Quid Pro Quo? Corporate Returns to Campaign Contributions∗
Anthony Fowler, University of Chicago Haritz Garro, Northwestern University Jo¨rg L. Spenkuch, Northwestern University
January 2019
Abstract Scholars, pundits, and political reformers have long worried that corporations distort public policy and subvert the will of the electorate by donating to politicians. Well-publicized anecdotes notwithstanding, whether and how much corporations actually benefit from supporting political candidates remains unknown. To systematically address this question, we utilize two complementary empirical approaches that isolate the monetary benefits a company derives from a favored candidate winning office. First, we use a regression discontinuity design exploiting close congressional, gubernatorial, and state legislative elections. Second, we leverage withincampaign changes in market beliefs about the outcomes of U.S. Senate races. We find no evidence that corporations benefit from electing candidates supported by their PACs, and we can statistically reject effect sizes greater than 0.3 percent of firm value. Our results suggest that corporate campaign contributions do not buy significant political favors—at least not on average.
∗We thank Scott Ashworth, Michael Barber, Chris Berry, Ethan Bueno de Mesquita, Tim Feddersen, Pablo Montagnes, Nicola Persico, Koleman Strumpf, and conference participants at PECA and W-PECO for helpful comments and suggestions. Correspondance can be addressed to anthony.fowler@uchicago.edu [Fowler], haritzgarro@@u.northwestern.edu [Garro], or j-spenkuch@@kellogg.northwestern.edu [Spenkuch].
In Citizens United v. FEC, U.S. Supreme Court Justice John Paul Stevens contends that
“[c]orporations with a large war chest to deploy on electioneering may find democratically
elected bodies becoming much more attuned to their interests” (2010, p. 65). Echoing this
suspicion, legions of pundits and political reformers allege that corporate money corrupts
politics. Corporate influence is, for instance, thought to alter election results in favor of
pro-business candidates and away from the preferences of the public. In addition, corporate
donations are often suspected to serve as a quid pro quo, affecting the policy choices of
candidates, who, once elected, may feel indebted to their benefactors or hope to receive
similar contributions in the future.
While the predominant view on the role of corporations in the electoral process is one of dis
tress and disapproval, scholars remain starkly divided on the question of how much influence
firms are actually able to exert through their donations. One influential school of thought
notes that corporate campaign contributions are far too small to be reasonably expected
to buy political favors (Tullock 1972; Milyo, Primo, and Groseclose 2000; Ansolabehere, de
Figueiredo, and Snyder 2003). According to this view, firms do not derive meaningful ben
efits from meddling in elections or else they would do more of it—at least as much as is
allowed under the law. Others however, have presented evidence that special interests give
strategically (see, e.g., Barber 2016; Fouirnaies and Hall 2014, 2018; Powell and Grimmer
2016), even buying access (Kalla and Broockman 2016). In the words of Powell and Grimmer
(2016, pp. 985), “business PAC contributions are consistent with a spot-market for short
term policy influence. [...] It appears that PACs and corporate interests use donations to
solicit favors and to change policy.” In this alternative view, firms give for nefarious reasons.
Yet, as the authors themselves acknowledge, the extant evidence creates only “an appearance
of corruption—the key word being appearance” (p. 986). The extent to which corporations
do, in fact, benefit from supporting political candidates remains unknown.
In this paper, we contribute to the literature on money in politics by asking whether and
how much firms gain when a candidate they supported wins office. After all, if campaign
1
contributions do indeed buy influence, then corporations ought to profit from their political
investments. To measure and monetize the benefits a company derives from its favored
candidate holding office, we utilize stock prices, which reflect the best available information
about a firm’s value, present and future. Thus, to the extent that corporations derive material
advantages from contributing to political candidates, we expect these to manifest themselves
in the value of the firm itself, i.e., its share price.
