View
2
Download
0
Category
Preview:
Citation preview
The Effect of Tax Incentives on U.S. Manufacturing:
Evidence from State Accelerated Depreciation Policies
Eric Ohrn*
July 2019
Abstract
Accelerated depreciation policies decrease the cost of new investments by allowing firms to
deduct the new investments from their taxable income more quickly. Countries around the
world use these policies to stimulate business investment. To date, almost all empirical evidence
of the effect of these policies relies on cross-industry variation in their generosity. Using a
modified difference-in-differences framework, this paper estimates the effects of state adoption of
federal accelerated depreciation policies on the U.S. manufacturing sector. The results based on
this alternative source of quasi-experimental variation reinforce findings from the cross-industry
literature and suggest accelerated depreciation policies have large and significant effects on
capital investment.
Keywords : tax incentives, investment behavior, depreciation policy
JEL Classification : H25; E22, H5, H71
*ohrneric@grinnell.edu. Department of Economics, Grinnell College. 1226 Park St. Grinnell, IA 50112. Iam very grateful for comments and suggestions from Max Farrell, Xavier Giroud, Jim Hines, Logan Lee, JuanCarlos Suarez Serrato, Joel Slemrod, and Eric Zwick, and from seminar participants at the NBER State Taxation ofBusiness Income Conference, National Tax Association Annual Meeting, International Institure for Public FinanceAnnual Congress, Grinnell College Data Seminar Series, and Wesleyan University Economics Seminar. All errors aremy own.
1
1 Introduction
Accelerated depreciation tax incentives decrease investment costs by allowing firms to deduct new
capital purchases from their taxable income more quickly. These policies are widespread and costly.
Steinmuller, Thunecke and Wamser (2019) document that 41 countries used accelerated depreci-
ation policies to stimulate investment during the years 2004–2016. The U.S. federal government,
which currently allows firms to immediately deduct all investment costs, is spending approximately
$20 billion per year on accelerated depreciation incentives (JCT, 2017). Given their extensive
use and potentially high costs, understanding how accelerated depreciation polices affect business
investment is important to economists and policy makers alike.
This paper estimates how state adoption of federal accelerated depreciation policies affects
investment in the U.S. manufacturing sector. Specifically, I focus on state adoption of two federal
policies: Section 179 expensing and bonus depreciation. Section 179 expensing allows firms to
immediately deduct all new investments below a dollar value threshold referred to as the Section
179 “allowance” from their taxable income. The benefit is phased out at higher levels of investment
meaning the Section 179 deduction is only available for smaller firms that do less investment. The
second policy, bonus depreciation, allows firms to deduct a “bonus” percentage – up to 100% –
of the purchase price of new capital in the year it is purchased. Because bonus depreciation is
available regardless of the level of investment, it decreases investment costs for larger firms.
During the sample period 1997–2013, the U.S. federal government significantly increased the
Section 179 allowance and offered bonus depreciation at rates varying from 30 to 100%. Because
state corporate tax bases are intimately tied to the federal base definition, when these federal policy
changes were made, U.S. states had to decide how to respond. Many states chose to adopt the
policies at their federal levels. Other states decided to only partially adopt the policies. Finally, a
portion of states did not respond to the federal changes and offered no local versions of the federal
tax incentives.
This paper uses a modified difference-differences empirical framework leveraging the quasi-
experimental variation in state policy adoption and aggregate industry-by-state data from the
Annual Survey of Manufactures (ASM) to estimate how investment responds to state accelerated
depreciation policies. The modified difference-in-differences design includes terms for each policy
plus their interaction to account for the fact that firms only apply bonus depreciation to capital
expenditures in excess of the Section 179 allowance. I find state-level bonus depreciation and Section
179 expensing both have large and significant effects on investment activity. However, consistent
with the interaction of the policies, the effect of either policy is tempered as the other is made
more generous. I estimate that an increase in the state Section 179 allowance of $100,000 increases
investment by 2%. State adoption of 100% bonus depreciation increases investment by 18%.
The assumption underlying the modified difference-in-differences empirical design is that state
2
adoption of either policy is not correlated with other coincident shocks affecting investment. I
support this assumption in several ways. First, I predict state adoption of the policies based on
state level characteristics. Changes in state economic, political, and demographic characteristics do
not predict policy adoption. Second, I graph the evolution of gross state product and state budget
gaps around the time of state policy decisions and find no concerning trends in these variables prior
to policy adoption or rejection. Third, the results are estimated in the presence of industry-by-time
fixed effects suggesting time-varying shocks to particular industries do not drive the results. Fourth,
I provide dynamic estimates illustrating how investment differs between adopting and non-adopting
states in each year during the sample. No differential investment pretrends are apparent in years
prior to initial implementation of the policies. Furthermore, the largest differences in investment
coincide with years in which the policies were most generous. Fifth, consistent with the nature
of each policy, I show that the effects of Section 179 expensing were concentrated among firms
that do less investment while the effects of bonus depreciation were concentrated among firms
that do more investment. While the assumption underlying the research design is fundamentally
untestable, these checks significantly limit the risk that the study’s findings are the result of a
spurious relationship.
The primary contribution of this study is that it uses a novel source of quasi-experimental
variation to reinforce the findings of a well established literature examining the investment effects
of accelerated depreciation policies. To date, nearly all empirical studies in this area (especially in
the U.S. context) identify the effects of accelerated depreciation policies by comparing investment
in industries that typically invest in “long-lived” assets that are depreciated more slowly for tax
purposes to investment in industries that invest in “short-lived” assets that are depreciated more
quickly.1 A number of recent papers in the area have used this strategy to measure the investment
effects of the bonus depreciation policy (see House and Shapiro, 2008; Desai and Goolsbee, 2004;
Edgerton, 2010; Zwick and Mahon, 2017; Ohrn, 2018).2 These papers, like this study, find that
accelerated depreciation policies have large effects on investment.3
Combining the state policy identification strategy with industry-by-state investment data has
two distinct advantages over the boilerplate long-lived versus short-lived methodology. First, effects
can be estimated in the presence of industry-by-time fixed effects. As a result, in contrast to other
studies, this paper’s findings cannot be explained by industry shocks or trends. Second, the state
policy strategy avoids concerns that intra-industry competition may amplify the estimated effects.
1This strategy dates back to Auerbach and Hassett (1991) and Cummins, Hassett, Hubbard et al. (1994) who firstexploited the fact that accelerated depreciation policies disproportionately decrease present value investment costsfor firms that invest in long-lived assets because deductions are accelerated from further in the future.
2Two notable exceptions are Maffini, Devereux and Xing (2019) who exploit an investment-level discontinuity toestimate the investment effects of first-year allowances in the U.K. and Tuzel and Zhang (2018), who, inspired bythis study, measure how computer investment responds to state Section 179 allowances.
3Garrett, Ohrn and Suarez Serrato (2019) measures a strong local employment response to bonus depreciation byinteracting this same industry-level identification strategy with county-level industry location data.
3
As an example, Patel and Seegert (2015) document that not-for-profit hospitals, which do not
benefit from bonus depreciation, increased investment in response to bonus depreciation-induced
investments made by for-profit hospitals operating in the same market.
The nature of the data used in this study yields a second substantive contribution. The industry-
by-state ASM data is aggregated from establishment-level – as opposed to firm-level – survey data.
As a result, investment responses to state accelerated depreciation policies consolidate both within-
establishment increases in investment and reallocation of investment activity from one state to
another. Because the majority of studies in the literature use firm-level data, they do not account
for capital mobility responses.4 Accordingly, the elasticities I estimate are slightly larger than those
based on panel data that capture only within-firm responses. Because the slightly-larger elasticities
I estimate capture both avenues of response, they are more relevant than within-firm elasticities
from the perspective of state policy makers hoping to stimulate investment.
By documenting, for the first time, that state accelerated depreciation policies affect investment,
I also contribute to a recently reinvigorated literature showing that state corporate taxation and
corporate tax provisions can have large effects on real economic outcomes. Seminal papers include
Helms (1985), Wasylenko and McGuire (1985), Papke (1991), and Bania, Gray and Stone (2007)
while more recent entries are exemplified by Wilson (2009), Suarez Serrato and Zidar (2016), Heider
and Ljungqvist (2015), Ljungqvist and Smolyansky (2016), Suarez Serrato and Zidar (2018), and
Giroud and Rauh (2019).
The remainder of the paper is organized as follows. In Section 2, I describe both accelerated
depreciation policies in detail and document how the policies evolved at the federal and state
levels during the sample period. Section 3 presents a simple Neoclassical investment model to
more formally predict how investment responds to the accelerated depreciation policies and their
interaction. The data sources and empirical design are described in Sections 4 and 5. Section
6 presents the baseline empirical results, robustness checks, and heterogeneity. In Section 7, I
explore effects on outcomes other than investment. Section 8 calculates investment elasticities and
compares them to other studies in the literature. Section 9 concludes.
2 Bonus depreciation and Section 179 expensing
Typically, businesses in the U.S. cannot immediately deduct the full cost of newly purchased assets
from their taxable income. Instead, businesses deduct the value of the assets over time according
to the Modified Accelerated Cost Recovery System (MACRS) (detailed in IRS Publication 946).
MACRS specifies the life and depreciation method for each type of potential investment. For
equipment, lives can be 5, 7, 10, 15, or 20 years and the depreciation method is called the “declining
4Giroud and Rauh (2019) use disaggregated establishment-level data to show how state corporate income tax ratesaffect reallocation of business activity.
4
balance switching to straight line deduction method.” In the absence of accelerated depreciation
policies, the present value of depreciation deductions associated with $1 of new equipment is equal
to
z0 =T∑t=0
1
(1 + r)tDt,
where T is the life of the asset, Dt is the portion of the dollar that is depreciated in year t, and r
is the rate used to discount future cash flows.
Bonus depreciation allows firms to immediately write off an additional “bonus” percentage
– which I denote as b – of the total cost of new equipment purchases in the first year. The
remaining 1 − b percent of the equipment cost is depreciated according to MACRS rules. With
bonus depreciation, the present value of $1 of depreciation deductions is b+ (1 − b)z0.5
Federal bonus depreciation was first enacted in 2001 at a rate of 30%. In 2003, the additional first
year deduction was increased to 50%. The incentive was discontinued in 2005, but was reinstated
in 2008 at a 50% rate. Bonus depreciation remained available to businesses at the 50% rate through
2018, but for in 2011 when the federal bonus depreciation rate was increased to 100% and firms
were allowed to immediately deduct or simply to “expense” all new equipment costs.
Section 179 of the U.S. Internal Revenue Code allows businesses to expense limited amounts
of equipment investments. Thus, Section 179 expensing is equivalent to 100% bonus depreciation
for qualifying assets. Firms are allowed to deduct new capital assets dollar-for-dollar up to the
“Section 179 allowance,” the maximum Section 179 deduction that a taxpayer may elect to take
in a year. Critically, the Section 179 deduction is reduced dollar-for-dollar after the “Section 179
limit” is reached.
From 1997 to 2001 the Section 179 allowance was modest and grew annually from $18,000 to
$24,000. The allowance jumped to $100,000 in 2003 and grew to $125,000 in 2007 before vaulting
to $250,000 in 2008. The allowance saw another big rise in 2010 when it was increased to $500,000
where it remained through 2018. Panels (a) and (b) of Figure 1 display the federal Section 179
allowance and federal bonus depreciation rates during the years 1997–2013.
To better understand how the Section 179 allowance and limit determine the maximum Section
179 deduction consider the deduction in 2003. The Section 179 allowance was $100,000 and the limit
was $400,000. Panel (c) of Figure 1 shows how the maximum Section 179 deduction varies with the
amount of equipment investment. For firms investing less than $100,000, all investment costs can
5Notice that bonus depreciation increases the present value of depreciation deductions by b(1−z0). The benefit ofbonus depreciation is, therefore, larger when z0 is smaller or when MACRS rules dictate that equipment has a longerlife and is depreciated more slowly. Based on this observation, the majority of the existing literature (referencedin Section 1) uses cross-industry differences in the long-livedness of average investments to identify the effects ofbonus depreciation and accelerated depreciation policies more generally. Appendix A provides an example furtherillustrating the effect of bonus depreciation on the present value of depreciation deductions.
5
be immediately deducted under Section 179 rules. Firms investing between $100,000 and $400,000,
can immediately deduct the full allowance, as well. But, for every dollar a firm invests more than
the $400,000 limit, the deduction is decreased by a dollar. As a result, no deduction is available
to firms investing more than $500,000 during the 2003 tax year. Because the allowance dictates
the amount of investment to which the deduction applies, I focus on this number in analyzing the
effects of the policy.
[Figure 1 about here]
When both Section 179 expensing and bonus depreciation are available, the policies interact.
Firms first take the Section 179 deduction, then apply bonus depreciation and MACRS rules to
the remaining investment. To see this interaction, take, for example, the Section 179 and bonus
depreciation deductions together in 2003 when bonus depreciation was offered at a 50% rate. Panel
(d) of Figure 1 shows the amount of bonus depreciation deduction available to a firm at different
investment levels assuming the firm takes full advantage of the Section 179 deduction. Firms that
invest less than the Section 179 allowance of $100,000 deduct all of their investment under Section
179 and, therefore, receive no benefit from bonus depreciation. In contrast, firms investing more
than $500,000 are not allowed to write off anything under Section 179 rules, but instead deduct
50% of their investment costs immediately under bonus depreciation. For firms that invest between
$100,000 and $500,000, they expense as much as they can under Section 179 rules and then use
bonus depreciation on the remaining investment.
