political activism and firm innovation · political activism and firm innovation 1. introduction a...
Post on 07-Feb-2020
4 Views
Preview:
TRANSCRIPT
1
Political activism and firm innovation
Alexei V. Ovtchinnikov
HEC Paris
Syed Walid Reza
Queensland University of Technology
and
Yanhui Wu
Queensland University of Technology
Abstract
Political activism positively affects firm innovation. Firms that support more politicians, politicians on
Congressional committees with jurisdictional authority over the firms’ industries and politicians who join
those committees innovate more. We employ instrumental variables estimation and a natural experiment
to show a causal effect of political activism on innovation. The results are consistent with the hypothesis
that political activism is valuable because it helps reduce policy uncertainty, which, in turn, fosters firm
innovation. Also consistent with this hypothesis, we show that politically active firms successfully time
future legislation and set their innovation strategies in expectation of future legislative changes.
Keywords: political contributions, innovation, investment policy, policy uncertainty
JEL Classification: D72, D80, G31, G38, O31, O38
August 11, 2014
Ovtchinnikov is the corresponding author. Please send correspondence to: ovtchinnikov@hec.fr or HEC Paris, 1
rue de la Libération, 78351 Jouy-en-Josas cedex, France. We thank Art Durnev, Bill Christie, Mara Faccio, Johan
Hombert, Elena Loutskina, David Parsley, and Philip Valta for helpful comments. The usual disclaimer applies.
2
Political activism and firm innovation
Abstract
Political activism positively affects firm innovation. Firms that support more politicians, politicians on
Congressional committees with jurisdictional authority over the firms’ industries and politicians who join
those committees innovate more. We employ instrumental variables estimation and a natural experiment
to show a causal effect of political activism on innovation. The results are consistent with the hypothesis
that political activism is valuable because it helps reduce policy uncertainty, which, in turn, fosters firm
innovation. Also consistent with this hypothesis, we show that politically active firms successfully time
future legislation and set their innovation strategies in expectation of future legislative changes.
1
Political activism and firm innovation
1. Introduction
A considerable body of research finds that firms are politically active and engage in building
long-term connections with politicians. These connections are typically broken into explicit connections
that arise when a politician joins the firm or its board of directors (or vice versa) and implicit connections
that arise when a firm makes political contributions to the politician’s (re)election campaign.1 There is
also growing consensus in the literature that political activism is valuable to shareholders of the
contributing firms.2 Although the relation between political activism and firm value appears well
established, our understanding of the exact mechanisms through which political activism creates value
and affects real economic outcomes is far from complete.3 In this paper, we contribute to this literature
by analyzing the effect of political activism on firm investment decisions, specifically investment in
innovation.
Our analysis is motivated by a long string of theoretical work on irreversible investment under
uncertainty (see, e.g., Bernanke (1983), Bertola and Caballero (1994), Abel and Eberly (1996), Hassler
(1996), Caballero and Engel (1999), Bloom, Bond and Van Reenen (2007)). One of the central tenets in
these models is that when investment is at least partially irreversible, uncertainty depresses investment
because uncertainty increases the value of the firms’ option to wait. Bhattacharya, et al. (2014) note that
the option to wait is especially important for investment in innovation, considering that innovation
represents the exploration of unknown approaches and untested methods (Holmstrom (1989), Aghion and
Tirole (1994)). Thus, firm decisions that reduce future uncertainty should lower the value of the option to
wait and stimulate current investment, especially investment in innovation.
Bhattacharya, et al. (2014) show that policy uncertainty, defined as uncertainty about future
government economic policy, significantly reduces firm innovation. This is consistent with the intuition
that politics is important for firm innovation because politicians’ decision making significantly affects the
operating environment of innovating firms. Furthermore, the negative correlation between policy
uncertainty and innovation implies that firms’ actions that reduce policy uncertainty should stimulate
investment in innovation.
1 See, e.g., Masters and Keim (1985), Zardkoohi (1985), Grier, Munger, and Roberts (1994), Kroszner and
Stratmann (1998, 2005), Faccio (2006), Goldman, Rocholl and So (2009), Cooper, Gulen and Ovtchinnikov (2010),
Akey (2014). 2 See, e.g., Fisman (2001), Faccio (2006), Ferguson and Voth (2008), Faccio and Parsley (2009), Goldman, Rocholl
and So (2009), Cooper, Gulen and Ovtchinnikov (2010), Ovtchinnikov and Pantaleoni (2012), Amore and
Bennedsen (2013), Akey (2014). Aggarwal, Meschke and Wang (2012) and Coates (2012) present evidence that
political connections may be value destroying. 3 Previous research finds that politically connected firms enjoy preferential access to external financing (Claessens,
Feijen and Laeven (2008)) and are more likely to receive government bailouts in financial distress (Faccio, Masulis
and McConnell (2006), Duchin and Sosyura (2012)). Politically connected firms are also more likely to receive
government procurement contracts (Tahoun (2012), Goldman, Rocholl and So (2013)) as well as government
subsidies and other support (Johnson and Mitton (2003)).
2
In this paper, we propose that firm political activism, defined as a strategy of making recurrent
political contributions to build long-term connections with politicians, facilitates investment in innovation
because it helps reduce policy uncertainty. First, political contributions purchase access to politicians
who may be willing to trade policy information for contributions (Stratmann (1998, 2002), Coate (2004)).
Second, political contributions allow firms to actively lobby for government policy (Stratmann (2005)),
thereby reducing future policy uncertainty. Restated in the framework of Pastor and Veronesi (2012),
political activism reduces political uncertainty, i.e. the uncertainty about whether government policy will
change in the future. In Pastor and Veronesi (2012), political uncertainty arises because the political cost
of changing government policy is unobservable to outsiders. We argue that politically active firms
engaged in repeat interactions with policymakers are better informed about the policymakers’ political
cost and, therefore, face lower political uncertainty. Lower political uncertainty decreases the volatility of
future cash flows from innovation, which, in turn, lowers the value of the option to wait and stimulates
investment in innovation. The main testable implication of this argument is that political activism
positively affects firm innovation.
In our empirical tests, we use firm patent activity to measure innovation and find strong support
for our hypothesis. Our sample comprises the intersection of the NBER patent citations data file and the
FEC detailed political contributions file for the period 1979 – 2004. In all, our study covers 813,692 hard
money political contributions made by 1,805 U.S. firms and 1,158,693 patents granted to 6,528 unique
U.S. firms as well as 9,673,067 patent citations over our sample period. Using hard money political
contributions data, we build three separate sets of measures of political activism designed to capture
different aspects of firm political activity.
Our empirical strategy follows a four-tiered approach to isolate a causal effect of political
activism on innovation. In each step, we set the identification bar progressively higher in order to rule out
alternative interpretations of the baseline results. In the first step, we employ panel OLS regressions to
establish a baseline correlation between our measures of firm political activism and future innovation.
The results show that political activism is strongly positively correlated with firm innovation one and
three years ahead. We measure firm innovation with patent applications and patent citations and find that
our political activism measures have a large economic and statistical impact on firm innovation. The
results are robust to alternative sample construction methodologies. We first estimate regressions on the
entire cross-section of CRSP/Compustat firms treating firms that are not politically active as firms with
zero political contributions. However, also recognizing that zero contributions for politically inactive
firms may not accurately capture the choice not to contribute, we re-estimate our baseline regressions on
the subsample of firms with observed political contributions. In the second set of regressions, we control
for the resulting self-selection bias with a two-stage Heckman (1979) approach. Regardless of which
3
sample we use, we find a strong positive correlation between firm political activism and future
innovation.
In the second tier of our tests, we conduct a wide range of subsample analyses that, in their
entirety, help address the omitted variable bias and thus move us closer to identifying a causal effect of
political activism on firm innovation. Our tactic is to decompose the sample of political contributions
into those that are predicted by our hypothesis to have a stronger effect on innovation and those
contributions predicted to have a weaker or no effect. If there is indeed a differential impact in the data,
this evidence would be consistent with our hypothesis but harder to explain by an omitted variable bias.
We find evidence consistent with our predictions. The effect of political contributions on innovation is
stronger for contributions to politicians who sit on Congressional committees with jurisdictional authority
over a firm’s industry and politicians who join such committees. Moreover, the effect is completely
concentrated in non-election years when Congressional committee appointments are unexpected.
However, we find no effect of political contributions on innovation for contributions to politicians outside
of Congressional committees with jurisdictional authority over a firm’s industry or for contributions to
politicians who leave such committees.
Next, we set the identification hurdle higher and employ instrumental variables estimation.
Reverse causality is a concern in our analysis, since it may well be the case that firm innovation actually
drives the demand for political activism. To rule out this explanation, we instrument for political
contributions with a politician’s margin of victory in the prior election. Intuitively (and we confirm this
in the data), a politician’s margin of victory in the prior election should not affect current firm innovation.
However, it should significantly affect the firm’s propensity to make political contributions to that
politician (we also confirm this in the data) because the marginal dollar of contributions matters more to a
vulnerable politician in a close election. In return, a vulnerable politician should promise more favors in
exchange for contributions to attract more campaign financing (Stratmann (1998, 2002), Coate (2004)).
Thus, we hypothesize that firms target politicians in close elections and expect more benefits, including
access to valuable information, in return for their political contributions. We isolate the exogenous
variation in firm political contributions with a politician’s margin of victory in the prior election and show
in the second stage that instrumented political contributions strongly predict future firm innovation.
In our final test, we move the identification bar still higher and perform a natural experiment that
exploits an exogenous shock to the value of political contributions that occurred during our sample
period. We use the surprise Republican win in the 1994 mid-term election and the consequent surprise
Newt Gingrich’s decision to appoint four junior Congressmen to committee chairman positions as an
exogenous shock to the value of political contributions for those firms that supported the four new
chairmen prior to the 1994 election. Because the chairman appointments were unexpected, firms could
not have adjusted their contribution strategies prior to the election. Thus, if political contributions to
4
committee chairmen are valuable (Cooper, Gulen and Ovtchinnikov (2010)) and affect firm innovation,
firm patent activity should increase following the election. Consistent with this, we show that patent
activity of the treatment group of firms increased significantly following the election compared to that of
the control group. Specifically, the number of patent applications for the treatment firms grew steadily
from an average of 4.8 applications in 1994 to 8.4 applications in 1999, a 75% increase, while the patent
activity of the control group stayed essentially flat over the same period. We replicate these results in
univariate and multivariate tests and show that the results are statistically and economically significant.
Taken together, the preponderance of evidence shows that political activism positively affects
firm innovation. A caveat is immediately in order. Even though each individual step of our identification
strategy is subject to understandable empirical limitations (for example, our use of a relatively small
sample in the natural experiment and the resulting difficulty of generalizing the results, or the challenge
of addressing the exclusion restriction in instrumental variables estimation), we draw our conclusions
based on the overall weight of combined evidence from our analysis.
Collectively, the results are consistent with our hypothesis that political activism reduces policy
uncertainty, which, in turn, fosters firm innovation. In the last section, we go further and provide
evidence consistent with the negative relation between political activism and policy uncertainty. Whether
firms use political contributions to purchase valuable policy information or actively lobby for government
policy, both arguments imply that politically active firms are able to successfully time future legislation
and set their innovation strategies in expectation of upcoming legislative changes. Empirically, this
means that changes in innovation by politically active firms predict future legislative changes. Consistent
with this hypothesis, we find that firm innovation, especially by politically active firms, strongly predicts
future legislative changes. Specifically, we show that politically active firms are able to successfully time
industry deregulation and start innovating in soon-to-be-deregulated industries prior to deregulation
legislation. In contrast, politically inactive firms do not exhibit this timing ability.
The analysis in this paper is important for three reasons. First, the results in this paper contribute
to our understanding of the sources of value from firm political activism. We show that political
activism, in addition to its impact on firm financing decisions (Claessens, Feijen and Laeven (2008)) and
operating performance (Cooper, Gulen and Ovtchinnikov (2010), Ovtchinnikov and Pantaleoni (2012),
Goldman, Rocholl and So (2013)), also significantly affects firm real investment decisions, such as
investment in innovation. These results are important for building a complete picture of firm political
activism and its impact on firm decisions and value.
Second, our results add to the growing literature exploring the drivers of innovation across firms.
Previous research finds that innovation is related to the firm’s organizational form (see, e.g., Spiegel and
Tookes (2008), Ferreira, Manso, and Silva (2014), Seru (2014)), corporate governance characteristics
(see, e.g., Atanassov (2013)), CEO characteristics (see, e.g., Hirshleifer, Low, and Teoh (2012)), VC
5
characteristics (Kortum and Lerner (2000), Tian and Wang (2013)), legal systems (see, e.g., Acharya and
Subramanian (2009), Fan and White (2003), Armour and Cumming (2008), Acharya, Baghai, and
Subramanian (2009)) and institutional and market settings (see, e.g., Aghion, Van Reenen and Zingales
(2013), Acharya and Subramanian (2009), He and Tian (2013), Fang, Tian, and Tice (2014)). We show
that firm political activism is an important determinant of innovation that adds considerable explanatory
power to traditional cross-sectional models of firm innovation. These results are especially significant
considering a critical role of innovation in driving competitive advantage and long-term economic growth
(Sollow (1957), Romer (1987, 1990)).
Third, our results add to the emerging literature on the effects of policy uncertainty on real
investment. A growing consent among researchers asserts that policy uncertainty depresses investment
(see, e.g., Barro (1991), Pindyck and Solimano (1993), Alesina and Perotti (1996), Bloom, Bond and Van
Reenen (2007), Julio and Yock (2012, 2014)).4 In this paper, we argue that political activism can mitigate
the negative effects of policy uncertainty and stimulate real investment of politically active firms. In turn,
if investment of politically active firms does not completely crowd out investment of other firms, political
activism can promote competitive advantage and long-term economic growth. These results have clear
and important policy implications. Many academics, political observers and policymakers have called for
major curbs on different forms of corporate political activism citing potential corruption and an unfair
capture of the legislative process by the corporate sector (see, e.g., Briffault (2011) for a discussion of
legislation actions surrounding the Citizens United Supreme Court case). While these concerns may be
valid in some instances, we caution against such one-sided blanket view of firm political activism. Our
results show that firm political activism can serve an important role in reducing policy uncertainty and
stimulating firm investment, which, in turn, promotes real economic growth.
Section 2 presents the sample and describes the construction of our measures of firm political
activism. Section 3 documents the effect of political activism on firm innovation. Section 4 documents
the effect of political activism on the firm’s ability to predict future legislative changes. Section 5
concludes.
