least some stakeholders will not completely adjust firms’ financial statement numbers for accounting

method differences (e.g., due to limited accounting information processing skills or having too small a stake

in the company to justify a thorough financial analysis). In these circumstances, Bowen et al. hypothesize

that implicit contract relationships motivate management to choose income increasing accounting methods

for the long-term.

The analysis is conducted for inventory and depreciation methods, considered separately

and jointly. The authors analyze five years in the 1981-93 period (with sample sizes for the

inventory choice tests ranging from 1,113 to 2,330 firms). Firms changing inventory or

depreciation methods within two years preceding the year selected are excluded as are regulated

firms and firms likely experiencing financial distress. Bowen et al. use variables to proxy for a

firm’s dependence on implicit claims with stakeholders. Proxies for implicit contracts with

customers include an indicator for durable goods manufacturers to capture the availability of parts

and service over a long time period, and R&D expense deflated by firm size to reflect uniqueness of

products. The authors use cost of goods sold interacted with a dummy variable for manufacturers

(versus non-manufacturers) to proxy for suppliers’ implicit claims. They proxy for employee

implicit claims by using a labor intensity variable and an indicator for the existence of defined

benefit pension plans. To capture the extent a firm depends on implicit claims of short-term

creditors, Bowen et al. use short-term notes payable scaled by firm size. Advertising expense is

used to proxy for reputational concerns that span all stakeholders.

In the analysis of inventory choice, Bowen et al. regress firms’ inventory choices (LIF0 = 0,

average costing = 0.5, FIFO = 1) against the implicit contracting variables and other variables that control

for leverage, bonus plans, firm size, and taxes. The results indicate that implicit claim proxies for

customers, employees, and short-term creditors are generally significant in predicted directions in the five

19

different test years, and advertising is significant much of the time. Cost of goods sold (CGS) is neither

significant nor positive as expected in its role as a proxy for suppliers’ implicit claims, although Bowen et

al. make the reasonable conjecture that it may proxy for tax benefits, which would imply a negative

coefficient. In a diagnostic test, they replace the tax variable, which is a dummy variable (DTLC) indicating

positive inventory and no tax loss carryforwards, with two new variables that multiply DTLC times the

CGS-manufacturing and CGS-non-manufacturing variables. The coefficients on the two new tax variables

are reliably negative in three of the five years, and the coefficients on the original CGS variables now

generally become reliably positive, as predicted in their role as proxies for supplier implicit claims. Overall,

the results support the hypothesis that implicit contracting variables are also significant determinants of

inventory choice. Bowen et al. acknowledge that some of the variables they use to proxy for implicit

claims may also capture other firm-specific characteristics. Future research can potentially refine these

proxies.

In summary, a large stream of research investigates the determinants of inventory accounting

choice. Initially, much of the research was motivated by the inconclusive and often negative share price

effects associated with LIFO adoptions and the associated LIFO puzzle that many firms that apparently

could benefit from the use of LIFO were not adopting it. The papers reviewed in this section provide

compelling evidence that taxes play a dominant role in the LIFO/FIFO choice,7 but that other factors are

important as well. We interpret the results of this research as saying that firms minimize taxes subject to

other considerations, which is consistent with the thesis of Scholes and Wolfson [1992]. We also

conclude that the evidence does not support a LIFO puzzle of any substantial proportion. Future research

can contribute to a more complete understanding of the non-tax factors that affect the LIFO/FIFO choice,

and their importance relative to taxation.

4.0 IS THERE ANOTHER WAY TO CONCEPTUALIZE THE CAPITAL MARKET

REACTION TO LIFO ADOPTIONS BEYOND A “SMART” VS. A “DUMB” MARKET?

For the most part, the capital market studies previously discussed in section 2 aim to assess the

capital market reaction associated with LIFO adoptions, and they document mixed and more often

negative share price effects. The design and interpretation of these studies are based on two alternative

7 Pincus and Wasley [1994] document high positive correlations between changes in producer prices and the frequency of LIFO adoptions over the 1969 to 1988 period they examine.

20

views of the world. The first view is the traditional tax benefit hypothesis, which predicts a positive stock

price effect should accompany LIFO adoptions due to enhanced cash flows. The alternative view is the

functional fixation or mechanistic hypothesis, which predicts a negative impact on share prices if investors

do not adjust for differing accounting methods but instead focus on the lower reported earnings that result

from using LIFO.

The capital market studies we discuss in next section also focus on LIFO adoptions. However,

they are structured (formally or informally) around a theoretical model (Kang [1993]) that offers a view of

the world that differs from the more simplistic view of a “smart” market versus a “dumb” market.

Moreover, these papers typically run a “horse race” among the competing predictions. Kang’s model and

the empirical work it spawned mark an important advance in research design and in our understanding of

the stock market’s response to LIFO adoptions.

Motivated by the well-documented and seemingly anomalous negative market reactions to LIFO

adoptions made during 1974-75, Kang [1993] models the market reaction to a firm’s LIFO adoption in

the context of managers’ receipt of private information about inflation and with explicit allowance for non-

trivial LIFO adoption costs.8 In the model (referred to as the “real value” model), investors form

expectations about the impact that inflation will have on a FIFO firm -- specifically, on the taxation of

inventory profits (i.e., realized inventory holding gains or losses) the FIFO firm generates. Managers act to

maximize the value of their firms and will adopt LIFO only if the present value of any LIFO tax benefits

exceed the costs of adopting LIFO. However, under LIFO the tax benefits are “benefits” in name only;

they merely offset incremental taxes that become due when a firm continues to use FIFO in an inflationary

environment. Adopting LIFO simply allows a firm to reduce its tax burden to the level it would have paid

had its input prices not been rising.

Moreover, even if it is optimal to adopt LIFO to avoid inflation-induced taxes on inventory profits,

the firm will nevertheless be worse off relative to a non-inflationary setting. This is because it must bear the

costs of adopting LIFO. On the other hand, if it is optimal to remain on FIFO -- i.e., if the expected

LIFO tax benefits are less than the expected LIFO adoption costs -- the firm is still worse off than if it

8 There are two versions of the Kang [1993] model and both versions assume asymmetric information between managers and external investors. In the first version the information asymmetry is about inflation and in the second version it is about adoption costs. We focus primarily on the first version since the empirical research is based on it.

21

were operating in a non-inflationary setting because it must pay taxes on inventory profits that arise by

using FIFO. Thus, the firm’s value will fall in the presence of rising prices regardless of whether or not it is

optimal to adopt LIFO.

Investors assess the probability a firm will adopt LIFO and incorporate this into the pricing of its

stock. The firm’s stock price is a weighted-average of the value of the firm if inflation is sufficiently high to

warrant a LIFO adoption and the value of the firm if inflation is lower and the firm remains on FIFO. If

management receives private information that the firm will face sufficiently high inflation, they will (instantly

announce a) switch to LIFO; otherwise, they will remain on FIFO.

To illustrate what happens to the firm’s stock price when its inventory method choice

becomes public information, consider the following simple example. Assume that if the firm faces

high inflation it will save $1.00 of taxes by adopting LIFO but must incur $0.40 of adoption costs.

Alternatively, if it faces low inflation it will incur $0.20 of additional taxes by remaining on FIFO.

Thus, it will be (will not be) optimal to adopt LIFO in the high (low) inflation case since the tax

benefit exceeds (does not exceed) the adoption costs. Also note that the value of the firm drops in

both cases relative to its non-inflationary value. If it adopts LIFO, its value falls by $0.40, the

amount of the adoption costs that must be incurred; the $1.00 of taxes the firm “saves” by using

LIFO simply recoups taxes the firm would not have to pay in a non-inflationary setting. If, instead,

it remains on FIFO, it pays $0.20 of additional taxes.

Assume (without loss of generality) that prior to the firm making its inventory method

decision the market assesses a probability of 0.5 that the firm will face high inflation and thus

switch. The market will price the firm’s stock to reflect the expected impact of inflation.