Our analysis begins by identifying donations from the political action committees (PACs)
of publicly traded companies to candidates running for Congress, governor, or state legisla
tor between 1980 and 2010. Corporations almost never support more than one candidate in
any given race, allowing us to easily determine firms’ preferred politicians. Causal inferences,
however, are complicated by selection into “who gives to whom.” Firms that are ex ante more
successful are more likely to support winners, meaning that a na¨ıve comparison of firms that
gave to winning and losing candidates would produce spurious results. To account for se
lection, we rely on two complementary empirical approaches. Our main analysis leverages
the quasi-random outcomes of very close elections through a regression discontinuity (RD)
design. Our second design uses within-campaign variation in market beliefs—as measured by
betting odds—about the outcomes of U.S. Senate elections from 2004 to 2010. Identification
comes from high-frequency changes in the probability that a corporation’s preferred candi
date wins the race. Both research designs yield substantively identical results. An electoral
victory of the supported candidate does not significantly benefit the typical firm.
Our results are not easily attributable to low statistical power. Ex ante power analyses
suggest that we would have reliably detected any effect greater than about a quarter of a
percent. Ex post, our confidence intervals allow us to statistically reject any purported effect
greater than 0.3 percent of firm value. In addition, we find no sign that a genuine effect is
masked by heterogeneity. In fact, we detect little variation across offices, time periods, the
size of firms, the size of donations, or economic sectors. If companies benefit from donating to
political candidates, then these effects must be small on average, and they are not detectable
2
even in the settings where we would most expect them.
Since the corporations in our sample are extremely large relative to the typical donation,
we cannot dismiss the possibility that campaign contributions are a worthwhile investment
for the firm—albeit a modest one in absolute terms. We can rule out, however, that political
donations are good investments for individual executives, and even the largest of our esti
mates are too small to support the jeremiads of pundits and reformers. In sum, our results
suggest that, on average, contributing firms “give a little and get a little,” as previously
argued by Ansolabehere, de Figueiredo, and Snyder (2003; p. 126).
Some readers are likely to be surprised by the finding that companies do not derive signif
icant benefits from a supported politician holding office. How can our results be so different
from public perception and well-known journalistic accounts? One potential explanation is
that our results are based on a broad sample of firms that supported many different politi
cians. If scholars and pundits focus on a few extreme cases in which corporations appear to
have benefited from political quid pro quos, they might conclude that the perverse conse
quences of money in politics are severe, even if they are negligible on average. By conducting
a large-scale study that aggregates over thousands of different firms and elections, our find
ings speak to the typical effect of a supported candidate rising to office. To be clear, our
contention is not that corporate money has no adverse effects; our point is that these effects
appear to be small in most circumstances. Evidence on both averages and outliers is needed
to sensibly discuss the consequences of corporate influence in American politics.
Before proceeding, we should clarify what we mean by corporate campaign contributions.
In all federal elections and in most state elections, corporations are prohibited from donating
directly to political candidates. However, firms can set up and fund the operation of PACs,
which, in turn, raise money from individuals—often managers and shareholders of the firm—
and give it to candidates. We use the term corporate campaign contributions as shorthand
for contributions from corporate PACs to political candidates and their campaigns.1 Since
1Naturally, individual managers can also donate directly to political candidates. Previous scholars have argued that money channeled to candidates through corporate PACs is used by companies to seek access
3
Citizens United and SpeechNow v. FEC, corporations can draw on their treasuries to give
unlimited amounts to independent expenditure-only committees, which in turn advocate for
or against particular candidates. Although these independent committees are officially pro
hibited from coordinating with campaigns, candidates are often alleged to solicit large sums
for the groups that support them (Lee, Ferguson, and Earley 2014). On one hand, Citizens
United makes the results of this paper all the more important because corporate election
eering expenditures are at an all-time high. On the other hand, one might worry that recent
changes in the political environment limit the generalizability of our results to the present
era. Nonetheless, our limited data for the post–Citizens United period are consistent with
our overall findings. Both before and after this landmark decision, an additional supported
candidate winning office does not, on average, benefit the firm.
