RPS 11 Poin 2

Journal of Accounting and Economics 48 (2009) 54–68
Contents lists available at ScienceDirect
Journal of Accounting and Economics

journal homepage: www.elsevier.com/locate/jae
Economic consequences of firms’ depreciation method choice:

Evidence from capital investments$
×
Scott B. Jackson a, , Xiaotao (Kelvin) Liu b, Mark Cecchini a
a
School of Accounting, Moore School of Business, University of South Carolina, Columbia, SC 29208, USA
b
Northeastern University, Boston, MA 02115, USA
a r t i c l e i n f o a b s t r a c t
Article history:
Received 28 January 2008 This study identifies several interrelated reasons why firms’ depreciation method choice
Received in revised form is likely to influence managers’ capital investment decisions. We find that firms that use
23 May 2009 accelerated depreciation make significantly larger capital investments than firms that
Accepted 2 June 2009 use straight-line depreciation. Further, we find that there has been a migration away
Available online 9 June 2009
from accelerated depreciation to straight-line depreciation over the past two decades.
JEL classification: Firms that make such accounting changes make smaller capital investments in the
G31 post- change periods than in the pre-change periods. These results suggest that a
M21 choice made for external financial reporting purposes influences managers’ capital
M4 investment decisions.
1
& 2009 Elsevier B.V. All rights reserved.
Keywords:
Depreciation method choice
Capital investments
Straight-line depreciation
Accelerated depreciation
1. Introduction
Capital investment decisions are among the most important decisions entrusted to managers because those decisions
influence shareholder wealth (McConnell and Muscarella, 1985), firms’ long-term prospects for survival (Klammer et al.,
1991), and the overall economic welfare of society (Harris and Raviv, 1996). The sheer magnitude of capital investments
is enormous. The United States Census Bureau (2007) reports that businesses spent an aggregate of approximately
$7.5 trillion on capital investments or an average of almost $1.1 trillion per year during the period 1999 through 2005.
Given the importance of capital investments to firms, shareholders, and society, efforts to understand the factors that
influence managers’ capital investment decisions are of considerable practical importance.
This study examines whether the choice that firms make between straight-line depreciation and accelerated
depreciation influences managers’ capital investment decisions. From a normative perspective, internal management
decisions should focus on the incremental costs and benefits associated with alternative courses of action. While choices
that firms make for external financial reporting purposes are normatively irrelevant factors in making capital investment
decisions (Garrison and Noreen, 2003; Horngren et al., 2005; Titard, 1993), some academics have speculated that firms’
depreciation method choice may nonetheless influence those decisions (Dearden, 1960; Hatfield, 1944; Titard, 1993; Zajac,
1995). Evidence confirming the existence of such an association would suggest that a seemingly inconsequential choice
$
We gratefully acknowledge the comments of S.P. Kothari (the editor), an anonymous referee, Kin Blackburn, Tom Canace, Marc Caylor, Dutch
Fayard, Victoria Glackin, Noah Jackson, Scott Whisenant, and Rich White.
×
Corresponding author. Tel.: +1803777 3100; fax: +1803777 0712.
E-mail address: scott.jackson@moore.sc.edu (S.B. Jackson).
0165-4101/$ - see front matter & 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jacceco.2009.06.001
S.B. Jackson et al. / Journal of Accounting and Economics 48 (2009)
5
that firms make for external financial reporting purposes influences one of the most important internal decisions entrusted
to managers.
We draw on research in accounting and psychology to identify four interrelated reasons why firms’ depreciation
method choice is likely to influence managers’ capital investment decisions (i.e., loss aversion, earnings consequences,
perceived utility, and waste avoidance). These reasons are based on theories of individual behavior and derive from the
observation that, at any point during the life of a depreciable asset, accelerated depreciation results in a lower accounting
book value than straight-line depreciation. Our main research hypothesis is that firms that use accelerated depreciation
make larger capital investments than firms that use straight-line depreciation, ceteris paribus.
To test this hypothesis, we estimate a two-stage ‘‘treatment effects’’ model ( Greene, 2003; Maddala, 1983) that accounts
for the endogenous nature of firms’ depreciation method choice. In the first stage, we estimate a multivariate probit
model in which the dependent variable is firms’ depreciation method choice and the independent variables include
previously identified economic determinants of that choice. In the second stage, we estimate a regression in which the
dependent variable is capital investments and the independent variables include (i) an indicator for whether the firm uses
accelerated depreciation (the variable of interest), (ii) previously identified economic determinants of capital
investments, and (iii) the inverse Mills ratio, which accounts for the self-selection nature of firms’ depreciation method
choice. By including a large array of economic determinants in the second-stage regression, we control for the factors
that drive capital investments, thereby enabling us to make inferences about the incremental effect of firms’
depreciation method choice.
Consistent with our expectations, we find that the coefficient on the indicator for whether the firm uses accelerated
depreciation is significantly positive and economically meaningful. This finding indicates that accelerated depreciation
is associated with higher levels of capital investments than straight-line depreciation. To help corroborate this finding, we
also examine investments in research and development (R&D), which is a type of investment that should have a non-
positive relation with the use of accelerated depreciation. Consistent with our expectations, we find that the coefficient on
the indicator for whether the firm uses accelerated depreciation is non-positive. Thus, we find a positive relation between
the use of accelerated depreciation and investments when we expect that relation to be positive (i.e., in the case of capital
investments), and we find a non-positive relation between the use of accelerated depreciation and investments when we
expect that relation to be non-positive (i.e., in the case of R&D investments).
Further, we find that there has been a migration away from accelerated depreciation to straight-line depreciation over
the past two decades. The frequency of firms using accelerated depreciation for all or some of their depreciable assets has
declined from approximately 31 percent in 1988 to approximately 14 percent in 2006. Firms that change from
accelerated depreciation to straight-line depreciation make significantly smaller capital investments in the post-change
periods than in the pre-change periods. Taken together, our results suggest that firms’ depreciation method choice
influences one of the most important decisions that managers make—their decisions about investing scarce capital
resources.
In a review of the accounting choice research published during the 1990s, Fields et al. (2001) conclude that
‘‘yaccounting research has made modest progress in advancing the state of knowledge beyond what was known in the
1970s and 1980s.’’ They contend that the rate of research progress slowed in the 1990s because researchers continued to
replicate well-known results in slightly different settings. Perhaps one of the reasons that research on depreciation
method choice has slowed is because of the perception that ‘‘ydepreciation is one accounting issue where the effects of
the different methods are obvious and well understood’’ (Ricks, 1982, p. 71). An implication of our study is that the array
of consequences associated with different depreciation methods may be broader than prior accounting choice research
contemplates.
Prior research on firms’ depreciation method choice has focused on the market-related consequences and contracting
consequences of that choice. Beaver and Dukes (1973) find no evidence that investors fixate on earnings prepared under
different depreciation methods (i.e., the market appears to adjust for depreciation-method-induced differences in
earnings). Archibald (1972) and Kaplan and Roll (1972) find that changes from accelerated depreciation to straight-line
depreciation have no discernable stock price effect even though earnings are greater under the new method. In the
absence of market-related consequences, firms’ depreciation method choice may have economic consequences because
that choice alters firms’ reported earnings, thereby altering how firms’ cash flows are divided among contracting parties
(Fields et al., 2001; Holthausen and Leftwich, 1983; Holthausen, 1981; Leftwich, 1981; Ricks, 1982; Watts and Zimmerman,
1986). Our study contributes to the accounting choice literature by providing archival evidence that firms’ depreciation
method choice has economic consequences even in the absence of market-related consequences and contracting
consequences.
Although archival researchers have used a variety of economic determinants of capital investments in their empirical
models (Adam and Goyal, 2006; Barro, 1990; Richardson, 2006; Shin and Kim, 2002), there has been limited effort devoted
to understanding whether real-world capital investment decisions are influenced by variables of a non-economic nature.
A number of experimental studies document that managers’ capital investment decisions are influenced by factors of a
non-economic nature (Kida et al., 2001; Moreno et al., 2002; Sawers, 2005; Staw, 1976), but it is not clear whether the
results of those studies persist in real-world settings that involve actual economic consequences to both managers and
their firms. This study identifies an additional determinant of managers’ capital investment decisions (i.e., firms’
depreciation method choice) that is of non-economic nature and provides archival evidence that this determinant
influences managers’ capital investment decisions.
The remainder of this study proceeds as follows. Section 2 provides the theory and formulates our research
hypothesis. Section 3 describes our research methodology, and Section 4 describes our sample selection procedures. Section 5
discusses the main results and additional analyses. The final section provides concluding comments.
5
2. Theory and hypothesis
2.1. Capital investment decisions
It is widely accepted that managers should make decisions based on the incremental costs and benefits associated with
alternative courses of action (Garrison and Noreen, 2003; Hilton, 2002; Horngren et al., 2005). Consistent with this
notion, surveys of chief financial officers (CFOs) and other company officials indicate that firms increasingly use
sophisticated capital budgeting techniques (i.e., net present value) to evaluate capital investment options (Klammer et al.,
1991; Sangster, 1993; Schall et al., 1978). Graham and Harvey (2001) find that 75 percent of CFOs always or almost always
evaluate capital investment options using the net present value method. However, evidence suggests that managers’
capital investment decisions are not solely determined by the outcome of capital budgeting techniques (Klammer and
Walker, 1984).
Experimental research indicates that various psychological factors influence managers’ capital investment decisions.
For example, Kida et al. (2001) find that affective reactions cause managers to select capital investment options that have
lower net present values than other investment options under consideration. Moreno et al. (2002) find that the effect of
gain/loss domains on managers’ capital investment decisions is overwhelmed by managers’ affective reactions. Sawers
(2005) finds that managers tend to delay making capital investment decisions in response to choice difficulty. There is
also evidence suggesting that managers may escalate their commitment to poorly performing projects despite feedback
that the projects should be discontinued (Brockner, 1992; Staw, 1976).
2.2. Depreciation methods and capital investment decisions
Accounting depreciation is not intended to measure the deterioration of an asset or changes in an asset’s market value.
In recognition of this fact, many accounting textbooks (i) emphasize that accounting depreciation is a method of cost
allocation, not asset valuation (Kieso et al., 2007) and (ii) implore managers to disregard accounting book values when
making decisions (Garrison and Noreen, 2003; Hilton, 2002; Horngren et al., 2005; Titard, 1993). Despite this, we identify
four interrelated reasons why accelerated depreciation is likely to be associated with higher levels of capital investments
than straight-line depreciation. Each of the four reasons derives from the observation that, at any point during the life of a
depreciable asset, accelerated depreciation results in a lower accounting book value than straight-line depreciation.
The first reason why accelerated depreciation is likely to be associated with higher levels of capital investments
than straight-line depreciation is because assets depreciated under straight-line depreciation are more likely to result
in financial statement losses on replacement than are assets depreciated under accelerated depreciation. One of the
central implications of prospect theory is loss aversion (Kahneman and Tversky, 1979; Tversky and Kahneman, 1991),
which suggests that managers may attempt to delay recognizing losses by continuing to use existing assets rather than
investing in replacement assets. The notion that managers find it distasteful to dispose of assets at a loss has been
expressed repeatedly for decades. For example, Hatfield (1944, p. 66) states
It is always disagreeable to write off an asset. It seems almost self-evident that if, by a narrow margin, it is
desirable to discard an old machine with a book value of $1000 the scales would be tipped the other way if the
book value were
$5000. One might, in order to secure the more efficient machine, be willing to sacrifice $1000 but not to sacrifice
$5000. This seems mere common sense, not a matter of accounting.
Along similar lines, Dearden (1960, p. 83) states
If the assets are replaced or scrapped before they are fully depreciated, the division may have to show a loss in the
period that this action is taken. Although the accounting loss on a piece of equipment is not a valid consideration in
making a replacement decision (except as it reflects the timing of income tax payments), it does affect the
division’s profit and could influence the division manager’s decision.
More recently, Titard (1993, p. 386) states
ymanagement should overcome its concern about recognizing losses when it disposes of equipment before
the original estimated life has been completed...the firm can recognize the loss now or depreciate the remaining
book value over the remaining life of the equipment. But in the latter case, it loses the advantage of having a
superior machine, with the related cost savings.
The second reason why accelerated depreciation is likely to be associated with higher levels of capital investments
than straight-line depreciation is because assets depreciated under accelerated depreciation, relative to assets depreciated
under straight-line depreciation, will have more positive (or less negative) financial statement effects on replacement.1 A
growing body of evidence suggests that earnings considerations may influence managers’ investment decisions.
Graham et al.
1
While the first and second reasons are similar, the second reason explicitly addresses situations in which both depreciation methods result in (i)
gains of differing magnitudes or (ii) losses of differing magnitudes. Loss aversion only addresses the situation in which one method results in a loss
on replacement and the other results in a gain on replacement.
5
(2005) find that earnings considerations often supersede cash flow considerations when the firm is in danger of missing
the consensus forecast and that the objective of reporting smooth earnings may cause managers to forego real economic
value. Bhojraj and Libby (2005) find that increased capital market pressure may cause managers to choose projects that
maximize short-term earnings rather than projects that maximize total cash flows. Evidence also suggests that earnings
considerations influence firms’ R&D investments (Bushee, 1998; Dechow and Sloan, 1991; Shehata, 1991).
Third, the magnitude of an asset’s book value may influence the utility that managers perceive it will deliver in the
future. Assets with higher (lower) book values may be perceived to be capable of providing greater (lesser) future utility
and are therefore less (more) likely to be replaced. With respect to this issue, Zajac (1995, p. 240) states
yconventional depreciation can lead to resource misallocation. Equipment may be worth nothing on the
firm’s books, which might suggest that it should be replaced, when in fact it has considerable market value, and
equipment whose book value is high might be technologically obsolete and might have a market value of zero.
The views of Zajac (1995) are consistent with the experimental evidence reported in Jackson (2008), who finds that
assets depreciated using straight-line depreciation (i.e., assets with high book values) are perceived to be capable of
providing greater future utility than assets depreciated using accelerated depreciation (i.e., assets with low book values).
In turn, Jackson (2008) finds that depreciation-method-induced differences in managers’ utility perceptions influence
their asset replacement decisions.
The fourth reason focuses on perceptions of wastefulness. An asset with a high magnitude accounting book value may
cause managers to believe that they are engaging in wasteful behavior by replacing the asset. The results of Arkes and
Blumer (1985) and Arkes (1996) suggest that individuals find it unpleasant when they do not fully utilize what has been
purchased, and the results of Heath and Fennema (1996) suggest that consumers may overuse a product to get their
money’s worth out of it. Similarly, the results of Okada (2001) suggest that a consumer’s mental book value for a
durable good, which is the psychological analog of a business asset’s accounting book value, is felt as a cognitive cost of
replacing a durable good. In turn, Okada (2001) finds that a higher mental book value for a durable good may cause
consumers to postpone replacing it. If an asset’s accounting book value serves as a heuristic for evaluating (i) whether
replacing an asset is wasteful and (ii) whether the firm has gotten its money’s worth out of an asset, then an asset’s high
accounting book value may impede its replacement, while an asset’s low accounting book value may promote its
replacement.
Although there are reasons why accelerated depreciation may be associated with higher levels of capital investments
than straight-line depreciation, it is also possible that accelerated depreciation may be associated with lower levels of
capital investments. Immediately following new capital investments, firms that use accelerated depreciation will record
larger increments to depreciation expense arising from those investments than will firms that use straight-line
depreciation. As a result, firms that use accelerated depreciation may experience a decline in earnings relative to firms
that use straight-line depreciation. In turn, managers of firms that use accelerated depreciation may forego new capital
investments due to earnings considerations.
However, aggregate earnings may not be affected by new capital investments in the manner described above.
Depreciation expense reported on a firm’s income statement is comprised of depreciation expense related to both prior
period and current period capital investments. Under the accelerated depreciation method, increments to depreciation
expense associated with new capital investments can be completely offset by rapidly declining depreciation expense
associated with prior period capital investments. Thus, when one focuses on aggregate depreciation expense associated
with all capital investments rather than depreciation expense associated with current period capital investments, the
expense-related consequences of accelerated depreciation may not trigger reductions in capital investments.
Taken as a whole, the discussion in this section suggests that firms are likely to make larger capital investments when
they use accelerated depreciation than when they use straight-line depreciation. Our primary research hypothesis, stated
in alternative form, is as follows:
Hypothesis. Ceteris paribus, firms that use accelerated depreciation tend to make larger capital investments than firms
that use straight-line depreciation.
3. Methodology
3.1. Overview
Generally accepted accounting principles allow firms to choose their depreciation method, and evidence suggests that
firms that select straight-line depreciation differ significantly from firms that select accelerated depreciation (Bowen et
al., 1995; Christie and Zimmerman, 1994). To control for the non-random nature of firms’ depreciation method choice,
we estimate a two-stage ‘‘treatment effects’’ model (Greene, 2003; Maddala, 1983; Shehata, 1991) that accounts for the
endogenous nature of firms’ depreciation method choice. The two-stage ‘‘treatment effects’’ model has been used in
accounting contexts by, for example, Kim et al. (2003), Leuz and Verrecchia (2000), and Shehata (1991). In the first-
stage regression, we estimate a multivariate probit model in which the dependent variable is firms’ depreciation method
choice and the independent variables are previously identified economic determinants of that choice. The key output of
the
5
first-stage probit model is the inverse Mills ratio, which serves as a control variable in the second-stage regression
(discussed next). For an explanation of the inverse Mills ratio, see Greene (2003), Maddala (1983), and Shehata (1991).
In the second stage, we estimate a multivariate regression using ordinary least squares (OLS). The dependent variable
in the second-stage regression is capital investments and the independent variables are previously identified economic
determinants of capital investments. In addition to these determinants, we include (i) an indicator for whether the firm
uses accelerated depreciation (the variable of interest) and (ii) the inverse Mills ratio (obtained from the first-stage probit
regression), which accounts for the self-selection nature of firms’ depreciation method choice. By including the inverse
Mills ratio in the second-stage regression, OLS yields consistent coefficients (Greene, 2003; Maddala, 1983). Also, because
most of the firms in our sample have repeated observations over time, we adjust standard errors for possible dependence
in the residuals using Rogers (1993) (i.e., clustered) standard errors.
3.2. First-stage regression
Bowen et al. (1995) is the most recently published study examining the economic determinants of firms’ depreciation
method choice. Our first-stage regression, patterned after Bowen et al. (1995), is specified as follows:
CHOICEit ¼ l0 þ l1DURit þ l2R&Dit þ l3LABORit þ l4DBENit