Specifically, its stock price will be lower than its non-inflationary value by $0.30 (= .5*-$0.40 +

.5*-$0.20). If the firm subsequently adopts LIFO, the probability of adoption moves to 1.0 and the

firm’s stock price moves down an additional $0.10 to reflect the full $0.40 reduction in value. If,

instead, the firm remains on FIFO, its stock price will move up. That is, the loss in value the

market was expecting when there was uncertainty about the impact of inflation (-$0.30) overstates

the loss due to higher taxes the firm will now incur in the low inflation environment (-$0.20).

Hence, Kang’s model predicts a negative stock price reaction to a LIFO adoption. This is not

because the market fails to adjust for the effect that alternative accounting methods have on

22

earnings, as is the case under the functional fixation/mechanistic hypothesis. Rather, it is a rational

response to the resolution of uncertainty about inflation’s impact on the firm’s value.

There are several aspects of Kang’s model that may prove troublesome when implementing

empirical tests. For example, he assumes inflation neutrality except for the taxation of inventory profits;

thus, firms facing rising input costs can pass those increases on to their customers. To the extent they

cannot do so, and the market appreciates this, a LIFO adoption may also reveal that the firm must absorb

higher than expected input costs and thus will have lower earnings. Another potential problem with the

model is that investors will most likely face uncertainty about both LIFO tax savings and adoption costs.

Uncertainty about both complicates the predictions. In particular, if investors expect that adoption costs

will exceed LIFO tax savings but are proved wrong by the firm switching to LIFO, then a positive market

reaction is predicted.

Notwithstanding its potential shortcomings, the model provides a new way to think about LIFO

adoptions and capital market responses to them. We now turn to several empirical papers that rely

directly or indirectly on the Kang model to structure their empirical analyses.

4.1 IS THERE SUPPORT FOR THE KANG [1993] MODEL?

The first study we examine is not motivated as a test of Kang’s model, but it does incorporate into

its empirical design many features that are in the spirit of the model. The paper is by Jennings, Mest, and

Thompson [1992]. They develop an empirical model to estimate the probability of a LIFO adoption, for

both subsequent adopters and non-adopters alike. The model uses publicly available information

concerning the likelihood that a firm will switch to LIFO. Any changes in the stock price as a result of a

subsequent LIFO adoption or non-adoption should reflect the revision in investors’ beliefs about the

probability of adoption (holding constant the expected magnitude of LIFO tax benefits). This revision in

investors’ beliefs is what the authors measure as the market’s response to the disclosure of LIFO

decisions. For non-adopters, the inventory decision occurs at the firm’s next regular earnings

announcement when no change in method is announced.

Jennings et al. use four groups of predictor variables to develop a model to estimate the

probability a firm will adopt LIFO. The four groups are (1) measures of LIFO benefits (current

and prior period holding gains and inflation rate), (2) financial and operating characteristics (15

variables including inventory turnover, inventory variability, total assets, ROA, leverage, NOLs),

23

(3) financial analysts’ data (e.g., Value Line information, if any, about a possible LIFO adoption

and forecasted change in FIFO-based annual earnings), and (4) industry information regarding the

extent of LIFO use and adoptions in the firm’s two-digit SIC industry. Two versions of the model

are used; one with no truncation of continuous variables and another using ranks for those

variables. Approximately 63% of firms that, in fact, adopt LIFO have probabilities of adoption

that exceed 0.5 and 70% of non-adopting firms have probabilities of 0.5 or less.

The authors regress size-adjusted security returns, accumulated over various intervals

during which a LIFO adoption or non-adoption becomes known to the public, on revisions of the

probability of adoption and unexpected earnings (based on the pre-change inventory method). For

LIFO adopters they also control for the unexpected earnings effect of the LIFO adoption. The

sample is comprised of 195 LIFO adopters and 292 non-adopters over the period June 1974 to May

1975. Thus, their sample is similar to the one used in Ricks (1982b).

Jennings et al. conclude that stock prices respond to the disclosures of LIFO decisions. The

primary focus is on their probability revision variable. Its coefficient is reliably positive for non-adopters,

which indicates a negative market reaction to not adopting LIFO (since the probability revision is

negative). However, for LIFO adopters the coefficient is either weakly positive (using their probability

model based on ranks) or insignificant (using the probability model without variable truncation).

Surprisingly, the coefficient on unexpected earnings (i.e., the earnings response coefficient) is insignificant

for non-adopting firms and only marginally significantly positive for adopters. Further, the coefficient on

the unexpected LIFO earnings effect is insignificant.

To make reliable inferences about security price movements associated with inventory decisions,

investors’ prior beliefs about the likelihood of a LIFO adoption must be specified (e.g., Lanen and

Thompson [1988]). This is a difficult task, and this variable likely is measured with error.

Hand [1993] attempts to simultaneously test the implications of the tax benefits hypothesis, a form

of the functional fixation/mechanistic hypothesis, and Kang’s real value model by focusing on the resolution

of uncertainty about whether or not a firm will adopt LIFO. Each model yields predictions about how

mean excess security stock returns will behave when uncertainty about the LIFO adoption decision is

resolved. Moreover, each yields testable implications for the cross-sectional relation between excess

stock returns and certain explanatory variables. It is in the cross-sectional analysis that Hand develops

24

predictions to differentiate among the competing hypotheses.

Hand uses a database of actual and proposed LIFO adoptions made between June 1974 and

May 1976 to conduct the empirical tests of the three models. In a rather clever approach, he identifies

firms that publicly indicated they were considering adopting LIFO but did not announce their decision

until their annual earnings announcement date. Hand estimates mean size-adjusted security returns

associated with the LIFO adoption/non-adoption announcements. He also regresses the excess returns on

several explanatory variables depending on the assumptions of the particular model under examination.

For non-adopters, the regression model includes two variables that reflect the modeling of

the assumption that the fixed costs of adoption are negatively related to firm size. Both variables

include the probability (p̂ ) of adopting LIFO. While complete details are not provided, Hand

develops his p̂ estimates using many of the variables Jennings et al. [1992] identify plus a

variable reflecting the ranking of the likelihood of adoption based on publicly available

information from either a financial analyst or the specific firm. The returns model also includes

p̂ *forecasted earnings reduction (a positive number), if LIFO were adopted, and unexpected

FIFO-based earnings. Both the tax benefit and real value models predict a negative coefficient on

the earnings reduction variable (due to the loss of forecasted tax savings), while the functional

fixation model predicts a positive coefficient (since earnings will be higher for non-adopters).

For LIFO adopters, the probability of adoption enters the regression as (1 - p̂ ), to capture

the uncertainty resolved by the LIFO adoption, unexpected earnings is measured on a LIFO basis,

and the empirical model includes an unexpected LIFO earnings effect variable. Both the tax

benefit and real value hypotheses predict a positive coefficient on the forecasted LIFO earnings

effect variable for adopters, while functional fixation predicts a negative coefficient (due to lower

earnings under LIFO). The tax benefit hypothesis also predicts a positive coefficient on the

unexpected LIFO earnings effect variable (which reflects higher than expected tax benefits), but

the real value model predicts a negative coefficient. Recall the real value model predicts that LIFO

tax benefits only recoup the opportunity cost of remaining on FIFO. Thus, under the real value

model news of higher LIFO tax benefits implies additional LIFO adoption costs, assuming adoption

costs are related to the magnitude of tax benefits. For adopters and non-adopters alike, all three

models predict a positive coefficient on unexpected earnings. Finally, only the real value model

25

has predictions about adoption costs; the fixed cost component’s coefficient should be negative

(positive) for adopters (non-adopters) and the variable cost component, which captures economies

of scale, should have a positive (negative) coefficient for adopters (non-adopters).

The adoption sample includes 96 firms, 58 of which meet the data requirements for the cross-

sectional tests. The non-adopters total 83 (59 having all the necessary data). Hand reports that, on

average, reliably negative excess returns accompany LIFO adoption announcements and reliably positive

excess returns accompany non-adoptions. The cross-sectional analysis provides evidence to support both

the tax benefit and real value models. That is, investors recognize and positively price the tax benefits of

adopting LIFO relative to the continued use of FIFO; this contradicts the prediction of the functional

fixation model. Consistent with the tax benefits model, but inconsistent with the real value model’s

predictions, investors positively price firms’ unexpected LIFO tax benefits. On the other hand, the results

regarding adoption costs are entirely consistent with the predictions of the real value model.