2. Sample
2.1. Data sources
Our sample comprises the intersection of the NBER patent citations data file described in Hall,
Jaffe, and Trajtenberg (2001) and the political contributions data file in Cooper, Gulen, and Ovtchinnikov
4 As a case in point, consider Italy, a country that has had 65 governments in the 69 years since the World War II.
Jones and Mackenzie (2014) argue that political instability created by the new governments regularly ripping up the
predecessors’ policies and starting new has had a significant negative impact on corporate investment and is one of
the central reasons for the country’s current economic decline.
6
(2010). The patent citations file consists of 3,279,509 patents granted to firms for the period January 1,
1963 – December 30, 2006 and 24,082,781 patent citations made by patents granted during the period
January 1, 1975 – December 30, 2006. From this file, we extract data on the identity of the patent
recipient, the patent application date, the patent technological category and sub-category, and the number
of patent citations received. After deleting data for private firms and firms with no identification codes,
we are left with 1,158,693 patents granted to 6,528 unique firms and 9,673,067 patent citations.
The political contributions data file is described in detail in Cooper, et al. (2010) and consists of
1,064,830 hard money political contributions made by firms through their political action committees
(PACs) to politicians running for the President, the Senate, and the House of Representatives during the
period January 1, 1979 – December 31, 2004. From this file, we extract data on the identity of the
contributing firm, the date and the amount of each contribution, and the identity of the receiving
candidate. For each receiving candidate, we collect data from the Federal Election Commission (FEC) on
the sought after public office, the state and the district for which the candidate is running, the candidate’s
party affiliation, the election outcome, and the percentage of votes received. In addition, for all
candidates already in office, we obtain data on their Congressional committee assignments and their party
rankings on each serving committee. This data is from Charles Stewart’s Congressional Data Page.5
After deleting contributions from private firms, subsidiaries of foreign firms and firms with no data in the
CRSP/Compustat merged file, we are left with 813,692 contributions made by 1,805 unique firms.
After intersecting the patent citations and the political contributions files, our final sample
includes all 1,805 politically active firms with 404,536 granted patents and 3,814,120 patent citations
during the period January 1, 1979 – December 31, 2004. During this period, the firms in our sample
contribute a total of $933,002,309 in 2005 dollars to 5,584 unique political candidates. Our sample of
political contributions covers 67.1% of the total dollar volume of all corporate contributions reported to
the FEC and 27.2% of all political candidates running for office.
The firms at the intersection of the patent citations and the political contributions files represent
9.1% of all firms with available data on assets and book equity from CRSP/Compustat and 59.3% of the
total market capitalization of those firms. It is clear that politically active firms are significantly larger
than politically inactive firms. Indeed, the average market capitalization of politically active firms places
them in the top 8th percentile of the NYSE market capitalization in 2004.
Table 1 reports summary statistics for various firm characteristics of politically active and
inactive firms. All variables are defined in Appendix A. In panel A, politically active firms are much
larger, older and more profitable than inactive firms. Politically active firms also have more tangible
assets, higher leverage, lower Q, and invest significantly less in R&D relative to their assets. In terms of
5 We thank Charles Stewart III for generously providing this data on his website
http://web.mit.edu/17.251/www/data_page.html.
7
sheer raw innovation, politically active firms receive more patents and patent citations, although this
result likely reflects the firm size effect. To address this issue, panel B reports the characteristics of
politically active and inactive firms within size-ranked deciles. We sort all firms in the patent citations
file by NYSE annually ranked size decile breakpoints and within each decile report the characteristics of
politically active and inactive firms.
The results in panel B show that politically active firms receive more patents compared to
inactive firms, but this result holds only for very large firms. For other firms, there is no discernible
pattern in patent activity between the two subsamples. Similarly, there appears no consistent differences
in patent citations between politically active and inactive firms once we control for firm size. As regards
other variables, the relation between firm size and political activism is particularly evident from the
distribution of politically active firms across size deciles. The percentage of politically active firms is
1.1% in the smallest decile but increases significantly to 14.7% for firms in decile five and 71.5% for the
largest decile. Controlling for firm size, politically active firms tend to be older and less profitable, with
more tangible assets and higher leverage but lower Q and R&D expenditures. Politically active firms also
operate in more competitive industries.
2.2. Measures of political activism
We construct three separate sets of measures of firm political activism. Our first set of measures
is similar to Cooper, et al. (2010) and is designed to capture the general level of firm political activism.
Specifically, we rely on prior evidence that firms build long-term relationships with politicians by making
frequent, roughly constant political contributions and define our first measure of political activism as the
total number of political candidates supported by a firm over a five-year period:
∑
(1)
where i and t index firms and years, respectively, and is an indicator variable equal to one if
firm i made at least one hard money political contribution to candidate j during the period t - 4 – t and
zero otherwise. We use a five-year window in Eq. 1 to capture the firms’ desire to develop long-term
relationships with politicians (Snyder (1992)). To match the timing of political contributions to the
timing of elections, which take place on the Tuesday following the first Monday of November of every
even year, we use election cycle years spanning 12 consecutive months from November of year t - 1 to
October of year t. Our first election cycle year starts in November 1979, and the number of supported
candidates in Eq. 1 is updated once a year from October 1984 to October 2004.
Given prior evidence that hard money political contributions are correlated with other ways in
which firms build relationships with politicians (see, e.g., Wright (1990), Milyo, Primo, and Groseclose
8
(2000), Ansolabehere, Snyder, and Tripathi (2002), Bombardini and Trebbi (2008)), the number of
supported candidates in Eq. 1 should be an accurate measure of firm political activism. For robustness,
we supplement this measure with the total dollar amount of political contributions made to political
candidates by a firm over a five-year period:
∑
(2)
where is the total amount of money that firm i contributed to candidate j during the period t
- 4 – t. We use the same timing convention to construct this measure as in Eq. 1.
Our next set of measures of political activism is based on the results in Ovtchinnikov and
Pantaleoni (2012) that show that politicians who sit on Congressional committees with jurisdictional
authority over a firm’s industry receive a greater level of political support from individuals who reside in
a vicinity of the industry firms. Moreover, Ovtchinnikov and Pantaleoni (2012) show that the higher
level of political support from those individuals is associated with better operating performance of local
firms. These results imply that relationships with politicians who sit on Congressional committees with
jurisdictional authority over a firm’s industry are more valuable. We hypothesize that such relationships
are also more relevant for firm innovation because they provide access to valuable legislative information
(that originates in relevant Congressional committees) thereby reducing policy uncertainty. For brevity,
we label Congressional committees with jurisdictional authority over a firm’s industry as influential
committees and politicians on those committees as influential politicians. We then build analogs to Eqs.
1 and 2 that include only the subset of influential politicians:
∑
(3)
∑
(4)
where is an indicator variable equal to one if firm i contributed to politician k who sits on
an influential committee during the period t - 4 – t and zero otherwise, and is the total
amount of money that firm i contributed to politician k over the same period. The mappings between
Congressional committees and their industry jurisdictions are from Ovtchinnikov and Pantaleoni (2012),
Appendix B. We also build compliments to Eqs. 3 and 4, and
, as the difference between Eqs. 1 and 3 and Eqs. 2 and 4, respectively. These
variables track a firm’s relationship with politicians outside of Congressional committees relevant to the
firm. We label these committees as outside committees and politicians on those committees as outside
politicians.
9
Our last set of measures of political activism is based on the argument in Kroszner and Stratmann
(1998) that membership on specialized Congressional committees facilitates repeated interactions
between special interest groups and politicians. It also fosters long-term relationships with politicians and
supports a reputational equilibrium in which contributions-for-service, including contributions-for-
information, contracts are enforceable. One implication of this argument is that contributions to
politicians who join a committee become more valuable for those interest groups that operate under the
jurisdiction of that committee. Consequently, politicians who join a particular committee should
experience an increase in the level of contributions from the affected interest groups. Using this logic, we
calculate the total number of politicians supported by a firm who join influential Congressional
committees as well as the total amount of political contributions made to those politicians:
∑
(5)
∑
(6)
where is an indicator variable equal to one if firm i contributed to politician m during the
period t - 4 – t who joined an influential Congressional committee in year t and zero otherwise, and
is the total amount of money that firm i contributed to politician m over the same period.
Finally, and
are defined analogously by focusing on contributions made to
politicians who change their Congressional committees from influential to outside committees.
Table 2 reports summary statistics on our political activism measures. On average, politically
active firms support 84 political candidates over any given 5-year period. This support is divided
between 13.5 candidates who serve on influential Congressional committees and 70.6 candidates who
serve on outside committees. The minimum number of supported candidates is one (315 firms in our
sample support a single candidate) and the maximum is 844 (AT&T in 1984). In terms of contribution
amounts, firms in our sample contribute on average a total of $226,050 over a 5-year period. These
contributions are divided between $44,082 contributed to members of influential Congressional
committees and $181,968 contributed to other politicians. These results show that members of influential
committees receive more in political contributions ($44,082/13.483 = $3,270 on average) compared to
other politicians ($181,968/70.619 = $2,577), which is consistent with Romer and Snyder (1994) and
Kroszner and Stratmann (2000). The minimum amount of political contributions is zero and the
maximum is $6,293,663 (AT&T in 1990).6
6 Zero contributions arise when a firm’s original contribution is refunded by a politician’s campaign, usually after an
election loss. We view them as valid contributions in the sense that the firm attempted to establish a relationship
with a politician based on considerations deemed important by the firm. However, the money was refunded for
reasons most likely outside of the firm’s control.
10
In panel B, we present a detailed account of political contributions made by each politically active
firm in each election cycle to members of each Congressional committee in our sample. Ranked by the
average contribution amount, members of the House Energy and Commerce committee collectively
receive the most in political contributions from a typical firm ($13,578 on average), while members of the
House Merchant Marine and Fisheries committee receive the least ($7,335). Similarly in the Senate,
members of the Commerce committee receive the most in political contributions ($9,479), while members
of the Environment and Public Works receive the least ($7,409). Other committees receive between
$7,000 and $10,000 on average from firms, with the Armed Services, Financial Services, and the House
Transportation committee receiving amounts at the top end of the distribution and the Agriculture and
Natural Resource committees at the bottom.
In addition to contribution totals, two other results stand out. First, firm contributions to
Congressional committees are significantly related to the rankings of powerful committees developed in
Edwards and Stewart (2006). Three out of seven House committees in our sample (Energy and
Commerce, Transportation and Infrastructure, and Armed Services/National Security), and two out of six
Senate committees (Commerce, Science, and Transportation and Armed Services) are on the Edwards and
Stewart (2006) list of the ten most powerful committees. Moreover, the correlation between contribution
totals in panel B and committee power rankings is 0.782 for the House committees and 0.538 for the
Senate committees. These correlations are significantly higher than those reported in Ovtchinnikov and
Pantaleoni (2012) who study individual political contributions and show that firms target powerful
committees to a much greater extent than individuals. The results are consistent with prior evidence that
committees with significant power over firms receive more money (see, e.g., Grier and Munger (1991),
Romer and Snyder (1994), Ansolabehere and Snyder (1999)).
Second, the differences in committee contribution totals reflect differences in the number of
committee members that firms choose to support rather than differences in the total amount of political
contributions made to each committee member. In fact, the latter is roughly constant, with approximately
$1,700 contributed to members of the House committees and $2,800 contributed to members of the
Senate committees. These results are in line with Cooper, et al. (2010) and show that the level of firms’
political activism is determined by the number of political candidates firms choose to support rather than
by the amount of money contributed to each candidate. Therefore, in our analysis below, we focus on
political activism measures based on the number of supported political candidates. We use political
activism measures based on the amount of contributions to political candidates as a robustness check.
3. Political contributions and firm innovation
In this section, we document the effect of political activism on firm innovation. Our
identification strategy follows a four-step approach. First, we establish a baseline positive correlation
11
between firm political contributions and future innovation. Second, we conduct a wide range of
subsample analyses that, in their entirety, provide stronger causal evidence of the effect of political
contributions on firm innovation. Third, we set the identification hurdle higher and employ instrumental
variables estimation. We instrument for political contributions with a politician’s margin of victory in the
prior election and show that instrumented contributions strongly predict future firm innovation. Fourth,
we move the identification bar even higher and exploit an exogenous shock to the value of political
contributions that occurred during our sample period. We show that subsequent innovation activity of the
treatment firms was significantly affected by the shock relative to that of the control firms. To preview
our results, the collective evidence below offers strong support for our hypothesis that political activism
facilitates firm investment in innovation.
3.1. Baseline results
Our baseline test of the hypothesis that political activism facilitates firm innovation is to establish
a positive correlation between our proxies for political activism and innovation. We estimate the
following panel OLS regression that relates political activism proxies to firm innovation:
(7)
where is a measure of innovation of firm i during the period ( { }) and
are industry and year fixed effects, respectively, is a measure of firm political activism,
is a vector of control variables, and is a random error term assumed to be possibly
heteroskedastic and correlated within firms (Petersen (2009)). All control variables are winsorized at the
upper and lower one-percentiles.
In our baseline tests, we use two measures of firm political activism, and
defined in Eqs. (1) and (2), respectively, and estimate two separate models of firm innovation. Following
extant literature (see, e.g., Hall, Jaffe, and Trajtenberg (2005), Atanassov (2013), He and Tian (2013),
Seru (2014), Fang, Tian, and Tice (2014)), our first measure of firm innovation, , is the
natural logarithm of the total number of successful patent applications for firm i during the period .
We use the patent application year because it is closer to the timing of actual innovation than the patent
grant year (Grilliches, Pakes, and Hall (1988)). Our second measure of innovation, , is the
natural logarithm of the total number of future patent citations received by each patent. Compared to
simple patent counts, this variable more accurately captures the significance and value of individual
patents and, therefore, may serve as a better proxy for innovation “success” (Hall, Jaffe, and Trajtenberg
(2005); Grilliches, Pakes, and Hall (1988) report that patent value is highly skewed, with most of the
value concentrated in a small number of patents). Because of the well-known truncation problem in the
patent data file, we follow Hall, Jaffe, and Trajtenberg (2005), Trajtenberg (2005), Atanassov (2013) and
12
Seru (2014) and estimate effective patents and patent citation counts by dividing the annual number of
patent applications and patent citations for a given firm in a given technology class by the average
number of patent applications and patent citations received by all firms in the same technology class in
the same year. Our results are robust to an alternative truncation adjustment used in He and Tian (2013)
and Fang, Tian, and Tice (2014).