þ l5MFGit CGSit þ l6NMFGit CGSit þ l7NPAYit þ l8ADVit þ l9LEV MVit
þ l10DROAit þ l11SALEit þ l12OGit SALEit þ zit (1)
where CHOICEit is an indicator variable equal to 1 for firms that use the accelerated depreciation method for all or some
of their assets, and 0 for firms that use the straight-line depreciation method, DURit is an indicator variable equal to 1 for
firms producing durable goods (SIC codes 150–179, 245, 250–259, 283, 301, and 324–399), and 0 otherwise (predicted
sign —), R&Dit is research and development expense scaled by ADJ_TAit (0 if R&D expense is missing; predicted sign —),
LABORit is labor intensity measured as one minus the ratio of gross property, plant, and equipment to ADJ_TA it (0 if
gross property, plant, and equipment is missing; predicted sign —), DBENit is an indicator variable equal to 1 for firms
with defined benefit pension plans, identified as those firms with a non-negative value for projected pension obligation or
assumed rate of return for pension benefits, and 0 otherwise (predicted sign +), MFGit is an indicator variable equal to 1
for manufacturing
firms (SIC codes 200–399), and 0 otherwise, NMFG it is an indicator variable equal to 1 for non-manufacturing firms (all SIC
codes except 200–399), and 0 otherwise, CGS it is cost of goods sold scaled by ADJ_TA it (predicted sign —), NPAYit is notes
payable scaled by ADJ_TAit (0 if notes payable is missing; predicted sign —), ADVit is advertising expense scaled by
ADJ_TAit (0 if advertising expense is missing; predicted sign —), LEVMVit is the ratio of long-term debt to market value
of common stock (0 if long-term debt is missing; predicted sign —), DROAit is an indicator variable equal to 1 for firms
with ROA in deciles 2–9, inclusive, where deciles are defined within two-digit SIC codes, and 0 otherwise (predicted sign
–), SALEit is the
natural logarithm of net sales in thousands (predicted sign +), OGit is an indicator variable equal to 1 for firms in the oil
and gas industry (SIC codes 131 and 291), and 0 otherwise (predicted sign +), ADJ_TA it is the average of beginning of year
and end of year adjusted total assets, where adjusted total assets are equal to total assets plus accumulated depreciation,
and i, t are the firm and year subscripts, respectively.
Several features of Eq. (1) require elaboration. First, following Bowen et al. (1995), we scale by average adjusted
total assets. The adjustment to total assets involves adding back accumulated depreciation, which allows us to avoid
scaling by a variable that is influenced by firms’ depreciation method choice. Second, in specifying the variable
CHOICEit, we combine firms that exclusively use accelerated depreciation with firms that use a combination of both
accelerated depreciation and straight-line depreciation. We do this because the frequency of firms that exclusively use
accelerated depreciation for all of their depreciable assets is quite low.2 Third, our coding of CHOICEit is opposite of that
in Bowen et al. (1995). We do this merely to facilitate expositional clarity when we discuss the results.
3.3. Second-stage regression
In the second-stage regression the dependent variable is capital investments and the independent variables are (i) an
indicator for whether the firm uses accelerated depreciation, (ii) economic determinants of capital investments, and (iii)
the inverse Mills ratio obtained from the first-stage regression. Our second-stage regression is specified as follows:
CAPXit ¼ d0 þ d1 CHOICEit þ d2 CAPXi t—1 þ d3 MBi t—1 þ d4 LEVTAi t—1