This example demonstrates three key insights. First, Section 179 applies mainly to smaller
firms doing less investment while bonus depreciation applies to larger firms that, on average, do
more investment. As a result, we would expect the investment effects of Section 179 allowances
to be concentrated among firms that do less investment and the effects of bonus depreciation to
be concentrated among firms that do more investment. The second insight is that increasing the
Section 179 allowance will mitigate the effect of bonus depreciation on investment because the base
on which bonus depreciation operates is smaller when the Section 179 allowance is higher. The
third, more nuanced insight is that the effect of increased Section 179 allowances is also mitigated
at higher bonus depreciation rates. When the Section 179 allowance is increased from I1 to I2, it
decreases the present value cost of all newly eligible investments (from I1 to I2). If those newly
eligible investments were already eligible for generous bonus depreciation, then the decrease in the
present value cost is smaller and the effect of increasing the allowance on investment should also
be smaller.
Zwick and Mahon (2017) estimate that 100% federal bonus depreciation – or, equivalently,
Section 179 expensing – decreases the present value cost of eligible investments by 5.46%. State
bonus depreciation and Section 179 expensing are inherently less valuable to firms than federal
incentives because the resulting deductions are taken against the state corporate tax rates which are
6
significantly lower than the 35% federal rate in place during the analysis period. Among adopting
states, the average state corporate income tax rate during the sample period was 7.0%. Because
the state tax rate is only 20% as high as the federal rate, state bonus depreciation adoption only
decreases the purchase price of new investments by 1.092% (= 0.2×5.46%). Appendix C calculates
the effect of state accelerated depreciation policies on the user cost of capital accounting for a
variety of federal and state tax provisions. The calculations show very similar effects of the policies
on the user cost of capital.
2.1 State adoption of federal accelerated depreciation policies
Because state corporate tax bases are intimately tied to the federal base definition, when the federal
policies were altered, states had to choose whether to adopt the federal policies. Figure 2 depicts
which states fully or partially adopted bonus depreciation during the sample period. When bonus
depreciation was first introduced, 17 states fully adopted the 30% federal rate. Five other states
adopted the federal policy at rates less than 30%. Three states, Wyoming, Nevada, and South
Dakota had no corporate income taxes during the sample period and, therefore, could not adopt
the federal incentive. The remaining 25 states did not offer bonus depreciation. In 2008, when
bonus depreciation was reintroduced, 12 states fully adopted while five partially adopted the 50%
rate. Overall, Figure 2 shows significant cross-state and within-state variation in adoption.6
[Figure 2 about here]
[Figure 3 about here]
Figure 3 depicts which states’ tax codes conformed to the federal Section 179 allowance. In 2001,
when the federal Section 179 limit was $24,000, nearly every state also allowed for full expensing of
investments up to the federal limit for state tax purposes. As the Section 179 allowance increased
during the years 2003–2011, most – but not all – states also increased their state Section 179
allowance and limits in step. The largest drops in the percentage of adopting states were in 2003,
when the federal allowance jumped from $24,000 to $100,000, and in 2010, when the allowance
increased from $250,000 to $500,000. Despite these large drops, by the end of the sample, more
than 60% of states still conformed to the federal Section 179 rules.
States that adopted the federal Section 179 rules are more likely to fully or partially adopt
bonus depreciation. As shown in Appendix Table A4, in 2004, of the 35 states that conformed to
federal Section 179 rules, 21 adopted bonus depreciation at some level. Of the the remaining states
that did not conform to the federal Section 179 rules, none of them offered bonus depreciation at
any rate. A very similar pattern is apparent during later years of the sample. The overlap between
the policies highlights the need to understand their interactive effects on investment.
6Appendix B provides more information detailing state adoption of both policies during the sample period.
7
3 Predicting responses
To better understand how the two policies and their interaction affect investment, in this section,
I present a stylized two period model in which a representative firm maximizes investment in the
presence of Section 179 expensing and bonus depreciation. The firm starts Period 1 with retained
earnings, X, and must decide how much to invest, I, and how much to pay out as a dividend,
D = X − I. I generates net profits according to the concave production function f(I). Profits are
taxed at rate τc.7 The investment economically depreciates at rate δ, but can only be depreciated
for tax purposes and deducted from taxable income at rate z. Investors can also purchase a
government bond that pays fixed rate r and therefore discount period 2 dividends by 1 + r. The
firm’s maximization problem can be written as
maxIV = (X − I) +
(1 − τc)f(I) + τczI + (1 − δ)I
1 + r.
Both bonus depreciation and Section 179 expensing affect z.
z = b+ (1 − b)z0
where b is the bonus depreciation rate (i.e. 0.3, 0.5, 1) and
z0 =
1 if I ≤ Section 179 allowance and
zMACRS if I > Section 179 allowance.
When the investment level is less than the Section 179 allowance, the full cost of the investment can
be depreciated in the first year and z = 1. When the investment level is greater than the Section
179 allowance, z0 is equal to the MACRS depreciation rate, zMACRS .8
The firm’s first order condition with respect to I is
f ′(I) =r + δ − τcz
1 − τc
and ∂I/∂z > 0, meaning that the profit maximizing level of investment increases as the investment
can be depreciated more quickly. How bonus depreciation and Section 179 expensing affect I
(through z) is slightly more complicated because the marginal effect of each policy depends on the
7τc is a generic corporate tax rate that can be construed to represent the federal, state, or a combined rate.8For simplicity, this specification ignores the Section 179 phase-out that occurs after I reaches the Section 179
limit. Because ASM data on firm-level investment is not very precise, any additional empirical predictions that couldbe made by adding the phase-out region are difficult to test empirically.
8
level of the other. The effect of bonus depreciation on investment is
∂I/∂b
= 0 if z0 = 1 when I ≤ Section 179 allowance), but
> 0 if z0 < 1 when I > Section 179 allowance).
When the investment level is less than the Section 179 allowance, bonus depreciation does not
increase z and does not incentivize investment. On the other hand, for marginal investments over
the Section 179 allowance, bonus depreciation increases z and incentivizes investment. Put simply,
investments under the Section 179 allowance are already immediately expensed and therefore cannot
benefit from bonus depreciation. At higher Section 179 allowances, bonus depreciation has no effect
for a larger portion of investment.
The effect of an increase in the Section 179 allowance on investment also depends on bonus
depreciation rate, b. When the Section 179 allowance is increased, z0 is now equal to 1 for newly el-
igible investments. For these newly eligible investments, ∂I/∂z0 > 0, but ∂2I/∂z0∂b < 0, meaning
the effect of increases in the Section 179 allowance are smaller at higher bonus rates. In sim-
pler terms, an increase in the Section 179 allowance simply isn’t worth as much to newly eligible
investments when a bonus percentage is already deducted in the first year.
To reinforce this intuition, imagine increasing the allowance from $100,000 to $250,000 when the
bonus depreciation rate is either 0 or 50%. When the bonus depreciation rate is 0%, investments
from $100,000 to $250,000 essentially go from no additional first year deduction to being fully
deductible. When the bonus depreciation rate is 50%, investments from $100,000 to $250,000 go
from being 50% deductible in the first year to 100% deductible. The effect of Section 179 expensing
on the present value cost of these newly eligible investments is lower when the bonus depreciation
rate is higher.
These comparative statics suggest two symmetric, empirically testable hypotheses that motivate
the study’s empirical design. First, state bonus depreciation will increase investment and the effect
will be concentrated among states that have low Section 179 allowances. Second, increases in state
Section 179 allowances will increase investment and the effect will be concentrated among states
that offer lower bonus depreciation rates.
Corollary hypotheses are generated by recasting these comparative statics in terms of firm
investment levels. The first corollary hypothesis is that state adoption of bonus depreciation will
increase investment, and the effect will be concentrated among firms that invest at levels beyond the
Section 179 allowance. Second, state conformity to Section 179 allowances will increase investment,
and the effect will be concentrated among firms that invest at levels below the Section 179 allowance.
Section 6.5 tests these corollary hypotheses by exploring heterogeneity using a proxy for investment
level.
Before moving on, I note a potential shortcoming of this simple model. The model predicts that
9
the effect of each policy increases in τc. However, this may not be the case if firms are choosing to
reallocate investments across state lines. In most instances, generous depreciation rules and high
tax rates will not be preferred to slower depreciation and lower corporate income tax rates. Because
the model is based on a representative firm operating in a single jurisdiction, it abstracts from this
reallocation effect. I further discuss this nuance and empirically explore how state corporate income
tax rates interact with the policies in Subsection 6.5.
4 Data sources
To estimate the effects of state adoption of bonus depreciation and Section 179 expensing, I rely
on business activity data from the ASM, state bonus depreciation data from Lechuga (2014),
hand collected state Section 179 allowance data, and time-varying state economic, political, and
demographic control variables from various sources. Table 1 provides descriptive statistics for the
analysis sample.
[Table 1 about here]
4.1 Manufacturing data
Measures of business activity come from the ASM and the Economic Census, both products of the
U.S. Census Bureau. The sample period is the years 1997-2013. The ASM is conducted annually in
all years except for years ending in 2 and 7. In those years, corresponding data are available from
the Economic Census. The ASM provides sample estimates and statistics for all manufacturing
establishments with one or more paid employees, which is the entire Economic Census sample. As
a result, statistics in all years are comparable.
The unit of observation in the study is state-level 3-digit North American Classification System
(NAICS) industries. The choice to use industry-by-state data (as opposed to state-level data) is
made for two reasons. First, the more granular data allows me to control for industry-by-time
shocks. This constitutes an improvement over nearly all other studies in the literature because
papers that use cross-industry differences in policy generosity to identify effects cannot control for
time-varying shocks that apply to one industry but not another. Second, the industry-by-state level
data allow me to explore heterogeneity in effects by investment level. As Section 179 expensing
should be most valuable for smaller firms that do less investment and bonus depreciation should
be most valuable for larger firms that do more investment, heterogeneity along this margin helps
to support the internal validity of the study.
There are 21 3-digit NAICS manufacturing industries and 1022 observational units used in the
baseline analysis.9 The primary business outcome I investigate is the log of real capital expenditure
9If each NAICS × State unit was represented there would be 1050 units. Some industries are either not representedin some states or there are too few establishments to report statistics while maintaining confidentiality.
10
which I denote Log(CapX). In Section 7, I also explore the effects of the policies on the log of
compensation per employee, the log of employees, and the log of real total shipments, a measure
of total output. From the ASM data, I also construct Invest Level equal to capital expenditure
divided by the number of establishments in each state-industry unit in the year 2002.10
4.2 State policy variables
Derived from Lechuga (2014), State Bonus is the state bonus depreciation rate. The variable can
be thought of as an interaction between the federal bonus rate and State Adoption, which takes
on values between 0 and 1 and describes the extent to which a state adopts the federal bonus rate.
State Adoption is equal to 0 if an observational unit is located in a state that fully rejects the policy
in a given year. State Adoption is set equal to 1 for states that adopted the policy at the federal
level. When state bonus is adopted at X% of the federal rate, State Adoption is set to X/100.
State 179, the state-level Section 179 allowance, was hand collected for each state in each year
2000–2014 from department of revenue documents, web pages, and interviews with state employees.
Like State Bonus, State 179 can be expressed as an interaction between the federal Section 179
allowance and State Conformity, which describes the extent to which a state’s Section 179
allowance matches the federal allowance. The State 179 variable is scaled so a one unit increase
denotes a $100,000 increase in the state Section 179 allowance.
4.3 Time-varying state controls
Time-varying state level data is used to explore any systematic differences between states that do
and do not adopt federal bonus depreciation and Section 179 rules. From The Book of States data, I
construct Corp Rev % (the percentage of total state revenue derived from state corporate income
taxes), Budget Gap (total state deficit as a fraction of total state revenue), Democratic Legis-
lator % (percentage of state legislators that identify as Democrats), and Democratic Governor
(an indicator equal to 1 if the governor is a Democrat). I take Corp Tax Rate (the top marginal
corporate income tax rate in each state) from The Tax Foundation. I take State Population from
Census and construct Real GSP per capita from Census and BEA data.
5 Empirical design
Because state adoption of each policy is correlated and because both are designed to spur in-
vestment, estimating the effect of either policy alone will lead to biased estimates. Furthermore,
10The number of establishments data needed to construct Invest Level is only available via the Economic Censuswhich is taken in years ending in 2 and 7. I rely on 2002 data because it is very close to the first implementationof bonus depreciation and prior to the first major increases in the Section 179 allowance. Because the year is soearly, it is unlikely that either policy will dramatically affect average investment per establishment, especially whenindustry-by-time fixed effects are included in the analysis.
11
following the logic laid out in Section 3, because the effect of one policy is predicted to decline when
the other is offered at more generous levels, the empirical specification must include the interaction
between the two policies to provide unbiased estimates of the effect of each policy.11 Accordingly,
I estimate the effect of both policies simultaneously using the following regression equation
ln(CapX)jst = β0 + β1[State Bonusst] + β2[State 179st] (1)
+ β3[[State Bonusst] × [State 179st]
]+ X′
stγ + σt + νjs + ζjt + ψs + εjst
where j denotes NAICS 3-digit industries, s denotes state, and t denotes time. In addition to the
policy variables, equation (1) includes an interaction between the two policies. The fully specified
model also includes industry-state fixed effects (νjs) to control for time invariant determinants
of business activity, year fixed effects (σt) to control for aggregate trends, time-varying state-
level controls (X′st), state-specific linear time trends (ψs) to account for state-level changes in
the business environment, and finally, industry-by-year fixed effects (ζjt) to control for changes
in business activity that occur at the industry-level.12 Because State Bonus is equal to State
Adoption interacted with the federal rate and State 179 is equal to State Conformity interacted
with federal Section 179 allowances, β1 and β2 can both be interpreted as difference-in-differences
(DD) estimates and equation (1), which contains State Bonus, State 179, and their interaction
represents a modified DD methodology.