Hand also estimates the lower bound of the gross cost of adopting LIFO to be about 6% of the

market value of equity for his sample firms. This amount seems large, but, unfortunately, little direct

evidence of the costs of implementing and maintaining LIFO inventory systems appears in the literature.

Recall, Hilke’s [1986] survey yields estimates of the average present value of adoption costs of 2.9% for

retailers.9 Hence, future research in documenting the costs of LIFO adoptions would be useful. LIFO

adoption costs potentially include fixed and variable costs of accounting system changes, more extensive

inventory management activities, higher inventory levels to minimize the occurrence or tax impact of LIFO

liquidations, a greater risk of violating contractual provisions dependent on inventory carrying values,

renegotiation of accounting-based contract provisions to continue to use non-LIFO-based inventory

valuations, incremental bookkeeping costs (including producing non-LIFO-based numbers required in

contracts and adjusting for the artificial liquidations that can occur when using the perpetual system with

LIFO), and the risk of escalating tax liabilities that coincide with reductions in the scale of operations.

Pending direct evidence on the costs of implementing and operating a LIFO system, and given the key role

adoption costs play in Kang’s model, one must interpret Hand’s evidence about adoption costs with

caution.

9 At a price-earnings ratio of 10, these costs would amount to no more than 0.3% of an average firm’s market value.

26

As is apparent, the majority of capital market (and non-capital market) LIFO studies use samples

that are dominated by 1974 LIFO adoptions. Sharp price increases, associated with an oil embargo and

the relaxation of wage and price controls, coupled with a recession arguably make 1974 a unique year.

Firms from many varied industries reaped large tax benefits by adopting LIFO at that time, and, in fact, the

largest number of LIFO adoptions occurred in 1974 (see Butters and Niland [1949]; Sunder [1973];

Pincus and Wasley [1994]). While focusing on 1974 LIFO adoptions means a researcher has a large

number of observations to analyze, which enhances statistical power, the possibility exists that the results

may not generalize to other time periods.

Moreover, most LIFO adoption research uses the date of the annual earnings announcement as

the date of the LIFO adoption announcement. Sometimes this is by assumption, but often it is, in fact,

what occurs. U.S. tax regulations allow a firm to make a LIFO adoption decision after year-end, so long

as it does so before the earlier of the date it reports to stockholders or files its tax return. Hence, firms can

be expected to exercise the option of making the LIFO decision as late as possible in order to know “how

much money is on the table” for the year just ended. However, the joint occurrence of both an earnings

announcement and a LIFO adoption announcement causes research design problems. The researcher

must isolate the LIFO adoption effect by developing satisfactory controls for the concurrent earnings

announcement. This means developing a measure of unexpected LIFO earnings even though the firm has

never before reported earnings on a LIFO basis. It turns out that for many fiscal 1974 adopters, this is not

too difficult; analysts were aware that many firms were considering a LIFO adoption and, in anticipation,

often developed earnings forecasts on a LIFO basis. Outside of fiscal 1974, however, that is not the case.

Pincus and Wasley [1996] further investigate the real value, tax benefit, and functional fixation

hypotheses by analyzing a sample of post 1974-75 LIFO adoptions that are disclosed at or before the

annual earnings announcement date. Hence, some of their sample firms resolve uncertainty about adopting

LIFO prior to announcing their annual earnings. The authors hypothesize that firms announcing their

decision to adopt prior to the annual earnings announcement date do so to signal earnings-related

information. That is, by forgoing the option to announce their adoption later, early announcers may be able

to credibly signal positive information about earnings growth.10 Such signaling is outside the real value

10 Evidence that LIFO adopters generally enjoy favorable earnings performance in the year of adoption is reported in Ricks [1982b] and Pincus and Wasley [1994].

27

model. Pincus and Wasley [1996] also conjecture that the stock exchange on which a firm is listed might

play a role in the market's reaction to LIFO adoption announcements. Hand [1995] observes differential

market reactions to LIFO announcements of NYSE/ASE firms relative to OTC firms. Although he offers

no explanation for the exchange-listing effect, it may proxy for firm size and thus differential LIFO adoption

costs, which, in the context of Kang’s model, can induce differing market reactions to LIFO adoptions.

Pincus and Wasley identify their sample from various data sources and search the Dow

Jones News Retrieval Service to pinpoint the first public announcement of a firm’s LIFO adoption.

They sample from mid-1979 through mid-1989, which is a time period characterized by varying

rates of price change. The sample consists of 190 LIFO adoptions; 146 are from the NYSE/ASE

and 44 from the OTC market. OTC firms have significantly larger median LIFO earnings effects

as a percentage of share prices compared to NYSE/ASE firms.11 Of the 190 LIFO adoptions, 112

are disclosed with annual earnings announcements, 27 with interim earnings announcements, and

51 on non-earnings announcement dates (of which 24 have no concurrent disclosures of any kind).

OTC firms that early announce their LIFO adoption exhibit significantly higher increases in

as-if and unadjusted earnings than OTC firms who disclose LIFO adoptions with annual earnings

announcement. This suggests that at least some firms use early announcement as a signal for

strong earnings performance.

The mean share price results indicate a positive reaction, which differs from the negative share

price effects documented for 1974-75 LIFO adoptions. Moreover, the market response to LIFO

adoptions disclosed on earnings announcement dates differs from the response to adoptions announced at

other times, and the reaction differs as a function of the exchange on which a firm’s shares are traded. The

results of the cross-sectional tests provide some support for the tax benefit hypothesis over the real value

model in that unexpected changes in tax benefits revealed in joint LIFO adoption/earnings announcements

are positively priced. Overall, however, the cross-sectional regressions are insignificant. This may reflect

measurement error in key variables such as the probability of adoption or the expected LIFO earnings

effects. Unlike 1974 LIFO adoptions, which occurred in what may be described as a “LIFO information-

rich environment” (e.g., Hand [1993]), the 1979-89 period is characterized by little public discussion or

11 By way of comparison, Hand [1993] reports larger magnitude earnings effects in his sample due to the very high inflationary rates in 1974-75.

28

anticipation of LIFO adoptions by financial analysts, and thus financial analysts’ assessments of possible

LIFO effects of adoption are not available for estimating these variables.

There are several questions that remain unanswered after considering the papers in this section.

First, the market’s generally negative reaction to 1974-75 LIFO adoptions stands in contrast to the

typically positive reactions in subsequent periods. However, the Kang [1993] model suggests the

following conjecture (based on the assumption that the magnitude of adoption costs is related to the

magnitude of tax benefits). Adoption costs may have been so large for the 1974-75 period, as Hand’s

[1993] evidence suggests, that the resolution of uncertainty associated with LIFO adoptions did, in fact,

reflect the final impounding of these costs. However, resolution of LIFO adoption uncertainty in later

periods may have signaled lower than expected adoption costs due to lower than expected inflation. In

such a case the Kang model predicts positive share price effects.

Second, to the extent that factors such as the timing of a LIFO adoption announcement and the

exchange on which a firm’s shares are traded help to explain market reactions to LIFO adoptions, then the

Kang model is incomplete. Further, it seems reasonable to believe that other factors a LIFO adoption

announcement may signal, such as the impact of inflation beyond the taxation of inventory profits, also

affect share prices. Overall, the empirical results suggest that extant theories are incomplete; none fully

explains share price reactions to firms’ LIFO adoptions.

4.2 HOW DID THE CAPITAL MARKET REACT TO THE TAX ALLOWNCE OF LIFO?

As previously discussed, problems exist with investigating whether LIFO adoptions have

significant valuation implications. For example, a researcher must develop proxies for the market's

expectations of a LIFO adoption and its earnings (and thus) tax effects. In addition, a LIFO adoption may

signal that earnings are growing faster than expected or perhaps that the impact of inflation is worse than

anticipated. These problems stem mostly from the self-selection bias that is inherent in all voluntary

accounting change studies.

Pincus [1997] examines the valuation implications of LIFO from a different perspective. He

analyzes the capital market response to LIFO’s introduction into the U.S. tax code. Like all regulatory

event studies, concerns can be raised about event dating, concurrent events, and the endogeneity of the

legislative process.12 Nevertheless, his research design mitigates many of the problems that plague LIFO

12 The endogeneity issue is the concern that legislative outcomes are not exogenous but rather reflect the demands of affected parties

29

adoption (i.e., voluntary accounting change) studies and allows for a clearer interpretation of how market

agents view the impact of LIFO.