Table 3 reports the results of estimating Eq. 7. Panel A presents the results for the total number
of patent applications; panel B presents the results for the total number of patent citations. Models 1 and
2 test whether political activism affects firm innovation during the year immediately following the year
when political contributions are measured. Models 5 and 6 allow for a longer three-year lag between
political contributions and their effect on subsequent innovation.
The results in panel A show that firm patent activity is strongly related to political activism. The
coefficient is positive and significant at the 1% level in all models, which implies that firms that
support more political candidates and firms that contribute more capital to political candidates innovate
more. The results are significant at both the one- and three-year horizons. It is important to note that all
models include firm size as a standard control variable, so our results do not simply reflect the firm size
effect (table 1). The inclusion of political activism variables improves the R2 statistic in model 1 from
0.295 to 0.326 and from 0.333 to 0.357 in model 5. These figures represent a 10.1% and a 7.2%
improvement in R2, respectively, and show that firm political activism adds significant explanatory power
to the baseline model of innovation.7 The results for Camount in models 2 and 6 are similar.
To get a better sense of economic magnitudes of our political activism variables, note from table
2 that firms in the 25th percentile of the number of supported political candidates support 14 candidates,
while firms in the 75th percentile of that distribution support 117 candidates. Thus, an interquartile
increase in the number of supported candidates is associated with an additional 0.489 patent applications
over the next year ( ([ ] ) ) and an additional 0.517 patent applications over the
next three years. Compared to the unconditional mean number of patent applications by politically active
firms (6.1 patents), the results appear economically large. As another benchmark, the impact on
innovation of an interquartile change in each control variable ranges from 0.357 additional patent
applications for leverage to 0.461 for total assets. Thus, the number of supported political candidates has
the largest economic effect on firm innovation, even larger than firm size. This is especially compelling
given prior evidence that firm size is one of the most dominant determinants of firm innovation (see, e.g.,
Atanassov (2013), Cornaggia, et al. (2014), He and Tian (2013)). The results for Camount are
7 The estimated coefficients on traditional control variables line up well with expectations and previous studies. In
particular, firm innovation is positively related to firm size, age, profitability, investment, and Q and negatively
related to firm leverage and the ratio of fixed assets to total assets. Firm innovation is higher in more concentrated
industries as measured by the Herfindahl index, although the relation is non-linear. Similar results are reported in
Atanassov (2013), Cornaggia, et al. (2014), He and Tian (2013), and Fang, Tian and Tice (2014).
13
economically equally significant and indicate that an interquartile increase in the amount of political
contributions is associated with an additional 0.419 (0.430) patent applications over the next year (three
years).
The results in panel B similarly show a strong positive association between firm patent activity,
as defined by patent citations, and political activism. The is positive and significant at the 1% level for
both measures of political contributions and at both time horizons. The results also appear economically
large. The magnitude of the coefficient implies that an interquartile increase in Pcand and Camount is
associated with an additional 0.381 and 0.374 patent citations over the next year and an additional 0.395
and 0.380 patent citations over the next three years. Compared to economic magnitudes of the control
variables, which range from 0.360 additional patent citations for an interquartile change in leverage to
0.370 for total assets, the results again indicate that political activism has the largest economic effect of
all variables on firm innovation.
In the current estimation, we use all firms with available data on Compustat to estimate Eq. 7,
irrespective of whether those firms have an established PAC. Firms without a PAC are assigned a value
of zero for political activism variables, which may be problematic for two reasons. First, a zero political
contribution implies a choice in the sense that a firm could have contributed money to a politician but
decided not to do so. This is not the case for firms without a PAC that decided not to participate in the
political process altogether. Second, firms do not randomly decide to get involved in politics, and this
choice introduces a potential self-selection bias into our observed sample. To control for this bias, we use
a two-stage Heckman (1979) approach in the remainder of table 3. In the first stage, we estimate a probit
model of whether a firm has an established PAC on determinants of PAC participation. These
determinants come from prior literature that analyzes the determinants of firm political involvement (see,
e.g., Masters and Keim (1985), Zardkoohi (1985), Grier, Munger, and Roberts (1994), Hart (2001)). The
first-stage probit results are presented and discussed in Appendix B. From the probit, we calculate the
inverse Mills ratio (IMR) and include it in second-stage regressions in the rest of table 3. The inclusion of
IMR in innovation regressions helps control for firm self-selection into the politically active group.
Models 3 and 4 and models 7 and 8 present the results for one-year ahead and three-year ahead innovation
measures, respectively, for the sample of politically active firms.
The results of the two-stage Heckman model show that firm innovation is strongly positively
related to political activism after controlling for self-selection. The coefficient remains significant at
the 1% level in both panels and at both time horizons despite the fact that our estimation utilizes only
politically active firms. Put differently, our baseline results do not simply reflect differences in
innovation between politically active and inactive firms. Rather, within politically active firms, those
firms that support more political candidates and contribute more capital, also innovate more. In terms of
economic significance, the Heckman results are quite similar to full sample results. An interquartile
14
increase in Pcand and Camount is associated with an additional 0.435 and 0.393 patent applications and
0.379 and 0.372 patent citations over the next year and an additional 0.453 and 0.399 patent applications
and 0.391 and 0.377 patent citations over the next three years. These results are quite similar in
magnitude to those reported for the whole sample. Firms that are more politically active innovate more.
3.2. Subsample analysis
Our baseline result establishes a strong positive correlation between political contributions and
firm innovation. Whereas this result is consistent with the hypothesis that political activism facilitates
firm innovation, this clearly is not the only possible interpretation. Firm political activism is plausibly
endogenous to innovation, so to comment on causality, we must tackle the omitted variable bias as well as
the simultaneity of political contributions and firm innovation (Roberts and Whited (2010)).
Despite including all standard determinants of firm innovation in table 3 (Atanassov (2013),
Cornaggia, et al. (2014), He and Tian (2013), Fang, Tian, and Tice (2014)), we cannot dismiss a
possibility that some omitted variable, correlated with political contributions, actually drives innovation.
For example, political contributions are correlated with firms’ access to external financing (Claessens,
Feijen, and Laeven (2008)) which may, in turn, predict innovation. Likewise, political contributions are
related to certain governance characteristics (Aggarwal, Meschke, and Wang (2009)), and it is possible
that those characteristics drive firm innovation (Atanassov (2013)). This problem is especially
challenging because the complete set of omitted variables is potentially infinite and, by construction,
unobservable to an econometrician. So, instead of relying on additional controls in Eq. 7, we go a
different route. Our second step toward identification is to decompose the sample of political
contributions into those that are predicted by our hypothesis to have a stronger impact on firm innovation
and those contributions predicted to have a weaker or no effect on innovation. If there is indeed a
differential impact in the data, this evidence would be consistent with our hypothesis that political
contributions facilitate innovation. However, it would be more difficult to interpret the results as driven
by an omitted variable since that interpretation would require that the omitted variable is correlated with
one particular subset of political contributions but not with others. This analysis is also important in
ruling out the possibility that the relation between political contributions and firm innovation is spurious.
Table 4 reports the results for various subsamples of political contributions. If political
contributions facilitate innovation because they grant access to politicians who, in turn, either supply
valuable information or trade legislative favors for contributions, the relation between political
contributions and firm innovation should be stronger for contributions to influential politicians. Because
these politicians sit on Congressional committees with jurisdictional authority over the firm’s industry,
they have direct legislative control over the firm and possess valuable private information about future
legislative changes. The evidence in table 2 shows that influential politicians receive more in political
15
contributions compared to other politicians.8 Hence, under our hypothesis, contributions to influential
politicians are more significant in predicting future innovation compared to contributions to outside
politicians.
Panels A and B present the results of estimating Eq. 7 for subsamples of contributions to
influential and outside politicians, respectively. All models include the full set of control variables in
table 3 and control for firm self-selection into the politically active group with IMR. Consistent with our
hypothesis, the results show a much stronger relation between political contributions to influential
politicians and firm innovation compared to the relation between political contributions to outside
politicians and firm innovation. In panel A, the coefficient on PcandCommittee and CamountCommittee is
positive and statistically significant, which implies that political contributions to influential politicians are
significant predictors of future firm innovation. This result holds for firm innovation measured by patent
applications and patent citations and at both the one- and three-year horizons. In contrast, the
coefficient on PcandNon-Committee and CamountNon-Committee in panel B is insignificant. It is actually
marginally negative in model 1, although this result is not robust to alternative specifications. The
differences in results across panels A and B are stark and provide further evidence that political
contributions, especially to influential politicians, are strongly related to firm innovation. This evidence
is consistent with our hypothesis but is more difficult to explain by an omitted variable bias. The latter
explanation would require that an omitted variable is positively correlated with political contributions but
only to influential politicians while uncorrelated with contributions to outside politicians.
In our next set of tests, we switch from levels to changes in firm political activism and focus on
politicians who join or leave influential Congressional committees. Kroszner and Stratmann (1998) argue
that membership on specialized Congressional committees facilitates a reputational equilibrium in which
politicians trade services to special interest groups for their political contributions. Under our hypothesis,
this argument implies that contributions to politicians who join (leave) influential committees become
more (less) significant in predicting future firm innovation after the committee change. On average, 144
(123) politically active firms in table 4 make at least one political contribution over any five-year period
to a politician who joins (leaves) an influential committee. These firms represent, respectively, 26.7%
and 22.8% of an average annual number of politically active firms in our sample. In panel C, we estimate
the following regression:
(8)
8 We analyze the relation between a politician’s status and firm contributions in greater detail. In unreported tests,
we find that firms (i) are more likely to contribute to influential politicians, (ii) contribute more frequently to
influential politicians, and (iii) contribute greater amounts to influential politicians. These results are available upon
request.
16
where and
are the number of politicians joining and leaving influential
Congressional committees defined in section 2.2 and the rest of the variables are as defined above.
Consistent with our hypothesis, the results show a significant positive relation between
contributions to politicians who join influential committees and firm innovation while no relation between
contributions to politicians who leave those committees and firm innovation. The coefficient on
is positive and significant in all specifications while the coefficient on
is either
insignificant or marginally negative. We also estimate a variant of Eq. 7 that uses and
in place of
and . The results are very similar, so we do not report them
in the interest of space.
The variables and
are clearly not mutually exclusive. In fact, 508 unique
firms in our sample simultaneously support politicians who join influential committees and politicians
who leave those committees. So, in the rest of panel C, we replace and
with
indicator variables that isolate firms that net gain or lose connections with influential politicians.
Specifically, we define ( ) as an indicator variable set to one for firms that support
a greater number of politicians joining influential committees than politicians leaving those committees.
Analogously, we define ( ) as an indicator variable set to one for firms that
support a lower number of politicians joining influential committees than politicians leaving those
committees. As an illustration, AT&T supported seven politicians during the 1998 – 2002 period who
joined influential committees following the 2002 election and one politician during that period who left
an influential committee. So, ( ) takes a value of one for AT&T in 2002. During
the same time period, Boeing supported eight politicians who joined influential committees and 14
politicians who left influential committees in 2002. So, ( ) takes a value of one
for Boeing in 2002.
Consistent with our hypothesis, the results show a strong positive effect of political contributions
on firm innovation for firms that support a greater number of politicians joining relevant committees than
politicians leaving those committees. The coefficient is positive and significant for both measures of
innovation and at both the one- and three-year horizons. In contrast, there is no effect of political
contributions on innovation for firms that support fewer politicians joining relevant committees than
politicians leaving those committees. The coefficient is negative when innovation is measured by the
number of patents and insignificant when innovation is measured by the number of patent citations. The
results are very similar if the indicator variables are computed based on the total amount of political
contributions to politicians joining and leaving influential committees. The results are available upon
request. Collectively, the results in panel C are consistent with our hypothesis but seem, at least to us,
more difficult to explain by an omitted variable bias.
17
In our last set of subsample tests, we interact ,
and indicators (
) and ( ) with election year and non-election year dummies. Election
outcomes and, consequently, many upcoming Congressional committee changes are often predictable
during election years, especially in weeks just prior to general elections. Mullins and Mundy (2010)
suggest that firms exploit this predictability and strategically shift their contributions to politicians likely
to gain influence in the upcoming elections. If political contributions are set optimally and firms
redistribute contributions during election years to politicians expected to join influential Congressional
committees, we expect a lower, if any, impact of those contributions on subsequent innovation.
Conversely, if firms make political contributions to politicians who unexpectedly join influential
committees during non-election years, we expect a stronger impact of those contributions on future
innovation.9
Panel D presents the results. On average, 65 (39) out of 539 politically active firms in table 4
make at least one contribution to a politician who joins (leaves) an influential committee during a non-
election year. In comparison, 217 (200) firms support at least one politician who joins (leaves) an
influential committee during an election year. Naturally, committee changes in non-election years are
rarer although far from infrequent. Consistent with our hypothesis, the entire effect of political
contributions on firm innovation is concentrated among politicians who join influential committees
during non-election years. The coefficient on is positive and significant but only in non-
election years. In contrast, the coefficient is insignificant in election years. The results hold for firm
innovation measured by patent applications and patent citations and at both the one- and three-year
horizons. The results are identical if is replaced with ( ) in Eq. 8.
Firms that support a greater number of politicians joining influential committees than politicians leaving
those committees subsequently innovate more but only for committee changes that take place in non-
election years. There is no effect of political support on firm innovation in election years. Also
consistent with our hypothesis, the coefficient is insignificant in all years. Our results are robust if all
variables are computed based on the total amount of political contributions to politicians joining and
leaving influential committees. The results are available upon request.
Overall, the analysis in table 4 shows that an omitted variable bias is not likely to explain our
baseline results. Such an explanation would require that an omitted variable is correlated with the overall
level of firm political activism because it is correlated with the firm’s support for politicians on influential
Congressional committees and for politicians who join such influential committees. Moreover, the
correlation between the omitted variable and the firm’s political support has to exist only in non-election
years. We are unaware of a reasonable candidate for such a variable.
9 Committee reassignments in non-election years take place because of a politician’s death, resignation from
Congress, or appointment to a different post.
18
3.3. Instrumental variables estimation
We next turn to analyses that represent a higher hurdle to identify the effect of political
contributions on firm innovation. While the results in the previous section serve an important first step
toward identification, we cannot yet completely rule out the existence of an omitted variable that causes
both firm innovation and political contributions. Moreover, the above analysis does little to address the
simultaneity of political contributions and firm innovation. It may well be the case that innovative firms
choose to make political contributions not only to access information from politicians but also to protect
their current innovative status. If the latter is true, the relation between political contributions and firm
innovation runs in the direction opposite to that predicted by our hypothesis. The cleanest example is
firms lobbying members of Congress to exert influence on various government officials, including those
in the U.S. Department of Commerce and its Patent and Trademark Office.10 If such lobbying is
successful, firms can protect their innovative status with political contributions by increasing the odds of
approval of their own future patents or decreasing the odds of approval of their competitors’ patents.