β1 is interpreted as the percentage increase in investment experienced by a state that fully adopts
100% federal bonus depreciation relative to the increase that a fully rejecting state experiences
when neither state allows for any Section 179 expensing. Similarly β2 is the impact of an additional
$100,000 in Section 179 allowances when there is no state bonus depreciation. The β3 coefficient
captures how much β1 or β2 change as State Bonus or State 179 are increased. More precisely
β3 is equal to the change in the effect of bonus depreciation adoption (β1) that occurs when state
Section 179 allowances increase by $100,000. Because state bonus depreciation is predicted to be
less effective when Section 179 allowances are high and Section 179 allowances are predicted to have
less effect when bonus depreciation is adopted at higher rates, the interaction term is predicted to
be negative.13
11Columns (5)–(7) of Appendix Table A6 show that, consistent with this logic, estimating the effect of either policyalone or the effect of both policies without the interaction produces downward biased estimates.
12The state-specific linear time trends are important to include for two reasons. First, because the accelerateddepreciation policies vary at the state-level over time, state-by-year fixed effects cannot be included in the model.Second, because Section 179 allowances increase in a linear fashion over time, estimates of the effect of State 179 willbe especially sensitive to linear trends in investment, positive or negative, generated by unobserved factors. Addingthe state-specific linear trends tempers this concern.
13In order to jointly estimate β1, β2, and β3, state bonus adoption and state Section 179 conformity cannot beperfectly collinear. Table A4 describes the overlap of the two policies during the first and second episodes of bonus.Because, there is significant variation in bonus adoption among Section 179 conforming states, all three coefficientscan be jointly estimated.
12
5.1 Predicting state adoption
The assumption underlying the modified DD identification strategy is that investment in states that
adopt the policies would follow the same path as investment in states that do not in the absence
of the policies. While this assumption is inherently untestable, the year, industry-by-time, and
state-by-year fixed effects as well as state-specific linear time trends included in most specifications
minimize the concern that this assumption is violated. Even in the presence of these fixed effects
and trends, a particular concern is that states that adopted the policies were observably different in
terms of economic, demographic, or political characteristics and that these differences are correlated
with investment patterns.
As a first pass in assessing whether adopting states are observably different, I use a linear
probability model to regress three adoption outcomes, (1) state adoption of bonus depreciation,
(2) state adoption of Section 179 expensing, and (3) state adoption of both policies on state eco-
nomic, demographic, and political characteristics. Table 2 presents the linear probability model
results. Specifications (1)–(3) include just year fixed effects to control for aggregate trends. Speci-
fications (4)–(6) add state fixed effects. Specifications (7)–(9) add state-specific linear time trends.
The results presented in specifications (1)–(3) suggest that, indeed, states that adopt the policies
have difference characteristics than than those that do not. However, when state fixed effects or
state fixed effects and state-specific time trends are included, these observable differences disap-
pear suggesting that changes in state productivity, population, corporate tax rates, budget gaps,
and political representation do not predict adoption of bonus depreciation, Section 179, or both
policies. Because the DD methodology, which includes observational-unit fixed effects, relies only
on changes in variables within a unit, the Table 2 results suggest economic, political, and demo-
graphic differences between adopting and non-adopting states will not undermine the underlying
DD assumption.
[Table 2 about here]
As a second check, in Appendix D, I create event study graphs that show how GSP growth
rates and state budget gaps evolve in three situations: when state bonus depreciation was initially
implemented in 2001 or reimplemented in 2008, when states choose not to conform to federal
Section 179 allowances, and when states adopt or reject federal bonus depreciation in years other
than 2001 or 2008. 9 of the 10 event study graphs included in the appendix show no divergence
in GSP growth rates or budget gaps prior to adoption or rejection. The one exception shows state
budget gaps were increasing in states that adopted bonus depreciation in years other than 2001 or
2008.14 While these diverging pretrends are concerning, budget gaps are likely negatively correlated
with investment. Therefore, this pattern is likely to downward bias the estimated effects of state
14The states were Florida in 2004, Illinois in 2011, Iowa in 2003, Maine in 2011, North Carolina in 2003, Oregonin 2010, and Pennsylvania in 2011.
13
adoption of bonus depreciation. Consistent with this logic, I find larger bonus depreciation effects
when I eliminate these states from the sample (specification (6), Table 3).
Overall, changes in state economic, political, and demographic characteristics do not seem to
predict state adoption of accelerated depreciation policies during the sample period. Furthermore,
GSP growth and budget gaps – for the most part – do not diverge between adopting and non-
adopting states. These analyses support the underlying DD assumption that investment in adopting
and non-adopting states would trend similarly in the absence of policy adoption. Despite this
conclusion, in Section 6.1, I use Entropy Balancing methods (Hainmueller (2012)) to verify that
level differences in state characteristics are not responsible for the estimated policy effects.
6 Effect of state accelerated depreciation policies on investment
Table 3 presents estimates of coefficients from equation (1) describing the effect of State Bonus,
State 179, and their interaction on investment. The baseline specification presented in column
(1) presents estimates of the effect of policy adoption in the presence of year, NAICS × State,
and NAICS × year fixed effects as well as state-specific linear time trends and time-varying state
controls.15 Standard errors in this specification and throughout the paper are clustered at the
state-level.16
The coefficient on State Bonus is 0.180 and statistically significant at the 95% level. The 0.180
parameter indicates that state adoption of 100% federal bonus depreciation increases investment by
18% when state Section 179 allowances are set to $0. The State 179 coefficient of 0.0206 is smaller,
but still significant at the 95% level, indicating that an increase in state Section 179 allowances of
$100,000 increases manufacturing investment by 2.06% when state bonus rates are set to 0%. While
this effect is smaller, consider that the federal Section 179 allowance has been set at $500,000 since
2010. As a result, fully conforming to federal Section 179 levels after 2010 increases manufacturing
investment just over 10% if no federal bonus depreciation is adopted.
[Table 3 about here]
The coefficient on the interaction term is also statistically significant but negative in sign,
meaning that, as hypothesized, an increase in the generosity of one policy undermines the effect of
the other. The -0.0420 magnitude means that for every $100,000 increase in Section 179 allowances,
15Specifications (1)–(4) of Table A6 presents models with varying levels of fixed effects and controls. Acrossall models, adoption of bonus depreciation has a large and significant effect on investment. Consistent with theobservation that federal Section 179 allowances increase at a nearly linear rate during the sample period, the effectof State 179 is most precisely estimated in the presence of state-specific linear time trends.
16Following Cameron and Miller (2015), because both State Bonus and State 179 vary at the state level and overtime, standard errors are clustered at the state level. Appendix G constructs p-values for the baseline specificationpolicy coefficients using randomization inferences methods. These situation-specific p-values are smaller than thosepresented in Table 3 suggested the p-values produced by the state-level clustering procedure are conservative.
14
adoption of 100% bonus depreciation stimulates 4.2% less investment. The interaction coefficient
can also be interpreted as the decrease in the effect of State 179 when state bonus depreciation
is increased from 0 to 100%. Thus, the interaction term suggests that the effect of Section 179
expensing decreases by 4.2% when 100% bonus depreciation is adopted by a state.17
6.1 Matching based on state characteristics
The analysis presented in subsection 5.1 showed that states that adopted the federal accelerated
depreciation policies are different in terms of economic, political, and demographic characteristics
than those that did not.18 To combat the concern that these differences may drive the empirical
results, I rely on the Entropy Balancing methods developed in Hainmueller (2012) to match adopting
and non-adopting units based on the suite of economic, political, and demographic controls used
throughout the paper. Entropy Balancing is especially useful in this context because the effect of
continuous treatment variables can be estimated after the sample is reweighted to match adopting
and non-adopting states.19
In specification (2) of Table 3, states that did not adopt bonus depreciation are reweighted
so that their characteristics match those of states that adopted bonus depreciation in at least 9
of the 11 years the policy was offered federally. In specification (3), states that adopted federal
Section 179 rules during the entire sample period are matched to those that did. In specification
(4), states that both adopted bonus depreciation in 9 of the 11 possible years and adopted federal
Section 179 rules during the entire sample period are matched to those that did not. Across all
three Entropy Balanced specifications, the coefficients on all three policy variables have the same
direction, are similar in magnitude, and maintain at least 95% statistically significance levels. In
sum, the Entropy Balanced results suggest that differences across states in economic, political,
or demographic characteristics are not responsible for the estimated effects of state accelerated
depreciation policies.
6.2 Lasso-selected controls
In specification (5) of Table 3, I use Lasso machine learning techniques (Tibshirani, 1996) to choose
time-varying state controls from an expanded suite of potential covariates that includes the baseline
economic, political, and demographic controls plus a full set of interactions between each. Policy
17In Appendix E, I use the baseline estimates to plot the marginal effect of adopting each policy at the federal levelduring each sample year assuming states adopt the other policy at the average adoption level.
18Recall, the same analysis showed that these differences disappear in the presence of state fixed effects and thereforedo not violate the underlying DD identification assumption.
19I rely on Entropy Balancing to match units rather than more traditional matching techniques because theidentification strategy relies on within-state changes in the generosity of the policies, not just whether a state adoptedthe policy or not. Because traditional matched difference-in-differences methods can only estimate the effect of binarytreatments, these techniques cannot accommodate the policy variation upon which this study is based.
15
estimates are stable in the presence of Lasso-selected controls suggesting the results are robust to
alternative control variable choices.
6.3 Eliminating abnormal bonus depreciation adopters
The analysis in Appendix D suggests that states that adopt bonus depreciation in years other than
2001 or 2008 experience increased budget gaps prior to adoption. As noted earlier, if anything, this
pattern is likely to downward bias the estimated effects of state adoption of bonus depreciation.
Specification (6) of Table 3 eliminates these abnormal adopters from the sample. Consistent with
the prediction of downward bias, the coefficients on all three policy variables increase in size relative
to the baseline specification. The State Bonus coefficient is now only marginally significant, but
this is likely due to the smaller sample size. The specification (6) result suggests that these states,
which may adopt bonus depreciation based on changes in their budget gaps, are not driving the
empirical results and do not substantial bias the baseline estimates.
6.4 Dynamic DD analysis
If state adoption of accelerated depreciation policies affects investment, then investment should
trend similarly in industry-state units that do and do not adopt the federal policies prior to adoption
and then diverge upon adoption. Furthermore, differences in investment patterns between adopting
and non-units should be largest in years when the policies are offered at more generous levels. To
test whether these patterns reinforce the baseline results, I present the following dynamic DD
analysis.
I estimate the impact of state adoption of bonus depreciation in each year by replacing the
State Bonus variable in specification (1) with an interaction between adoption and a series of year
indicators:
2013∑k=1997
βk1[State Adoptions,t × 1[Yeark]
].
The coefficients β19971 –β20141 are the difference in log investment between adopting and non-adopting
units in each year.20 Panel (a) of Figure 4 presents these coefficients.21 In panel (b), these coeffi-
cients are grafted onto the average investment trend for context. Panels (c) and (d) focus just on
2001–2004 and 2008–2010 adoption, respectively.
20To create comparable pretrends, adopting and non-adopting states in years 1997–2000 and 2005 is set equal tothe average State Adoption observed in years 2001–2004 and State Adoption in years 2005–2007 is set equal to theaverage State Adoption observed in years 2008–2010.
21Notice, these estimates differ from those in Figure A5 in two ways. First, they are not marginal effects. Thatis, they represent the impact of each policy assuming the other is set to zero. Second, a single estimate is not usedto predict responsiveness in each year. Instead, the responsiveness of adopting to non-adopting units is estimatedseparately in each year.
16
[Figure 4 about here]
Panels (a) and (b) show that, in years 1997–2000, before federal bonus depreciation was first
implemented, there is no difference in investment behavior between adopting and non-adopting
units. In 2002, investment begins to increase for adopting units. The divergence is even stronger
in years 2003–2004 when the bonus rate is increased to 50%. When bonus depreciation is turned
off in 2005–2007, investment differences decrease. In 2008, when federal bonus depreciation is
reinstated, investment by units in adopting states diverges sharply from investment by units in
non-adopting states. The largest divergence between adopting and non-adopting units is in 2011,
when federal bonus depreciation is offered at 100%. When bonus depreciation is scaled back in 2012,
the difference again converges. Panels (c) and (d) which focus on adoption of bonus depreciation in
2001 and 2008 separately each show similar investment pretrends and diverging investment behavior
upon adoption. Overall, consistent with the baseline estimates, this dynamic analysis suggests state
adoption of bonus depreciation has a substantial effect on investment.
I alter the estimation procedure in two ways to examine the effects of state Section 179 adoption.
First, instead of replacing State Bonus in Specification (1), I now replace State 179 with
2014∑k=2003
βk2[State Conformitys,t × 1[Yeark]
].
Second, I limit the analysis to years after 2003 because prior to 2003 nearly all states fully conformed
to the federal Section 179 allowance level.
[Figure 5 about here]
Panel (a) of Figure 5 presents the coefficient estimates. Panel (b) adds these coefficients to the
average investment trend. Investment trends similarly in units in adopting and non-adopting states
prior to 2008 when the federal allowance was set near $100,000. Coefficients are larger (although not
always statistically different from zero) after 2008 when the federal allowance increased to $250,000
and then $500,000. While the dynamic evidence is less strong than in the case of bonus depreciation,
overall, Figure 5 shows state adoption of federal Section 179 allowances led to increased investment
by industries in adopting states after 2008.
6.5 Heterogeneity in effects of state accelerated depreciation policies
The simple model presented in Section 3 demonstrates that bonus depreciation only affects firms
that invest beyond the Section 179 allowance level. Additionally, Section 179 expensing should
only affect marginal investment decisions for firms that invest below the Section 179 limit. To
implement a simple test of these predictions, I reestimate the baseline specification (column (1) of
17
Table 3) including interactions between State Bonus, State 179, and State Bonus × State 179 with
Invest Level, a proxy for the level of investment undertaken by each firm. Invest Level is equal to
the 2002 per establishment level of investment for each NAICS × State observation. Invest Level
is scaled so each unit translates to a $100,000 increase in investment per establishment.