Based primarily on the legislative history developed in Pincus [1989], ten critical events in

the legislative process leading to the statutory acceptance of LIFO for tax purposes are identified.13

These events are associated with the Revenue Acts of 1938 and 1939 and are points when the

probability LIFO would enter the tax code can be expected to have changed.

Pincus [1997] identifies 93 firms that likely would benefit from the allowance of LIFO (e.g.,

firms from industries in which inventory profits are common) and for which security price and

dividend data could be obtained. For 60 of those firms, Pincus derives estimates of the LIFO tax

benefits they likely could capture. These estimates are computed by adapting the estimation

algorithms in Biddle [1980] and Dopuch and Pincus [1988] and using financial statement data

obtained from Moody’s and Poor’s industrial manuals and Butters and Niland [1949].

For firms having tax benefits that exceed the sample median, the results from mean share price

tests indicate reliably net positive abnormal security returns associated with LIFO’s entry into the law.

These results reflect positive market reactions to news of Congressional actions occurring early in the

LIFO legislative process, and also a predicted price reversal in response to more restrictive LIFO

provisions than expected that emerged from a joint House-Senate Conference Committee. There is also

supportive evidence from tests examining the correlation between excess returns and estimated tax

benefits. Overall, the results support the conclusion that the market revalued upward the securities of firms

likely to avail themselves of the opportunity to use LIFO and defer taxes on inventory profits. Moreover,

the results suggest that potential tax benefits of LIFO adoptions were impounded into share prices at the

time LIFO entered the tax law. This implies that firms’ values subsequent to that time reflect the option to

adopt LIFO. Accordingly, even if it is possible to isolate the valuation effect of LIFO at the time a firm

switches, that effect would only capture the final resolution of uncertainty, if any, about whether the firm

who bear the costs to lobby for relief from taxation of inventory profits. While receiving tax relief is good news to affected firms, it may also signal that such firms are most in need of tax relief, which is bad news. An example of this is that firms may be unable to pass on higher input costs to customers. Unfortunately, it is difficult to assess ex ante the relative strengths of the simultaneous good and potentially bad news effects and thus to predict the net market reaction. Pincus [1997] adopts the conventional wisdom of the tax benefits hypothesis and assumes a net positive effect. He assumes that any negative implications from lobbying activities are captured in firms’ share prices prior to the critical events of the LIFO legislative process. 13 A by-product of the development of the legislative history is a reexamination of arguments for and against LIFO use made by academics and practitioners.

30

would exercise its option to adopt LIFO. This provides additional insight into why prior research has had

only limited success in gauging investor reactions associated with LIFO adoptions.

5.0 WHAT IMPACT DOES INVENTORY METHOD CHOICE HAVE ON THE SECURITY

PRICES?

Researchers have also investigated the impact, if any, alternative inventory methods have on the

security prices. The authors of the leading intermediate accounting text assert that stock “prices are the

same and, in some cases, even higher under LIFO in spite of lower reported earnings” [Kieso and

Weygandt, 1998, 420]. However, the research literature is more equivocal.

5.1 IS THERE A RELATION BETWEEN INVENTORY METHODS AND P/E RATIOS?

One approach taken in the literature is to examine price-earnings ratios. Beaver and Morse

[1978, 65] had earlier concluded that “the most likely explanation of the evident persistence in price-

earnings ratios is not growth or risk, but differences in accounting method.” Lee [1988] argues that since

earnings under LIFO should be lower than FIFO-based earnings, and further, since the lower earnings

under LIFO lead to lower tax payments and presumably higher stock prices, price-earnings ratios should

be higher for LIFO firms than for FIFO firms.

Lee [1988] examines cross-sectional differences in earnings-price (EP) ratios of LIFO and

FIFO firms using a sample of 3,419 firm/years in which firms cannot have changed methods during

a five-year period centered on the event year. He uses EP ratios instead of PE ratios because the

latter suffer from a problem of interpretation when negative earnings are in the denominator. He

predicts LIFO firms will have lower EP ratios compared to FIFO firms.

Lee bases his analysis on the Gordon-Shapiro [1956] valuation equation, which controls for risk

and earnings growth, and he also considers controls for factors that may possibly confound the impact of

using LIFO on the EP ratio (e.g., firm size and industry). Lee finds that LIFO firms have higher EP ratios

than FIFO firms do in all 16 years he examines (significantly so in 11 years). These results contradict his

expectations, and he cautions that some unidentified economic phenomenon or an error in the model

design may explain what he terms an anomalous result.

Biddle [1988] argues that Lee misstates the intuition of the EP ratio as it relates to LIFO firms.

First, because Lee makes no adjustments for the impact that different inventory methods have on reported

income, real earnings growth may be obscured. In addition, Biddle points out firms using LIFO and firms

31

using FIFO may differ in fundamental ways. Lacking a complete theory about joint inventory method and

EP ratio determinants, omitted variables could explain the observed differences in EP ratios. Furthermore,

Biddle suggests that causality could run in the opposite direction; i.e., firms with higher earnings (and EP

ratios) may have incentives to adopt LIFO. Overall, he views Lee’s EP results as evidence of another

dimension along which LIFO and FIFO firms differ systematically.14

5.2 DO INVENTORY METHODS IMPACT EARNINGS RESPONSE COEFFICIENTS?

Another approach to examining the relation of inventory choice and share prices is to examine

earnings response coefficients (ERCs), which capture the impact of unexpected earnings on unanticipated

security returns. An early paper in this area is Collins and Salatka [1983] who study the impact of SFAS

No. 52 versus SFAS No. 8 (i.e., foreign currency translation adjustment alternatives) on the “quality” of

earnings as measured by their ERCs. Pincus [1993] investigates whether differing market reactions to

earnings announcements are associated with voluntary accounting method choices. With regard to

inventory methods, the conventional wisdom is that since LIFO approximates the matching of current costs

against current revenues, it yields a less “noisy” measure of operating income than FIFO-based earnings.

On the other hand, inventory reductions induce noise under LIFO since revenues and costs of goods sold

can move in opposite directions.15 Furthermore, FIFO-based earnings includes realized inventory profits.

Although they are undoubtedly more transitory in nature than the components of LIFO-based operating

income, they reflect management’s purchasing (as distinct from selling) performance, which may be

informative but is excluded from LIFO earnings (see Drake and Dopuch [1965]). Thus, alternative

inventory methods may differentially affect the market’s assessment of and reaction to earnings.

Pincus’s [1993] sample is a set of firms that did not change their accounting methods for

inventories, depreciation, investment tax credits, and treatment of leases over a six-year period. There is

little support for individual accounting method choices having a pervasive first-order effect on firm-specific

ERCs. However, there is some indication that firms using all conservative accounting methods, which

includes LIFO, have larger ERCs than firms with accounting method portfolios comprised of all income

accelerating choices, including FIFO.

14 Ceteris paribus, EP ratios would be higher for LIFO firms given the Kang model’s prediction of lower share prices due to the implementation and maintenance costs of a LIFO system. 15 There is a literature on LIFO liquidations in the capital markets arena that we do not review here. Papers include Stober [1986] and Tse [1990].

32

In two unpublished papers, Carroll, Collins, and Johnson [1993] and Pincus and Wasley [1993]

investigate whether changes in firms’ inventory methods are associated with a shift in their ERCs. Pincus

and Wasley [1993] analyze 250 LIFO adopting firms from 1973-87, of which 63% are from 1974. They

construct a matched control sample of non-adopters on the basis of pre-change inventory method,

industry, and size. Based on a cross-sectional regression, they find results indicating a significant decline,

both absolutely and relative to a matched control sample of non-LIFO adopters, in the cross-sectional

average ERC of LIFO adopters in the period following the change. This is opposite the conventional

wisdom that LIFO-based earnings should be more informative. In a sensitivity analysis, they partition the

sample into 1974 adopters and non-1974 adopters. ERCs of the 1974 firms decrease significantly after

adoption, while the mean ERC on non-1974 adopters increases after the change.