Under this explanation, innovation drives the supply of political contributions.
To tackle both issues, we turn to instrumental variables (IV) estimation. We use a politician’s
margin of victory in the prior election as an instrument for firms’ political contributions to that politician
in the current election cycle. Intuitively, a politician’s margin of victory in the prior election is unlikely to
affect current firm innovation. However, it should significantly affect the firm’s propensity to make
political contributions to that politician because the marginal dollar of contributions matters more to a
vulnerable politician in a close election. Consistent with this, several studies show theoretically and
empirically that politicians in close elections raise more campaign financing (see, e.g., Jacobson (1980,
1985), Kau, Keenan, and Rubin (1982), Poole and Romer (1985), Stratmann (1991)). Because the value
of a marginal dollar of campaign funds is higher in close elections, a politician running in a close race
ought to promise more favors in return for contributions to attract more campaign financing. The
assumption that a politician trades favors in exchange for contributions is common in theoretical
arguments including Coate (2004) and has received empirical support (see, e.g. Stratmann (1998, 2002)).
Thus, we hypothesize that firms target politicians in close elections and expect more benefits, including
access to information valuable for future innovation, in return for their political contributions.
Firms typically support multiple political candidates per year, so we calculate our instrument as
follows. For every year and for every firm, we estimate a first-stage logit regression that relates a firm’s
choice to contribute to a politician to her margin of victory in the prior election. The coefficient on the
10 Wiley (1988) argues that Congress indeed has significant influence and oversight of government agencies
including control by statute, the power of the purse, the continuing oversight of standing committees, and pressures
from individual politicians and congressional staff members.
19
margin of victory measures the sensitivity of the firm’s decision to contribute to the politician’s prior
election result. A low (in fact, negative) sensitivity identifies firms that choose to support politicians with
a low prior margin of victory who, we argue, should be more vulnerable in the current election cycle and,
therefore, more likely to trade political favors for contributions going forward. Conversely, a high
sensitivity identifies firms that choose to support politicians firmly entrenched in their Congressional
positions.
Because of the well-known “errors-in-the-variables” problem with the first-stage sensitivity
estimates, we sort firms into 20 equal- and value-weighted portfolios based on their estimated sensitivities
to the margin of victory. Firms with the lowest sensitivity are placed in portfolio one and firms with the
highest sensitivity are placed in portfolio 20. Fama and MacBeth (1973) show that as long as the errors in
the estimated first-stage coefficients are less than perfectly correlated, the portfolio coefficients are much
more precise estimates of the true coefficients. In the second-stage regression, we then estimate
predictive regressions that relate portfolio average growth in innovation over different time horizons to
the lagged average portfolio sensitivity to the margin of victory. Our hypothesis predicts that firms that
target politicians with a low margin of victory subsequently benefit from a faster growth in innovation.
So, we expect a negative relation between firm sensitivity to the margin of victory and future innovation
growth.
Table 5 reports the results of IV estimation. Panel A presents the first-stage logit results, with
emphasis on the instrument relevance condition. The results show that firm political contributions are
strongly related to a politician’s margin of victory. The average coefficient on the margin of victory is
negative and significant, which indicates that firms are much more likely to support politicians with a low
margin of victory in prior elections. The analysis of the distribution of the first-stage coefficients shows
that 11,439 out of 17,873 coefficients on the margin of victory are negative (64% of firm-year
coefficients) and 1,024 of the negative coefficients are significant at the 5% level or higher (9% of all
negative coefficients). At the same time, only 297 positive coefficients are statistically significant (4.6%
of all positive coefficients). These results imply that the entire distribution of the first-stage coefficients
is centered below zero and exhibits a significantly fatter left tail than what would be expected by chance.
This evidence is consistent with our hypothesis that firms target vulnerable politicians with a low margin
of victory in prior elections.
To address the exclusion restriction, in unreported tests, we sort firms into deciles based on the
average margin of victory of their supported politicians and analyze future innovation growth across the
different deciles. Irrespective of the time horizon considered, we find no consistent differences in
innovation growth across the deciles. We also regress our instrument on lagged innovation growth and
find a positive but insignificant coefficient (p-value > 0.3). Although not a formal proof of the exclusion
restriction, these results are an important building block in our IV estimation. They corroborate our
20
intuition that a politician’s margin of victory has no impact on firm innovation except through its impact
on political contributions. Together with panel A results, this evidence shows that the margin of victory
is a valid instrument.
The remainder of table 5 reports the results of the second-stage regressions that relate our
instrument to firm innovation. Panel B presents the results for innovation measured by the number of
patent applications; panel C presents the results for innovation measured by the number of patent
citations. The results in panel B show that firm innovation is strongly related to its propensity to support
politicians with a low margin of victory. The coefficient on the sensitivity to the margin of victory is
negative and significant at the 1% level in regressions that employ both equal- and value-weighted
portfolios and for innovation growth measured at one-, two- and three-year horizons. These results imply
that firms that choose to support vulnerable politicians who narrowly beat their opponents in prior
elections subsequently experience a faster growth in patent applications. This evidence is consistent with
our argument that vulnerable politicians are more likely to trade favors for political contributions; hence,
firms that target vulnerable politicians obtain higher tangible benefits, including access to information
valuable for future innovation.
To gauge economic magnitudes of our results, the interquartile range in the sensitivity to the
margin of victory is 2.9% in our sample, so an interquartile decrease in the sensitivity to the margin of
victory is associated with a 3%, 5.1% and 7.4% (2.5%, 4.3% and 6.1%) faster growth rate in patent
applications over the next year, two years, and three years respectively using EW (VW) portfolios. The
results are economically important.
The results in panel C appear statistically weaker. Nevertheless, the sign of the coefficients in all
but one specification is consistent with our hypothesis that changes in firm patent citations are positively
related to the propensity to support politicians with a low margin of victory. In addition, the results for
three year changes in patent citations are statistically significant at the 5% level. In terms of economic
significance, an interquartile decrease in the sensitivity to the margin of victory is associated with a 0.17%
(0.14%) faster growth in patent citations over the next three years using EW (VW) portfolios.
Overall, the results in table 5 provide stronger evidence of a causal effect of firm political
contributions on innovation. By instrumenting political contributions with a politician’s margin of
victory, our analysis isolates the exogenous variation in firm political contributions, which, in turn, allows
us to more precisely pin down the causal effect of contributions on innovation. The IV results are
consistent with our hypothesis that political activism facilitates firm innovation.
3.4. Difference-in-difference estimation
Our next set of tests moves the identification bar still higher and exploits a significant exogenous
change in the value of political contributions that occurred during our sample period. The 1994 midterm
21
Congressional election was a dramatic and unexpected victory for Republicans that not only put them in a
leadership position in the House of Representatives but also resulted in significant changes to the rules
and practices of the House. Aldrich and Rohde (1997) detail the changes, the most important of which for
us was Newt Gingrich’s unexpected departure from the traditional Republican practice of assigning
Congressional committee chairman positions based on committee seniority to that based on party loyalty.
Within a week of the November 8 election, the new leadership in the House announced the appointment
of four junior members of Congress to committee chairman positions, thereby bypassing the expected
senior candidates for the posts.11 Aldrich and Rohde (1997) note that this was the most significant
departure from seniority in the selection of committee chairs since 1974, when four senior conservative
Southern Democrats were stripped of their chairs following the election of the Democratic class of 1974
known as the “Watergate babies”.
The 1994 event represents a particularly convenient setting for us to analyze the importance of
political contributions, especially to powerful politicians with chairman roles, on firm innovation. First,
the election results and the new procedures for assigning committee chairman positions were completely
unexpected. Note Aldrich and Rohde (1997):
“Well before the Republican Conference could meet, and before any formal change in the method of
choosing committee chairs could be considered, Gingrich and his allies simply asserted the power to
choose the chairs of the committees that were most important to them.” (p. 550).
Second, the changes in committee chairmanships were clearly exogenous to firm innovation. Third,
various firms were impacted differently by the event. So, our treatment group comprises 126 firms that
supported the four junior politicians who unexpectedly became committee chairs in November 1994. To
be included in the treatment group, a firm must have contributed to at least one of the four would-be
committee chairs during the 1990 – 1994 period. We use two control groups to identify the treatment
effect. The first group, firms with chair contributions, includes 475 firms that supported one or more
existing committee chairs during the 1990-1994 period. The second control group, firms without chair
contributions, includes 197 firms that supported no existing committee chairs during the 1990-1994
period. If political contributions to powerful politicians are important for firm innovation, we expect an
increase in innovation for the treatment firms relative to the control firms without chair contributions
following the 1994 election. We also expect the level of innovation of the treatment firms to converge to
that of the control firms with chair contributions following the 1994 election.
11 During the week following the election, Robert Livinston (LA), ranked fifth in committee seniority, was promoted
to the chairman position on the Appropriations committee, bypassing the second ranking John Myers (IN). Henry
Hyde (IL) and Thomas Bliley (VA) were promoted to the chairman positions on the Judiciary and Energy
committees, respectively, bypassing a higher ranking Carlos Moorhead (CA). David McIntosh (IN), a House
freshman was appointed the chair of the National Economic Growth, Natural Resources, and Regulatory Affairs
subcommittee of the Government Reform and Oversight committee.
22
Figure 1 presents the results. The solid line tracks the treatment firms. The dashed line tracks the
control firms without chair contributions. The dashed line with a triangular marker tracks the control
firms with chair contributions. The shaded area designates the November 1994 election. The first
important result in figure 1 is the dynamics of patent activity of the treatment and control firms prior to
the election. It is clear that the level of patent applications varies considerably across the three
subsamples. However, no clear differences in trends are evident across the groups. This result, while not
a formal test of the “parallel trends” assumption, nevertheless suggests that in the absence of treatment,
the average trend in patent activity would have been the same across the treatment and control firms. The
differences in levels of patent activity is not a cause for concern as those will be differenced out in our
formal difference-in-difference tests below.
The post-election results in figure 1 show that the treatment firms significantly increase their
patent activity following the 1994 election. The number of patent applications hovers at roughly five
prior to 1994 but increases steadily to 5.3 applications in 1995, 5.8 in 1996, 7.1 in 1997, 7.5 in 1998 and
8.4 applications in 1999. The almost two-fold increase in patent activity of the treatment firms is
particularly striking given a relatively constant patent activity of the control firms without committee
chair contributions. Those firms apply for slightly fewer than two patents per year on average throughout
the event window. Thus, firms that supported politicians who were unexpectedly appointed committee
chairmen in 1994 subsequently significantly increased their patent activity relative to firms that did not
support committee chairs. The results in figure 1 also show that patent activity of the treatment firms
converges to that of the control firms with chair contributions after the 1994 election. Indeed, while a
large difference in patent applications exists between the two groups at the time of the election, it is
completely wiped out by 1999. Firms that supported would-be committee chairs increased their
innovation exactly to the level of innovation of firms that supported committee chairs.
Table 6 reports formal tests of the difference-in-difference estimates. Columns 1 – 3 analyze
differences in levels of patent activity between the treatment and control firms; columns 4 – 6 analyze
differences in changes in patent activity. The results show that both the level and the growth in patent
applications is significantly higher for the treatment firms following the 1994 election compared to the
pre-1994 period. The differences between the pre- and post-1994 estimates are positive and economically
large for both measures of patent activity. This stands in sharp contrast to the dynamics of patent activity
for the control firms, which shows no discernible trend in patent applications or patent application growth
from before to after 1994. The bottom two rows show that the resulting difference-in-difference estimates
are significantly positive in tests using both sets of the control firms and for patent activity measured in
levels and changes.
Table 7 decomposes the post-1994 period by year. We estimate the following regression:
23
( ) ( )
(9)
where { } is a measure of patent activity of firm i
during the year t+1, is the firm fixed effect, ( ) is an indicator time variable that identifies post-
1994 years, is an indicator variable set to one for firms in the treatment group and zero
otherwise and is an error term. Note that the indicator by itself is not in Eq. 9
because it is perfectly collinear with the firm fixed effect .
Panel A (panel B) presents the results comparing patent activity of the treatment firms with that
of the control firms with committee chair contributions (without committee chair contributions). Models
1 and 3 corroborate our findings in figure 1 and table 6 and show a significant treatment effect in the post-
1994 period. Models 2 and 4 show that when the post-event window is decomposed by year, the
treatment effect is concentrated in the post-1996 period. The lag between the committee chair
appointments and subsequent patent activity makes intuitive sense given that it takes considerable time
and resources for firms to adjust their innovation activity to new information (Atanassov (2013)). Finally,
we repeat the analysis in this section by measuring patent activity with patent citations rather than patent
applications and obtain qualitatively similar results.
Overall, the results in this section provide possibly the strongest evidence of a causal effect of
firm political contributions on innovation. By identifying an exogenous shock to the value of political
contributions, our analysis isolates firms that were significantly impacted by the shock and shows that
those firms responded with greater patent activity in subsequent years. We interpret these results as
robust evidence that political activism facilitates firm innovation.
4. Firm innovation and future legislative changes
The evidence in the previous section offers strong support for our hypothesis that political
activism positively affects firm innovation. This is consistent with the view that political activism
reduces policy uncertainty, which, in turn, fosters firm innovation. In this section, we explore in greater
detail how political activism may reduce policy uncertainty and stimulate firm innovation. We discuss
two channels, first introduced in section 1, and offer evidence consistent with both of them. We note
right away that the channels we describe are not mutually exclusive. Albeit the evidence below is
consistent with both explanations, we believe it is important to devise empirical tests that could isolate the
relative importance of each channel. This exercise, however, falls outside the scope of the current study
so we leave it to future work.
24
4.1. How does political activism reduce policy uncertainty?
We put forward two channels through which political contributions may reduce policy
uncertainty and stimulate firm innovation. First, firms may use political contributions to purchase access
to politicians who, in return for contributions, supply legislative information relevant for firm innovation.
For example, a politically active auto manufacturer will benefit significantly from its connections to
members of Congress who reveal private information to the firm that Congress plans to take a hard stance
on environmental or product safety issues. This information is clearly valuable to the firm and allows it
to adjust its innovation strategy in anticipation of upcoming legislative changes.