[Table 4 about here]
Column (1) of Table 4 presents the Invest Level heterogeneity results. The State Bonus coeffi-
cient is now statistically indistinguishable from zero, but the State Bonus × Invest Level interaction
is positive and statistically significant suggesting that the effect of adoption of state bonus deprecia-
tion is, as expected, larger for units that do more investment. The State 179 coefficient is now larger
than in the baseline case and the State 179 × Invest Level coefficient is negative and marginally
statistically significant. Together these coefficients imply that, as predicted, the effects of state
Section 179 allowances are concentrated among firms that do less investment. The State Bonus
× State 179 × Investment Level coefficient is negative and statistically significant suggesting the
interaction between the policies is concentrated among firms that do more investment. The sign
of this coefficient is consistent with the logic that Section 179 allowances only mitigate the effect
of bonus depreciation at high allowance levels for firms that do lots of investment. Overall, the
investment level heterogeneity results are entirely consistent with the nature of the two policies.
The adoption of bonus depreciation has larger investment effects for firms that do more investment
while conformity to federal Section 179 allowances mainly affects firms that do less investment.
This result further reinforces the internal validity of the study.
A second heterogeneity prediction of the Section 3 model is that the effects of both policies
should be concentrated in states with higher corporate income tax rates. However, as noted in
Section 3, accelerated depreciation provisions are unlikely to cause firms to reallocate investments
from low tax to high tax states. Therefore, reallocated investment is more likely in low tax states.
If reallocation effects constitute a substantial portion of the investment response, then they may
swamp out the predicted tax rate heterogeneity.
To explore state tax rate heterogeneity, in specification (2) of Table 4, I interact State Bonus,
State 179, and State Bones × State 179 with High Tax, an indicator equal to 1 if the maximum
state corporate income tax rate is equal to or above 7%, the average (and median) rate in the
analysis sample. All three interaction coefficients are statistically insignificant implying the effect
of state accelerated depreciation policies are not concentrated among states with high corporate
income tax rates. This result suggests that reallocation of investment across state lines makes up
a non-negligible amount of the measured investment response.
18
7 Other outcomes
Overall, the results presented in Section 6 show that adoption of both accelerated depreciation
policies increase investment and that the effect of either policy on investment is mitigated as the
generosity of the other increases. I now turn to exploring the effects of state accelerated depreciation
policies on other outcomes of interest. Columns (3)–(5) of Table 4 show the effects of State Bonus,
State 179, and State Bonus × State 179 on log employment, log compensation per worker, and log
output. The estimates suggest that bonus depreciation has no effect on employment or output.
However, bonus depreciation does have an effect on compensation per employee. Adoption of 100%
state bonus depreciation with no Section 179 allowance is estimated to increase compensation per
worker by just under 2%, although the result is only statistically significant at the 10% level.
Increasing state Section 179 allowances by $100,000 decreases this effect by just over 0.5%. A
natural explanation for these three results is that there are short-run adjustment costs to hiring
more workers. As a result, the same number of employees may work longer hours, resulting in more
compensation. If this is the case, the fact that only bonus depreciation increases compensation
may be because larger firms are more able respond on this hours / compensation margin. While
investment may increase output in the long-run, a null effect in the short-run is plausible especially
if installing new machinery temporarily slows production.
Interestingly, the employment results contrast with Garrett, Ohrn and Suarez Serrato (2019),
who show that federal bonus depreciation had significant positive effects on employment in local
labor markets. The divergence in findings is likely due to differences in response timing. Garrett,
Ohrn and Suarez Serrato (2019) show that almost all of the local employment response due to
industry-level differences in bonus depreciation generosity happened in the 2001–2005 period. If
state bonus depreciation adoption also increased employment only in the 2001-2005 period, the
modified DD methodology that I use would likely show no effects as adoption of the federal policy
at higher rates during the later period had no effect on employment.
8 Constructing and comparing investment elasticities
The primary contribution of this study is to use a new source of quasi-experimental variation to
asses whether and to what extent accelerated depreciation policies affect investment. Thus far, the
study has provided a resounding ‘yes’ to the question of whether accelerated depreciation policies
affect investment. To answer the question of extent, I now calculate the elasticity of investment
with respect to its the net of tax rate, 1 − τf − τsz.22. I then compare these estimates to those
based on the cross-industry literature and attempt to explain discrepancies.
I begin by calculating the bonus depreciation investment elasticity. To construct the elasticity, I
22This is equivalent to the present-value after tax cost of $ of investment
19
divide the baseline estimated percentage increase in investment due to state adoption of 100% bonus
depreciation of 18.0% by the percentage change in net of tax rate due to 100% bonus depreciation.
As described in Section 2, 100% state bonus depreciation decreases the net of tax rate by 1.092
cents per dollar of investment and decreases 1− τf − τsz by 1.883% assuming a federal rate of 35%,
a state corporate rate of 7%, and z = 0.88 from Zwick and Mahon (2017). Dividing the 18% by
1.883% yields an elasticity of investment with its after-tax cost of 9.55.
Section 179 expensing also decreases the after-tax present value cost of eligible investments by
1.883%.23 Simply dividing the 2.06% State 179 effect by the 1.833% decrease in investment costs
yields an elasticity of 1.12. This understates the real effect of the policy because each $100,000
increase in the Section 179 allowance from $0 to $500,000 only hits about 13% of the units in the
sample.24 Thus, we can multiply the 1.12 by 1/0.13 to scale the elasticity to measure the effect
of decreasing the cost of investment by 1.833% for the full sample. This yields a state Section
179 elasticity of 8.62, which is relatively close to the 9.55 state bonus depreciation elasticity. The
remaining disparity is likely due to the fact that plants that do more investment are more likely
to be part of large multi-state corporations that can reallocate investment across states to take
advantage of state accelerated depreciation policies.
Because of the additional complications introduced in scaling the Section 179 elasticity, I prefer
to focus on the bonus depreciation elasticity in comparing this study to others. Three recent papers
use variation in accelerated depreciation policies to measure the elasticity of investment with respect
to the net of tax rate. Using industry-level variation created by federal bonus depreciation, Zwick
and Mahon (2017) report an elasticity of 7.2 for all U.S. firms. Ohrn (2018) uses the same estimation
strategy but focuses on publicly traded firms and reports an elasticity of 3.98. Maffini, Devereux
and Xing (2019) exploit changes in the definition of Small Medium Enterprises in the UK, which
bestowed medium sized firms with more generous first year depreciation allowances, and find an
elasticity of 8.7.
Of these three studies, the Zwick and Mahon (2017) elasticity is most comparable to the 9.55
bonus depreciation elasticity. Both studies estimate responses to the same type of accelerated
depreciation policy and over roughly the same time period. The 9.55 elasticity I estimate is 32.6%
larger than the 7.2 elasticity from Zwick and Mahon (2017). Because the Zwick and Mahon (2017)
estimate is based on firm-level tax return data and an industry-level treatment, it does not capture
reallocation effects as I do in this study. Therefore, if all of the difference were due to reallocation
responses, then contrasting the two elasticities would suggest 32.6% of the investment response to
state accelerated depreciation policies is due to reallocation of investment across state lines.25
23Recall Section 179 expensing is akin to 100% bonus depreciation for eligible investments.24In 7,765 of the 11,979 units for which Invest Level is available, investment is less than $500,000 per year. Therefore
13% of units invest in each $100,000 interval from 0 to $500,000.25Giroud and Rauh (2019) find a slightly higher ratio of within firm reallocation to total response (approxi-
mately 50%) when they estimate the effect of changes in state corporate tax rates on employment using linkedfirm-establishment data.
20
This reallocation-effect difference between the two elasticities highlights the use of aggregated
state-industry level data based on establishment-level surveys in this study. The use of this aggre-
gated data means within-establishment responses and reallocation of investment are combined into
a single response. On the one hand, this does not allow for a disentangling of reallocation and non-
reallocation effects or estimation of deeper structural parameters of the firm. On the other hand,
from the perspective of the state policy maker whose goal is to attract mobile capital, disentangling
the effects is largely unnecessary. Whether the response is generated by within-firm increases or
reallocation does not matter with regard to state economic growth or state tax revenues. As a
result, the more aggregate elasticity represents a sufficient statistic, at least in most settings, for
the state policy maker.
9 Conclusion
This study has used a novel source of quasi-experimental variation – state adoption of federal
incentives – to measure the effects of accelerated depreciation policies on investment in the U.S.
manufacturing sector. I find large effects of state adoption of both bonus depreciation and Section
179 expensing on investment. Consistent with the fact that firms can only apply bonus depreciation
to investment in excess of the Section 179 allowances, I find the effect of either policy is mitigated
as the other is made more generous. By demonstrating significant investment responses to the
policies, the results presented herein corroborate a large number of papers that use industry-
level differences in the generosity of depreciation policies to identify investment effects (Auerbach
and Hassett, 1991; Cummins, Hassett, Hubbard et al., 1994; House and Shapiro, 2008; Desai and
Goolsbee, 2004; Edgerton, 2010; Zwick and Mahon, 2017; Ohrn, 2018).
While a litany of checks were used to verify the validity of the empirical design and robustness
of the findings, the study is limited in several ways. First, because the study uses state-by-industry
data, the methodology cannot distinguish between within-establishment investment responses and
reallocation of investment from one state to another. Comparing the elasticity of investment with
respect to the net-of-tax rate derived from this study’s results to the Zwick and Mahon (2017)
elasticity suggests approximately a third of the response to state accelerated depreciation policies is
driven by reallocation. Combining the state-level policy variation with matched firm-establishment
data could yield more insights into the relative size of each type of investment response.
A second limitation is that because the identifying variation is cross-sectional in nature, the
estimates provided – like those using industry-level variation – do not speak to the aggregate
effects of the accelerated depreciation policies. However, if anything, the aggregate effects seem
to be larger than cross-sectional elasticities would suggest. Edge and Rudd (2011) study bonus
depreciation in the context of a linearized New Keynesian model and find aggregate elasticities
21
exceed the cross-sectional elasticities found by House and Shapiro (2008).26
A final limitation is that the study focuses exclusively on investment behavior in the man-
ufacturing sector. Although the manufacturing sector represents the largest share of equipment
investment among any sector in the U.S. (BEA, 2019), knowing how state accelerated depreciation
policies affect other sectors is crucial in designing effective and well-targeted tax incentives.
Despite these limitations, this study’s results have several important implications for policy
makers. First, accelerated depreciation incentives work to stimulate business investment. As a
result, these policies can be used to promote economic growth and to counter business cycles. Sec-
ond, expensing provisions that phase out at higher-levels of investment effectively target firms that
do less investment and are, ostensibly, smaller. Policies like Section 179 expensing are, therefore,
well-suited to promote business investment among “main-street” businesses as opposed to larger
corporations. Third, due to the high mobility of capital, accelerated depreciation provisions can
have extra-large effects when implemented at the subnational-level. Therefore, accelerated depreci-
ation policies are especially cost-effective tools for state and local policy makers looking to stimulate
investment.
26Focusing on the first episode of bonus depreciation (2001–2005), House and Shapiro (2008) find investmentelasticizes between 6 and 14.
22
10 Figures
Figure 1: Federal Accelerated Depreciation Deductions
(a) Section 179 Expensing over time
010
020
030
040
050
0
1997 2001 2007 2010 2013
(b) Bonus Depreciation over time
020
4060
8010
0
1997 2001 2003 2008 2011 2013
(c) Section 179 by Investment Level, 2003
010
050
0S
ectio
n 17
9 E
xpen
sing
Ded
uctio
n ($
1,00
0)
0 100 400 500 1000Equipment Investment ($1,000)
(d) Bonus Depreciation by Investment Level, 2003
015
025
050
0B
onus
Dep
reci
atio
n D
educ
tion
($1,
000)
0 100 400 500 1000Equipment Investment ($1,000)
Notes: Panels (a) and (b) show the federal Section 179 allowance and federal bonus depreciation rates during the years1997–2013. Panel (c) plots the maximum federal Section 179 expensing deduction that a firm can elect by varyingamounts of eligible investment as governed by 2003 tax law, when the Section 179 allowance was $100,000 and theSection 179 limit was $400,000. Panel (d) of Figure 1 plots the maximum federal bonus depreciation deduction that afirm can elect at varying amounts of eligible investment assuming they take full advantage of Section 179 expensing.The bonus depreciation rate in 2003 was 50%.
23
Figure 2: State Adoption of Federal Bonus Depreciation
WyomingWisconsin
West VirginiaWashington
VirginiaVermont
UtahTexas
TennesseeSouth Dakota
South CarolinaRhode IslandPennsylvania
OregonOklahoma
OhioNorth Dakota
North CarolinaNew York
New MexicoNew Jersey
New HampshireNevada
NebraskaMontanaMissouri
MississippiMinnesota
MichiganMassachusetts
MarylandMaine
LouisianaKentucky
KansasIowa
IndianaIllinoisIdaho
HawaiiGeorgiaFlordia
DelawareConnecticut
ColoradoCaliforniaArkansas
ArizonaAlaska
Alabama
2001 2002 2003 2004 2008 2009 2010 2011 2012
Notes: Figure 2 describes adoption of bonus depreciation by states during the years 2001–2004 and 2008–2012. Darkblue dots represent full adoption. Light blue dots represent less than full adoption. Florida adopted at 1/7th federalrates after 2008. Firms in Maine had to postpone their 2002 bonus deductions by 2 years. Maine adopted at 5% ofthe federal rate in 2003 and 2004 and 10% of the federal rate in 2011 and 2012. Minnesota adopted at 20% federalrates. Nebraska adopted at 32% federal rates prior to 2008. North Carolina adopted at 30% of federal rates in 2003and 2004 and at 15% of federal rates in 2008–2012. Ohio adopted at 1/6th the federal rate before the corporatefranchise tax was phased out in 2010. Oklahoma adopted at 20% of federal rates in 2001–2002 and in 2008–2009.Pennsylvania adopted at 3/7th of the federal rates prior to 2008. Source: Bloomberg BNA
24
Figure 3: State Conformity to Federal Section 179 Allowances
WyomingWisconsin
West VirginiaWashington
VirginiaVermont
UtahTexas
TennesseeSouth Dakota
South CarolinaRhode IslandPennsylvania
OregonOklahoma
OhioNorth Dakota
North CarolinaNew York
New MexicoNew Jersey
New HampshireNevada
NebraskaMontanaMissouri
MississippiMinnesota
MichiganMassachusetts
MarylandMaine
LouisianaKentucky
KansasIowa
IndianaIllinoisIdaho
HawaiiGeorgiaFlordia
DelawareConnecticut
ColoradoCaliforniaArkansas
ArizonaAlaska
Alabama
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Notes: Figure 3 describes conformity to federal Section 179 rules by states during the years 2001–2012. A dotrepresents conformity to federal rules. States that always conformed to federal rules are depicted in dark blue. Statesthat only conformed during a subsample of the period are depicted in light blue. States that never conformed aregiven no dots.