Related Literature
Political economists have extensively theorized about the role of corporate campaign contri
butions in the policy-making process. Many models conceptualize elections as a competitive
market for private benefits, whereby firms or other special interests can give money to curry
favor with politicians (e.g., Baron 1989; Denzau and Munger 1986; Grossman and Helpman
2001). Since elected officials have many levers through which to provide benefits to a firm,
companies have an incentive to identify sympathetic politicians and increase their chances
of being elected. A corporation might also use campaign contributions to ingratiate itself
with a candidate who is likely to win anyway. Such a connection could be valuable because
elected officials may propose grants and procurement contracts that can directly benefit the
firm, change their roll-call vote on an important piece of legislation, alter the content of a bill
through amendments or committee work, or pressure the bureaucracy to achieve favorable
regulatory outcomes.
(e.g., Fouirnaies and Hall 2014; Barber 2016), while individual contributions may be due to various other motivations, even when carried out by CEOs (Bonica 2016, Barber, Canes-Wrone, and Thrower 2017). We, therefore, rely on the contribution patterns of corporate PACs rather than individual donations to identify firms’ preferred candidates.
4
In light of these theoretical concerns, empiricists have actively looked for evidence on
the perverse influence of political donations. The most common strategy involves regressing
the roll-call votes of legislators on campaign contributions. One study, for instance, asks
whether members of Congress who received contributions from dairy producers are more
likely to vote in favor of a dairy subsidy (Welch 1982). Even if the correlations between
campaign contributions and subsequent legislator behavior were clear cut, such regressions do
not necessarily uncover causal relationships. If corporations support politicians who already
agree with them, then campaign contributions and roll-call votes will be correlated regardless
of whether the former have any impact on the latter. Consistent with this account, the
estimated effects of corporate contributions on roll-call votes tend to shrink dramatically
with the inclusion of district or member fixed effects (Ansolabehere, de Figueiredo, and
Snyder 2003; Wawro 2001). As a result, some dismiss the idea that campaign contributions
affect legislators’ votes. Others, however, dispute their conclusions. Stratmann’s (2005) meta
analysis of the literature, for instance, rejects the null hypothesis of no effect. Similarly,
Roscoe and Jenkins (2005) claim that about one in three roll-call votes show influence of
campaign contributions.
Taking a contrarian stance, Tullock (1972), Milyo, Primo, and Groseclose (2000), and An
solabehere, de Figueiredo, and Snyder (2003) note that there is not nearly as much corporate
money in politics as one might expect if public policy were actually for sale. If dairy pro
ducers could really generate a billion dollars in price supports through a million dollars in
campaign contributions, why would the dairy industry not give more? And why can members
of Congress be bought so cheaply? Since the vast majority of corporate interests give less
than the statutory limit on campaign contributions, legal restrictions on political donations
are unlikely to be the answer to this puzzle.
Other scholars have looked beyond roll-call votes (e.g., Hall and Wayman 1990). Recently,
Gordon and Hafer (2005) argue that firms contribute in order to signal their willingness to
fight regulators and thereby achieve more favorable treatment from government agencies.
5
Gordon, Hafer, and Landa (2007) find that executives whose compensation is more closely
tied to corporate earnings are more likely to contribute to political candidates, and they
suggest that contributions are best understood as purchases of good will. Kalla and Broock
man (2016) report that activists are more likely to secure a meeting with a senior staffer
of a member of Congress when they reveal themselves to be donors rather than merely
constituents.
Even if political donations do buy access, the ultimate consequences of such meetings
remain unknown. If for-profit companies make campaign contributions to curry favor, and
if these favors actually affect policy, then their value ought to be reflected in the financial
worth of the firm. Thus, for publicly traded companies, we can utilize stock prices to gauge
the pecuniary value of quid pro quos.
Outside the U.S., politically connected firms trade at a premium relative to unconnected
ones (Faccio 2006; Fisman 2001). Within the U.S., Do, Lee, and Nguyen (2015) report
that corporations profit from having a director who went to school with a governor, and
Goldman, Rocholl, and So (2009) find benefits from appointing former politicians and high
ranking bureaucrats to the company’s board. Similarly, Brown and Huang (2017) argue that
corporate executives’ meetings with key White House staff lead to positive abnormal returns.