þ d5 CASHi t—1 þ d6 D CASHit þ d7 AGEi t—1 þ d8 SIZEi t—1 þ d9 RETi t—1
þ d10 CFOit þ d11 DSALEit þ d12 MILLSit þ zit (2)
where CAPXit is capital investments scaled by ADJ_TAit, CHOICEit is an indicator variable equal to 1 for firms that use
the accelerated depreciation method for all or some of their assets, and 0 for firms that use the straight-line depreciation
method (predicted sign +), CAPXi t—1 is the lag of CAPXit (predicted sign +), MBi t—1 is the lag of MBit, where MBit is the
2
In Section 5 we explore various alternative specifications of CHOICEit.
5
market value of common equity plus book value of assets minus book value of equity, with the resulting amount scaled
by ADJ_TAit (predicted sign +), LEVTA i t—1 is the lag of LEVTA it, where LEVTAit is the book value of short-term debt plus book
value of long-term debt, with this sum scaled by ADJ_TA it (predicted sign —), CASHi t—1 is the lag of CASH it, where CASHit is
cash and short-term investments scaled by ADJ_TA it (predicted sign +), D CASHit is the year-to-year change in CASH
(predicted sign —), AGEi t—1 is the lag of AGEit, where AGEit is the log of number of years firm has been listed on CRSP
(predicted sign —), SIZEi t—1 is the lag of SIZEit, where SIZEit is the log of ADJ_TAit (predicted sign +), RETi t—1 is the lag of RETit,
where RETit is stock returns (predicted sign +), CFOit is cash flows from operations scaled by ADJ_TAit (predicted sign
+), DSALEit is the year-to-year change in net sales scaled by ADJ_TA it (predicted sign +), MILLSit is the inverse Mills ratio
from estimation of Eq. (1),3 ADJ_TAit is the average of beginning of year and end of year adjusted total assets, where
adjusted total assets are equal to total assets plus accumulated depreciation, and i, t are firm and year subscripts,
respectively4.
Consistent with our first-stage regression, the scaling variable is average adjusted total assets.5 The variables in Eq.
(2) come from Shin and Kim (2002) (MBi t—1, DCASHit, CFOit, and DSALEit), Richardson (2006) (MBi t—1, LEVTAi t—1, CASHi t—1,
AGEi t—1, SIZEi t—1, and RETi t—1), and McNichols and Stubben (2008) (CAPXi t—1, MBi t—1, CFOit). The rationale behind the
inclusion of various economic determinants in Eq. (2) (other than CHOICE it and MILLSit) is discussed in Adam and Goyal
(2006), Barro (1990), McNichols and Stubben (2008), Richardson (2006), and Shin and Kim (2002), among others. By
including a large array of economic determinants in Eq. (2), we control for the factors that drive capital investment,
thereby enabling us to make inferences about the incremental effect of CHOICEit.
4. Sample selection procedures
Our sample period covers the 19-year period beginning with 1988 and extending through 2006. We start with 1988
because that is the first year that one of our economic determinants (operating cash flows) is available directly from the
statement of cash flows. We exclude regulated industries (SIC codes between 4000 and 4999) and financial institutions (SIC
codes between 6000 and 6999). Firms in these industries are likely to face constraints and incentives that differ from
firms in other industries. We require firms to have the needed financial data in Compustat to construct all variables in
Eqs. (1) and (2).6 We also require firm years to have positive sales and stockholders’ equity. Finally, to mitigate the
influence of outliers and extreme values on our regression results, we winsorize all variables (except for indicator
variables and AGEit) at the 1st and 99th percentiles.
5. Descriptive statistics and primary results
5.1. Descriptive statistics and univariate tests
Table 1 provides the depreciation method choice profile of our sample by both year (Panel A) and industry (Panel B).
Panel A shows that our sample is comprised of 61,096 total firm year observations, with 48,591 firm years (79.53 percent
of the sample) using straight-line depreciation and 12,505 firm years (20.47 percent of the sample) using accelerated
depreciation. This panel also reveals that the popularity of straight-line depreciation increases monotonically over time
while the popularity of accelerated depreciation decreases monotonically over time. Panel B shows that our sample is
comprised of a broad cross-section of different industries, with only one industry comprising more than 10 percent of our
sample firms (SIC code 73, which comprises 14.94 percent of our sample firms). Also, with the exception of SIC codes 10, 13,
and 29, the predominant depreciation method choice is the straight-line method. 7 There are 7843 different firms in our
sample.
Table 2 provides descriptive statistics for the variables in Eqs. (1) and (2), including CAPX it, which are shown for all
firms, straight-line firms, and accelerated firms. 8 As a starting point in our analyses, we conduct univariate tests
of differences in CAPXit between firms that use straight-line depreciation and firms that use accelerated depreciation.
The mean (median) value of CAPXit is 0.044 (0.031) for firms that use straight-line depreciation, while the mean
(median) value of CAPXit is 0.060 (0.039) for firms that use accelerated depreciation. The mean value of CAPXit is
significantly greater
for firms that use accelerated depreciation than for firms that use straight-line depreciation (t-statistic ¼ 25.97, po0.001).
3
The inverse Mills ratio is defined as fðZ i gÞ=FðZ i gÞ for an accelerated firm and —fðZ i gÞ=ð1 — FðZ i gÞÞ for a straight-line firm, where f and F are the
standard normal probability density function and standard normal cumulative density function, respectively, Z is the row vector of explanatory variables
in Eq. (1), and g the column vector of coefficients as estimated in Table 3.
4
Industry indicator variables, defined at the two-digit SIC code level, are also included in Eq. (2) but are not shown.
5
We discuss the effect of using alternative scaling variables in Section 5.
6
In order to construct the variable DROAit in Eq. (1), we exclude firm years in which there are not at least ten firms within a two-digit SIC code.
7
Of the 12,505 firm years using accelerated depreciation, 10,490 firm years use a combination of straight-line depreciation and accelerated
depreciation and 1565 firm years use accelerated depreciation only. Of the 1565 firm years that use accelerated depreciation only, approximately 30
percent of those firm years are in SIC code 10, 13, and 29. Our inferences and conclusions are unaltered if we exclude these three industries from our
analyses.
8
In untabulated results, we note that most of the univariate correlations between CHOICEit (CAPXit) and the economic determinants of
CHOICEit (CAPXit) are in the predicted direction and highly significant (po0.001). Although some of the regression variables are highly correlated with
one another, multicollinearity is not an econometric problem.
6
Table 1
Depreciation method choice and industry profile of sample.
Year Straight-line depreciation Accelerated depreciation Year Straight-line depreciation Accelerated depreciation
# % # % # % # %
Panel A: depreciation method choice profile by year

1988 1797 69.22 799 30.78 1998 2903 80.84 688 19.16
1989 1925 70.49 806 29.51 1999 3040 82.03 666 17.97
1990 2030 72.16 783 27.84 2000 3013 82.05 659 17.95
1991 2082 73.49 751 26.51 2001 2902 83.01 594 16.99
1992 2108 73.73 751 26.27 2002 2984 84.70 539 15.30
1993 2189 74.66 743 25.34 2003 3050 85.08 535 14.92
1994 2393 77.22 706 22.78 2004 2976 85.94 487 14.06
1995 2536 78.27 704 21.73 2005 2821 85.80 467 14.20
1996 2759 78.90 738 21.10 2006 2205 85.93 361 14.07
1997 2878 79.81 728 20.19 All years 48,591 79.53 12,505 20.47
SIC code SIC code description # % of total % SL % ACC
Panel B: depreciation method choice profile by industry