Carroll et al. [1993] perform both time-series and cross-sectional tests. They consider both LIFO

adoptions (n = 30, of which 17 are from 1974) and LIFO abandonments (n = 30) over the 1970-87

period. They also form corresponding control samples matched by industry and time-period and control

for the amount of inventory holding gains imbedded in FIFO earnings. They find ERCs decrease

significantly after firms abandon LIFO and ERCs of firms adopting LIFO increase significantly after the

switch. These results support the conjecture that LIFO earnings are of higher quality than non-LIFO

earnings. Further, more of the variation in returns is explained when LIFO is used than under a FIFO

regime. They also observe smaller market reactions when LIFO earnings reflect liquidation profits. The

conflicting results in the two studies, which appear to be centered on the 1974 adopters, may be due to

differing sample sizes and/or empirical models.

However, one can ask whether the ERC changes these studies observe reflect something other

than (or in addition to) the change in the way companies measure earnings. It is likely that firms change

inventory accounting methods because some fundamental characteristics of their operations have changed;

e.g., changes in product mix or input prices. This, of course, is simply a restatement of the self-selection

(i.e., endogeneity) problem that plagues the study of discretionary corporate actions in general. If this is

what has occurred, then changes in ERCs may be driven by the changes in underlying economic factors

and not by the change in the accounting method.16 At one level, documenting that the market’s

16 The question is when the market would price the economic factors that contribute to the accounting change. Both Carroll et al. [1993] and Pincus and Wasley [1993] estimate post-change ERCs over a number of earnings announcement dates. Hence, for the post-change ERCs to reflect the underlying factors contributing to the accounting change, one would have to argue that the pricing

33

interpretation of earnings changes coincident with changes in underlying economic fundamentals -- and the

accounting change -- is reassuring. However, the challenge for future research is to design tests to isolate

the change due to economic factors and the change, if any, due to the informational effects of a change in

accounting methods.

It may seem that as-if calculations would be sufficient to address the endogeneity issue. As we

have previously noted, an important aspect of many inventory papers is the use of as-if calculations to

adjust reported financial statement amounts for the effects of an alternative accounting method. However,

these calculations can require numerous simplifying assumptions, and they additionally assume that

operating, investing, and financing activities remain the same (i.e., that they would be unaffected by a

change of inventory methods). It has long been recognized (e.g., Ball [1972, 1980]; Sunder [1973]) that

such a ceteris paribus assumption is problematic because the choice of method is endogenous and firms

self-select into LIFO or FIFO samples.

The difficulty is that it is not possible to observe what managers’ decisions would be under

inventory costing methods different from the ones their firms currently use, and models of such managerial

action are unavailable.17 The as-if assumptions imply restatements are not based on a model of what

managers’ decisions would have been had they used the alternative inventory method; rather, operating,

investing, and financing decisions are assumed not to change. Furthermore, while the use of the LIFO

Reserve disclosure, if available, reduces measurement error relative to estimation algorithms, it does not

eliminate the endogeneity problem. The LIFO Reserve is itself an as-if restatement since it is based on the

inventory levels of a firm that uses LIFO. Accordingly, care must be taken in comparisons of LIFO firms’

reported and as-if numbers, and even more caution should be exercised when comparing LIFO firms and

FIFO firms (even when using as-if numbers) since they likely differ in fundamental ways. The lack of

models of managerial action means the researcher cannot verify how successful as-if adjustments are in

restating firms’ financial statement numbers to amounts they would report if they actually used the

alternative inventory method.

5.3 WHAT IS THE IMPACT OF INVENTORY METHODS ON FIRM VALUE?

Guenther and Trombley [1994] attempt to isolate the effect of inventory method choice on firm

effects of such factors corresponds with the subsequent earnings announcements. 17 Biddle [1980] examines managers’ inventory operating decisions before and after adopting LIFO. Also see Biddle and Martin [1985].

34

value. Their analysis is structured using the difference in book value from using LIFO versus FIFO. This

is reflected in the LIFO Reserve, which is the difference between the replacement (i.e., current or FIFO)

cost of inventory and the LIFO carrying value of inventory. Guenther and Trombley investigate the

incremental informativeness of LIFO Reserve amounts by developing hypotheses based on the existence

of unrecorded (“unbooked”) assets as well as the functional fixation hypothesis and Kang’s [1993] real

value model. Firm value is assumed to be based solely on reported balance sheet numbers under the

functional fixation hypothesis. Hence, this hypothesis predicts no relation between the LIFO Reserve and

firm value. The unrecorded asset hypothesis views the LIFO Reserve as a hidden or unbooked asset. By

adding it to a firm’s recorded current assets one can derive the firm’s total value, and the unrecorded asset

hypothesis predicts a positive relation between the LIFO Reserve and firm value. Finally, the authors

interpret Kang’s model as predicting a negative relation between the LIFO Reserve and firm value; e.g., to

the extent LIFO adoption costs are related to LIFO tax benefits, the LIFO Reserve may proxy for

adoption costs since tax benefits are a function of the LIFO Reserve.

Guenther and Trombley examine 2,240 firm/years using a modified version of Ohlson’s

[1991] model in which market value of equity is a linear function of the book value of equity,

current earnings, current dividends, and a variable representing other value-relevant factors.

Guenther and Trombley add the LIFO Reserve as one of these other value-relevant factors, and

also allow for different coefficients on book value of equity and earnings for FIFO and LIFO firms.

The results from annual and pooled cross-sectional regressions indicate that the coefficient on the

LIFO Reserve is significantly negative. The coefficients of all other variables in the model have the

signs predicted under the Ohlson framework, with LIFO firms having a reliably larger earnings

coefficient and smaller book value coefficient.

Guenther and Trombley’s finding that firms with larger LIFO Reserves have lower stock prices is

inconsistent with investors treating the LIFO Reserve as an unrecorded asset and with the functional

fixation hypothesis. One possible interpretation of these results is based on their observation that LIFO

Reserve amounts are positively related to LIFO firms’ inventory levels. Bernard and Noel [1991] find that

large build-ups of inventory are negatively related to future earnings. If the magnitude of the LIFO

Reserve reflects inventory build-ups, there could be a negative relation between the LIFO Reserve and

firm value. In the same vein, investors may view inventory increases as reflecting poor inventory

35

management. As noted, another possibility is the LIFO reserve may proxy for the costs of implementing

and maintaining a LIFO system. Additionally, the negative valuation of the LIFO Reserve may represent a

firm’s tax vulnerability to future LIFO layer liquidations and/or price declines.

Jennings, Simko, and Thompson [1996] also investigate the influence of inventory methods on firm

valuation for a sample of LIFO firms. They examine whether reported LIFO income statements explain

more of the cross-sectional variation in equity values of LIFO firms than do as-if non-LIFO income

statements derived from LIFO Reserve disclosures. Similarly, they examine whether as-if non-LIFO

balance sheet components have greater explanatory power for share prices than reported LIFO balance

sheet amounts.

They estimate annual cross-sectional regressions over the 1976-91 period, using an average

of 544 LIFO firms per year. They regress LIFO firms’ market values against income statement

components based (1) on reported LIFO amounts and (2) on non-LIFO amounts computed using

the firms’ LIFO Reserve disclosures. They perform the same kind of analysis using major balance

sheet categories as independent variables. Adjusted R2s for the reported LIFO income statement

variables are slightly (but significantly) higher than those for the same firms’ regressions based on

as-if non-LIFO income statement amounts, as expected, with the greatest differences coming

during the more inflationary period of 1978-81. The results are consistent with the hypothesis that

LIFO cost of goods sold is a somewhat more useful indicator of future resource outflows than non-

LIFO measures.

The authors analyze the components of LIFO cost of goods sold; i.e., as-if non-LIFO cost of

goods sold and the change in the LIFO Reserve. Both prove to have similar (and significant) pricing

implications and thus their aggregation into reported LIFO cost of goods sold does not result in a loss of

information. Note that, in part, the change in the LIFO Reserve reflects the change in current period

inventory input prices, which is often seen as being a more transitory component of non-LIFO-based

earnings. Yet Jennings et al.’s [1996] results indicate the change in the LIFO Reserve is priced in the

same way as as-if non-LIFO cost of sales.