There is surprisingly little evidence in the existing literature whether firm political contributions
purchase access to politicians purely for information acquisition reasons. Hojnacki and Kimball (2001)
note that many studies define access to politicians as an opportunity to (i) persuade or make a case or
lobby, (ii) secure a legislator’s attention, (iii) meet with a legislator to communicate policy views, or (iv)
influence a legislator to consider and act on arguments. Yet, theirs is the only study of which we are
aware that considers a broader view of access that includes not only active lobbying but also purely
passive acquisition of information. Hojnacki and Kimball (2001) find that special interest groups with
established PACs contact more members of Congressional committees compared to groups without
established PACs. This is consistent with the view that political contributions purchase access, including
access to legislative information. If access to such information is valuable for firm innovation because it
reduces policy uncertainty and the value of the firms’ option to wait, we expect politically active firms to
successfully time future legislation and set their innovation strategies in expectation of upcoming
legislative changes.
The second way in which political contributions may reduce policy uncertainty and stimulate
innovation is through active lobbying. Firm political contributions and lobbying expenditures are
positively correlated (see, e.g., Wright (1990), Ansolabehere, Snyder, and Tripathi (2002), Akey (2014)),
so firms that contribute more also lobby more. There is also evidence that political contributions affect
legislative outcomes (see Stratmann (2005) for a review of the literature), which implies that lobbying is
successful.12 When applied to firm innovation, this argument suggests that firms may use political
contributions to lobby for legislation that increases their returns to innovation. For example, Aghion, et
al. (2005) present evidence that product market competition and innovation are linked. So, firm lobbying
for legislative changes that impact industry competition, such as foreign trade policy legislation, may
impact returns to innovation and subsequent supply of innovation. Under this explanation, we expect
politically active firms to have an advantage in timing their innovation to their own lobbying efforts and
the resultant future legislative changes.
12 Ansolabehere, et al. (2003) argue that endogeneity issues make inferences regarding the effects of political
contributions on legislative outcomes difficult.
25
Note that even though the two explanations are not mutually exclusive, they are distinct in one
important way. The “acquisition of information” explanation implies that firms are passive with respect
to future legislation and simply use political contributions to learn about upcoming legislative changes.
In contrast, the “active lobbying” explanation implies that firms are actively engaged in the legislative
process and use their political capital to shape future legislation. We believe it is important in future work
to report evidence on the relative importance of the two explanations. However, in this study, we simply
note that under both explanations, political contributions positively affect firm innovation. This relation
exists because under both explanations, politically active firms successfully time future legislation and set
their innovation strategies in expectation of upcoming legislative changes. Empirically, this implies that
changes in innovation by politically active firms predict future legislative changes. We now turn to
testing this prediction.
4.2. Political activism, firm innovation and future legislative changes
In predicting legislative changes, we focus on major deregulatory initiatives affecting the
petroleum and natural gas, utilities, telecommunications, and transportations industries as defined by the
Fama-French 49-industry definitions. Those deregulatory initiatives are from Ovtchinnikov (2010, 2013)
and are listed in table 8, panel A. We focus on economic deregulation, defined as deregulation of entry,
exit, price, and quantity, because it represents a major change in the operating environment of U.S.
industries in the 1980s and 1990s (Winston (1993)) and, therefore, serves a useful laboratory to study the
effects of firm political activism on subsequent innovation. Ovtchinnikov (2010) reports that deregulation
significantly affected firms’ decision making including their financing and investment choices and
resulted in significant changes in operating performance.
Given the impact of deregulation on the industry’s operating environment, we expect that firms
invest heavily in information about the timing of different deregulatory steps. Importantly, the incentive
to invest in information should arise not only among existing industry firms but also among potential new
entrants. If successful, access to legislation timing information gives politically active firms competitive
advantage because it allows them to set their innovation strategies in expectation of upcoming
deregulation.
In addition to acquiring information, we expect that firms, especially new entrants, actively lobby
for deregulation. The incentive to lobby should be particularly high in industries with positive producer
rents that are shielded by entry regulation. In those industries, new entrants have the most to gain if entry
barriers are removed and, therefore, ought to lobby aggressively for deregulation. If politically active
firms expect their lobbying efforts to be successful, their innovation strategies are set in expectation of
upcoming deregulation.
26
To analyze whether firm innovation predicts legislative changes, we focus on firms that are new
innovators in soon-to-be-deregulated industries and test whether innovation of those firms predicts future
deregulation. We expect this relation to hold but only for politically active firms. New innovator firms
are defined as follows. Every year and for every firm in our sample, we record all technology classes for
which the firm submits new patent applications. For every new technology class in a particular year, we
track the firm’s patent activity in that class over the remaining sample period. If the duration to the
subsequent patent application is no longer than that of the average firm in the same technology class, we
classify the firm in question as a new innovator in that technology class in the first year of new patent
activity. The explanatory variable in our tests below, New innovatort, is the sum of new innovator firms
in each technology class in year t scaled by the total number of firms with patent activity in that
technology class in year t-1. The technology classes are then mapped into 4-digit SIC codes using the
Silverman (1996, 1999) concordance procedure.
An example may be useful to fix ideas and outline the intuition for our tests. Northrop Grumman
Corp., a politically active firm, which itself operates in the Search, Detection, Navigation, Guidance,
Aeronautical, and Nautical Systems and Instruments industry (SIC 3812), unexpectedly began innovating
in the Railroads, Line-Haul Operating industry (SIC 4011) in 1994. Two years later, the firm continued
its innovation activity in the new industry. Because the Northrop Grumman’s two-year duration between
successive patent applications was shorter than the industry average of 3.43 years, we define Northrop
Grumman as a new innovator in the railroad industry in 1994. The Northrop Grumman’s timing of
innovation in the railroad industry is particularly noteworthy given that the following year Congress
passed the ICC Termination Act that abolished the Interstate Commerce Commission and fully
deregulated the railroad industry. In our view, the Northrop Grumman’s strategy to start innovating in the
railroad industry just prior to its full deregulation is quite suggestive and consistent on the anecdotal level
with the hypothesis that political contributions are valuable because they allow firms to set their
innovation strategies in anticipation of future legislative changes.
In table 8, panel B, we build on the above example and analyze whether firm innovation,
especially that of politically active firms, systematically predicts future legislative changes. We estimate
industry-year probit models that relate the year of industry deregulation to the inflow of new innovator
firms into each industry during the prior year. The baseline results in model 1 show no relation between
the inflow of new innovators into an industry and the likelihood of its subsequent deregulation. However,
when we split the sample of new innovator firms into those that are politically active and inactive, the
results in model 2 show that the inflow of new innovator politically active firms strongly predicts future
deregulation. In terms of economic magnitudes, an interquartile increase in the inflow of politically
active new innovators increases the probability of deregulation by 6.84%. In contrast, the inflow of new
innovators that are not politically active does not predict future deregulation. These results are consistent
27
with our hypothesis and show that changes in innovation of politically active firms predicts future
legislative changes.
In models 3 – 6, we split the sample of deregulatory initiatives into those passed by Congress and
those adopted by the Executive Order. We hypothesize that politically active firms have significant
advantage predicting legislative changes adopted by Congress (because of a direct connection to members
of Congress) compared to those adopted by the Executive Branch. Consistent with this, we find that
changes in innovation of politically active firms strongly predict deregulatory Congressional Acts (models
3 and 4) but have little ability to predict deregulatory Executive Orders (models 5 and 6). In economic
terms, an interquartile increase in the inflow of politically active new innovators increases the probability
of a deregulatory Congressional Act by 10.21%. Compared to the unconditional probability of
deregulation of 21%, these results are economically significant. In a series of robustness tests, we repeat
the analysis with different lags between firm innovation and subsequent deregulation. We find that
changes in innovation of politically active firms predict industry deregulation, especially deregulation
adopted by Congress, two and three years in the future.
In sum, the results in table 8 show that firm innovation, especially by politically active firms,
strongly predicts future legislative changes. This result is consistent with the view that political
contributions allow firms to acquire information that is relevant for innovation decisions and with the
view that political contributions allow firms to lobby for legislation that impacts returns to innovation.
Both effects are consistent with our hypothesis that political activism facilitates investment in innovation
because it helps reduce policy uncertainty.
5. Conclusion
In this paper, we show that firm political activism positively affects investment in innovation.
Using data on political contributions over the period 1979 – 2004, we show a positive and economically
significant correlation between different measures of firm political activism and innovation. In a series of
further empirical tests, we establish a casual positive effect of political activism on innovation. These
results are consistent with the hypothesis that political activism allows firms to reduce policy uncertainty,
which, in turn, stimulates firm investment, specifically investment in innovation. We also find that
politically active firms are able to successfully time future legislation and set their innovation strategies in
expectation of future legislative changes. These results are consistent with the negative relation between
political activism and policy uncertainty. Overall, the results add significantly to our understanding of the
sources of value from political activism and indicate that political activism is an important determinant of
firm innovation.
An important question that arises at this point concerns the impact of political activism on
aggregate investment. Our results indicate that political activism encourages innovation of politically
28
active firms. Whether this translates into higher aggregate investment depends on how investment of
politically active firms interacts with that of inactive firms. If the former does not completely crowd out
the latter, political activism positively affects aggregate investment. This is clearly an important question
but it falls outside the scope of this study. We hope that future work will address this question in detail.
The results will be important for our understanding of the effects of political activism on aggregate
investment, competitive advantage and long-term economic growth.
29
References
Abel, A. B. and J. C. Eberly, 1996. Optimal investment with costly reversibility. Review of Economic
Studies 63, 581-593.
Acharya, V. V. and K. Subramanian, 2009. Bankruptcy codes and innovation. Review of Financial
Studies 22, 4949-4988.
Acharya, V. V., R. Baghai, and K. Subramanian, 2009. Labor laws and innovation. Working paper.
Aggarwal, Rajesh, Felix Meschke, and Tracy Wang, 2012, Corporate Political Donations: Investment or
Agency, Business and Politics 14.
Aghion, P. and J. Tirole, 1994. On the management of innovation, Quarterly Journal of Economics 109,
1185-1209.
Aghion, P. N. Bloom, R. Blundell, R. Griffith, and P. Howitt, 2005. Competition and innovation: An
inverted U relationship. Quarterly Journal of Economics 120, 701-728.
Aghion, P., J. Van Reenen, and L. Zingales, 2013. Innovation and institutional ownership. American
Economic Review 103, 277-304.
Akey, P., 2014. Valuing Changes in Political Networks: Evidence from Campaign Contributions to Close
Congressional Elections. Working paper.
Aldrich, J. H. and D. W. Rohde, 1997. The transition to Republican rule in the House: Implications for
theories of Congressional politics. Political Science Quarterly 112, 541-567.
Alesina, A. and R. Perotti, 1996. Income distribution, political instability, and investment. European
Economic Review 40, 1203-1228.
Amore, M. D. and M. Bennedsen 2013. The value of local political connections in a low-corruption
environment. Journal of Financial Economics 110, 387-402.
Ansolabehere, S. and J. M. Snyder Jr., 1999. Money and institutional power. Texas Law Review 77,
1673-1704.
Ansolabehere, Stephen, J. M. de Figueiredo, and J. M. Snyder Jr., 2003, Why is there so little money in
U.S. Politics?, Journal of Economic Perspectives 17, 105-130.
Ansolabehere, J. M. Snyder Jr., and M. Tripathi, 2002. Are PAC contributions and lobbying linked?
Business and Politics 4.
Armour, J. and D. J. Cumming, 2008. Bankruptcy law and entrepreneurship. American Law and
Economics Review 10, 303-350.
Atanassov, J., 2013. Do hostile takeovers stifle innovation? Evidence from antitakeover legislation and
corporate patenting. Journal of Finance 68, 1097-1131.
Barro, R. J., 1991. Economic growth in a cross-section of countries. Quarterly Journal of Economics
106, 407-443.
Bernanke, B., 1983. Irreversibility, uncertainty, and cyclical investment. Quarterly Journal of Economics
98, 85-106.
30
Bertola, G. and R. Caballero, 1994. Irreversibility and aggregate investment. Review of Economic
Studies 61, 223-246.
Bhattacharya, U., P. Hsu, Tian, H., and Y. Xu, 2014. What affects innovation more: policy or policy
uncertainty? Working paper.
Bloom, N., S. Bond, and J. Van Reenen, 2007. Uncertainty and investment dynamics. Review of
Economic Studies 74, 391-415.
Bombardini, M. and Trebbi, F., 2008. Votes or money? Theory and evidence from the US Congress.
Working paper.
Briffault, R., 2011. Corporations, corruption, and complexity: campaign finance after Citizens United.
Symposium: Citizens United V. Federal Election Commission: Implications for the American Electoral
Process.
Caballero, R., and E. Engel, 1999. Explaining investment dynamics in U.S. manufacturing: a generalized
(S,s) approach. Econometrica 67, 783-826.
Claessems, S., E. Feijen and L. Laeven, 2008. Political connections and preferential access to finance:
the role of campaign contributions. Journal of Financial Economics 88, 554-580.
Coate, S., 2004. Pareto improving campaign finance policy. American Economic Review 94, 628-655.
Coates, J. C., 2012. Corporate politics, governance, and value before and after Citizens United. Journal
of Empirical Legal Studies 9, 657-696.
Cooper, M. J., H. Gulen, and A. V. Ovtchinnikov, 2010. Corporate political contributions and stock
returns. Journal of Finance 65, 687-724.
Cornaggia, J., Y. Mao, X. Tian, and B. Wolfe, 2014, Does banking competition affect innovation?
Journal of Financial Economics, forthcoming.
Duchin, R. and D. Sosyura, 2012. The politics of government investment. Journal of Financial
Economics 106, 24-48.
Edwards, K. and C. Stewart, III, 2006. The value of committee assignments in Congress since 1994.
Working paper.
Faccio, M., 2006, Politically connected firms, American Economic Review 96, 369-386.
Faccio, M., R. W. Masulis, and J. J. McConnell, 2006. Political connections and corporate bailouts.
Journal of Finance 61, 2597–2635.
Faccio, M. and D. C. Parsley, 2009. Sudden deaths: taking stock of geographic ties, Journal of Financial
and Quantitative Analysis 44, 683-718.
Fama E. F. and J. MacBeth, 1973. Risk, return, and equilibrium: empirical tests. Journal of Political
Economy 81, 607-636.
Fan, W. and M. White, 2003. Personal bankruptcy and the level of entrepreneurial activity. Journal of
Law and Economics 46, 543-568.