While political connections of this kind appear to be valuable—especially connections to the
executive—it is a priori unclear whether campaign contributions can be used to forge similar
bonds.
There also exists a sizable literature on the impact of presidential elections on different
sectors of the economy as well as the stock market as a whole (e.g., Snowberg, Wolfers,
and Zitzewitz 2007a; Wolfers and Zitzewitz 2016). Knight (2007), for instance, defines a
group of politically sensitive firms and uses prediction market data to show that presidential
policy platforms are capitalized into equity prices. Our second empirical approach mirrors
this research design, but we find no evidence that the election of senators affects the firms
that supported them. One likely explanation for the discrepancy is that the president wields
6
much more influence over policy than individual legislators.
The studies most closely related to our own are Boas, Hidalgo, and Richardson (2014),
and Akey (2015). Using a regression discontinuity design, Boas et al. show that public-works
firms in Brazil receive more government contracts when their favored candidates rise to office.
Akey (2015) analyzes thirteen close U.S. congressional races and reports that the stock price
of a firm increases by about three percent when a candidate to which it contributed narrowly
wins. However, in the appendix, we use the author’s own data to show that, given the small
number of elections, his findings are highly sensitive specification. In particular, we show
that the results reported in Akey (2015) are not robust—in terms of sign and magnitude—to
using either reasonable alternative bandwidths or different-order polynomials in the running
variable. In fact, eleven out of thirty-two combinations of bandwidth and polynomial yield
negative point estimates. As a result, Akey’s estimates fail one of the standard robustness
checks recommended by Lee and Lemieux (2010) (see the Online Appendix for additional
detail).
In sum, whether corporations benefit from making campaign contributions in a mature
democracy like the U.S. remains unknown. Our subsequent analyses provide systematic evi
dence on this question by drawing on data from thousands of closely contested elections for
governor, Congress, and state legislatures.
Why Give?
There are at least six different reasons a corporation might contribute money to political
campaigns, with each one corresponding to varying degrees of concern about the health of
democracy.
First, corporations might give with the goal of altering election outcomes. If one candidate
is more likely to support policies that would benefit a firm, that company has an incentive
to contribute in order to help the aligned candidate win. Many studies show that campaign
spending can influence vote shares, although the substantive size of these effects is typically
small (e.g., Green and Gerber 2015). Experimental evidence from get-out-the-vote studies
7
suggests that each additional vote costs between 100 and 200 dollars, so influencing the
outcome of most large elections would be expensive. For example, consider the 2014 guber
natorial election in Illinois, which, ex ante, was thought to be an extremely competitive race.
Ultimately, Bruce Rauner (R) won by over 142,000 votes. Had special interests wanted to
tip the scale in favor of Rauner’s opponent, Pat Quinn (D), they would have had to spend at
least $14 million, even if there was no equilibrium response from contributors on the other
side. Since most campaign contributions are three or four orders of magnitude smaller and
go to candidates in less competitive races, one may question whether influencing election
outcomes is firms’ primary motivation.
Second, companies might give to influence the behavior of a candidate who would have
been elected regardless of their help. Perhaps, once in office, elected officials offer quid pro
quos to companies that supported them, or they might dole out favors in order to secure
similar contributions in future campaigns. This mechanism appears to be the most prominent
one in the academic literature as well as the one that especially troubles observers.
Third, corporations might give with the goal of obtaining information from elected officials.
Even if contributions affect neither elections nor public policy, they might create a connection
between a firm and an official that benefits the company in other ways—say, by enabling it to
better anticipate regulatory changes. Companies may be willing to pay for such information,
even if they cannot directly affect policy choices.