10 Metal mining 96 1.22 10.89 89.11
13 Oil and gas extraction 379 4.83 17.39 82.61
15 Building construction general contractors 58 0.74 79.37 20.63
20 Food and kindred products 197 2.51 86.05 13.95
22 Textile mill products 55 0.70 84.75 15.25
23 Apparel and other finished products made from fabrics 92 1.17 77.88 22.12
25 Furniture and fixtures 62 0.79 80.60 19.40
26 Paper and allied products 101 1.29 63.25 36.75
27 Printing, publishing, and allied industries 118 1.50 81.75 18.25
28 Chemicals and allied products 680 8.67 84.78 15.22
29 Petroleum refining and related industries 56 0.71 38.89 61.11
30 Rubber and miscellaneous plastics products 112 1.43 79.83 20.17
32 Stone, clay, glass, and concrete products 58 0.74 66.22 33.78
33 Primary metal industries 134 1.71 76.19 23.81
34 Fabricated metal products 135 1.72 81.58 18.42
35 Industrial and commercial machinery and equipment 655 8.35 81.75 18.25
36 Electronic and other electrical equipment and components 736 9.38 83.67 16.33
37 Transportation equipment 182 2.32 74.75 25.25
38 Measuring, analyzing, and controlling instruments 615 7.84 82.07 17.93
39 Miscellaneous manufacturing industries 108 1.38 76.03 23.97
50 Wholesale trade durable goods 262 3.34 75.96 24.04
51 Wholesale trade non-durable goods 131 1.67 78.47 21.53
53 General merchandise stores 60 0.77 95.08 4.92
54 Food stores 54 0.69 92.73 7.27
56 Apparel and accessory stores 73 0.93 93.33 6.67
57 Home furniture, furnishings, and equipment stores 57 0.73 93.22 6.78
58 Eating and drinking places 154 1.97 93.71 6.29
59 Miscellaneous retail 186 2.37 87.24 12.76
73 Business services 1,172 14.94 86.86 13.14
78 Motion pictures 87 1.11 51.55 48.45
79 Amusement and recreation services 101 1.29 83.93 16.07
80 Health services 204 2.60 91.55 8.45
87 Engineering, accounting, management, and related services 203 2.59 81.98 18.02
Other SIC codes 470 6.00 73.91 26.09
Total 7843 100.00
In both panels, ‘‘accelerated depreciation’’ captures firms that use accelerated depreciation for all or some of their assets. In Panel B, ‘‘Other SIC codes’’
captures 15 two-digit SIC codes containing fewer than 50 firms each.
Similarly, the median value of CAPXit is significantly greater for firms that use accelerated depreciation than for firms
that use straight-line depreciation (Z-statistic ¼ 18.44, po0.001). Although these findings are consistent with our main
research hypothesis, univariate tests do not control for the self-selection nature of firms’ depreciation method choice
or for the
economic determinants on CAPXit.
5.2. First-stage probit regression results
The first-stage pooled probit regression results for Eq. (1) are provided in Table 3. These results are similar to those
reported in Bowen et al. (1995). The coefficients on R&Dit, LABORit, MFGit CGSit, NMFGit CGSit, ADVit, and LEVMVit are
6
Table 2
Descriptive statistics for regression variables.
Variables All firms (n ¼ 61,096) Straight-line firms (n ¼ 48,591) Accelerated firms (n ¼ 12,505)
MeanMedianMeanMedianMeanMedian
CHOICEit 0.205 0.000 0.000 0.000 1.000 1.000

DURit R&Dit LABORit 0.450 0.000 0.473 0.000 0.360 0.000
0.042
DBENit MFGit CGSit NMFGit CGSit NPAYit 0.003 0.047 0.007 0.021 0.000
ADVit LEVMVit 0.602 0.633 0.634 0.659 0.476 0.493
0.335
DROAit SALEit OGit SALEit CAPXit MBit—1 0.000 0.309 0.000 0.434 0.000
LEVTAit—1 0.367 0.264 0.377 0.278 0.328 0.158
CASHi t—1 DCASHit AGEi t—1 0.331 0.000 0.353 0.000 0.242 0.000
0.026 0.000 0.026 0.000 0.027 0.000
SIZEit—1
0.011 0.000 0.012 0.000 0.007 0.000
RETit—1 0.382 0.096 0.380 0.086 0.389 0.131
CFOit 0.844 1.000 0.839 1.000 0.860 1.000
D SALEit 11.869 11.861 11.857 11.855 11.916 11.892
CAPXit ($millions) 0.523 0.000 0.061 0.000 2.320 0.000
0.047 0.033 0.044 0.031 0.060 0.039
1.612 1.131 1.694 1.180 1.292 0.974
0.176 0.144 0.177 0.143 0.173 0.144
0.152 0.070 0.163 0.077 0.108 0.050
—0.004 —0.001 —0.004 —0.001 —0.005 —0.001

2.277 2.303 2.235 2.197 2.441 2.485
5.145 4.991 5.056 4.932 5.493 5.331
0.172 0.000 0.182 0.135 0.000
0.030 0.057 0.029 —0.002 0.036 0.058
0.082 0.060 0.087 0.057 0.062 0.039
84.004 5.289 58.677 0.066 182.415 9.179
4.747
ADJ_TAit ($billions)1.8250.1561.3670.1463.6030.222
Variables are defined as follows: CHOICEit ¼ indicator variable equal to 1 for firms that use the accelerated depreciation method for all or some of their
assets, and 0 for firms that use the straight-line depreciation method; DUR it ¼ indicator variable equal to 1 for firms producing durable goods (SIC codes
150–179, 245, 250–259, 283, 301, and 324–399), and 0 otherwise; R&D it ¼ research and development expense scaled by ADJ_TA it (0 if R&D expense is
missing); ABORit ¼ labor intensity measured as one minus the ratio of gross property, plant, and equipment to ADJ_TAit; DBENit ¼ indicator variable
equal to 1 for firms with defined benefit pension plans, identified as those firms with a non-negative value for projected pension obligation or assumed rate
of return for pension benefits, and 0 otherwise; MFG it ¼ indicator variable equal to 1 for manufacturing firms (SIC codes 200–399), and 0 otherwise;
NMFGit ¼ indicator variable equal to 1 for non-manufacturing firms (all SIC codes except 200–399), and 0 otherwise; CGSit ¼ cost of goods sold scaled
by ADJ_TAit (0 if cost of goods sold is missing); NPAY it ¼ notes payable scaled by ADJ_TAit (0 if notes payable is missing); ADV it ¼ advertising expense
scaled by ADJ_TAit (0 if advertising expense is missing); EVMVit ¼ ratio of long-term debt to market value of common stock; DROAit ¼ indicator
variable equal to 1 for firms with ROA in deciles 2–9, inclusive, where deciles are defined within two-digit SIC codes, and 0 otherwise; SALEit ¼ natural
logarithm of net sales in thousands; OG it ¼ indicator variable equal to 1 for firms in the oil and gas industry (SIC codes 131 and 291), and 0 otherwise;
CAPXit ¼ capital investments scaled by ADJ_TAit; CHOICEit ¼ indicator variable equal to 1 for firms that use the accelerated depreciation method for all
or some of their assets, and 0 for firms that use the straight-line depreciation method; MB it—1 ¼ lag of MBit, where MBit is the market value of common
equity plus book value of assets minus book value of equity, with the resulting amount scaled by ADJ_TAit; EVTAit—1 ¼ lag of LEVTAit, where LEVTAit
is the book value of short-term debt plus book value of long-term debt, with this sum scaled by ADJ_TA it; CASHit—1 ¼ lag of CASHit, where CASHit is
cash and short-term investments scaled by ADJ_TAit; DCASHit ¼ year-to-year change in CASH; AGEit—1 ¼ lag of AGEit, where AGEit is the log of number
of years firm has been listed on CRSP; SIZEit—1 ¼ lag of SIZEit, where SIZEit is the log of ADJ_TA it; RETit—1 ¼ lag of RETit, where RETit is stock returns; CFOit
¼ cash flows from operations scaled by ADJ_TAit; D SALEit ¼ year-to-year change in net sales scaled by ADJ_TA it; ADJ_TAit ¼ average of beginning of year
and end of year adjusted total assets, where adjusted total assets are equal to total assets plus accumulated depreciation; i and t ¼ firm and year
subscripts, respectively.
negative and significant (pr0.005), consistent with expectations. The coefficients on DBENit and OGit SALEit are
positive and significant (po0.001), consistent with expectations. The coefficients on DUR it and DROAit are insignificant
(p40.10). The coefficients on NPAYit and SALEit are significant but in the opposite direction (pr0.037). The opposite
signs on these variables are not surprising because they had inconsistent signs across years in Bowen et al. (1995).
Finally, the pseudo-R2 is 14.69 percent and the model is highly significant (po0.001).
5.3. Second-stage regression results
The second-stage OLS regression results for Eq. (2) are reported in Table 4. The coefficients on CAPXi t—1, MBi t—1, RETi t—
1, CFOit, and D SALEit are positive and significant (po0.001), consistent with expectations. The coefficients on LEVTAi t
—1, D CASHit, and AGEi t—1 are negative and significant (po0.001), consistent with expectations. However, the coefficient on
CASHi t—1 is significant in the opposite direction (po0.001) and the coefficient on SIZE i t—1 is insignificant. The opposite sign
on CASHi t—1 appears to arise because of our relatively narrow definition of the dependent variable. In other studies that
include CASHi t—1 as an economic determinant of capital investments (e.g., Richardson, 2006), the dependent variable
is defined more broadly (i.e., the sum of capital investments, R&D, and acquisition expenditures). When we define our
dependent variable in this alternative manner, the coefficient on CASHi t—1 is positive and highly significant. The coefficient
on the self-selection variable, MILLSit, is highly significant (po0.001), which confirms the importance of using a two-
stage
6
Table 3
First-stage probit regression of depreciation method choice on economic determinants.
Variables Predicted sign Coefficient Z-statistic p-value
Intercept ? —0.035 —0.28 0.778