Turning to the balance sheet analysis, the regression results reveal the unexpected result of slightly

(but significantly) smaller adjusted R2s for the non-LIFO balance sheets vis-à-vis the balance sheets based

on reported LIFO amounts. Moreover, consistent with Guenther and Trombley, LIFO firms’ equity

36

values vary inversely with the levels of the LIFO Reserve, and further analysis makes clear that information

is lost when the LIFO Reserve and the LIFO inventory carrying value are aggregated into an estimate of

the current cost of inventory. Additional analyses yield some evidence that the coefficient on the LIFO

Reserve is greater for those firms more able to pass on price increases compared to firms classified as less

able to do so.

In particular, Jennings et al. [1996] develop a simple model that suggests that the

coefficient on the LIFO Reserve in the balance sheet regression can be interpreted as capturing the

ability a firm has to respond to input price changes by changing its output prices. They rank their

sample firms into groups of firms “more able” or “less able” to pass on price increases using three

proxies: variance of the gross profit ratio; correlation of gross profit ratio with the CPI; and

average cost response estimates adapted from Hopwood and Schaefer [1989]. They re-run the

market value on balance sheet components regression taking account for a firm’s ability to pass

through price increases.

Future research is needed to investigate further the result of a negative relation between firm value

and the LIFO Reserve. The suggestion that the LIFO Reserve proxies for (ongoing) adoption costs

seems to ignore economies of scale aspects of inventory management and underscores how little is known

about the costs of operating LIFO inventory systems. Similarly, the extent to which LIFO Reserve levels

represent inventory build-ups has not been examined, and it raises the question how build-ups by FIFO

firms are reflected and priced by the market. Finally, the claim that the coefficient on the LIFO Reserve

captures a firm’s ability to pass on cost increases to customers needs to be more rigorously studied.

6.0 WHAT IS THE RELATION OF LIFO TO INCOME SMOOTHING?

Because LIFO effectively purges realized inventory profits from reported earnings, it yields

(roughly) a matching of current costs and current revenues. This means that gross margin percentages

likely exhibits less variability over time when measured using LIFO versus if measured using FIFO. Thus,

LIFO should dampen fluctuations in -- i.e., smooth -- reported earnings. This has long been asserted.

For example, Paton [1938] argued that LIFO “represents nothing more nor less than a major device for

equalizing earnings, to avoid showing in the periodic reports the severe fluctuations which are inherent in

certain business fields…[I]t is not good accounting to issue reports for a copper company, for example,

which make it appear that the concern has the comparative stability of earning power of the American

37

Telephone and Telegraph Co.” [1938, 199-200].18 Thus, it is reasonable to expect that if firms wish to

smooth earnings, they will consider adopting LIFO. What makes this potentially even more interesting is

that we know there are tax consequences from adopting LIFO. Thus, the smoothing and tax aspects of

LIFO may be reinforcing or conflicting.19

Results in Healy, Kang, and Palepu [1987] and Abdel-kahlik [1985] are inconclusive as to

whether compensation committees of corporations’ boards of directors make compensation decisions

using reported earnings or earnings adjusted for accounting changes. If adjustments are not made,

managers have an incentive to manage earnings to achieve some desired objective with regard to their

compensation. One such objective may be a desire to minimize variation in their compensation function,

and often this can be accomplished through minimizing the variation in their company’s earnings. Moses

[1987] hypothesizes that there are various firm-specific factors that influence management to engage in

income smoothing techniques. He examines accounting changes and how they can act as smoothing

mechanisms. He considers firm size, the existence of a bonus plan, and the divergence of actual earnings

from expected levels. He examines the relation between smoothing techniques and proxies generally used

for these firm-specific economic factors.

When attempting to smooth income, manager presumably prefer a less conspicuous to a

more conspicuous technique. Discretionary accounting changes undoubtedly fall in the latter

category since they must be disclosed along with their year-of-change earnings effect, and if

sufficiently material, they will also be mentioned in the auditor’s report. Of course, discretionary

changes are not necessarily made with the intent to smooth income. The smoothing techniques

Moses considers include LIFO adoptions and changes in pension methods or assumptions,

amortization or depreciation methods, and expense versus capitalization policies. The initial

sample comes from Accounting Trends and Techniques for the years 1975-80. Using a random

walk model, Moses calculates expected earnings and assumes managers smooth pre-change

earnings toward expected earnings. He develops a measure of smoothing as the difference

18 Copper and other nonferrous metals firms lobbied actively for LIFO’s acceptance for tax purposes and were, along with petroleum companies, among the earliest adopters of LIFO. 19 In addition, having adopted LIFO means managers have additional opportunities to manage earnings. For example, purchases made or postponed at year-end directly affect cost of goods sold and thus income under LIFO (e.g., Frankel and Trezevant [1994]). This is not the case under FIFO. Managers potentially can also manipulate the composition of LIFO pools or more generally allow for or prevent inventory liquidations to achieve earnings management goals.

38

between the absolute value of pre-change earnings minus expected earnings and the absolute value

of reported earnings minus expected earnings. Positive values are consistent with smoothing.

The literature has identified many factors as being associated with smoothing in addition to

compensation incentives. Dampening fluctuations in reported earnings may allow firms to avoid calling

unnecessary attention to themselves from regulatory bodies. This is a form of the political costs

hypothesis. Similarly, smoothing may occur when a firm’s market share increases to a level that indicates

possible antitrust activity or perhaps to discourage takeovers. In his analysis, Moses uses various proxies

for these factors since they too may motivate smoothing behavior.

The accounting changes in the sample of 212 firms are first separated by whether they

qualify as smoothing or not; i.e., whether the smoothing measure is positive. Moses finds about

65% of the accounting changes qualify as smoothing changes. His multivariate test consists of an

OLS regression with the measure for smoothing behavior as the dependent variable. The proxies

for the economic factors and three earnings variables measuring deviation from expected earnings,

variability in earnings, and directional impact of the accounting change are included as

independent variables. The sample is divided into high and low pre-change earnings deviation

groups, and regressions are run on the full sample and by type of accounting change.

Results for both univariate and multivariate analyses indicate that smoothing behavior is significantly

more common in firms that are large, have bonus plans, have large divergences in pre-change earnings,

and make income-decreasing accounting changes. When the sample is partitioned by accounting changes,

the independent variables align themselves by smoothing techniques. With particular reference to LIFO

adoptions, firm size, existence of bonus plans, and the directional impact of the change are all significant.

Moreover, due to the large sample size and the fact that the earnings effect is a decrease in

earnings for almost all firms, Moses runs additional tests on the LIFO subsample. He partitions

firms with pre-change earnings into above or below expected earnings groups and finds that larger

earnings reductions occur for firms with pre-change earnings above expectations. This result is

consistent with a preference for adjusting toward rather than away from expected earnings.

Consistent with larger deviations influencing larger downward adjustments, the earnings impact is

found to increase with the magnitude of unexpected earnings.

Moses points out that he only considers smoothing in a single period and the independent variables

39

he uses are imprecise proxies for the factors that likely affect smoothing behavior.

Hunt, Moyer, and Shevlin [1996] test the hypothesis that managers use LIFO to obtain sometimes

conflicting tax and financial reporting benefits. They analyze a sample of long-time LIFO users by

developing a two-period model using a simultaneous equations approach. Hunt et al.’s model is

developed assuming managers have specific objectives in mind when they choose LIFO. The objectives

modeled are smoothing reported earnings, minimizing debt-related costs, and minimizing taxes. Other

objectives are assumed to be inconsequential and accounting adjustments to other than inventory, other

current assets, or depreciation are assumed to be prohibitively costly.

The sample includes 240 U.S. firms who by 1991 had used LIFO as their primary inventory

method continuously for at least 15 years. The average sample firm is larger and more profitable

than the average firm in the Compustat population. The authors use a three-equation

simultaneous regression model. The dependent variables are what they refer to as adjustable

accounting measures. These are accounting adjustments for (1) inventory (specifically, the LIFO

Reserve), (2) current accruals other than inventory, and (3) depreciation. These variables are also

one of the independent variables in each of the other’s regression functions. For periods one and

two, the difference in net income, shareholders equity, and as-if net income and shareholder’s

equity had the accounting adjustments not been made are also included as independent variables.