Fang, V., X. Tian, and S. Tice, 2014. Does stock liquidity enhance or impede firm innovation? Journal
of Finance, forthcoming.
31
Ferguson T. and H. Voth, 2008. Betting on Hitler – the value of political connections in Nazi Germany.
Quarterly Journal of Economics 123, 101-137.
Ferreira, D., G. Manso, and A. Silva, 2014. Incentives to innovate and the decision to go public or
private. Review of Financial Studies 27, 256-300.
Fisman, R., 2001. Estimating the value of political connections. American Economic Review 91, 1095-
1102.
Goldman, E., J. Rocholl, and J. So, 2009. Do politically connected boards affect firm value? Review of
Financial Studies 22, 2331-2360.
Goldman, E., J. Rocholl, and J. So, 2013. Political connections and the allocation of procurement
contracts. Review of Finance 13, 1617-1648.
Grier, K. B. and M. C. Munger, 1991. Committee assignments, constituent preferences, and campaign
contributions. Economic Inquiry 29, 24-43.
Grier, K. B., M.C. Munger, and B. E. Roberts, 1994. The determinants of industry political activity,
1978-1986. American Political Science Review 88, 911-926.
Grilliches, Z., A. Pakes, and B. Hall, 1988. The value of patents as indicators of inventive activity.
Working paper.
Hall, B., A. Jaffe, and M. Trajtenberg, 2001. The NBER patent citation data file: lessons, insights and
methodological tools. Working paper.
Hall, B., A. Jaffe, and M. Trajtenberg, 2005. Market value and patent citations. The RAND Journal of
Economics 36, 16-38.
Hart, D. M., 2001. Why do some firms give? Why do some give a lot?: High-tech PACs, 1977-1996.
Journal of Politics 63, 1230-1249.
Hassler, J. A., 1996. Variations in risk and fluctuations in demand: a theoretical model. Journal of
Economic Dynamics and Control 20, 1115-1143.
He, J., and X. Tian, 2013. The dark side of analyst coverage: the case of innovation. Journal of Financial
Economics 109, 856-878.
Heckman, J. J., 1979. Sample selection bias as a specification error. Econometrica 47, 153-161.
Hirshleifer, D., A. Low, and S. H. Teoh, 2012. Are overconfident CEOs better innovators? Journal of
Finance 67, 1457-1498.
Hojnacki, M. and D. C. Kimball, 2001. PAC contributions and lobbying access in Congressional
committees. Political Research Quarterly 54, 161-180.
Holmstrom, B., 1989. Agency costs and innovation. Journal of Economic Behavior & Organization 12,
305–327.
Jacobson, G. C., 1980. Money in Congressional elections. New Haven: Yale University Press.
Jacobson, G. C., 1985. Money and votes reconsidered: Congressional elections 1972-1982. Public
Choice 47, 7-62.
32
Johnson, S. and T. Mitton, 2003. Cronyism and capital controls: evidence from Malaysia. Journal of
Financial Economics 67, 351-382.
Jones, G. and J. Mckenzie, 2014. Special report: how Italy became a submerging economy. Reuters.
Julio, B. and Y. Yook. 2012. Political uncertainty and corporate investment cycles. Journal of Finance
67, 45-83. Julio, B. and Y. Yook. 2014. Policy uncertainty, irreversibility, and cross-border flows of capital.
Working paper.
Kau, J. B., D. Keenan, and P. H. Rubin, 1982. A general equilibrium model of Congressional voting.
Quarterly Journal of Economics 97, 271-293.
Kortum, S. and J. Lerner, 2000. Assessing the contribution of venture capital to innovation. RAND
Journal of Economics 31, 674-692.
Kroszner, R. S. and T. Stratmann, 1998. Interest-group competition and the organization of Congress:
theory and evidence from financial services’ political action committees. American Economic Review 88,
1163-1187.
Kroszner, R. S. and T. Stratmann, 2000. Congressional committees as reputation-building mechanisms:
repeat PAC giving and seniority on the house banking committee. Business and Politics 2, 35-52.
Kroszner, R. S. and T. Stratmann, 2005. Corporate campaign contributions, repeat giving, and the
rewards to legislator reputation. Journal of Law and Economics 48, 41-71.
Masters, M. F. and G. D. Keim, 1985. Determinants of PAC participation among large corporations.
Journal of Politics 47, 1158-1173.
Milyo, J., D. Primo, and T. Groseclose 2000. Corporate PAC campaign contributions in perspective.
Business and Politics 2, 75-88.
Ovtchinnikov, A. V., 2010. Capital structure decisions: evidence from deregulated industries. Journal of
Financial Economics 95, 249-274.
Ovtchinnikov, A. V., 2013. Merger waves following industry deregulation. Journal of Corporate
Finance 21, 51-76.
Ovtchinnikov, A. V. and E. Pantaleoni, 2012. Individual political contributions and firm performance.
Journal of Financial Economics 105, 367-392.
Pastor, L. and P. Veronesi, 2012. Uncertainty about government policy and stock prices. Journal of
Finance 64, 1219-1264.
Petersen, M., 2009. Estimating standard errors in finance panel sets: comparing approaches. Review of
Financial Studies 22, 435-480.
Pindyck, R. S. and A. Solimano, 1993. Economic instability and aggregate investment. NBER
Macroeconomics Manual 8, 259-318.
Poole, K. and T. Romer, 1985. Patterns of political action committee contributions to the 1980
campaigns for the U.S. House of Representatives. Public Choice 47, 63-112.
33
Romer, P. 1987. Growth based on increasing returns due to specialization. American Economic Review
77, 56-62.
Romer, P. 1990. Endogenous technological change. Journal of Political Economy 98, 71-102.
Romer, T. and J. M. Snyder Jr., 1994. An empirical investigation of the dynamics of PAC contributions.
American Journal of Political Science 38, 745-769.
Seru, A., 2014. Firm boundaries matter: evidence from conglomerates and R&D activity. Journal of
Financial Economics 111, 381-405.
Snyder, J. M. Jr., 1992 Are contributors rational? Untangling strategies of Political Action Committees.
Journal of Political Economy 10, 647-664.
Solow, R. M., 1957. Technical change and the aggregate production function. Review of Economics and
Statistics 39, 312-320.
Spiegel, M. and H. Tookes, 2008. Dynamic competition, innovation and strategic financing. Working
paper.
Stratmann, T., 1991. What do campaign contributions buy? Deciphering causal effects of money and
votes. Southern Economic Journal 57, 606-620.
Stratmann, T., 1998. The market for congressional votes: is timing of contributions everything? Journal
of Law and Economics 41, 85-113.
Stratmann, T., 2002. Can special interests buy congressional votes? Evidence from financial services
legislation. Journal of Law and Economics 45, 345-374.
Stratmann, T., 2005. Some talk: money on politics. A (partial) review of the literature. Public Choice
124, 135-156. Tahoun, Ahmed, 2014, The role of stock ownership by U.S. members of Congress on the market for
political favors, Journal of Financial Economics 111, 86-110.
Tian, X. and T. Wang, 2013. Tolerance for failure and corporate innovation. Review of Financial
Studies, forthcoming.
Wiley, R. E., 1988. “Political” influence at the FCC. Duke Law Journal 1988, 280-285.
Winston, C., 1993. Economic deregulation: days of reckoning for microeconomists. Journal of
Economic Literature 31, 1263-1289.
Wright, J. R., 1990. Contributions, lobbying, and committee voting in the U.S. House of Representatives.
The American Political Science Review 84, 417-438.
Zardkoohi, A., 1985. On the political participation of the firm in the electoral process. Southern
Economic Journal 51, 804-817.
34
Appendix A
Variables construction
Variable Definition Source
Innovation measures
Cpatent[t+1, t+3] Natural logarithm of one plus firm i’s total number of citations per patent applied for
in years t+1, t+2 and t+3. The citations of each patent are adjusted for truncation
bias by scaling citations of a given patent by the mean number of citations received
by all patents in that year in the same technological class as the patent.
NBER
Cpatentt+1 Natural logarithm of one plus firm i’s total number of citations per patent applied for
in year t+1, adjusted for truncation bias.
NBER
Effpatent[t+1, t+3] Natural logarithm of one plus firm i’s total number of patents in application years
t+1, t+2 and t+3. Each patent is adjusted for truncation bias, which is corrected by
dividing each patent for each firm-year by the mean number of patents of all firms
for that year in the same technology class as the patent.
NBER
Effpatentt+1 Natural logarithm of one plus firm i’s total number of patents in application year t+1,
adjusted for truncation bias.
NBER
Control variables
R&DAssets R&D expenditure (XRD) divided by book value of total assets (AT) measured at the
end of fiscal year, set to 0 if missing. We adjust for inflation.
COMPUSTAT
ROA Operating income before depreciation (OIBDP) divided by book value of total assets
(AT), measured at the end of fiscal year. We adjust for inflation.
COMPUSTAT
PPEAssets Property, Plant & Equip (PPENT) divided by book value of total assets (AT),
measured at the end of fiscal year. We adjust for inflation.
COMPUSTAT
Leverage Book value of debt (DLTT+DLC) divided by book value of total assets (AT),
measured at the end of fiscal year. We adjust for inflation.
COMPUSTAT
CapexAssets Capital expenditure (CAPX) divided by book value of total assets (AT), measured at
the end of fiscal year. We adjust for inflation and set it to 0 if missing.
COMPUSTAT
HI Herfindahl index in year t constructed based on sales at the 4-digit SIC. COMPUSTAT
HISquare HI x HI COMPUSTAT
TobinQ A firm's market-to-book ratio at the end of fiscal year, calculated as book value of
total assets plus market value of equity minus book value of equity minus balance
sheet deferred taxes (txdb, set to 0 if missing) divided by book value of total assets,
i.e. (AT+CSHOcPRCC_F-CEQ-txdb)/AT.
COMPUSTAT
Size Natural logarithm of book value of total assets (AT) measured at the end of fiscal
year.
COMPUSTAT
Age Firm age. It equals the number of years since its first appearance on the CRSP
database.
CRSP
35
Appendix B
Determinants of firms’ political activity, 1984–2004: First-stage probit model
We use individual firm and industry characteristics shown in the prior literature to be related to the
likelihood of a firm having a political action committee (PAC) to analyze the determinants of firms’
political activity. Relevant prior studies include Masters and Keim (1985), Zardkoohi (1985), Grier,
Munger, and Roberts (1994), Hart (2001). We use firm size, leverage, sales, cash flows, the number of
employees, percentage of industry employees that are unionized, the number of business and geographical
segments, sales concentration, market share, a regulated industry indicator, the amount of government
purchases from an industry, and the number of politically active firms in an industry as the determinants
of firms’ political activity. Consistent with prior studies, we find that larger and more levered firms, with
more sales, more employees, and a higher percentage of unionized employees are more likely to be
politically active. Also, there is an increased likelihood of political activism as the number of business
segments and market share increases and book-to-market and cash flows decrease. Finally, the likelihood
of political activism is positively related to government purchases and to the regulated industry indicator
and negatively related to sales concentration, the number of geographic segments, and the number of
politically active industry firms.
Variable
Probit Model
(1 = active; 0 = not active)
Coefficient
Ln(size) 0.187
(0.005)
a
Ln(sales) 0.136
(0.007)
a
Ln(employees) 0.215
(0.007)
a
No. business segments 0.029
(0.004)
a
No. geographic segments -0.062
(0.005)
a
BM -0.000
(0.000)
a
Leverage 0.274
(0.025)
a
CF -0.026
(0.010)
b
Market share 1.840
(0.312)
a
(Market share)2 -3.206
0.645
a
Herfindahl index -0.342
(0.104)
a
Regulation indicator 0.773
(0.017)
a
Government purchases 0.848
(0.102)
a
No. politically active firms -0.000
(0.000)
a
Pct. employees unionized 2.014
(0.062)
a
Log likelihood -30,662
N 143,274
36
Figure 1
Corporate innovation surrounding the 1994 midterm Congressional election
The political contributions data are from the FEC detailed contribution files. We exclude all noncorporate
contributions, contributions from private firms and subsidiaries of foreign firms, as well as contributions from firms
with insufficient data on CRSP/Compustat. The final sample consists of 813,692 political contributions to 5,584 unique
political candidates made by 1,805 unique firms. This figure presents the time-series dynamics of patent applications of
different firms surrounding the 1994 midterm Congressional election. The solid line tracks the treatment firms. The
dashed line tracks the control firms without chair contributions. The dashed line with a triangular marker tracks the
control firms with chair contributions. The shaded area designates the November 1994 election. The portfolio of
treatment firms comprises 126 firms that supported four junior politicians who unexpectedly became committee chairs
in November 1994. The portfolio of control firms with chair contributions comprises 475 firms that supported one or
more existing committee chairs during the 1990 – 1994 period. The portfolio of control firms without chair
contributions comprises 197 firms that did not support existing committee chairs during the 1990 – 1994 period.
0
1
2
3
4
5
6
7
8
9
1991
1992
1993
1994
1995
1996
1997
1998
1999
Num
ber
of
pat
ents
Year
37
Table 1
Characteristics of politically active and inactive firms, 1979 – 2004
This table presents formation period summary statistics for various firm characteristics of politically active and politically inactive firms. The subsample of
politically active firms consists of 1,805 unique firms with an established political action committee (PAC) and political contributions data from the Federal Election
Commission (FEC) for the period January 1, 1979 – December 31, 2004. The subsample of politically inactive firms consists of all other firms in the
CRSP/Compustat merged database with nonmissing values of the firm characteristics in this table. The numbers in each cell are time-series averages of yearly cross-
sectional median statistics. N is the average annual number of firms in each subsample. t-test is the t-statistic computed under the null hypothesis that the difference
between politically active and inactive firms is zero. Panel A reports characteristics of politically active and inactive firms. Panel B reports characteristics of
politically active and inactive firms based on NYSE annually ranked size decile breakpoints. All numbers, except Size, Age, Patents, and Patent citations are in
decimal form, i.e. 0.01 is 1%. All variables are fully defined in Appendix A.