DURit — 0.033 0.67 0.505
R&Dit — —2.385 —7.76 o0.001
LABORit — —0.855 —10.41 o0.001
DBENit + 0.188 4.48 o0.001
MFGit CGSit — —0.179 —3.21 0.001
NMFGit CGSit — —0.160 —3.81 o0.001
NPAYit — 1.170 4.83 o0.001
ADVit — —2.065 —2.82 0.005
LEVMVit — —0.068 —3.93 o0.001
DROAit — —0.018 —0.63 0.528
SALEit + —0.020 —2.09 0.037
OGit SALEit + 0.152 14.21 o0.001
Observations Pseudo R2 (%) Model p-value 61,096

14.69
o0.001
CHOICEit ¼ l0 þ l1 DURit þ l2 R&Dit þ l3 LABORit þ l4 DBENit þ l5 MFGit CGSit þ l6 NMFGit CGSit

þ l7 NPAYit þ l8 ADVit þ l9 LEVMVit þ l10 DROAit þ l11 SALEit þ l12 OGit SALEit þ zit
Variables are defined as follows: CHOICEit ¼ indicator variable equal to 1 for firms that use the accelerated depreciation method for all or some of their
assets, and 0 for firms that use the straight-line depreciation method; DUR it ¼ indicator variable equal to 1 for firms producing durable goods (SIC codes
150–179, 245, 250–259, 283, 301, and 324–399), and 0 otherwise; R&D it ¼ research and development expense scaled by ADJ_TA it (0 if R&D expense is
missing); ABORit ¼ labor intensity measured as one minus the ratio of gross property, plant, and equipment to ADJ_TA it (0 if gross property, plant, and
equipment is missing); DBENit ¼ indicator variable equal to 1 for firms with defined benefit pension plans, identified as those firms with a non-
negative value for projected pension obligation or assumed rate of return for pension benefits, and 0 otherwise; MFG it ¼ indicator variable equal to 1
for manufacturing firms (SIC codes 200–399), and 0 otherwise; NMFG it ¼ indicator variable equal to 1 for non-manufacturing firms (all SIC codes except
200–399), and 0 otherwise; CGSit ¼ cost of goods sold scaled by ADJ_TAit; NPAYit ¼ notes payable scaled by ADJ_TAit (0 if notes payable is missing);
ADVit ¼ advertising expense scaled by ADJ_TAit (0 if advertising expense is missing); EVMVit ¼ ratio of long-term debt to market value of common stock (0
if long-term debt is missing); DROA it ¼ indicator variable equal to 1 for firms with ROA in deciles 2–9, inclusive, where deciles are defined within two-
digit SIC codes, and 0 otherwise; SALE it ¼ natural logarithm of net sales in thousands; OG it ¼ indicator variable equal to 1 for firms in the oil and gas
industry (SIC codes 131 and 291), and 0 otherwise; ADJ_TA it ¼ average of beginning of year and end of year adjusted total assets, where adjusted total
assets are equal to total assets plus accumulated depreciation; i and t ¼ firm and year subscripts, respectively. Significance tests use clustered
standard errors.
‘‘treatment effects’’ model to control for the endogenous nature of firms’ depreciation method choice. Finally, the adjusted
R2 is 49.58 and the model is highly significant (po0.001).
The variable of interest in Eq. (2) is CHOICEit, which we use to test our research hypothesis. The coefficient on CHOICEit
is positive and highly significant (t-statistic ¼ 12.81, po0.001), which supports the hypothesis that firms that use
accelerated depreciation make larger capital investments than firms that use straight-line depreciation. The
interpretation of the
coefficient on CHOICEit is that capital investments, expressed as a percentage of average adjusted total assets, are on
average
3.8 percentage points higher for firms that use accelerated depreciation than for firms that use straight-line depreciation.
The magnitude of the coefficient suggests that association between CHOICEit and CAPXit is economically meaningful. At the
same time, we also acknowledge that the coefficient magnitude is large given the second-order role that firms’ depreciation
method choice is likely to play in managers’ capital investment decisions. While we caution against literal interpretation
of the coefficient on CHOICEit, our results continue to hold when we conduct various robustness tests (see Section 5.7).
5.4. Partitions of depreciation method choice
As discussed previously, the variable CHOICEit combines firms that exclusively use accelerated depreciation with
firms that use a combination of accelerated depreciation and straight-line depreciation. We do this because the
frequency of firms that exclusively use accelerated depreciation for all of their depreciable assets is quite low. On
average, 82 firms per year exclusively use accelerated depreciation during our sample period. This number drops to 30
firms that exclusively use accelerated depreciation by 2006 (we elaborate on depreciation method changes in the next
section). With this sample size limitation in mind, we estimate regressions using the following alternative specifications
of CHOICEit:
1. CHOICEit is an indicator variable equal to 1 for firms that use the accelerated depreciation method only, and 0 for
firms that use the straight-line depreciation method only (i.e., CHOICE it excludes firms that use both the accelerated
depreciation method and the straight-line depreciation method).
2. CHOICEit is an indicator variable equal to 1 for firms that use both the accelerated depreciation method and the
straight- line depreciation method, and 0 for firms that use the straight-line depreciation method only (i.e., CHOICE it
excludes firms that use the accelerated depreciation method only).
6
Table 4
Second-stage OLS regression of capital investments on depreciation method choice, economic determinants, and inverse Mills ratio.
Variables Predicted sign Coefficient t-statistic p-value

Intercept ? 0.015 4.25 o0.001
CHOICEit + 0.038 12.81 o0.001
CAPXit—1 + 0.536 70.31 o0.001
MBi t—1 LEVTAi t—1 CASHi t—1 DCASHit + 0.002 12.80 o0.001 o0.001 o0.001 o0.001 o0.001 0.847
AGEi t—1 SIZEi t—1 RETi t—1 — —0.017 —13.27 o0.001
+ —0.007 —6.33
— —0.029 —15.60
— —0.001 —6.83
+ 0.000 0.19
+ 0.003 15.60
CFOit + 0.022 14.00 o0.001
D SALEit + 0.018 26.17 o0.001
MILLSit ? —0.021 —13.09 o0.001
Observations 61,096
Adjusted R2 (%) 49.58
Model p-value o0.001
CAPXit ¼ d0 þ d1 CHOICEit þ d2 CAPXit þ d3 MBit—1 þ d4 LEVTAit—1 þ d5 D CASHit—1