25
Figure 4: Dynamic Effects of State Bonus Depreciation Adoption on Investment
(a) State Bonus Coefficients
0.5
11.
52
Sta
te B
onus
regr
essi
on c
oeffi
cien
t
1997 2001 2005 2008 2013year
(b) Investment Trends + State Bonus Coefficients
10.52
10.81
11.1
11.39
11.68
1997 2001 2008 2013
30% Bonus 50% Bonus 1 50% Bonus 2100% Bonus Bonus Adopting States No Bonus States
(c) Trends + Coefficients 1997–2004
10.7
10.8
10.9
1111
.1Ln
CA
PX
1997 2001 2004
Bonus Adopting StatesNo Bonus States
(d) Trends + Coefficients 2005–2012
10.6
10.8
1111
.211
.411
.6Ln
CA
PX
2005 2008 2012
Bonus Adopting StatesNo Adoption States
Notes: To produce the coefficients presented in panel (a), coefficients from equation (1) are reestimated after replacing
State Bonus with∑2014
k=1997 βk1
[State Adoptions,t × 1[Yeark]
]. The time-varying coefficients are then plotted by year.
Panel (b) adds these coefficients the average investment trends. Panel (c) and (d) are constructed in the same wayas panel (b) but focus on adopting and non-adopting states in 2001–2004 and 2008–2010 respectively.
26
Figure 5: Dynamic Effects of State Section 179 Allowance on Investment
(a) State 179 Coefficients
-.10
.1.2
Sta
te 1
79 re
gres
sion
coe
ffici
ent
2003 2008 2010 2013year
(b) Investment Trends + State 179 Coefficients
10.8
1111
.211
.4
2003 2008 2010 2013
$100,000 S179 $250,000 S179 $500,000 S179S179 Conforming States Non-Conforning States
Notes: To produce the coefficients presented in panel (a), coefficients from equation (1) are reestimated after replacing
State 179 with∑2014
k=2003 βk2
[State Conformitys,t × 1[Yeark]
]. Years prior to 2003 are eliminated from the analysis.
The time-varying coefficients are then plotted by year. Panel (b) adds these coefficients the average investmenttrends.
27
11 Tables
Table 1: Descriptive Statistics
mean std dev 25th Percentile 75th Percentile obs
Policy V ariables
State Bonus (percent) 6.403 17.045 0.000 0.000 13,780
State 179 (thousands) 125.185 166.799 20.000 125.000 13,780
Outcomes
Investment(millions) 188.765 409.914 17.890 196.998 13,780
Comp per Emp (thousands) 44.663 12.712 36.412 51.278 13,636
Employment (thousands) 16.407 23.956 2.873 20.506 13,780
Output (millions) 2808.085 5017.870 363.884 3046.083 13,618
Control V aribles
Real GSP per capita(billions) 40.828 9.980 33.686 46.400 13,780
Population (millions) 3.846 5.812 0.311 5.374 13,780
Corp Tax Rate 6.523 2.871 5.500 8.500 13,780
Corp Tax % 0.055 0.036 0.038 0.067 13,780
Budget Gap -0.019 0.237 -0.137 0.036 13,780
Dem Legislature % 51.431 15.188 40.404 60.000 13,780
Dem Governor 0.443 0.497 0.000 1.000 13,780
Heterogeneity V aribles
Invest Level (millions) 0.726 1.339 0.143 0.794 12,316
High Tax 0.506 0.500 0.000 1.000 13,780
Notes: Table 1 presents descriptive statistics for each variable used in the analysis. The unit of observation is aNAICS × State manufacturing industry. State Bonus is the state bonus depreciation rate. State 179 is the stateSection 179 allowance. Investment is the total value of capital expenditure. Compensation is total salary dividedby the number of employees. Employment is the total number of employees, Output is the total value of shipments(sales). Real GSP per capital is the real gross state product divided by state population. Population is the state’spopulation. Corp Tax Rate is the state’s to marginal corporate income tax rate. Corp Rev % is the percentage of astate’s revenue derived from state corporate income taxes. Budget Gap is the total state deficit as a fraction of totalstate revenue. Democratic Legislator % is the percentage of democratic state legislators that identify as Democrats.Democratic Governor is an indicator equal to 1 if the governor is a Democrat. Invest Level is total capital expendituredivided by the number of establishments. High Tax is an indicator equal to 1 if the state’s top marginal corporateincome tax is above 7%, the median value for the sample.
28
Tab
le2:
Pre
dic
ting
Sta
teA
dop
tion
ofA
ccel
erat
edD
epre
ciat
ion
Pol
icie
s;L
PM
Model
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
1(B
onu
s)1
(Sec
t.17
9)1
(Both
)1
(Bonu
s)1
(Sec
t.179)
1(B
oth
)1
(Bonu
s)1
(Sec
t.179)
1(B
oth
)
Log
(GS
P/P
op)
-0.1
160.
359
0.2
87
-0.2
88
0.5
97
0.2
13
-0.1
51
0.1
67
0.0
0878
(0.3
16)
(0.2
98)
(0.3
37)
(0.2
09)
(0.4
16)
(0.3
07)
(0.1
93)
(0.3
99)
(0.4
35)
Log
(Pop
)-0
.045
5-0
.145∗∗
-0.1
77∗∗∗
0.6
68
-1.3
70
-0.8
00
1.9
68
-0.3
96
3.5
36
(0.0
564)
(0.0
574)
(0.0
511)
(0.5
81)
(1.0
38)
(0.9
50)
(2.4
51)
(2.2
62)
(4.5
13)
Cor
pT
ax%
-0.6
49-2
.620
-1.4
92
-0.8
85
-0.5
73
-1.3
14
-0.8
84
-0.4
14
-0.8
76
(1.6
15)
(1.7
10)
(1.4
69)
(0.7
33)
(0.5
26)
(0.7
98)
(1.2
27)
(0.7
41)
(1.1
43)
Bu
dge
tG
ap-0
.184
-0.1
82∗∗
-0.1
43
-0.1
08
-0.0
561
-0.0
937
-0.1
09
-0.0
857
-0.1
16
(0.1
11)
(0.0
823)
(0.0
872)
(0.0
923)
(0.0
947)
(0.0
963)
(0.1
000)
(0.1
00)
(0.1
13)
Cor
pR
ate
4.16
4∗∗
-4.3
78-1
.994
-0.3
38
-0.4
50
0.5
59
0.3
67
-2.2
22
0.5
63
(1.5
89)
(2.9
26)
(2.3
97)
(1.0
15)
(1.0
83)
(1.3
07)
(1.2
89)
(2.3
73)
(1.6
35)
Dem
Leg
%-0
.006
26∗∗
-0.0
0722
-0.0
0667∗
-0.0
0428
0.0
0288
-0.0
0448
-0.0
0429
-0.0
0467
-0.0
0647
(0.0
0293
)(0
.004
64)
(0.0
0350)
(0.0
0310)
(0.0
0439)
(0.0
0313)
(0.0
0273)
(0.0
0480)
(0.0
0482)
Dem
Gov
0.10
60.
0918
0.1
86∗∗
0.0
286
-0.0
488
0.0
150
-0.0
361
0.0
312
-0.0
241
(0.0
707)
(0.0
759)
(0.0
814)
(0.0
363)
(0.0
422)
(0.0
349)
(0.0
304)
(0.0
421)
(0.0
608)
Yea
rF
EX
XX
XX
XX
XX
Sta
teF
EX
XX
XX
XS
tate
Tre
nd
sX
XX
Ob
serv
atio
ns
487
348
348
487
348
348
487
348
348
Ad
j.R
20.
0774
0.16
20.1
63
0.8
13
0.8
36
0.7
84
0.8
69
0.9
07
0.8
43
Notes:
Table
2use
slinea
rpro
babilit
ym
odel
sto
pre
dic
tth
eadopti
on
of
acc
eler
ate
ddep
reci
ati
on
polici
esbase
don
state
econom
ic,
politi
cal,
and
dem
ogra
phic
chara
cter
isti
cs.
The
outc
om
eva
riable
insp
ecifi
cati
ons
(1),
(4),
and
(7)
isth
euse
of
bonus
dep
reci
ati
on.
The
outc
om
eva
riable
insp
ecifi
cati
ons
(2),
(5),
and
(8)
isco
nfo
rmit
yto
the
feder
al
Sec
tion
179
allow
ance
.T
he
outc
om
eva
riable
insp
ecifi
cati
ons
(2),
(5),
and
(8)
isb
oth
use
of
bonus
dep
reci
ati
on
and
confo
rmit
yto
the
feder
al
Sec
tion
179
allow
ance
.A
llsp
ecifi
cati
ons
incl
ude
yea
rfixed
effec
ts.
Sp
ecifi
cati
ons
(4)–
(6)
add
state
fixed
effec
ts.
Sp
ecifi
cati
ons
(7)–
(9)
add
state
-sp
ecifi
clinea
rti
me
tren
ds.
Sta
ndard
erro
rsare
pre
sente
din
pare
nth
eses
and
clust
ered
at
the
state
level
.*
p<
0.1
0**p<
0.0
5***p<
0.0
1
29
Tab
le3:
The
Eff
ect
ofSta
teA
ccel
erat
edD
epre
ciat
ion
Pol
icie
son
Inve
stm
ent
(1)
(2)
(3)
(4)
(5)
(6)
Log
(Cap
X)
Log(C
ap
X)
Log(C
ap
X)
Log(C
ap
X)
Log(C
ap
X)
Log(C
ap
X)
Sta
teB
onus
0.180∗∗
0.2
01∗∗
0.1
78∗∗
0.2
70∗∗
0.1
75∗∗
0.2
12∗
(0.0
788)
(0.0
929)
(0.0
756)
(0.1
21)
(0.0
766)
(0.1
06)
Sta
te17
90.
0206∗∗
0.0
270∗∗
0.0
269∗∗∗
0.0
383∗∗
0.0
217∗∗
0.0
233∗∗
(0.0
0782)
(0.0
101)
(0.0
0877)
(0.0
152)
(0.0
0888)
(0.0
104)
Sta
teB
onu
s×
Sta
te17
9-0
.0420∗∗
-0.0
487∗∗
-0.0
429∗∗
-0.0
571∗∗
-0.0
431∗∗
-0.0
457∗∗
(0.0
173)
(0.0
203)
(0.0
172)
(0.0
252)
(0.0
171)
(0.0
220)
Entr
opy
Bal
anci
ng;
Bon
us
XE
ntr
opy
Bal
anci
ng;
S17
9X
Entr
opy
Bal
anci
ng;
Bot
hX
Las
soS
elec
ted
Con
trol
sX
No
Ab
nor
mal
Ad
opte
rsX
Sta
teN
AIC
SG
rou
ps
1022
1022
1022
1022
1022
875
Ob
serv
atio
ns
13047
13047
13047
13047
13108
10849
Ad
j.R
20.
0407
0.0
338
0.0
403
0.0
385
0.0
430
Notes:
Table
3pre
sents
coeffi
cien
tses
tim
ate
sfr
om
spec
ifica
tions
inth
efo
rmof
(1)
des
crib
ing
the
effec
tof
Sta
teB
onus,
Sta
te179,
and
thei
rin
tera
ctio
non
inves
tmen
t.A
llsp
ecifi
cati
ons
incl
ude
yea
r,N
AIC
S×
Sta
te,
and
NA
ICS×
yea
rfixed
effec
tsas
wel
las
state
-sp
ecifi
clinea
rti
me
tren
ds.
Sp
ecifi
cati
on
(1)
incl
ude
tim
e-va
ryin
gst
ate
contr
ols
vari
able
s.Sp
ecifi
cati
ons
(2)
–(4
)use
the
Entr
opy
Bala
nci
ng
tech
niq
ue
intr
oduce
din
Hain
muel
ler
(2012)
tom
atc
hadopti
ng
and
non-a
dopti
ng
state
sbase
don
econom
ic,
politi
cal,
and
dem
ogra
phic
chara
cter
isti
cs.
Insp
ecifi
cati
on
(2),
state
that
adopt
feder
al
bonus
dep
reci
ati
on
are
matc
hed
tost
ate
sth
at
did
not.
Insp
ecifi
cati
on
(3),
state
sth
at
adopt
feder
al
Sec
tion
179
rule
sare
matc
hed
tost
ate
sth
at
did
not.
Insp
ecifi
cati
on
(4),
state
sth
at
adopt
both
feder
al
bonus
dep
reci
ati
on
and
feder
al
Sec
tion
179
rule
sare
matc
hed
tost
ate
sth
at
did
not.