They also control for the earnings effects of non-discretionary inventory changes.

The hypothesis tested is aligned with achieving managers’ objectives. Thus, managers are

expected to adjust inventories, other current accruals, and depreciation to minimize the deviation from a

targeted level of stockholders’ equity that minimizes debt-related costs. Similarly, managers will adjust

inventories, other current accruals, and depreciation to smooth earnings, and they are predicted to increase

LIFO inventories and other current accruals to minimize the present value of tax payments. Moreover,

reversals of inventory and other current accrual adjustments affect taxes in the subsequent period. Hence,

Hunt et al. expect that the larger the subsequent period’s effective marginal tax rate, the less likely

managers are to decrease accounting measures to lower the current period’s taxable income. With regard

to depreciation, they assume different methods are used for tax and book purposes, and thus book

adjustments for depreciation have no effect on tax payments.

Hunt et al. find that inventory and other current accruals are used to smooth income. The results

40

indicate that management anticipates a reversal of their actions in the subsequent period. This

interpretation is correct assuming the model adequately controls for the predetermined or non-

discretionary levels of each adjustable accounting measure. The results also indicate that the average

long-time LIFO user does not adjust depreciation to smooth current period income. However,

when the depreciation adjustments are positively autocorrelated, there is evidence of income

smoothing. Managers also adjust inventories and other current accruals to lower debt-related

costs, but again it appears depreciation adjustments are not used for this purpose.

Other current accruals are adjusted to modify current tax payments. However the coefficient on

inventory is insignificant, implying that inventories are not adjusted to obtain tax savings beyond those

already being reaped by having adopted the LIFO method. This result differs from the results of prior

research on LIFO liquidations.20 Other current accruals and inventories are significant in one another’s

regressions, implying that they interact to achieve the same objective. Both are also significant in the

depreciation regression; however, depreciation is not significant in either of the other regressions. This

suggests that other current accruals and inventory influence depreciation adjustments, but the reverse is not

true.

The simultaneous equation approach allows for the analysis of multiple accounting measures to

achieve multiple objectives. The assumptions that adjustment of accounting measures excluded from the

model is prohibitively costly and that failing to meet objectives excluded from the model is costless seem

stringent. Nevertheless, the results suggest that firms finding it optimal for tax purposes to use LIFO

exploit the discretion allowed under LIFO to trade-off additional tax savings in favor of achieving a

smoother time-series in earnings and lower debt-related costs.

7.0 SUMMARY AND CONCLUSIONS

In this paper, we survey research of the past decade or so to update our understanding of the

choice and impact of the LIFO/FIFO decision. The major research areas we consider are capital market

reactions to LIFO adoptions, theoretical and empirical examinations of non-tax explanations of the LIFO

20 Several studies find that tax minimization, earnings management, and reducing the likelihood of debt covenant violations all seem to influence the timing of LIFO liquidations. For example, Dhaliwal, Frankel, and Trezevant [1994] use a multivariate tobit model to regress the amount of LIFO liquidation income on certain explanatory variables, and they also use a probit model where the dependent variable is a binary variable equal to 1 if a LIFO liquidation has occurred and 0 otherwise. They observe the following. The lower the tax status of the firm, the higher the probability and magnitude of a LIFO liquidation. The probability of LIFO liquidation is positively related with the negative change in earnings and the variability of earnings. There is also a negative

41

puzzle, Kang’s [1993] model and the capital market research it inspired, share price behavior surrounding

events leading to LIFO’s entry into the tax code, inventory method choices and the price-earnings relation

and firm value, and income smoothing.

Significant advances have been made and our understanding of LIFO’s impact on stock prices

and of the LIFO puzzle is very different from what it was just a few years ago. Research now suggests

that (a) there is not really much of a LIFO puzzle, (b) the capital market reaction to LIFO adoptions is

more complex than the analysis of the traditional tax benefits versus functional fixation hypotheses suggests

and may not yield meaningful assessments of LIFO’s impact on share prices, and (c) the implications of

LIFO use for the price--earnings relation are less straight-forward than what one might expect.

Throughout the paper we identify and discuss a number of research issues and make suggestions for future

research. One final research suggestion is that the sustained lower rates of price increase over the past

decade coupled with the drive to reduce inventory levels through the use of just-in-time inventory methods,

should make LIFO use less attractive. One would expect, for example, to observe increasing levels of

LIFO abandonments in industries where such trends are most pronounced.

association between LIFO liquidation and current debt to total asset ratio and sales growth. Also see Tse [1994] and Cottell [1986].

42

REFERENCES

Abdel-khalik, A.R. 1985. The effect of LIFO-switching and firm ownership on executives’ pay.

Journal of Accounting Research (Autumn): 427-47.

--------. 1992. Editor’s comment. The Accounting Review (April): 319.

-------- and J.C. McKeown. 1978. Understanding accounting changes in an efficient market:

Evidence of a differential reaction. The Accounting Review (October): 851-68.

Accounting Standards Executive Committee. 1984. Issues Paper: Identification and Discussion of

Certain Financial Accounting and Reporting Issues Concerning LIFO Inventories (New

York: American Institute of CPAs).

Aharony, J. and S. Bar-Yosef. 1987. Test of the impact of LIFO adoption on stockholders: A

stochastic dominance approach. Contemporary Accounting Research (Spring): 430-44.

Amershi, A.H. and S. Sunder. 1987. Failure of stock prices to discipline managers in a rational

expectations economy. Journal of Accounting Research (Autumn): 177-95.

Ball, R. 1972. Changes in accounting techniques and stock prices. Journal of Accounting Research

(Supplement): 1-38.

--------. 1980. Discussion of Accounting for research and development costs: The impact of research

and development expenditures. Journal of Accounting Research (Supplement): 27-37.

Bar-Yosef, S. and P.K. Sen. 1992. On optimal choice of inventory accounting method. The

Accounting Review (April): 320-36.

Beaver, W. and D. Morse. 1978. What determines price-earnings ratios? Financial Analysts Journal

(July/August): 65-76.

Bernard, V. and J. Noel. 1991. Do inventory disclosures predict sales and earnings? Journal of

Accounting, Auditing, and Finance (Spring): 145-81.

Biddle, G.C. 1980. Accounting methods and management decisions: The case of inventory costing

and inventory policy. Journal of Accounting Research (Supplement): 235-80.

--------. 1988. Discussion of Inventory accounting and earnings/price ratios: A puzzle.

Contemporary Accounting Research (Fall): 389-96.

-------- and F. Lindahl. 1982. Stock price reactions to LIFO adoptions: The association between

excess returns and LIFO tax savings. Journal of Accounting Research (Autumn II): 551-88.

43

-------- and R.K. Martin. 1985. Inflation, taxes, and optimal inventory policies. Journal of

Accounting Research (Spring): 57-83.

-------- and W.E. Ricks. 1988. Analysts forecast errors and stock price behavior near the earnings

announcement dates of LIFO adopters. Journal of Accounting Research (Autumn ): 169-94.

Bowen, R.M, L. DuCharme, and D. Shores. 1995. Stakeholders’ implicit claims and accounting

method choice. Journal of Accounting and Economics (December): 255-95.

Brown, R.M. 1980. Short range market reaction to changes to LIFO accounting using preliminary

earnings announcement dates. Journal of Accounting Research (Spring): 38-63.

Bull, C. 1987. The existence of self-enforcing implicit contracts. Quarterly Journal of Economics

(February): 147-59.

Butters J. and P. Niland. 1949. Effects of Taxation: Inventory Accounting and Policies (Cambridge,

MA: Riverside Press).

Carroll, T.J., W.B. Johnson, and D.W. Collins. 1993. The LIFO/FIFO choice and the quality of

earnings signals. Working paper, The University of Iowa.

Collins, D.W., and W. Salatka. 1993. Noisy accounting signals and earnings response coefficients:

The case of foreign currency accounting. Contemporary Accounting Research (Fall): 119-

59.

Cottell Jr., P.G. 1986. LIFO layer liquidation: Some empirical evidence. Journal of Accounting,

Auditing, and Finance (Winter): 30-45.