38
Table 1
Firms Size Age ROA PPE / Assets Lev Q Herfindahl R&D / Assets CAPX / Assets Patents Patent citations N
Panel A: Comparison of politically active and inactive firms
Politically active 1.289 22.346 0.118 0.338 0.261 1.119 0.154 0.012 0.052 6.106 0.309 740
Inactive 0.068 7.000 0.085 0.200 0.190 1.268 0.193 0.038 0.041 0.752 0.171 5405
t-test (difference) 8.14 21.38 6.99 8.13 8.75 -4.47 -4.96 -11.04 2.90 19.07 9.79
Panel B:Comparison of politically active and inactive firms by size deciles
Small Active 0.030 10.404 0.074 0.232 0.289 1.018 0.203 0.007 0.033 0.320 0.111 29
Inactive 0.018 6.240 0.049 0.188 0.202 1.149 0.204 0.041 0.034 0.193 0.094 2665
t-test 4.95 6.36 2.99 1.65 5.07 -5.42 -0.03 -11.63 -0.14 1.29 0.94
Decile 2 0.080 14.115 0.084 0.235 0.258 1.031 0.195 0.008 0.032 0.560 0.167 30
0.076 6.780 0.089 0.191 0.171 1.269 0.191 0.043 0.042 0.335 0.197 688
0.50 9.30 -0.60 2.18 5.39 -6.77 0.29 -10.48 -2.75 1.66 -1.42
Decile 3 0.136 13.077 0.091 0.249 0.265 1.050 0.172 0.007 0.036 0.525 0.176 35
0.134 7.080 0.104 0.191 0.164 1.326 0.190 0.038 0.044 0.714 0.217 490
0.15 7.67 -1.94 3.17 8.04 -7.27 -1.97 -9.97 -2.08 -1.36 -1.31
Decile 4 0.216 14.442 0.098 0.291 0.246 1.061 0.144 0.008 0.043 0.594 0.192 46
0.217 7.560 0.113 0.194 0.169 1.355 0.177 0.034 0.045 0.960 0.223 401
-0.02 7.29 -2.50 6.36 7.10 -8.78 -3.42 -9.17 -0.69 -2.68 -1.60
Decile 5 0.337 17.423 0.105 0.317 0.255 1.067 0.149 0.008 0.047 1.218 0.179 57
0.34 8.380 0.124 0.209 0.176 1.432 0.178 0.029 0.048 1.079 0.265 333
0.038 9.12 -3.38 6.36 7.98 -9.92 -3.30 -8.37 -0.24 0.42 -4.21
Decile 6 0.526 18.904 0.112 0.344 0.270 1.093 0.130 0.008 0.049 0.928 0.190 70
0.52 9.760 0.130 0.241 0.179 1.458 0.173 0.026 0.052 1.198 0.266 264
0.030 8.84 -3.20 5.04 10.17 -8.97 -5.79 -7.20 -0.69 -1.48 -3.27
Decile 7 0.832 21.481 0.118 0.397 0.276 1.098 0.145 0.008 0.053 2.011 0.248 84
0.84 11.260 0.140 0.268 0.201 1.503 0.173 0.024 0.057 1.562 0.310 213
-0.043 9.92 -4.74 5.92 8.86 -10.28 -3.38 -9.62 -0.83 1.67 -2.71
Decile 8 1.411 24.577 0.119 0.387 0.270 1.122 0.149 0.011 0.054 2.968 0.281 104
1.40 13.460 0.145 0.273 0.195 1.542 0.171 0.023 0.058 2.594 0.333 166
0.067 10.82 -5.17 4.41 9.26 -7.80 -2.54 -8.15 -0.69 0.98 -2.22
Decile 9 2.850 29.519 0.122 0.379 0.273 1.137 0.145 0.012 0.058 5.192 0.356 125
2.73 17.000 0.154 0.267 0.191 1.697 0.171 0.023 0.059 3.158 0.378 121
0.335 12.28 -6.28 4.71 14.10 -8.11 -3.40 -6.59 -0.09 4.84 -0.78
Big 9.433 36.769 0.146 0.345 0.245 1.300 0.166 0.022 0.064 20.947 0.559 160
7.326 22.100 0.162 0.289 0.184 1.861 0.189 0.028 0.065 14.835 0.507 64
1.46 10.66 -2.50 2.07 8.33 -4.82 -2.47 -3.17 -0.16 4.21 1.38
39
Table 2
Measures of firm political activism, 1984 – 2004
This table presents data from the FEC detailed contribution files on political contributions to House, Senate, and Presidential elections. We exclude all noncorporate
contributions, contributions from private firms and subsidiaries of foreign firms, as well as contributions from firms with insufficient data on CRSP/Compustat. The final
sample consists of 813,692 political contributions to 5,584 unique political candidates made by 1,805 unique firms. Individual contributions are combined into ten firm-year
level measures of political activism as described in Eqs. 1 – 6 in section 2.2. Panel A presents the descriptive statistics for each political activism measure. Total in row 1
equals Pcand (Eq. 1) in columns 1 – 6 and Camount (Eq. 2) in columns 7 – 12. Committee in row 2 equals PcandCommittee (Eq. 3) in columns 1 – 6 and CamountCommittee (Eq.
4) in columns 7 – 12. Non-committee in row 3 equals PcandNon-Committee (defined in section 2.2) in columns 1 – 6 and CamountNon-Committee (defined in section 2.2) in column 7
– 12. ΔY+ in row 4 equals ΔPcand+ (Eq. 5) in columns 1 – 6 and ΔCamount+ (Eq. 6) in columns 7 – 12. ΔY- in row 5 equals ΔPcand- (defined in section 2.2) in columns 1 –
6 and ΔCamount- (defined in section 2.2) in columns 7 – 12. Panel B presents election cycle firm political contribution totals for each Congressional committee in our
sample. The FEC data on political contributions is for the period January 1, 1979 – December 31, 2004. Since we require five years of data to compute each measure of
political activism, the variables are computed at the end of October of each year, from 1984 to 2004, resulting in 16,065 firm-year observations for each variable.
Pcand Camount
Variable Mean Min 25th Per Median 75th Per Max Mean Min 25th Per Median 75th Per Max
Panel A: Measures of political activism
Total 84.102 1 14 39 117 844 $226,050 0 $20,948 $73,135 $238,656 $6,293,663
Committee 13.483 0 0 3 18 172 44,082 0 0 5,339 35,911 1,574,725
Non-Committee 70.619 0 11 32 97 791 181,968 0 16,142 58,342 192,408 5,354,005
ΔY+ 0.422 0 0 0 0 15 1,126 0 0 0 0 88,615
ΔY- 0.367 0 0 0 0 15 1,306 0 0 0 0 123,913
Panel B: Political contributions to members of Congressional committees
House Congressional committees
Energy & Commerce 7.623 1 2 4 11 51 $13,578 0 $1,593 $4,787 $14,908 $518,973
Financial Services 5.640 1 1 3 6 62 10,637 0 1,284 3,346 9,436 380,510
Transportation 5.942 1 1 3 7 66 10,511 0 1,399 3,910 10,751 393,013
Armed Services 5.462 1 1 3 6 57 10,216 0 1,292 3,305 9,031 327,122
Agriculture 4.930 1 1 3 6 48 8,423 0 1,292 3,264 8,942 325,162
Natural Resources 4.622 1 1 3 6 40 7,886 0 1,272 3,381 9,019 196,176
Merchant Marine 4.915 1 1 3 6 42 7,335 0 1,295 3,328 8,267 173,512
Senate Congressional committees
Commerce 3.241 1 1 2 4 17 $9,479 0 $1,753 $4,726 $12,044 $133,323
Banking 3.165 1 1 2 4 17 9,008 0 1,673 4,457 10,950 109,078
Agriculture 3.138 1 1 2 4 16 8,945 0 1,747 4,501 11,581 108,511
Armed Services 3.042 1 1 2 4 20 8,872 0 1,673 4,185 10,742 139,483
Energy 3.211 1 1 2 4 17 8,639 0 1,587 4,333 10,987 126,395
Environment 2.717 1 1 2 4 15 7,409 0 1,523 3,875 9,597 97,661
40
Table 3
Political contributions and corporate innovation, 1984 – 2004
The political contributions data are from the FEC detailed contribution files. We exclude all noncorporate contributions, contributions
from private firms and subsidiaries of foreign firms, as well as contributions from firms with insufficient data on CRSP/Compustat. The
final sample consists of 813,692 political contributions to 5,584 unique political candidates made by 1,805 unique firms. This table
presents the results of estimating Eq. 7 on the sample of politically active and inactive firms for the period November 1984 – October
2004. Models 1, 2, 5, and 6 present the results for the all firms with available data on CRSP/Compustat and include 123,531 firm-year
observations. Models 3, 4, 7, and 8 present the results for politically active firms only and include 16,065 firm-year observations. We
control for firm self-selection in models 3, 4, 7, and 8 by including the inverse Mills ratio (IMR) from the first-stage probit model of
whether a firm has an established PAC on determinants of PAC participation. The first stage probit results are presented in Appendix B.
Panel A presents the results for the total number of patent applications. Panel B presents the results for the total number of patent
citations. All regressions include industry and year fixed effects. All variables are defined in section 2.2 and Appendix A. Standard
errors are in parentheses and are adjusted for heteroskedasticity and correlated within firms. R2 is the adjusted R-squared in the
regression. R2 controls only is the adjusted R-squared from the model that includes only the control variables but does not include
measures of political activism. N is the total number of firm-year observations. a, b, c, indicates significance at the 1%, 5%, and 10%
level, respectively.
Effpatentt+1 Effpatentt+3
Variable 1 2 3 4 5 6 7 8
Panel A: Number of patents
Pcand/102 0.276 a 0.162 a 0.331 a 0.202 a
(0.036) (0.038) (0.042) (0.045)
Camount/106 0.602 a 0.298 a 0.712 a 0.369 a
(0.099) (0.093) (0.117) (0.107)
R&D / Assets 0.689 a 0.692 a 11.197 a 11.226 a 1.139 a 1.142 a 14.074 a 14.111 a
(0.055) (0.056) (1.791) (1.802) (0.077) (0.078) (2.077) (2.087)
ROA 0.120 a 0.100 a 0.711 a 0.710 a 0.246 a 0.222 a 1.003 a 1.003 a
(0.019) (0.019) (0.266) (0.270) (0.026) (0.026) (0.335) (0.340)
PPE / Assets -0.054 b -0.042 c -0.283 b -0.242 c -0.067 c -0.052 -0.346 b -0.294 c
(0.026) (0.026) (0.128) (0.127) (0.034) (0.034) (0.161) (0.160)
Leverage -0.132 a -0.140 a -0.159 c -0.166 c -0.222 a -0.231 a -0.221 b -0.229 b
(0.017) (0.017) (0.086) (0.087) (0.022) (0.022) (0.106) (0.107)
Capex / Assets 0.394 a 0.382 a 0.511 c 0.475 0.556 a 0.542 a 0.611 c 0.566
(0.045) (0.046) (0.292) (0.291) (0.061) (0.061) (0.366) (0.366)
HI 0.342 a 0.365 a 0.212 0.333 0.527 a 0.555 a 0.497 0.649
(0.098) (0.099) (0.391) (0.391) (0.124) (0.125) (0.467) (0.467)
HI2 -0.235 c -0.254 b 0.208 0.081 -0.399 b -0.422 a -0.034 -0.192
(0.125) (0.127) (0.471) (0.474) (0.157) (0.159) (0.551) (0.554)
Q 0.026 a 0.027 a 0.030 0.033 0.041 a 0.043 a 0.037 0.040
(0.002) (0.002) (0.022) (0.022) (0.003) (0.003) (0.027) (0.027)
Assets 0.098 a 0.107 a 0.127 a 0.145 a 0.138 a 0.149 a 0.133 a 0.155 a
(0.006) (0.005) (0.027) (0.027) (0.007) (0.007) (0.033) (0.034)
LnAge 0.050 a 0.060 a 0.088 a 0.101 a 0.071 a 0.083 a 0.111 a 0.126 a
(0.006) (0.006) (0.023) (0.023) (0.008) (0.008) (0.030) (0.030)
IMR -0.127 c -0.149 b -0.224 a -0.252 a
(0.065) (0.065) (0.084) (0.084)
R2 0.326 0.316 0.574 0.570 0.357 0.349 0.597 0.594
R2 controls only 0.296 0.296 0.563 0.563 0.333 0.333 0.587 0.587
N 123,531 123,531 16,065 16,065 123,531 123,531 16,065 16,065
41
Table 3 – continued
Cpatentt+1 Cpatentt+3
Variable 1 2 3 4 5 6 7 8
Panel B: Patent citations
Pcand/102 0.035 a 0.028 a 0.070 a
0.060 a
(0.006) (0.007) (0.011)
(0.014)
Camount/106 0.075 a 0.053 a 0.148 a 0.109 a
(0.016) (0.017) (0.030) (0.032)
R&D / Assets 0.500 a 0.500 a 2.270 a 2.274 a 0.937 a 0.938 a 3.952 a 3.963 a
(0.029) (0.029) (0.292) (0.291) (0.053) (0.053) (0.529) (0.528)
ROA 0.097 a 0.094 a 0.169 a 0.169 a 0.210 a 0.205 a 0.287 b 0.287 b
(0.008) (0.008) (0.062) (0.063) (0.016) (0.016) (0.121) (0.123)
PPE / Assets -0.022 a -0.021 b -0.063 b -0.056 c -0.037 b -0.034 b -0.103 -0.087
(0.008) (0.008) (0.032) (0.032) (0.016) (0.016) (0.064) (0.065)
Leverage -0.068 a -0.069 a -0.028 -0.030 -0.145 a -0.147 a -0.075 -0.078 c
(0.006) (0.006) (0.025) (0.025) (0.012) (0.012) (0.047) (0.047)
Capex / Assets 0.141 a 0.140 a 0.025 0.019 0.272 a 0.269 a 0.131 0.118
(0.017) (0.017) (0.077) (0.077) (0.032) (0.032) (0.154) (0.155)
HI 0.108 a 0.111 a 0.200 b 0.220 a 0.219 a 0.225 a 0.490 a 0.535 a
(0.027) (0.027) (0.085) (0.085) (0.053) (0.053) (0.165) (0.165)
HI2 -0.094 b -0.096 a -0.128 -0.149 -0.185 a -0.190 a -0.364 c -0.411 b
(0.033) (0.033) (0.098) (0.098) (0.063) (0.064) (0.189) (0.189)
Q 0.014 a 0.014 a 0.007 0.007 0.027 a 0.027 a 0.016 0.017
(0.001) (0.001) (0.005) (0.005) (0.002) (0.002) (0.011) (0.011)
Assets 0.030 a 0.032 a 0.007 0.010 0.057 a 0.060 a 0.009 0.016
(0.001) (0.001) (0.006) (0.006) (0.002) (0.002) (0.012) (0.012)
LnAge 0.011 a 0.013 a 0.019 a 0.021 a 0.024 a 0.027 a 0.032 a 0.037 a
(0.002) (0.002)
(0.006) (0.006)
(0.004) (0.004) (0.012) (0.012)
IMR -0.057 a -0.061 a
-0.119 a -0.127 a
(0.017) (0.017)
(0.033) (0.034)
R2 0.203 0.202 0.385 0.384 0.275 0.274 0.464 0.462
R2 controls only 0.201 0.201 0.381 0.381 0.272 0.272 0.458 0.458
N 123,531 123,531 16,065 16,065 123,531 123,531 16,065 16,065
42
Table 4
Political contributions to Congressional Committees and corporate innovation, 1984 – 2004
The political contributions data are from the FEC detailed contribution files. We exclude all noncorporate contributions, contributions
from private firms and subsidiaries of foreign firms, as well as contributions from firms with insufficient data on CRSP/Compustat. The
final sample consists of 813,692 political contributions to 5,584 unique political candidates made by 1,805 unique firms. This table
presents the results of estimating Eq. 7 various subsamples of firms and political contributions measures. Models 1 – 4 present the
results for the total number of patent applications. Models 5 – 8 present the results for the total number of patent citations. All
regressions include industry and year fixed effects, all control variables from table 3 and control for firm self-selection into the
politically active group with the inverse Mills ratio (IMR). Panel A presents the results for political contributions to members of
influential Congressional committees defined in section 2.2. Panel B presents the results for political contributions to members of
outside committees defined in section 2.2. Panel C presents the results for political contributions to members who join or leave
influential committees. Panel D presents the results for political contributions to members who join or leave influential committees
during election and non-election years. Standard errors are in parentheses and are adjusted for heteroskedasticity and correlated within
firms. a, b, c, indicates significance at the 1%, 5%, and 10% level, respectively.