þ d6 D CASHit þ d7 AGEit—1 þ d8 SIZEit—1 þ d9 RETit—1 þ d10 CFOit þ d11 D SALEit þ d12 MILLSit þ zit
Variables are defined as follows: CAPX it ¼ capital investments scaled by ADJ_TA it; CHOICEit ¼ indicator variable equal to 1 for firms that use the
accelerated depreciation method for all or some of their assets, and 0 for firms that use the straight-line depreciation method; CAPX it—1 ¼ lag of CAPXit;
MBit—1 ¼ lag of MBit, where MBit is the market value of common equity plus book value of assets minus book value of equity, with the resulting amount
scaled by ADJ_TAit; EVTAit—1 ¼ lag of LEVTAit, where LEVTAit is the book value of short-term debt plus book value of long-term debt, with this sum scaled
by ADJ_TAit; CASHit—1 ¼ lag of CASHit, where CASHit is cash and short-term investments scaled by ADJ_TAit; D CASHit ¼ year-to-year change in CASH;
AGEit—1 ¼ lag of AGEit, where AGEit is the log of number of years firm has been listed on CRSP; SIZE it—1 ¼ lag of SIZEit, where SIZEit is the log of ADJ_TA it;
RETit—1 ¼ lag of RETit, where RETit is stock returns; CFOit ¼ cash flows from operations scaled by ADJ_TAit; D SALEit ¼ year-to-year change in net sales
scaled by ADJ_TAit; MILLSit ¼ inverse Mills ratio from estimation of Eq. (1); ADJ_TAit ¼ average of beginning of year and end of year adjusted total
assets, where adjusted total assets are equal to total assets plus accumulated depreciation; i and t ¼ firm and year subscripts, respectively. Industry
indicator variables, defined at the two-digit SIC code level, are included in the regression but are not shown. Significance tests use clustered standard
errors.
3. CHOICEit is an indicator variable equal to 1 for firms that use the accelerated depreciation method only, and 0 for
firms that use both the straight-line depreciation method and accelerated depreciation method (i.e., CHOICEit
excludes firms that use the straight-line depreciation method only).
Panels A–C of Table 5 provide the second-stage regression results using each of the above definitions of CHOICEit. In
Panel A, we find that CHOICE it is positive and highly significant (coefficient ¼ 0.026, t-statistic ¼ 3.90, po0.001), which
indicates that firms that exclusively use accelerated depreciation make larger capital investments than firms that
exclusively use straight-line depreciation. In Panel B, we find that CHOICE it is positive and highly significant
(coefficient ¼ 0.039, t-statistic ¼ 12.47, po0.001), which indicates that firms that use both accelerated depreciation and
straight-line depreciation make larger capital investments than firms that exclusively use straight-line depreciation. In
Panel C, we find that CHOICEit is insignificant (coefficient ¼ 0.005, t-statistic ¼ 0.35, p ¼ 0.729), which indicates that
firms that exclusively use accelerated depreciation make capital investments that are indistinguishable from firms
that use both
accelerated depreciation and straight-line depreciation. As a result, it appears that the use of accelerated depreciation for at
least some of a firm’s depreciable assets yields the effect predicted by our main research hypothesis.
5.5. Depreciation method changes
Depreciation method changes are somewhat frequent during our sample period. There are 507 firms that changed
from accelerated depreciation to straight-line depreciation, 92 firms that changed from straight-line depreciation to
accelerated depreciation, and 94 firms that made multiple depreciation method changes. 9 It is interesting to note that
there has been a migration away from accelerated depreciation to straight-line depreciation. As shown in Panel A of
Table 1, approximately 69 percent (31 percent) of our sample firms used straight-line depreciation (accelerated
depreciation) in 1988, while approximately 86 percent (14 percent) of our sample firms used straight-line depreciation
(accelerated depreciation) in 2006. The frequency of firms using accelerated depreciation is significantly lower in the
second half of our sample period
than in the first half of our sample period (po0.001). However, the temporal decline in accelerated depreciation may be the
result of newly public companies using straight-line depreciation rather than existing public companies migrating away
from straight-line depreciation.
9
Depreciation method changes are fairly evenly dispersed across two-digit SIC codes.
6
Table 5
Second-stage OLS regressions of capital investments on partitions of depreciation method choice, economic determinants, and inverse Mills ratio.
Panel A: CHOICEit is an indicator variable equal to 1 for firms that use the accelerated depreciation method only, and 0 for firms that use the straight-line
depreciation method only (i.e., CHOICE excludes firms that use both the accelerated depreciation method and the straight-line depreciation method)
Intercept ? 0.019 5.97 o 0.001
CHOICEit + 0.026 3.90 o 0.001
CAPXit—1 + 0.541 60.62 o 0.001
MBit—1 + 0.002 10.69 o 0.001
LEVTAit—1 — —0.016 —12.11 o 0.001
CASHit—1 + —0.012 —10.04 o 0.001
DCASHit o 0.001
— —0.030 —15.77
AGEit—1 — —0.001 —3.98 o 0.001
SIZEit—1 + 0.000 0.36 0.718
+ 0.003 14.19 o 0.001
RETit—1
CFOit + 0.021 12.92 o 0.001
D SALEit + 0.018 24.30 o 0.001
MILLSit ? —0.011 —4.13 o 0.001
Observations 50,156
Model p-value o 0.001
Panel B: CHOICEit is an indicator variable equal to 1 for firms that use both the accelerated depreciation method and the straight-line depreciation
method,
and 0 for firms that use the straight-line depreciation method only (i.e., CHOICE excludes firms that use the accelerated depreciation method only)
Intercept ? 0.014 4.37 o 0.001
CHOICEit + 0.039 12.47 o 0.001
CAPXit—1 + 0.535 68.52 o 0.001
MBit—1 + 0.002 12.66 o 0.001
LEVTAit—1 — —0.017 —13.53 o 0.001
CASHit—1 + —0.008 —6.86 o 0.001
D CASHit o 0.001
— —0.029 —15.50
AGEit—1 — —0.001 —7.16 o 0.001
SIZEit—1 + 0.000 0.42 0.674
RETit—1 + 0.003 15.31 o 0.001
CFOit + 0.021 13.51 o 0.001
D SALEit + 0.018 26.06 o 0.001
MILLSit ? —0.022 —11.78 o 0.001
Observations 59,531
Panel C: CHOICEit is an indicator variable equal to 1 for firms that use the accelerated depreciation method only, and 0 for firms that use both the
straight-
line depreciation method and accelerated depreciation method (i.e., CHOICE excludes firms that use the straight-line depreciation method only)
Intercept ? 0.002 0.22 0.824
CHOICEit + 0.005 0.35 0.729
CAPXit—1 + 0.548 40.85 o 0.001
+ 0.003 5.53 o 0.001
MBit—1
LEVTAit—1 — —0.021 —5.69 o 0.001
CASHit—1 + —0.008 —2.12 0.034
D CASHit —0.050 —6.93 o 0.001
—
AGEit—1 — —0.002 —3.88 o 0.001
SIZEit—1 + 0.000 1.19 0.234
+ 0.006 8.43 o 0.001
RETit—1
CFOit + 0.037 7.77 o 0.001
D SALEit + 0.013 7.52 o 0.001
MILLSit ? —0.003 —0.44 0.659
Observations 12,505
CAPXit ¼ d0 þ d1 CHOICEit þ d2 CAPXit þ d3 MBit—1 þ d4 LEVTAit—1 þ d5 D CASHit—1

Variables are defined as follows: CAPX it ¼ capital investments scaled by ADJ_TA it; CHOICEit ¼ definition provided at the beginning of each panel;
CAPXit—1 ¼ lag of CAPXit; MBit—1 ¼ lag of MBit, where MBit is the market value of common equity plus book value of assets minus book value of equity,
with the resulting amount scaled by ADJ_TAit; EVTAit—1 ¼ lag of LEVTAit, where LEVTAit is the book value of short-term debt plus book value of long-
term debt, with this sum scaled by ADJ_TAit; CASHit—1 ¼ lag of CASHit, where CASHit is cash and short-term investments scaled by ADJ_TAit; D CASHit ¼
year-to- year change in CASH; AGEit—1 ¼ lag of AGEit, where AGEit is the log of number of years firm has been listed on CRSP; SIZEit—1 ¼ lag of SIZEit, where
SIZEit is the log of ADJ_TAit; RETit—1 ¼ lag of RETit, where RETit is stock returns; CFOit ¼ cash flows from operations scaled by ADJ_TAit; D SALEit ¼ year-to-
year change in net sales scaled by ADJ_TAit; MILLSit ¼ inverse Mills ratio from estimation of Eq. (1); ADJ_TAit ¼ average of beginning of year and end
of year adjusted total assets, where adjusted total assets are equal to total assets plus accumulated depreciation; i and t ¼ firm and year subscripts,
respectively. Industry indicator variables, defined at the two-digit SIC code level, are included in the regression but are not shown. Significance tests
use clustered
standard errors.
6
Table 6
Capital investments partitioned by years before the change to straight-line depreciation and years of and after the change to straight-line depreciation.
CAPXit Years before change Years of and after change Difference in means Difference in medians
t-statistic p-value Z-statistic p-value
Mean 0.047 0.039 7.50 o0.001

Median 0.033 0.028 7.10 o0.001
Std. dev. 0.049 0.406
First quartile 0.018 0.014
Third quartile 0.060 0.050
n 2812 3270
CAPXit ¼ capital investments scaled by ADJ_TAit; ADJ_TAit ¼ average of beginning of year and end of year adjusted total assets, where adjusted total
assets are equal to total assets plus accumulated depreciation; i and t ¼ firm and year subscripts, respectively.
To address this issue, we restrict our analysis to the sample of firms that have available data in all 19 years of our
sample period (there are 611 such firms). Approximately 68 percent (32 percent) of these firms used straight-line
depreciation (accelerated depreciation) in 1988 while approximately 83 percent (17 percent) of these firms used straight-
line depreciation (accelerated depreciation) in 2006 (descriptive statistics for the constant sample are not tabulated).
Again, the frequency of firms using accelerated depreciation is significantly lower in the second half of our sample
period than in the
first half of our sample period (po0.001).
The fairly large number of firms that change from accelerated depreciation to straight-line depreciation (507 firms and
6082 firm years) enables us to compare capital investments in the periods before the change in depreciation method to
capital investments in the periods after the change in depreciation method.10 Table 6 provides capital investments in these
periods. Mean (median) CAPXit is 0.047 (0.033) before the change to straight-line depreciation and 0.039 (0.028) after
the change. The decline in CAPXit is highly significant at both the mean (t-statistic ¼ 7.50, po0.001) and median (Z-
statistic ¼ 7.10, po0.001).
To control for the economic determinants of CAPX it, we estimate a modified version of Eq. (2) using the 507 firms in our
sample that change from accelerated depreciation to straight-line depreciation. In this modified regression, we remove
the variable CHOICEit and replace it with a variable called CHANGE it, which is an indicator variable equal to 1 for the period
of and periods following the change to straight-line depreciation, and 0 for the periods before the change to straight-line
depreciation. There are 2812 firm years before the change and 3270 firm years of and after the change. 11 If capital
investments decline after firms change from accelerated depreciation to straight-line depreciation then the coefficient
on CHANGEit should be negative. Table 7 shows that the coefficient on CHANGEit is negative and significant (t-statistic ¼
—5.80, po0.001), which indicates that firms make smaller capital investments after the change than before the change.
5.6. Research and development investments
Although our theory suggests that there will be a positive relation between capital investments and CHOICEit, it
makes no prediction about the relation between R&D and CHOICE it. The reason that our theory makes no such
prediction is because R&D is expensed immediately and there are no book values generally associated with such
expenditures (e.g., salaries paid to scientists for inventive activities). Thus, we expect a non-positive relation between
R&D and CHOICEit. To examine this relation, we substitute R&D it in place of CAPXit in Eq. (2). In turn, we estimate this
modified version of Eq. (2) using two-stage least squares.12 Table 8 provides the results. Consistent with expectations, we
find that the coefficient on CHOICEit is non-positive (t-statistic ¼ —3.15, p ¼ 0.002). However, a large negative,
significant coefficient is quite surprising.13 Nonetheless, we find a non-positive relation between CHOICEit and
investments when we expect that relation
10
We do not analyze capital investments of firms that change from straight-line depreciation to accelerated depreciation because of the small
number of firms that make such changes (92 firms).
11
Our inferences and conclusions are unaltered when our analyses focus on the three or four annual periods before and after the change in firms’
depreciation method.
12
We use two-stage least squares because (i) research and development is an independent variable in the first equation and is the dependent
variable in the second equation and (ii) depreciation method choice is the dependent variable in the first equation and is an independent variable in
the second equation.
13
The negative coefficient on CHOICE it implies that firms that use accelerated depreciation make smaller R&D investments than firms that use
straight-line depreciation. One explanation for this negative coefficient is that firms that use accelerated depreciation are less likely to engage in inventive
activities. Consistent with this conjecture, we find that approximately 56 percent of accelerated depreciation firms report no R&D expense
whatsoever, compared to only 45 percent of straight-line depreciation firms. Another explanation is that firms that use accelerated depreciation
classify some R&D expenses as capital investments. This explanation, if valid, would help to jointly explain the results in Section 5.3 (i.e., firms that
use accelerated depreciation make larger capital investments than firms that use straight-line depreciation) and the results noted in this section (i.e.,
firms that use accelerated depreciation make smaller investments in R&D than firms that use straight-line depreciation). While there is some
ambiguity and judgment inherent in classification decisions between capital investments and R&D expense, we believe that this classification
explanation is, at best, a partial
6
Table 7
Second-stage OLS regression of capital investments on change in depreciation method choice, economic determinants, and inverse Mills ratio using the
depreciation method change sample.
Intercept ? 0.002 0.24 0.810