Insp
ecifi
cati
on
(5),
the
lass
om
ethod
isuse
dto
sele
ctco
ntr
ol
vari
able
sfr
om
the
state
econom
ic,
politi
cal,
and
dem
ogra
phic
contr
ols
and
inte
ract
ions
bet
wee
nea
chof
thes
eco
ntr
ols
.In
spec
ifica
tion
(6),
state
sth
at
adopt
bonus
dep
reci
ati
on
inan
abnorm
al
yea
rare
elim
inate
dfr
om
the
sam
ple
.Sta
ndard
erro
rsare
clust
ered
at
the
state
level
.*p<
0.1
0,
**p<
0.0
5,
***p<
0.0
1
30
Table 4: Heterogeneity and Effects on Other Outcomes
(1) (2) (3) (4) (5)
Log(CapX) Log(CapX) Log(Emp) Log(Comp) Log(Output)
State Bonus 0.0779 0.210∗∗ 0.00252 0.0186∗ 0.0111
(0.0896) (0.103) (0.0279) (0.00945) (0.0248)
State 179 0.0236∗∗ 0.0248∗∗ -0.00586 0.000512 0.00314
(0.00902) (0.00942) (0.00600) (0.00157) (0.00503)
State Bonus × State 179 -0.0249 -0.0520∗∗ 0.00513 -0.00590∗∗ -0.00683
(0.0167) (0.0213) (0.00786) (0.00245) (0.00787)
State Bonus × Invest Level 0.135∗∗∗
(0.0403)
State 179 × Invest Level -0.00542∗
(0.00286)
Interaction × Invest Level -0.0267∗∗
(0.0129)
State Bonus × High Tax -0.0832
(0.113)
State 179 × High Tax -0.00969
(0.0129)
Interaction × High Tax 0.0238
(0.0276)
State NAICS Groups 811 1022 1022 1022 1022
Observations 11979 13047 13020 13000 12793
Adj. R2 0.0427 0.0407 0.0464 0.0562 0.0823
Notes: Table 4 presents coefficients estimates from specifications in the form of equation (1). All specifications includeyear, NAICS × State, and NAICS × year fixed effects as well as state-specific linear time trends and time-varyingstate controls variables. The outcome variable in specifications (1) and (2) is Log(CapX). The outcome variables inspecification (3), (4), and (5) are Log(Emp), Log(Comp), and Log(Output). Specification (1) includes interactionsbetween State Bonus, State 179, and State Bonus × State 179 with Invest Level. Specification (2) includes interactionsbetween State Bonus, State 179, and State Bonus × State 179 with High Tax. Standard errors are clustered at thestate level. * p < 0.10 , ** p < 0.05 , *** p < 0.01
31
References
Auerbach, Alan J, and Kevin Hassett. 1991. “Recent US investment behavior and the tax reform act
of 1986: A disaggregate view.” Vol. 35, 185–215, Elsevier.
Bania, Neil, Jo Anna Gray, and Joe A Stone. 2007. “Growth, taxes, and government expenditures:
growth hills for US states.” National Tax Journal, 193–204.
BEA. 2019. “Table 3.7E. Investment in Private Equipment.” U.S. Bureau of Economic Analysis
https://apps.bea.gov/iTable/iTable.cfm?ReqID=10&step=2, accessed 7.29.2019.
Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2004. “How Much Should We Trust
Differences-In-Differences Estimates?” The Quarterly Journal of Economics, 119(1): 249–275.
Cameron, Colin A, and Douglas L Miller. 2015. “A practitioner’s guide to cluster-robust inference.”
Journal of Human Resources, 50(2): 317–372.
Chetty, Raj, Adam Looney, and Kory Kroft. 2009. “Salience and Taxation: Theory and Evidence.”
The American Economic Review, 99(4): 1145–1177.
Chirinko, Robert S, Steven M Fazzari, and Andrew P Meyer. 1999. “How Responsive is Business
Capital Formation to its User Cost?: An Exploration with Micro Data.” Journal of public economics,
74(1): 53–80.
Cummins, Jason G, Kevin A Hassett, R Glenn Hubbard, et al. 1994. “A Reconsideration of In-
vestment Behavior Using Tax Reforms as Natural Experiments.” Brookings Papers on Economic Activity,
25(2): 1–74.
Desai, Mihir A, and Austan Goolsbee. 2004. “Investment, overhang, and tax policy.” Brookings Papers
on Economic Activity, 2004(2): 285–355.
Edge, Rochelle M, and Jeremy B Rudd. 2011. “General-equilibrium effects of investment tax incen-
tives.” Journal of Monetary Economics, 58(6-8): 564–577.
Edgerton, Jesse. 2010. “Investment incentives and corporate tax asymmetries.” Journal of Public Eco-
nomics, 94(11-12): 936–952.
Garrett, Daniel G, Eric C Ohrn, and Juan Carlos Suarez Serrato. 2019. “Tax Policy and Local
Labor Market Behavior.” American Economic Review: Insights.
Giroud, Xavier, and Joshua Rauh. 2019. “State taxation and the reallocation of business activity:
Evidence from establishment-level data.” Journal of Political Economy, 127(3): 1262–1316.
Gordon, Roger, and John D Wilson. 1986. “An examination of multijurisdictional corporate income
taxation under formula apportionment.” Econometrica: Journal of the Econometric Society, 1357–1373.
Hainmueller, J. 2012. “Entropy Balancing: A Multivariate Reweighting Method to Produce Balanced
Samples in Observational Studies.” Political Analysis, 20(1): 25–46.
32
Hall, Robert E, and Dale W Jorgenson. 1967. “Tax Policy and Investment Behavior.” The American
Economic Review, 57(3): 391–414.
Heider, Florian, and Alexander Ljungqvist. 2015. “As certain as debt and taxes: Estimating the tax
sensitivity of leverage from state tax changes.” Journal of Financial Economics, 118(3): 684–712.
Helms, L Jay. 1985. “The Effect of State and Local Taxes on Economic Growth: A Time Series–Cross
Section Approach.” The Review of Economics and Statistics, 574–582.
House, Christopher L, and Matthew D Shapiro. 2008. “Temporary Investment Tax Incentives: Theory
with Evidence from Bonus Depreciation.” The American Economic Review, 98(3): 737–768.
JCT. 2017. “Estimated Budget Effects of the Conference Agreement for H.R. 1, the ”Tax Cuts and Jobs
Act”.” The Joint Committee on Taxation JCX-67-17.
Lechuga, Jessica. 2014. “State Conformity with Federal Bonus Depreciation Rules.”
Ljungqvist, Alexander, and Michael Smolyansky. 2016. “To Cut or Not to Cut? On the Impact of
Corporate Taxes on Employment and Income.”
Maffini, G, MP Devereux, and J Xing. 2019. “The impact of investment incentives: Evidence from
UK corporation tax returns.” American Economic Journal: Economic Policy.
Mead, Charles Ian. 2001. “State user costs of capital.” Federal Reserve Bank of Boston.
Ohrn, Eric. 2018. “The Effect of Corporate Taxation on Investment and Financial Policy: Evidence from
the DPAD.” American Economic Journal: Economic Policy, 10(2): 272–301.
Papke, Leslie E. 1991. “Interstate Business Tax Differentials and New Firm Location: Evidence from Panel
Data.” Journal of public Economics, 45(1): 47–68.
Patel, Elena, and Nathan Seegert. 2015. “Does Market Power Encourage or Discourage Investment?
Evidence from the Hospital Market.” Evidence from the Hospital Market (October 28, 2015).
Paz, Lourenco S, and James E West. 2019. “Should We Trust Clustered Standard Errors? A Comparison
with Randomization-Based Methods.” National Bureau of Economic Research.
Steinmuller, Elias, Georg U Thunecke, and Georg Wamser. 2019. “Corporate income taxes around
the world: a survey on forward-looking tax measures and two applications.” International Tax and Public
Finance, 1–39.
Suarez Serrato, Juan Carlos, and Owen Zidar. 2016. “Who Benefits from State Corporate Tax Cuts?
A Local Labor Markets Approach with Heterogeneous Firms.” American Economic Review, 106(9): 2582–
2624.
Suarez Serrato, Juan Carlos, and Owen Zidar. 2018. “The structure of state corporate taxation and
its impact on state tax revenues and economic activity.” Journal of Public Economics, 167: 158 – 176.
The Book of States. 2000–2014. The Book of States. The Council of State Governments.
33
Tibshirani, Robert. 1996. “Regression shrinkage and selection via the lasso.” Journal of the Royal Statis-
tical Society: Series B (Methodological), 58(1): 267–288.
Tuzel, Selale, and Miao Ben Zhang. 2018. “Economic stimulus at the expense of routine-task jobs.”
Wasylenko, Michael, and Therese McGuire. 1985. “Jobs and Taxes: The Effect of Business Climate
on States’ Employment Growth Rates.” National Tax Journal, 497–511.
Wilson, Daniel J. 2009. “Beggar thy Neighbor? The In-state, Out-of-state, and Aggregate Effects of R&D
Tax Credits.” The Review of Economics and Statistics, 91(2): 431–436.
Zwick, Eric, and James Mahon. 2017. “Tax Policy and Heterogeneous Investment Behavior.” The
American Economic Review, 107(1): 217–248.
34
Appendix A Bonus depreciation example
Table A1 examines the impact of 50% bonus depreciation on the cost of a $100 investment that has a 7-year
life. MACRS specifies that $25 of the total investment may be deducted in the first year, then $21.43 in the
second, etc. With a federal tax rate of 35%, this leads to tax savings of $8.75 in the first year, then $7.50
in the second. Over the course of the 7 year life, all $100 of the investment cost are deducted from taxable
income, generating $35 in total nominal tax shields. However, because the entire cost is not deducted from
taxable income in the first year, the present value of tax savings associated with the investment are only
worth $28.79.27 The after-tax present value cost of the investment is $71.21 (=100-28.79).
50% percent bonus depreciation allows 50 additional dollars to be deducted in the first year the investment
is made. The remaining $50 of cost is then deducted according to the original 7 year MACRS schedule. With
50% bonus depreciation there are now tax savings associated with the investment of $21.88 in the first year,
$3.75 in the second year, etc. Thus, bonus depreciation accelerates the deduction of the investment and
tax savings. The 50% bonus rate increases the present-value of tax shields and symmetrically decreases the
present-value after-tax cost of the investment by $3.10 or 3.1 percentage points.
Table A1: Effect of Federal 50% Bonus Depreciation on Present Value Deductions
Year 1 2 3 4 5 6 7 8 Total
MACRS Deduction 25 21.43 15.31 10.93 8.75 8.74 8.75 1.09 100
τf x Deduction 8.75 7.50 5.36 3.83 3.06 3.06 3.06 0.38 35
PV(τf x Deduction) 28.79
50% Bonus Ded. 62.5 10.72 7.65 5.47 4.37 4.37 4.37 0.545 100
τf x Deduction 21.88 3.75 2.68 1.91 1.53 1.53 1.53 0.19 35
PV(τf x Deduction) 31.89
Notes: Table A1 calculates the present value of federal tax deductions for a $100 investment under both a traditional7-year accelerated depreciation regime and under a 50% bonus depreciation regime. The federal corporate tax rateis assumed to be 35%. The discount rate is assumed to be 10%.
State bonus depreciation is inherently less valuable to firms than federal bonus depreciation because all
state corporate tax rates are significantly lower than the 35% federal rate. Among all states, the average
state income corporate tax rate during the sample period 7.2% (for states that collected corporate income
taxes). Table A2 shows the impact of state 50% bonus depreciation on the present value of tax depreciation
allowances when the corporate income tax rate is 7.2%. The 50% bonus depreciation decreases the after tax
cost of the $100 by 0.064% or $0.64.
27The $28.79 is a function of the assumed discount rate of 10%. At higher discount rates, the present value of thetax shield will be lower. 10% is used in the example because it is often the rate used in corporate net present valuecalculations.
35
Table A2: Effect of State 50% Bonus Depreciation on Present Value Deductions
Year 1 2 3 4 5 6 7 8 Total
MACRS Deduction 25 21.43 15.31 10.93 8.75 8.74 8.75 1.09 100
τf x Deduction 1.8 1.54 1.10 0.79 0.63 0.63 0.63 0.08 7.2
PV(τf x Deduction) 5.92
50% Bonus Ded. 62.5 10.72 7.65 5.47 4.37 4.37 4.37 0.545 100
τf x Deduction 4.5 0.77 0.55 0.39 0.32 0.32 0.32 0.04 7.2
PV(τf x Deduction) 6.56
Notes: Table A2 calculates the present value of federal and state tax deductions for a $100 investment under both atraditional 7-year accelerated depreciation regime and under a 50% bonus depreciation regime. The federal corporatetax rate is assumed to be 35% and the state corporate tax rate is assumed to be 7.2%. The discount rate is assumedto be 10%.
36
Appendix B Additional description of state policy adoption
Appendix B provides further description of state adoption of federal bonus depreciation and Section 179
expensing. Section B.1 maps adoption of the policies. Section B.2 provides adoption rates, federal bonus
rates, and federal Section 179 allowances during the sample period. Section B.3 describes how the state
policies covary in 2004 and 2010.
B.1 Mapping state policy adoption
Figure A1: Mapping State Bonus Adoption & Section 179 Conformity
(a) Bonus Adoption 2001
Full Adoption
Partial Adoption
No Corp. Tax
No Adoption
(b) Bonus Adoption 2008
Full Adoption
Partial Adoption
No Corp. Tax
No Adoption
(c) Section 179 Conformity 2004
Conformers
Non-Conformers
No Corp. Tax
(d) Section 179 Conformity 2010
Conformers
Non-Conformers
No Corp. Tax
Notes: Panels (a) and (b) of Figure A1 map bonus depreciation adoption in 2001 and 2008. Panels (c) and (d) mapSection 179 conformity in 2004 and 2010.