Cushing, B.E. and M.J. LeClere. 1992. Evidence on the determinants of inventory accounting policy

choices. The Accounting Review (April): 355-66.

Dhaliwal, D.S., M. Frankel, and R. Trezevant. 1994. The taxable and book income motivations for a

LIFO layer liquidation. Journal of Accounting Research (Autumn): 278-89.

Dopuch, N. and M. Pincus. 1988. Evidence on the choice of inventory accounting methods: LIFO

versus FIFO. Journal of Accounting Research (Spring): 28-59.

Drake, D.F. and N. Dopuch. 1965. On the case for dichotomizing income. Journal of Accounting

Research (Autumn): 192-205.

Elliott, J.A. and D.R. Philbrick. 1991. Accounting changes and earnings predictability. The

Accounting Review (January): 157-74.

44

Fellingham, J. 1988. Discussion of The LIFO/FIFO choice: An asymmetric information approach.

Journal of Accounting Research (Supplement): 59-62.

Frankel, M. and R. Trezevant. 1994. The year-end LIFO purchasing decision: An empirical test. The

Accounting Review (April): 382-98.

Gonedes, N.J. and N. Dopuch. 1974. Capital market equilibrium, information production, and

selecting accounting techniques: Theoretical framework and review of empirical work.

Journal of Accounting Research (Supplement): 48-129.

Gordon, M.J. and E. Shapiro. 1956. Capital equipment analysis: The required rate of profit.

Management Science (October): 102-10.

Guenther, D.A. and M.A. Trombley. 1994. The LIFO reserve and the value of the firm: Theory and

empirical evidence. Contemporary Accounting Research (Spring): 433-52.

Hagerman, R.L and M.E. Zmijewski. 1979. Some economic determinants of accounting policy

choice. Journal of Accounting and Economics (August): 141-61.

Hand, J.R.M. 1993. Resolving LIFO uncertainty: A theoretical and empirical reexamination of

1974-75 LIFO adoptions and nonadoptions. Journal of Accounting Research (Spring): 21-

49.

--------. 1995. 1974 LIFO excess stock return and analysts forecast error anomalies revisited.

Journal of Accounting Research (Spring): 175-91.

Healy, P.M., S. Kang, and K.G. Palepu, 1987. The effect of accounting procedure changes on

CEO’s cash salary and bonus compensation. Journal of Accounting and Economics

(April): 7-34.

Hilke, J.C. 1986. Regulatory compliance costs and LIFO: No wonder small companies haven’t

switched. Journal of Accounting, Auditing, and Finance (Winter): 17-29.

Hopwood, W. and T. Schaeffer. 1989. Firm-specific responsiveness to input price changes and the

incremental information in current cost income. The Accounting Review (April): 313-28.

Hughes, P.J. and E.S. Schwartz. 1988. The LIFO/FIFO choice: An asymmetric information

approach. Journal of Accounting Research (Supplement): 41-58.

--------, --------, and A.V. Thakor. 1994. Continuous signaling within partitions: Capital structure

and the LIFO/FIFO choice. Journal of Accounting, Auditing, and Finance (Winter): 1-19.

45

Hunt, A., S.E. Moyer, and T. Shevlin. 1996. Managing interacting accounting measures to meet

multiple objectives: A study of LIFO firms. Journal of Accounting and Economics

(June): 339-74.

Hunt III, H.G. 1985. Potential determinants of corporate inventory accounting decisions. Journal of

Accounting Research (Autumn): 448-67.

Jennings, R., D.P. Mest, and R.B. Thompson II. 1992. Investor reaction to disclosures of 1974-75

LIFO adoption decisions. The Accounting Review (April): 337-54.

--------, P.J. Simko, and R.B. Thompson II. 1996. Does LIFO inventory accounting improve the

income statement at the expense of the balance sheet? Journal of Accounting Research

(Spring): 85-109.

Johnson, W.B. and D.S. Dhaliwal. 1988. LIFO abandonment. Journal of Accounting Research

(Autumn): 236-72.

Jung, W-O. 1989. Strategic choice of accounting methods. Contemporary Accounting Research

(Fall): 1-25.

Kang, S-H. 1993. A conceptual framework for the stock price effects of LIFO tax benefits. Journal

of Accounting Research (Spring): 50-61.

Kieso, D.E. and J.J. Weygandt. 1998. Intermediate Accounting, 9th edition (New York: John

Wiley).

Lanen, W.N. and R. Thompson. 1988. Stock price reactions as surrogates for net cash flow effects

of corporate policy decisions. Journal of Accounting and Economics (December): 311-34.

Lee, C-W.J. 1988. Inventory accounting and earnings/price ratios: A puzzle. Contemporary

Accounting Research (Fall): 371-88.

-------- and D.A. Hsieh. 1985. Choice of inventory accounting methods: Comparative analyses of

alternative hypotheses. Journal of Accounting Research (Autumn): 468-85.

Lev, B. and J.A. Ohlson. 1982. Market-based empirical research in accounting: A review,

interpretations, and extension. Journal of Accounting Research (Supplement): 249-322.

Lindahl, F.W. 1989. Dynamic analysis of accounting choice. Journal of Accounting Research

(Autumn): 201-26.

--------, C. Emby, and R.H. Ashton. 1988. Empirical research on LIFO: A review and analysis.

46

Journal of Accounting Literature (Spring): 310-32.

Morse, D. and G. Richardson. 1983. The LIFO/FIFO decision. Journal of Accounting Research

(Spring): 106-27.

Moses, O.D. 1987. Income smoothing and incentives: Empirical tests using accounting changes.

The Accounting Review (April): 358-77.

Niehaus, G.R. 1989. Ownership structure and inventory method choice. The Accounting Review

(April): 269-84.

Ohlson, J. 1991. Earnings, book values, and dividends in security valuation. Working paper,

Columbia University.

Paton, W. 1938. Comments on A statement of accounting principles. Journal of Accountancy

(March): 196-207.

Pincus, M. Legislative history of the allowance of LIFO for tax purposes. Accounting Historians

Journal (June): 23-55.

--------. 1993. Accounting methods and differential stock market response to the announcement of

earnings. Journal of Accounting, Auditing, and Finance (Summer): 221-48.

--------. 1997. Stock price effects of the allowance of LIFO for tax purposes. Journal of Accounting

and Economics (November): 283-308.

-------- and C. Wasley. 1993. LIFO adoptions and the informativeness of earnings. Working paper,

Washington University (St. Louis).

-------- and --------. 1994. The incidence of accounting changes and characteristics of firms

making accounting changes. Accounting Horizons (June): 1-24.

-------- and --------. 1996. Stock price behavior associated with post-1974-75 LIFO adoptions

announced at alternative disclosure times. Journal of Accounting, Auditing, and Finance

(Fall): 535-64.

Purtill, J.S. LIFO Inventory: Tax and Accounting Issues (New York: American Institute of CPAs).

Ricks, W.E. 1982a. Market assessment of alternative accounting methods: A review of the empirical

evidence. Journal of Accounting Literature (Spring): 59-102.

--------. 1982b. The market’s response to the 1974 LIFO adoptions. Journal of Accounting

Research (Autumn I): 367-87.

47

Scholes, M. and M. Wolfson. 1992. Taxes and Business Strategy (Englewood Cliffs, NJ: Prentice-

Hall).

Skinner, D.J. 1993. The investment opportunity set and accounting procedure choice: Preliminary

evidence. Journal of Accounting and Economics (October): 407-45.

Stober, T.L. 1986. The incremental information content of financial statement disclosures: The case

of LIFO inventory liquidations. Journal of Accounting Research (Supplement): 138-60.

Sunder, S. 1973. Relationship between accounting changes and stock prices: Problems of

measurement and some empirical evidence. Journal of Accounting Research (Supplement):

1-45.

Tesler, L. 1980. Theory of self-enforcing agreements. Journal of Business (January): 27-44.

Tse, S. 1990. LIFO liquidations. Journal of Accounting Research (Spring): 229-38.

Watts, R.L. and J.L. Zimmerman. 1986. Positive Accounting Theory (Englewood Cliffs, NJ:

Prentice-Hall).

-------- and --------. 1990. Positive accounting theory: A ten year perspective. The Accounting

Review (January): 131-56.