CHANGEit — —0.050 —5.80 o 0.001
CAPXit—1 + 0.536 21.78 o 0.001
MBit—1 + 0.002 5.21 o 0.001
—0.023 —6.30 o 0.001
LEVTAit—1 —
CASHi t—1 D CASHit AGEi t—1 + —0.009 —2.63 0.009
SIZEit—1 — —0.022 —4.04 o 0.001
RETit—1 0.004
— —0.002 —2.93
0.001
+ 0.001 3.38
o 0.001
+ 0.003 4.55
CFOit + 0.021 3.97 o 0.001
D SALEit + 0.014 6.46 o 0.001
MILLSit ? 0.036 5.57 o 0.001
Observations 6,082
CAPXit ¼ d0 þ d1 CHANGEit þ d2 CAPXit þ d3 MBit—1 þ d4 LEVTAit—1 þ d5 D CASHit—1

Variables are defined as follows: CAPXit ¼ capital investments scaled by ADJ_TAit; CHANGEit ¼ indicator variable equal to 1 for the period of and
periods following the change to straight-line depreciation, and 0 for the periods before the change to straight-line depreciation; CAPXit—1 ¼ lag of
CAPXit; MBit—1 ¼ lag of MBit, where MBit is the market value of common equity plus book value of assets minus book value of equity, with the resulting
amount scaled by ADJ_TAit; EVTAit—1 ¼ lag of LEVTAit, where LEVTAit is the book value of short-term debt plus book value of long-term debt, with this sum
scaled by ADJ_TAit; CASHit—1 ¼ lag of CASHit, where CASHit is cash and short-term investments scaled by ADJ_TAit; D CASHit ¼ year-to-year change in
CASH; AGEit—1 ¼ lag of AGEit, where AGEit is the log of number of years firm has been listed on CRSP; SIZE it—1 ¼ lag of SIZEit, where SIZEit is the log of
ADJ_TAit; RETit—1 ¼ lag of RETit, where RETit is stock returns; CFOit ¼ cash flows from operations scaled by ADJ_TAit; D SALEit ¼ year-to-year change in net
sales scaled by ADJ_TAit; MILLSit ¼ inverse Mills ratio from estimation of Eq. (1) using CHANGE it as the dependent variable; ADJ_TA it ¼ average of
beginning of year and end of year adjusted total assets, where adjusted total assets are equal to total assets plus accumulated depreciation; i and t ¼
firm and year
subscripts, respectively. Industry indicator variables, defined at the two-digit SIC code level, are included in the regression but are not shown. Significance
tests use clustered standard errors.
to be non-positive (i.e., in the case of R&D investments) and we find a positive relation between CHOICEit and
investments when we expect that relation to be positive (i.e., in the case of capital investments).14
5.7. Robustness tests
In this section, we consider a variety of robustness tests relating to scaling, methodology, data transformation, and lags.
5.7.1. Scaling
We scale variables in Eqs. (1) and (2) by average adjusted total assets rather than average total assets to avoid scaling
by a variable that is itself influenced by firms’ depreciation method choice. However, our inferences and conclusions are
unaffected when we scale by average total assets (results not tabulated). Alternatively, we could scale by a variable
that needs no adjustment—net sales. When we scale by net sales, our inferences and conclusions are unaffected (results
not tabulated).
5.7.2. Median regression

Eq. (2) can be estimated using median regression rather than OLS. Median regression finds the regression plane that
minimizes the sum of the absolute residuals rather than the sum of the squared residuals. In some contexts median
regression is preferred to OLS regression because it is less susceptible to extreme values. When we estimate Eq. (2)
using median regression, our inferences and conclusions are unaffected (results not tabulated).
5.7.3. Maximum likelihood estimation

We also estimate Eqs. (1) and (2) using the full-information maximum likelihood estimator to assess whether our
results are sensitive to the econometric procedure chosen. 15 We find that our inferences and conclusions are unaffected
when we use the full-information maximum likelihood estimator (results not tabulated).
(footnote continued)
explanation for our results. Classification mistakes are unlikely to occur on a large scale for a large number of firms, particularly when the firms are
routinely audited. Further, it seems unlikely that classification mistakes are systematically clustered in firms that use accelerated depreciation.
14
We thank an anonymous referee for suggesting this analysis.
15
Puhani (2000) indicates that the full-information maximum likelihood estimator is preferable to the two-stage ‘‘treatment effects’’ model, although
the latter approach generally produces reasonable results.
6
5.7.4. Fama–Macbeth standard errors
To adjust the standard errors for possible dependence in the residuals, we use Rogers’ (1993) (i.e., clustered) standard
errors. An alternative approach is to estimate year-by-year regressions and use Fama and MacBeth (1973) standard
errors. While both of these approaches have been used extensively in the accounting and finance literatures, Peterson
(2009) finds that they may produce different results in certain circumstances. However, we find that our inferences and
conclusions are unaffected when we use Fama and MacBeth (1973) standard errors rather than Rogers (1993) standard
errors.
5.7.5. Ranked data

The theoretically correct form of the relation between the variables in Eq. (2) is unknown. As noted by Lang and
Lundholm (1993), if the relation between the dependent variable and independent variables is monotonic, a higher
ranked dependent variable will correspond to a higher ranked independent variable regardless of the precise functional
form of the relation between the variables. When we estimate Eq. (2) using rank transformed data, our inferences and
conclusions are unaffected (results not tabulated).16
5.7.6. Multiple lags of capital investments