43
Table 4
Effpatentt+1
Effpatentt+3 Cpatentt+1
Cpatentt+3
Variable 1 2
3 4 5 6
7 8
Panel A: Contributions to influential Congressional committees
PcandCommittee/102 0.538 a 0.667 a 0.076 b 0.158 a
(0.148) (0.177) (0.031) (0.061)
CamountCommittee/106 0.972 a 1.238 a 0.209 a 0.413 a
(0.343) (0.412) (0.075) (0.141)
R2 0.560 0.560 0.588 0.589 0.350 0.391 0.465 0.468
N 11,323 11,323 11,323 11,323 11,323 11,323 11,323 11,323
Panel B: Contributions to outside committees
PcandNon-Committee/102 -0.965 c -0.788 -0.036 0.009
(0.524) (0.693) (0.174) (0.313)
CamountNon-Committee/106 -1.267 -0.634 0.184 0.775
(1.243) (1.686) (0.506) (0.871)
R2 0.617 0.613 0.646 0.645 0.413 0.411 0.499 0.499
N 994 994 994 994 994 994 994 994
Panel C: Contributions to politicians joining/leaving influential committees
ΔPcand + 0.068 a 0.085 a 0.011 a 0.023 a
(0.014) (0.017) (0.004) (0.006)
ΔPcand - 0.007 0.002 0.001 -0.004
(0.013) (0.016) (0.004) (0.006)
I (ΔPcand + > ΔPcand -) 0.086 a 0.124 a 0.023 a 0.049 a
(0.027) (0.033) (0.009) (0.015)
I (ΔPcand + < ΔPcand -) -0.049 c -0.071 b 0.007 -0.003
(0.027) (0.034) (0.009) (0.015)
R2 0.559 0.554 0.587 0.583 0.389 0.387 0.465 0.464
N 11,323 11,323 11,323 11,323 11,323 11,323 11,323 11,323
Panel D: Contributions to politicians joining/leaving influential committees during election and off-election years
Election year coefficients
ΔPcand + 0.036 0.059 -0.001 0.014
(0.039) (0.049) (0.010) (0.014)
ΔPcand - 0.041 0.025 0.022 0.015
(0.037) (0.047) (0.015) (0.024)
I (ΔPcand + > ΔPcand -) 0.040 0.082 0.013 0.037
(0.046) (0.057) (0.015) (0.025)
I (ΔPcand + < ΔPcand -) -0.050 -0.098 0.030 0.020
(0.053) (0.069) (0.021) (0.031)
Non-election year coefficients
ΔPcand + 0.071 a 0.088 a 0.012 a 0.024 a
(0.013) (0.016) (0.004) (0.006)
ΔPcand - 0.004 -0.000 -0.000 -0.005
(0.013) (0.016) (0.004) (0.006)
I (ΔPcand + > ΔPcand -) 0.107 a 0.144 a 0.027 a 0.053 a
(0.031) (0.038) (0.010) (0.016)
I (ΔPcand + < ΔPcand -) -0.043 -0.060 0.003 -0.008
(0.031) (0.039) (0.010) (0.017)
R2 0.559 0.554 0.587 0.583 0.389 0.387 0.465 0.464
N 11,323 11,323 11,323 11,323 11,323 11,323 11,323 11,323
44
Table 5
Political contributions and corporate innovation: Instrumental variables estimation, 1984 – 2004
The political contributions data are from the FEC detailed contribution files. We exclude all noncorporate contributions, contributions from private firms and
subsidiaries of foreign firms, as well as contributions from firms with insufficient data on CRSP/Compustat. The final sample consists of 813,692 political
contributions to 5,584 unique political candidates made by 1,805 unique firms. This table presents the results of instrumental variables (IV) estimation. Panel
A presents the results of the first-stage regression that relates a firm’s choice to contribute to a politician to her margin of victory in the prior election. This
regression is estimated separately every year and for every firm in our sample. Average sensitivity is the average coefficient on the politician’s margin of
victory in the prior election. Average R2 is the average R-squared of all 17,873 firm-year first-stage regressions. Panels B and C present the results of the
second stage regressions that relate the first-stage sensitivity to the margin of victory to subsequent innovation growth. We sort all firms into 20 equal- and
value-weighted portfolios based on their estimated sensitivities to the margin of victory in the first-stage regression. We then regress average portfolio
innovation growth on the average sensitivity to the margin of victory. Columns 1 – 3 present the results for equal-weighted portfolios. Columns 4 – 6 present
the results for value-weighted portfolios. In panel B, innovation growth is measured by the growth in the number of patent applications. In panel C, innovation
growth is measured by the growth in the number of patent citations. Standard errors are in parentheses and are adjusted for heteroskedasticity. a, b, c, indicates
significance at the 1%, 5%, and 10% level, respectively.
Panel A: First-stage regression. Tests for relevant instrument
Average
sensitivity Average R2
Total first-stage
estimates
Positive first-stage
estimates
Negative first-stage
estimates
Positive and significant
first-stage estimates
Negative and significant
first-stage estimates
Variable -0.0295 a
(0.004)_
0.004 17,873 6,434 11,439 297 1,024
Panel B: Second stage regression. Dependent variable - number of patents
EW results VW results
Variable Δ Effpatentt+2 Δ Effpatentt+3
Δ Effpatentt+4 Δ Effpatentt+2
Δ Effpatentt+3 Δ Effpatentt+4
Sensitivity to prior margin of -1.028 a -1.747 b -2.508 a -0.851 a -1.446 b -2.078 a
victory (0.285) (0.623) (0.850) (0.236) (0.517) (0.706)
R2 0.123 0.103 0.120 0.123 0.104 0.120
N 20 20 20 20 20 20
Panel C: Second stage regression. Dependent variable - patent citations
EW results VW results
Variable Δ Cpatentt+2 Δ Cpatentt+3
Δ Cpatentt+4 Δ Cpatentt+2
Δ Cpatentt+3 Δ Cpatentt+4
Sensitivity to prior margin of -0.018 0.011 -0.058 b -0.015 -0.009 -0.048 b
victory (0.020) (0.027) (0.026) (0.016) (0.022) (0.021)
R2 0.006 0.002 0.004 0.006 0.002 0.040
N 20 20 20 20 20 20
45
Table 6
Political contributions and corporate innovation: Difference-in-difference estimation
The political contributions data are from the FEC detailed contribution files. We exclude all noncorporate
contributions, contributions from private firms and subsidiaries of foreign firms, as well as contributions from firms
with insufficient data on CRSP/Compustat. The final sample consists of 813,692 political contributions to 5,584 unique
political candidates made by 1,805 unique firms. This table presents the results of the difference-in-difference
estimation. The portfolio of treatment firms comprises 126 firms that supported four junior politicians who
unexpectedly became committee chairs in November 1994. The portfolio of control firms with chair contributions
comprises 475 firms that supported one or more existing committee chairs during the 1990 – 1994 period. The
portfolio of control firms without chair contributions comprises 197 firms that did not support existing committee chairs
during the 1990 – 1994 period. Columns 1 – 3 present the average number of patent applications submitted by the
treatment and the control firms during the pre- and post-1994 period. Columns 4 – 6 presents the average growth in
patent applications submitted by the treatment and the control firms during the pre- and post-1994 period. The bottom
two rows present the difference-in-difference results. T-statistics for the difference-in-difference estimates are in
parentheses.
Effpatentt+1 Δ Effpatentt+1
Firms Pre-1994 Post-1994 Difference Pre-1994 Post-1994 Difference
(1)Treatment firms 4.851 7.152 2.301 0.120 0.670 0.550
(2) Control firms with chair contributions 8.172 8.309 0.137 0.014 -0.287 -0.301
(3) Control firms w / o chair contributions 1.850 1.938 0.087 0.061 -0.024 -0.085
DiD (1 – 2) 2.164
(2.21)
0.851
(4.06)
DiD (1 – 3) 2.214
(2.39)
0.635
(3.21)
46
Table 7
Difference-in-difference estimation: Regression results
The political contributions data are from the FEC detailed contribution files. We exclude all noncorporate
contributions, contributions from private firms and subsidiaries of foreign firms, as well as contributions from firms
with insufficient data on CRSP/Compustat. The final sample consists of 813,692 political contributions to 5,584
unique political candidates made by 1,805 unique firms. This table presents the results of estimating Eq. 9. The
portfolio of treatment firms comprises 126 firms that supported four junior politicians who unexpectedly became
committee chairs in November 1994. The portfolio of control firms with chair contributions comprises 475 firms that
supported one or more existing committee chairs during the 1990 – 1994 period. The portfolio of control firms
without chair contributions comprises 197 firms that did not support existing committee chairs during the 1990 –
1994 period. Panel A presents the results for the treatment firms and the control firms with chair contributions.
Panel B present the results for the treatment firms and the control firms without chair contribution. I(post-1994) is an
indicator variable set to one for years 1995 – 1999 and zero otherwise. I(Year 1995) is an indicator variable set to
one for 1995 and zero otherwise. I(Year 1996) is an indicator variable set to one for 1996 and zero otherwise.
I(Post-1996) is an indicator variable set to one for years 1997-1999 and zero otherwise. Standard errors are in
parentheses and are adjusted for heteroskedasticity and correlated at the firm-post-event level. a, b, c, indicates
significance at the 1%, 5%, and 10% level, respectively.
Variable Effpatentt+1 Effpatentt+1
Δ Effpatentt+1 Δ Effpatentt+1
Panel A: Treatment vs. control firms with chair contributions
I(Post-1994) -5.491 a -5.387 a -1.594 a -1.585 a
(1.116) (1.117) (0.471) (0.489)
Treatment firm × I(Post-1994) 2.342 b 0.832 a
(0.914) (0.240)
Treatment firm × I(Year 1995) 0.036 0.066
(0.818) (0.459)
Treatment firm × I(Year 1996) 1.044 0.413
(0.844) (0.433)
Treatment firm × I(Post-1996) 2.905 a 0.967 a
(1.021) (0.364)
N 4,354 4,935 4,354 4,935
R2 0.917 0.919 0.169 0.167
Panel B: Treatment vs. control firms w / o chair contributions
I(Post-1994) -1.221 -1.382 b -0.641 b -0.663 a
(0.741) (0.764) (0.317) (0.337)
Treatment firm × I(Post-1994) 1.672 b 0.540 a
(0.761) (0.162)
Treatment firm × I(Year 1995) 0.089 0.299
(0.691) (0.414)
Treatment firm × I(Year 1996) 0.518 0.463
(0.765) (0.373)
Treatment firm × I(Post-1996) 2.203 a 0.560 b
(0.840) (0.261)
N 2,245 2,560 2,245 2,886
R2 0.890 0.906 0.173 0.189
47
Table 8
Political contributions and subsequent legislative dynamics, 1984 – 2004
The political contributions data are from the FEC detailed contribution files. We exclude all noncorporate
contributions, contributions from private firms and subsidiaries of foreign firms, as well as contributions from firms
with insufficient data on CRSP/Compustat. The final sample consists of 813,692 political contributions to 5,584
unique political candidates made by 1,805 unique firms. This table presents the results of the probit regression that
relates the year of industry deregulation to the inflow of new innovator firms into each industry during the prior
year. The deregulatory legislation and the year of the passage of each legislation are listed in panel B. New
innovators are defined as firms that begin innovating in an industry and innovate more often that the average
industry firm. Politically active are firms with an established political action committee (PAC). Standard errors are
in parentheses and are adjusted for heteroskedasticity. a, b, c, indicates significance at the 1%, 5%, and 10% level,
respectively.
Panel A: Major deregulatory initiatives
Year Deregulatory Initiative
Petroleum and Natural Gas
1989 Natural Gas Wellhead Decontrol Act
1992 FERC Order 636
Telecommunications
1988 Proposed rules on price caps (FCC)
1996 Telecommunications Act
Utilities
1988 Proposed rules on natural gas and electricity (FERC)
1992 Energy Policy Act
1996 FERC order 888
1999 FERC order 2000
Transportation
1986 Trading of airport landing rights
1987 Sale of Conrail
1993 Negotiated Rates Act
1994 Trucking Industry and Regulatory Reform Act
1995 ICC Termination Act
Panel B: Regression results
All deregulation Congressional Acts Executive orders
Variable 1 2 3 4 5 6
New innovator 0.002 0.004 b -0.001
(0.002) (0.002) (0.005)
New innovator × Politically active 0.006 b 0.008 a 0.001
(0.002) (0.002) (0.007)
New innovator × Politically inactive -0.004 -0.004 -0.004
(0.007) (0.007) (0.008)
Log likelihood -233.78 -232.95 -122.82 -122.01 -135.61 -135.51
Pseudo R2 0.001 0.004 0.004 0.010 0.000 0.001
N 8,812 8,812 8,812 8,812 8,812 8,812
top related