37
B.2 Accelerated depreciation adoption rates and generosity over time
Table A3: State Bonus Depreciation Adoption and Section 179 Conformity
(A) Bonus Depreciation (B) Section 179
Year Bonus Rate (%) Adopters Percent Allowance ($1,000) Conformers Percent
Pre 2000 0 – – ≤20 45 100
2001 30 20 44.4 24 45 100
2002 30 18 40.0 24 45 100
2003 50 19 42.2 100 34 75.6
2004 50 20 44.4 102 35 77.8
2005 0 – – 105 35 77.8
2006 0 – – 108 35 77.8
2007 0 – – 125 34 75.6
2008 50 16 35.6 250 31 68.8
2009 50 17 37.8 250 30 66.7
2010 50 16 35.6 500 28 62.2
2011 100 19 42.2 500 29 64.4
2012 50 18 40.0 500 29 64.4
2013 50 18 40.0 500 29 64.4
2014 50 17 37.8 500 30 66.6
Notes: Table A3 describes state adoption of federal bonus depreciation and state conformity to federal Section 179allowances during the years 1997–2014. Bonus depreciation rates and Section 179 allowances are taken from IRSForm 4562 1997–2014. Adopters are the number of states that fully or partially adopted federal bonus depreciation.Conformers are states that fully conform to federal Section 179 rules. The bonus adoption rate and Section 179conformity rates are calculated only for states that collect corporate income taxes.
38
B.3 Policy Overlap
Table A4: Overlapping Adoption of Bonus and Conformity to 179
2004
S179 Non-Conformers S179 Conformers Total
Bonus Rejecter 10 21 31
Bonus Adopter 0 14 14
Total 10 35 45
2010
S179 Non-Conformers S179 Conformers Total
Bonus Rejecter 17 16 33
Bonus Adopter 0 12 12
Total 17 28 45
Notes: Table A4 presents cross-tabulations of states that partially or fully adopt bonus depreciation and states thatfully conform to federal Section 179 rules in 2004 and in 2010.
Appendix C State tax-adjusted user cost of capital calculations
Appendix C incorporates bonus depreciation and Section 179 allowances into the state tax-adjusted user
cost framework of Mead (2001). The adjusted user costs are then calibrated and used to calculate how
different levels of federal and state bonus depreciation and federal and state Section 179 allowances affect
investment costs. The appendix also shows how state apportionment can be incorporated into the state
tax-adjusted user cost. The apportionment-adjusted user cost demonstrates the challenges in measuring
how state apportionment amplifies or mitigates the effect of state accelerated depreciation policies.
Several conclusions can be drawn from the user cost calculations. First, both state bonus depreciation
and state Section 179 allowances decrease the user cost of capital. Second, as predicted in Section 3, the
decrease in the average user cost due to state bonus depreciation is smaller in states with high Section 179
allowances because bonus depreciation has no effect on user costs for investments can be expensed under
Section 179 rules. Third, the effect of increasing state Section 179 allowances on the user cost of newly
eligible investments is decreasing in the state bonus depreciation rate. Fourth, adopting 100% state bonus
depreciation for investments that cannot be expensed under Section 179 rules, decreases the user cost of
capital by just under 1%. Increasing state Section 179 allowances when no bonus depreciation is offered also
decreases the user cost of capital on newly eligible investments by just under 1%. These percent changes
in the user cost of capital are very close to the percent change estimates in the after-tax present value cost
investment costs presented in Section 2, reinforcing the use of the after-tax present value cost in calculating
investment elasticities, as in Section 8.
39
C.1 State tax-adjusted user costs
The canonical Hall and Jorgenson (1967) user cost of capital describes the after-tax present value cost of $1
of new investment. In a model that consider only federal taxes, Hall and Jorgenson (1967) derive the federal
user cost of capital, uccf as
uccf = (r + δ)1 − ξf − τfzf
1 − τf
where r is the after-tax return on capital demanded by investors and δ is the rate of economic depreciation.
τf is the federal corporate income tax rate, ξf is the federal investment tax credit, and zf is the present
value of federal depreciation deductions as defined in Section 3.
Mead (2001) extends the classic user cost of capital framework to incorporate state taxes and tax in-
centives and their interactions with the federal statutes. First, consider a firm that operates in only one
state (all property, payroll, and sales occur only in a single state). Further assume that there are no local
(sub-state) taxes. Under these assumptions, the state-specific user cost of capital, uccs can be written as
uccs = (r + δ)1 − ξf − ξs − Tfzf − (1 − Tf )τszs
1 − Tf (1 − τs)
where τs is the state corporate income tax rate, ξs is the state investment tax credit, and zs is the present
value of depreciation deductions allowed by the state. Tf is the effective federal tax rate accounting for any
federal taxes deducted from the state tax bill.
Tf = 1 − 1 − τf1 − τfτsDs
where Ds is the percentage of federal corporate income taxes that may be deducted from the state corporate
income tax bill.28
To incorporate accelerated depreciation incentives, we can rewrite zf and zs as
zf = bf + (1 − bf )z0f and
zs = bs + (1 − bs)z0s
where bf and bs are federal and state bonus depreciation rates, respectively. As in Section 3, assume that
for both z0f and z0s ,
z0 =
1 if I ≤ Section 179 allowance and
zMACRS if I > Section 179 allowance.
Calibration
I begin by assuming r = 0.07 following Zwick and Mahon (2017) and δ = 0.10 following Chirinko, Fazzari
and Meyer (1999). The federal corporate tax rate during the sample period was 35% so τf = 0.35. zMACRS
28While federal deductibility has some effect on the user cost, because it scales both the numerator and denominator,it does little to mitigate the effect of state accelerated depreciation policies on the user cost.
40
for the average investment is equal to 0.88 (Zwick and Mahon, 2017). No federal investment tax credit was
in place during the sample period so ξf is set to 0.
For the sample of state-years in which either state bonus depreciation was available or states conformed
to federal Section 179 allowances in 2003 or later, the average state corporate income tax rate was 7%,
the average state investment tax credit was 1.8%, and average state allowed firms to deduct 13.5% of their
federal income taxes from their state taxes; τs = 0.07, ξs = 0.018, and Ds = 0.135. State investment tax
credit and deductibility data is taken from Suarez Serrato and Zidar (2018).
Calculating the effect of accelerated depreciation polices on ucc
Table A5 calculates the effect of federal and state depreciation policies on the state user cost of capital.
Panel A. focuses on the effect of federal and state bonus depreciation on ucc. Panel B. focuses on the effect
of federal and state Section 179 expensing allowances.
Bonus depreciation
In policy combination (A), no bonus depreciation is offered at either the federal or state level (bf = bs = 0).
In the absence of bonus depreciation, the present value of depreciation allowances at both the federal and
state levels is 0.88 (Zwick and Mahon, 2017). In (B), 50% federal bonus depreciation is implemented. This
decreases the user cost of capital by 3.283%. In (C), state 50% bonus is adopted on top of the federal bonus
rate, decreasing the user cost of capital decreases by an additional 0.445%. Although it is not represented
in Table A5, implementing 100% federal bonus depreciation decreases the user cost by 6.57%. Adopting
100% bonus depreciation at the state level on top of the federal 100% bonus decreases the state user cost by
0.922%.
Policy combination (D) and (E) shows how implementing federal and state bonus depreciation at a
50% rate affects the user cost of capital for investments below the federal and state Section 179 allowance,
those for which z0 = 1. Implementing bonus depreciation for investments that can already be immediately
expenses has no effect on the user cost of capital.
Overall, both federal and state bonus depreciation decrease the user cost of capital and incentivize
investment. The effect of state bonus depreciation is smaller than the effect of federal bonus depreciation.
The effect of state or federal bonus depreciation is zero for investments that qualify for Section 179 expensing.
As a result, the effect of state bonus depreciation implementation on the use cost will be smaller in states
will higher Section 179 allowances.
Section 179 expensing
Policy combination (F) repeats policy combination (A). Policy combination (G) adds federal Section 179
allowances as indicated by z0f = 1 to marginal investments. This decreases the user cost of capital by 7.428%.
Policy combination (H) adds state Section 179 expensing (z0s = 1) on top of federal expensing, decreasing
the user cost of capital by an additional 0.922%.
Policy combinations (I), (J), and (K) repeat the (F), (G), (H) progression of federal and state Section
179 expensing in the presence of 50% federal and state bonus depreciation. Federal Section 179 expensing
41
decreases the user cost of capital by 3.410%. State Section 179 expensing decreases the user cost an additional
0.463%.
Overall, Section 179 Expensing decreases the user cost of capital for investments below the allowance
level. The decrease in user costs decreases in magnitude as bonus depreciation is made more generous.
Therefore, the effect of increases in state Section 179 allowances on the ucc will be smaller in states with
higher bonus depreciation rates.
Table A5: User Cost of Capital and Accelerated Depreciation Incentives
Panel A. Federal and State Bonus Depreciation
Policy Combination z0f z0s bf bs uccs % change
(A) 0.88 0.88 0 0 0.1598 –
(B) 0.88 0.88 0.5 0 0.1545 (A) to (B): -3.283%
(C) 0.88 0.88 0.5 0.5 0.1538 (B) to (C): -0.445%
(D) 1 1 0 0 0.1479 –
(E) 1 1 0.5 0.5 0.1479 (D) to (E): 0%
Panel B. Federal and State Section 179 Expensing
Policy Combination z0f z0s bf bs uccs % change
(F) 0.88 0.88 0 0 0.1598 –
(G) 1 0.88 0 0 0.1493 (F) to (G): -7.428
(H) 1 1 0 0 0.1479 (G) to (H): -0.922
(I) 0.88 0.88 0.5 0.5 0.1538 –
(J) 1 0.88 0.5 0.5 0.1486 (I) to (J): -3.410%
(K) 1 1 0.5 0.5 0.1479 (J) to (K): -0.463%
Notes: Table A5 calculates the state user cost of capital at different levels of federal and state Section 179 expensingand bonus depreciation policies.
C.2 Apportionment
Many firms are located in or make sales in more than one state. In these cases, states rely on apportionment
formulas to determine how much of a firm’s total business income is taxable in a given state. Gordon and
Wilson (1986) show the percentage of total profits that are apportioned to state s can be written as
αs = aKsKs
K+ aWs
Ws
W+ aSs
Ss
S
where Ks/K is fraction of total property located in state s, Ws/W is fraction of total payroll located in
state s, and Ss/S is fraction of total sales that take place in state s. aKs , aWs , and aSs are the apportionment
weights given to property, payroll, and sales.
Mead (2001) shows that incorporating state apportionment leads to the following user cost of capital
formulation:
42
uccs = (r + δ)
1 −∑s∈S
αsξs − Tf [bf + (1 − bf )zf ] − (1 − Tf )∑s∈S
αsτs[bs + (1 − bs)zs]
1 − Tf (1 −∑s∈S
αsτs)
where the average effective federal corporate income tax rate taking into account state deductions is
Tf = 1 − 1 − τf1 − τf
∑s∈S
αsτsDs.
The key insight in terms of state accelerated depreciation policy is that the effect of bs or zs will have the
most effect when αs is large. While time-varying state-level data on weights aKs , aWs , and aSs are available,
data on the portion of capital, wages, and sales made in each state by the ASM sample are not. Therefore, it
is hard to empirically explore how state apportionment interacts with state accelerated depreciation policies.
Appendix D Adoption and rejection event study graphs
Appendix D presents a series of event study graphs that show how trends in gross state product (GSP) per
capita growth rates and state budget gaps differ between adoption and non-adopting groups of states before
and after adoption or rejection decisions.
The first series of graphs, presented in Figure A2, shows how trends in state GSP and state budget
gap outcomes differ between states that adopted bonus depreciation during the years 2001–2004 and those
that did not (in panels (a) and (b)) and between states that adopted bonus depreciation during the years
2008–2010 and those that did not (panels (c) and (d)). The coefficients in all panels are produced by
regressing the outcome on year indicators interacted with a bonus depreciation adoption indicator and year
fixed effects. State fixed effects are included when estimating coefficients in panels (c) and (d) because the
outcome variable is measured in levels, as opposed to changes as in panels (a) and (b).
The second series of graphs, presented in Figure A3, shows how trends in GSP and state budget gaps
differ between states that stopped conforming to federal Section 179 expensing allowances and those that
did not around the time of the non-conformity decision. To produce the coefficients in both panels, the
outcome is regressed on year relative to non-conformity indicators and year fixed effects. State fixed effects
are included when estimating the coefficients in panel (b) because the outcome variable is measured in levels.
Coefficients on the relative year indicators are plotted.
The third series of graphs, presented in Figure A4, shows how trends in GSP and state budget gaps
differ between states that adopted or rejected bonus depreciation in years other than 2001 or 2008. The
states that adopted bonus depreciation in years other than 2001 or 2008 are Florida in 2004, Illinois in 2011,
Iowa in 2003, Maine in 2011, North Carolina in 2003, Oregon in 2010, Pennsylvania in 2011. The states that
rejected bonus depreciation in years other than 2001 or 2008 are Florida in 2003, Idaho in 2010, Illinois in
2012, New Jersey in 2002, New York in 2003, Oklahoma in 2003, Pennsylvania in 2012, and Tennessee in
2003. To produce the coefficients in all four panels, the outcome is regressed on year relative to adoption
or rejection decision indicators and year fixed effects. State fixed effects are included when estimating the
coefficients in panels (c) and (d) because the outcome variable is measured in levels. Coefficients on the
43
relative year indicators are plotted.
Coefficients in 9 of the 10 panels show no divergence in state GSP per capita or state budget gap prior
to adoption or rejection decisions indicating changes in state economic health are unlikely to cause states
to adopt or reject the federal policies. The one exception is panel (b) of Figure A4 which shows that state
budget gaps increase just prior to adoption of bonus depreciation in years other than 2001 or 2008. Analysis
presented in column (6) of Table 3 suggests these pretrends downward bias the baseline policy estimates.