Eq. (2) has one lag of capital investments. It is possible that additional lags will cause the coefficient on CHOICE it to
decline and perhaps become insignificant. While we find that the coefficient on CHOICEit declines due to the inclusion of
one lag of capital investments (from about 0.083 with no lags to 0.038 with one lag), it declines by small increments with
each additional lag but is always highly significant. Indeed, when Eq. (2) has five lags of capital investments (results not
tabulated), the coefficient on CHOICEit is still highly significant (t-statistic ¼ 8.54, po0.001) and economically
meaningful (coefficient 1
7 ¼ 0.024).
6. Conclusion
The prescription that managers should make capital investment decisions based on the incremental costs and benefits
associated with alternative courses of action is non-controversial. However, our results indicate that accelerated
depreciation is associated with significantly higher levels of capital investments than straight-line depreciation even
though firms’ depreciation method choice is a normatively irrelevant consideration. We also find that there has been a
migration away from accelerated depreciation to straight-line depreciation over the past two decades. Firms that change
from accelerated depreciation to straight-line depreciation make significantly smaller capital investments in the post-
change periods than the pre-change periods. An important implication of our study is that a seemingly inconsequential
choice that firms make for external financial reporting purposes influences how firms invest scarce capital resources.
Research examining the economic consequences of firms’ depreciation method choice has been somewhat dormant in
recent decades. It is possible that such research has been viewed as being unprofitable because prior research has already
documented many of the market-related and/or contractual consequences of firms’ depreciation method choice.
However, research has largely overlooked at least one class of consequences associated with firms’ depreciation method
choice—its influence on internal management decisions. We see no conceptual, theoretical, or practical reason why
researchers should not examine the economic consequences of firms’ depreciation method choice in contexts outside of
the market-based and contracting domains.
Finally, this study is subject to certain limitations. First, our methodology does not permit us to conclude that
accelerated depreciation causes higher levels of capital investments. Instead, our methodology only permits us to
conclude that accelerated depreciation is associated with higher levels of capital investments. Second, although our
empirical results strongly suggest that there is a statistically significant and economically meaningful association
between the use of accelerated depreciation and capital investments, we cannot confidently assess the economic
significance of this association. In fact, the magnitude of the coefficient on CHOICEit seems to be somewhat large given
the second-order role that firms’ depreciation method choice is likely to play in managers’ capital investment decisions.
References
Adam, T., Goyal, V., 2006. The investment opportunity set and its proxy variables: theory and Evidence. Working paper, Massachusetts Institute of
Technology.
Archibald, T., 1972. Stock market reaction to the depreciation switch-back. The Accounting Review 47 (1), 22–30.
Arkes, H., Blumer, C., 1985. The psychology of sunk costs. Organizational Behavior and Human Decision Processes 35 (1), 124–140.
Arkes, H., 1996. The psychology of waste. Journal of Behavioral Decision Making 9 (3), 213–224.
Barro, R., 1990. The stock market and investment. Review of Financial Studies 3 (1), 115–131.
Beaver, W., Dukes, R., 1973. Interperiod tax allocation and d-depreciation methods: some empirical results. The Accounting Review 48 (3), 549–559.
Bhojraj, S., Libby, R., 2005. Capital market pressure, disclosure frequency-induced earnings/cash flow conflict, and managerial myopia. The Accounting
Review 80 (1), 1–20.
16
See Hollander and Wolf (1999, pp. 415–457) and Iman and Conover (1979, p. 499) for a discussion of this non-parametric regression technique.
17
The addition of five lags results in sample attrition of approximately 33 percent. We find that the decline in the coefficient magnitude is
attributable to the inclusion of lags and not sample attrition.
6
Bowen, R., DuCharme, L., Shores, D., 1995. Stakeholders’ implicit claims and accounting method choice. Journal of Accounting and Economics 20 (3),
255–295.
Brockner, J., 1992. The escalation of commitment to a failing course of action: toward theoretical progress. Academy of Management Review 17 (1), 39–61.
Bushee, B., 1998. The influence of institutional investors on myopic R&D investment behavior. The Accounting Review 73 (3), 305–333.
Christie, A., Zimmerman, J., 1994. Efficient and opportunistic choices of accounting procedures: corporate control contests. The Accounting Review 69 (4),
539–566.
Dearden, J., 1960. Problem in decentralized profit responsibility. Harvard Business Review 38 (3), 79–86.
Dechow, P., Sloan, R., 1991. Executive incentives and the horizon problem. Journal of Accounting and Economics 14 (1), 51–89.
Fama, E., MacBeth, J., 1973. Risk, return and equilibrium: empirical tests. Journal of Political Economy 81 (3), 607–636.
Fields, T., Lys, T., Vincent, L., 2001. Empirical research on accounting choice. Journal of Accounting and Economics 31 (1–3), 255–307.
Garrison, R., Noreen, E., 2003. Managerial Accounting, 10th ed. McGraw-Hill, New York, NY.
Graham, J., Harvey, C., 2001. The theory and practice of corporate finance: evidence from the field. Journal of Financial Economics 60 (2–3), 187–243.
Graham, J., Harvey, C., Rajgopal, S., 2005. The economic implications of corporate financial reporting. Journal of Accounting and Economics 40 (1–3), 3–73.
Greene, W., 2003. Econometric Analysis, fifth ed. Prentice-Hall, Englewood Cliffs, NJ.
Harris, M., Raviv, A., 1996. The capital budgeting process: incentives and information. The Journal of Finance 51 (4), 1139–
1174. Hatfield, H., 1944. Replacement and book value. The Accounting Review 19 (1), 66–67.
Heath, C., Fennema, M., 1996. Mental depreciation and marginal decision making. Organizational Behavior and Human Decision Processes 68 (2), 95–
108. Hilton, R., 2002. Managerial accounting: creating value in a dynamic business environment, fifth ed. McGraw-Hill, New York, NY.
Horngren, C., Sundem, G., Stratton, W., 2005. Introduction to management accounting, 13th ed. Pearson Prentice Hall, Upper Saddle River, NJ.
Hollander, M., Wolf, D., 1999. Nonparametric Statistical Methods, second ed. Wiley, New York, NY.
Holthausen, R., 1981. Evidence on the effect of bond covenants and management compensation contracts on the choice of accounting techniques: the case
of the depreciation switch-back. Journal of Accounting and Economics 3 (1), 73–109.
Holthausen, R., Leftwich, R., 1983. The economic consequences of accounting choice: implications of costly contracting and monitoring. Journal of
Accounting and Economics 5, 77–117.
Iman, I., Conover, W., 1979. The use of the rank transform in regression. Technometrics 21 (4), 499–509.
Jackson, S., 2008. The effect of firms’ depreciation method choice on managers’ capital investment decisions. The Accounting Review 83 (2), 351–376.
Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk. Econometrica 47 (2), 263–291.
Kaplan, R., Roll, R., 1972. Investor evaluation of accounting information: some empirical evidence. Journal of Business 45 (2), 225–257.
Kieso, D., Weygandt, J., Warfield, T., 2007. Intermediate Accounting, 12th ed. Wiley, New York, NY.
Kida, T., Moreno, K., Smith, J., 2001. The influence of affect on managers’ capital budgeting decisions. Contemporary Accounting Research 18 (3), 477–494.
Kim, J., Chung, R., Firth, M., 2003. Auditor conservatism, asymmetric monitoring, and earnings management. Contemporary Accounting Research 20 (2),
323–359.
Klammer, T., Koch, B., Wilner, N., 1991. Capital budgeting practices—a survey of corporate use. Journal of Management Accounting Research 3, 113–130.
Klammer, T., Walker, M., 1984. The continuing increase in the use of sophisticated capital budgeting techniques. California Management Review 27 (1),
137–148.
Lang, M., Lundholm, R., 1993. Cross-sectional determinants of analyst ratings of corporate disclosures. Journal of Accounting Research 31 (2), 246–271.
Leftwich, R., 1981. Evidence on the impact of mandatory changes in accounting principles on corporate loan agreements. Journal of Accounting and
Economics 3 (1), 3–36.
Leuz, C., Verrecchia, R., 2000. The economic consequences of increased disclosure. Journal of Accounting Research 38 (Suppl.), 91–124.
Maddala, G., 1983. Limited Dependent and Qualitative Variables in Econometrics. Cambridge University Press, New York, NY.
McConnell, J., Muscarella, C., 1985. Corporate capital expenditure decisions and the market value of the firm. Journal of Financial Economics 14 (3),
399–422.
McNichols, M., Stubben, S., 2008. Does earnings management affect firms’ investment decisions? The Accounting Review 6 (83), 1571–1603.
Moreno, K., Kida, T., Smith, J., 2002. The impact of affective reactions on risky decision making in accounting contexts. Journal of Accounting Research 40
(5), 1331–1349.
Okada, E., 2001. Trade-ins, mental accounting, and product replacement decisions. Journal of Consumer Research 27 (4), 433–446.
Peterson, M., 2009. Estimating standard errors in finance panel data sets: comparing approaches. The Review of Financial Studies 22 (1), 435–480.
Puhani, P., 2000. The Heckman correction for sample selection and its critique. Journal of Economic Surveys 14 (1), 53–68.
Richardson, S., 2006. Over-investment of free cash flow. Review of Accounting Studies 11 (2–3), 159–189.
Ricks, W., 1982. Market assessment of alternative accounting methods: a review of the empirical evidence. Journal of Accounting Literature 1, 59–102.
Rogers, W., 1993. Regression standard errors in clustered samples. Stata Technical Bulletin 13, 19–23.
Sangster, A., 1993. Capital investment appraisal techniques: a survey of current usage. Journal of Business Finance and Accounting 20 (3), 307–332.
Sawers, K., 2005. Evidence of choice avoidance in capital investment judgments. Contemporary Accounting Research 22 (4), 1063–1092.
Schall, L., Sundem, G., Geijsbeek, W., 1978. Survey and analysis of capital budgeting methods. The Journal of Finance 33 (1), 281–287.
Shehata, M., 1991. Self-selection bias and the economic consequences of accounting regulation: an application of two-stage switching regression to SFAS
No. 2. The Accounting Review 66 (4), 768–787.
Shin, H., Kim, Y., 2002. Agency costs and efficiency of business capital investment: evidence from quarterly capital expenditures. Journal of Corporate
Finance 8 (2), 139–158.
Staw, B., 1976. Knee-deep in the big muddy: a study of escalating commitment to a chosen course of action. Organizational Behavior and Human
Performance 16 (1), 27–44.
Titard, P., 1993. Managerial Accounting. Dame Publications, Houston, TX.
Tversky, A., Kahneman, D., 1991. Loss aversion in riskless choice: a reference-dependent model. Quarterly Journal of Economics 106 (4), 1039–1061.
United States Census Bureau, 2007. 2007 Capital Spending Report: US Capital Spending Patterns 1999–2005. United States Department of Commerce,
Economics and Statistics Administration.
Watts, R., Zimmerman, J., 1986. Positive Accounting Theory. Prentice-Hall, Englewood Cliffs, NJ.
Zajac, E., 1995. Political Economy of Fairness. MIT Press, Cambridge, MA.

RPS 11 Poin 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

RPS 11 Poin 2

Uploaded by

Copyright:

Available Formats

Journal of Accounting and Economics 48 (2009) 54–68

Contents lists available at ScienceDirect

Journal of Accounting and Economics

Economic consequences of ﬁrms’ depreciation method choice:

2.1. Capital investment decisions

2.2. Depreciation methods and capital investment decisions

Along similar lines, Dearden (1960, p. 83) states

More recently, Titard (1993, p. 386) states

3.2. First-stage regression

CHOICEit ¼ l0 þ l1DURit þ l2R&Dit þ l3LABORit þ l4DBENit

3.3. Second-stage regression

CAPXit ¼ d0 þ d1 CHOICEit þ d2 CAPXi t—1 þ d3 MBi t—1 þ d4 LEVTAi t—1

4. Sample selection procedures

5. Descriptive statistics and primary results

5.1. Descriptive statistics and univariate tests

Panel A: depreciation method choice proﬁle by year

SIC code SIC code description # % of total % SL % ACC

Panel B: depreciation method choice proﬁle by industry

Total 7843 100.00

5.2. First-stage probit regression results

CHOICEit 0.205 0.000 0.000 0.000 1.000 1.000

—0.004 —0.001 —0.004 —0.001 —0.005 —0.001

5.3. Second-stage regression results

Variables Predicted sign Coefﬁcient Z-statistic p-value

Intercept ? —0.035 —0.28 0.778

Observations Pseudo R2 (%) Model p-value 61,096

CHOICEit ¼ l0 þ l1 DURit þ l2 R&Dit þ l3 LABORit þ l4 DBENit þ l5 MFGit CGSit þ l6 NMFGit CGSit

5.4. Partitions of depreciation method choice

Variables Predicted sign Coefﬁcient t-statistic p-value

CAPXit ¼ d0 þ d1 CHOICEit þ d2 CAPXit þ d3 MBit—1 þ d4 LEVTAit—1 þ d5 D CASHit—1

5.5. Depreciation method changes

Variables Predicted sign Coefﬁcient t-statistic p-value

CAPXit ¼ d0 þ d1 CHOICEit þ d2 CAPXit þ d3 MBit—1 þ d4 LEVTAit—1 þ d5 D CASHit—1

t-statistic p-value Z-statistic p-value

Mean 0.047 0.039 7.50 o0.001

5.6. Research and development investments

Variables Predicted sign Coefﬁcient t-statistic p-value

Intercept ? 0.002 0.24 0.810

CAPXit ¼ d0 þ d1 CHANGEit þ d2 CAPXit þ d3 MBit—1 þ d4 LEVTAit—1 þ d5 D CASHit—1

5.7. Robustness tests

5.7.2. Median regression

5.7.3. Maximum likelihood estimation

5.7.5. Ranked data

5.7.6. Multiple lags of capital investments

You might also like