Figure A2: GSP Growth Rates and Budget Gaps, Bonus Adopters vs. Non-Adopters
(a) GSP Growth Rates, 2001–2004 Adoption
-.03
-.02
-.01
0.0
1.0
2D
iffer
ence
in G
SP
Grw
oth
Rat
es
1997 1998 1999 2000 2001 2002 2003 2004
(b) GSP Growth Rates, 2008–2010 Adoption
-.04
-.02
0.0
2.0
4D
iffer
ence
in G
SP
Gro
wth
Rat
es
2005 2006 2007 2008 2009 2010 2011 2012
(c) Budget Gap, 2001–2004 Adoption
-.2-.1
0.1
Diff
eren
ce in
Bud
get G
aps
1997 1998 1999 2000 2001 2002 2003 2004
(d) Budget Gap, 2008–2010 Adoption
-.2-.1
0.1
Diff
eren
ce in
Bud
get G
aps
2005 2006 2007 2008 2009 2010 2011 2012
Notes: Panel (a) depicts how GSP growth rates differ between states that adopted bonus depreciation during theyears 2001–2004 to those that did not. Panel (b) depicts how GSP growth rates differ between states that adoptedbonus depreciation during the years 2008–2010 to those that did not. Panel (c) depicts how state budget gaps differbetween states that adopted bonus depreciation during the years 2001–2004 to those that did not. Panel (d) depictshow state budget gaps differ between states that adopted bonus depreciation during the years 2008–2010 to those thatdid not. To estimate the coefficients in all panels, the outcome is regressed on year indicators interacted with a bonusdepreciation adoption indicator and year fixed effects. State fixed effects are included when estimating coefficient inpanels (c) and (d).
44
Figure A3: GSP Growth Rates and Budget Gaps: Section 179 Conformers vs. Non-Conformers
(a) State GSP ∆%, 2001–2004 Adopters
-.03
-.02
-.01
0.0
1.0
2D
iffer
ence
in G
SP
Gro
wth
Rat
es
-3 -2 -1 0 1 2 3Years Relative to Section 179 Rejection
(b) State GSP ∆%, 2008–2012 Adopters
-.2-.1
0.1
.2D
iffer
ence
Bud
get G
aps
-3 -2 -1 0 1 2 3Years Relative to Section 179 Rejection
Notes: Panel (a) depicts how GSP per capita growth rates differ between states the continue to conform to the federalSection 179 allowance against states that choose to stop conforming in year 0. Panel (b) depicts how state budgetgaps differ between states that continue to conform to the federal Section 179 allowance against states that chooseto stop conforming in year 0. To produce the coefficients in both panels, the outcome is regressed on year relative toyear 0 indicators for non-conforming states and year fixed effects. State fixed effects are included when estimatingcoefficients in panels (b)because the outcome variable is measured in levels, as opposed to changes (as in panel (a)).
45
Figure A4: GSP Growth Rates and Budget Gaps: Abnormal Adopters and Rejecters
(a) GSP Growth Rate; Abnormal Adopters
-.04
-.02
0.0
2.0
4D
iffer
ence
in G
SP
Gro
wth
Rat
es
-3 -2 -1 0 1 2 3Years Relative to Bonus Depreciation Adoption
(b) Budget Gap; Abnormal Adopters
-.20
.2.4
Diff
eren
ce B
udge
t Gap
s-3 -2 -1 0 1 2 3
Years Relative to Bonus Depreciation Adoption
(c) GSP Growth Rate; Abnormal Rejecters
-.04
-.02
0.0
2.0
4D
iffer
ence
in G
SP
Gro
wth
Rat
es
-3 -2 -1 0 1 2 3Years Relative to Bonus Depreciation Discontinuation
(d) Budget Gap; Abnormal Rejecters
-.2-.1
0.1
.2D
iffer
ence
Bud
get G
aps
-3 -2 -1 0 1 2 3Years Relative to Bonus Depreciation Discontinuation
Notes: Panel (a) plots the difference in GSP growth rates between states that adopted bonus depreciation in yearsother than 2001 or 2008 relative to other states around the time of adoption. Panel (b) plots difference in statebudget gaps between states that adopted bonus depreciation in an abnormal year relative to other states around thetime of adoption. The states that adopted bonus depreciation in years other than 2001 or 2008 are Florida in 2004,Illinois in 2011, Iowa in 2003, Maine in 2011, North Carolina in 2003, Oregon in 2010, Pennsylvania in 2011. Panel (c)plots the difference in GSP growth rates between states that rejected bonus depreciation in years other than 2001 or2008 relative to other states around the time of rejection. Panel (d) plots the difference in state budget gaps betweenstates that rejected bonus depreciation in years other than 2001 or 2008 relative to other states around the time ofrejection. The states that rejected bonus depreciation in years other than 2001 or 2008 are Florida in 2003, Idaho in2010, Illinois in 2012, New Jersey in 2002, New York in 2003, Oklahoma in 2003, Pennsylvania in 2012, and Tennesseein 2003. To estimate the coefficients in all panels, the outcome is regressed on relative year indicators interacted withan adopter or rejecter indicator and year fixed effects. State fixed effects are included when estimating coefficients inpanels (c) and (d) because the outcome variable is in levels, as opposed to changes (as in panels (a) and (b)).
46
Appendix E Marginal Effects
To further understand the interaction between the two policies and their effect on investment, Figure A5
presents estimates of the effect of both policies during each of the years 2000–2013 while controlling for the
effect of the other policy. Panel (a) plots the predicted impact of state adoption of bonus depreciation at the
federal level assuming that the state has the average observed state Section 179 allowances during that year
(see Table A3 for average levels in a given year). Therefore, the estimated effect of bonus depreciation is
large when bonus depreciation is high but is tempered as state Section 179 allowances become more generous
over time. As one might expect, for the average state, the impact of bonus depreciation was the largest in
2003, when the federal bonus level was high – 50% – but federal Section 179 allowances were still small –
only $100,000. According to the estimates, bonus depreciation had a statistically significant impact on state
investment in years 2001–2004 and a nearly significant impact in years 2008–2009. After federal Section 179
allowances increased to $500,000 in 2010, state adoption of bonus depreciation had no marginal effect on
investment.
Panel (b) presents estimates of the impact of state Section 179 allowances. Here, the estimates are
interpreted as the impact of conforming to federal Section 179 allowances assuming that the state has
adopted federal bonus depreciation at the average observed level in each year. These estimates are much less
affected by bonus depreciation than the bonus depreciation estimates are by Section 179 conformity because
fewer states adopt bonus depreciation than conform to section 179 allowances. Therefore, these estimates
closely mirror the rise in federal Section 179 allowances. However, bonus depreciation significantly reduces
the State 179 effect in 2011 when bonus depreciation was set to 100% and a larger proportion of states than
usual adopted federal bonus depreciation.
Figure A5: Estimated Effect of State Adoption of Accelerated Depreciation Policies
(a) Effect of Bonus Depreciation Adoption (b) Effect of Section 179 Adoption
Notes: Figure A5 uses the baseline estimates presented in Table ?? specification (1) to predict the investment impactof adopting each accelerated depreciation policy at the federal rate in a given year while adopting the other at theaverage rate observed in the data. Panel (a) depicts the effect of adoption of federal bonus depreciation on investmentwhile granting the average rate of Section 179 allowances in each year 2000–2013. Panel (b) depicts effect of adoptionof federal Section 179 allowances on investment while granting bonus depreciation at the average rate in each year2000–2013. Vertical bars represent 95% confidence intervals. Standard errors are computed using the delta method.
47
Ap
pen
dix
FA
ltern
ati
ve
Sp
eci
fica
tion
s
Tab
leA
6:E
ffec
tof
Sta
teA
ccel
erat
edD
epre
ciat
ion
Pol
icie
s;A
lter
nat
eSp
ecifi
cati
ons
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Log
(Cap
X)
Log(C
ap
X)
Log(C
ap
X)
Log(C
ap
X)
Log(C
ap
X)
Log(C
ap
X)
Log(C
ap
X)
Sta
teB
onus
0.18
4∗∗
0.2
16∗∗∗
0.2
01∗∗∗
0.1
80∗∗
0.0
711∗
0.0
588
(0.0
785)
(0.0
744)
(0.0
732)
(0.0
788)
(0.0
403)
(0.0
424)
Sta
te17
9-0
.002
600.0
0487
0.0
163∗
0.0
206∗∗
0.0
165∗∗
0.0
141∗
(0.0
140)
(0.0
103)
(0.0
0821)
(0.0
0782)
(0.0
0770)
(0.0
0810)
Sta
teB
onus×
Sta
te17
9-0
.023
4-0
.0437∗∗
-0.0
474∗∗∗
-0.0
420∗∗
(0.0
238)
(0.0
197)
(0.0
160)
(0.0
173)
Yea
rF
EX
XX
XX
XX
Sta
teN
AIC
SF
EX
XX
XX
XX
Con
trol
sX
XX
XX
XN
AIC
SY
ear
FE
XX
XX
XS
tate
Tre
nd
sX
XX
XS
tate
NA
ICS
Gro
up
s10
221022
1022
1022
1022
1022
1022
Ob
serv
atio
ns
1304
713047
13047
13047
13047
1304
713047
Ad
j.R
20.
0005
840.0
221
0.0
453
0.0
407
0.0
398
0.0
399
0.0
401
Notes:
Table
A6
pre
sents
coeffi
cien
tes
tim
ate
sfr
om
spec
ifica
tions
inth
efo
rmof
equati
on
(1).
Sp
ecifi
cati
on
(1)
incl
udes
yea
r(t
ime)
and
NA
ICS×
Sta
te(o
bse
rvati
on-l
evel
)fixed
effec
ts.
Sp
ecifi
cati
on
(2)
adds
tim
e-va
ryin
gst
ate
contr
ols
.Sp
ecifi
cati
on
(3)
adds
state
-sp
ecifi
clinea
rti
me
tren
ds.
Sp
ecifi
cati
on
(4)
adds
NA
ICS×
yea
rfixed
effec
tsand
iseq
uiv
ale
nt
toth
ebase
line
spec
ifica
tion,
(colu
mn
1,
Table
3).
Sp
ecifi
cati
ons
(5),
(6),
and
(7)
esti
mate
only
the
effec
tof
Sta
teB
onus,
only
the
effec
tof
Sta
te179,
and
Sta
teB
onus
and
Sta
te179
wit
hout
thei
rin
tera
ctio
nin
the
pre
sence
of
the
full
suit
eof
contr
ol
vari
able
s,fixed
effec
ts,
and
state
-sp
ecifi
clinea
rti
me
tren
ds.
Sta
ndard
erro
rsare
clust
ered
at
the
state
level
.*p<
0.1
0,
**p<
0.0
5,
***p<
0.0
1
48
Appendix G Randomization Inference
A concern difference-in-differences studies is that they may over-reject the null hypothesis when error terms
are serially correlated (Bertrand, Duflo and Mullainathan, 2004). To combat these concerns I perform a
randomization inference test similar to the one used in Chetty, Looney and Kroft (2009).29
I begin by randomly assignment units from one state another state’s annual series of State Bonus, State
179, and State Bonus × State 179 during from 1997–2013. I assign each state’s policy adoption series without
replacement. This effectively randomizes the state tax policies while maintaining the underlying structure
of the data. I then replicate the baseline analysis from column (1) of Table 3. I store the State Bonus, State
179, State Bonus × State 179 coefficients from the regression then repeat the process another 1,999 times.
Panel (a) of Figure A6 displays an empirical CDF of the 2,000 placebo State Bonus coefficients. No
parametric smoothing is applied; the CDF looks smooth because of the large number of points used to
construct it. The vertical black line represents the real State Bonus effect size from the baseline specification.
Only 1 out of the 2,000 placebo coefficients is larger than the original difference-in-differences estimate of
0.180, implying a randomization-based p-value of 0.0005. As this estimate suggests statistical significance
at the 1% level, this ranomization-inference test suggests the state-level clustering generates conservative
standard error estimates.
Panel (b) of Figure A6 presents the empirical CDF of the 2,000 placebo State 179 coefficients. Here,
76 of the 2,000 placebo coefficients are larger than the baseline State 179 estimate of 0.0206, implying a
randomization-based p-value of 0.036. This randomization based p-value is smaller than the baseline State
179 p-value.
Panel (c) of Figure A6 presents the empirical CDF of the 2,000 placebo State Bonus × State 179
coefficients. 6 of the 2,000 placebo coefficients are more negative than the baseline State Bonus × State
179 estimate, implying a randomization-based p-value of 0.003. Again, this randomization based p-value is
smaller than the baseline State Bonus × State 179 p-value.
29Paz and West (2019) show randomization-based inference tests dramatically outperform conventional clusteringmethods difference-in-differences specifications with unbalanced panels and a small number of treated groups.
49
Figure A6: Randomization Inference CDFs
(a) State Bonus
0.2
.4.6
.81
Cum
ulat
ive
Dis
tribu
tion
Func
tion
-.2 -.1 0 .1 .2Placebo State Bonus Coefficient
p-value = 0.0005
(b) State 179
0.2
.4.6
.81
Cum
ulat
ive
Dis
tribu
tion
Func
tion
-.04 -.02 0 .02 .04Placebo State 179 Coefficient
p-value = 0.038
(c) State Bonus × State 179
0.2
.4.6
.81
Cum
ulat
ive
Dis
tribu
tion
Func
tion
-.05 0 .05Placebo State Bonus x State 179 Coefficient
p-value = 0.003
Notes: Each panel of Figure A6 plots an empirical distributions of placebo coefficients from regressions in whichthe outcome is regressed on randomly assigned State Bonus, State 179, and State Bonus × State 179. The verticallines show the treatment effect estimate reported in column (1) of Table 3. Non-parametric p-values represent theprobability an observation is greater in magnitude than the baseline effect estimates.
50
Recommended