Fourth, corporations might give to sitting incumbents in order to change their behavior
before the next election. This mechanism is consistent with the observation that firms tend to
target aligned incumbents. However, the fact that most contributions come toward the end
of an official’s term—when it is too late to enact meaningful policy change before the next
election—casts doubt on such an explanation. Similarly, few corporations make contributions
in support of incumbents that are expected to lose.
Fifth, corporations might give to signal their type. In the model of Gordon and Hafer
(2005), firms have an incentive to signal their willingness to fight regulation. In the model
8
of Schnakenberg and Turner (2016), corporations use campaign contributions to signal com
pliance with the law and thereby reduce their chances of being audited.
Sixth and last, corporations might give for consumption purposes without receiving any
tangible benefits in return. This is the preferred explanation of Ansolabehere, de Figueiredo,
and Snyder (2003) for individual contributions, and it is plausible the same argument applies
to corporations. That is, individual managers may personally know candidates running for
office, enjoy attending glamorous fundraisers, or derive utility from supporting their preferred
candidates. Agency problems within large firms might allow the same executives to (mis)use
some of the company’s resources for political purposes. Given that the typical donation is
minuscule relative to the operating budgets of even medium-sized corporations, and as evi
denced by the mixed results in the academic literature, it would be difficult for shareholders
to determine whether campaign contributions are, in fact, a good investment.
To the extent that any of these mechanisms operate, our subsequent research designs
estimate the combined impact of the first three. Put differently, the benefits a company
derives from a supported candidate winning the election include the value of having someone
hold office who would intrinsically pursue policies that benefit the firm, the value of any
political quid pro quos, as well as the value of insider information. If any of these mechanisms
are substantively important, the successful election of a favored candidate should increase a
firm’s stock price.
The last three mechanisms are not reflected in our estimates. The fourth mechanism pro
duces immediate benefits, which do not directly depend on electoral outcomes. In the signal
ing theories, the value of campaign contributions is independent of election results. In fact,
to signal their type, companies might rationally support the candidate who is ex post worse.
And again, in the consumption account, who wins is irrelevant to the firm because there are
no meaningful direct effects.
Our impression is that most of the worry about corporate influence in politics is due to
the first two possibilities, i.e., its effects on election results and political quid pro quos.
9
Since our empirical tests capture most of what observers find concerning about corporate
contributions, our results directly contribute to the current normative debates about money
in politics.
Data and Research Design
Our regression discontinuity (RD) analysis relies on general election results for the U.S.
Senate, U.S. House, governor, and state legislatures from 1980 to 2010.2 Information on
campaign contributions over this period comes from the Database on Ideology, Money in
Politics, and Elections (Bonica 2013). We identify corporate PACs in these data, match
them to their publicly traded parent companies, and aggregate all contributions (within a
particular election cycle) from the same corporation to the candidates in these elections
(see Appendix for details). Supplemental data on firm size, profitability, sector, etc. come
from the CRSP/Compustat Merged Database. Our final data set includes 2,939 corporations
and 164,525 firm–candidate–election pairings—our unit of observation—across 16 two-year
election cycles with 18,907 individual races.3
In our data, the median (mean) firm spends about $4,000 ($24,000) per cycle and supports
4 (21) candidates. A small number of companies, however, spend as much $1.5 million and
contribute to nearly 1,000 candidates. As we show in the appendix, corporate contributions
skew toward Republicans but not dramatically so. Corporate PACs support Democratic
candidates in about 40 percent of the firm-elections, and Republicans in the remaining 60
percent. Important for our purposes, most but by no means all donations go to ex post
winners.” As we show in the Appendix, corporate contributions skew toward Republicans but
not dramatically so. Corporate PACs support Democratic candidates in about 40 percent of
the firm-elections, and Republicans in the remaining 60 percent. Important for our purposes,
2These data were kindly provided by Jim Snyder and represent an extended version of the data set used in Ansolabehere and Snyder (2002). 3As we explain in the Appendix, we restrict our sample to firm–election pairs in which more than 90% of the contributions from a particular company went to one candidate. As a consequence, we discard 1.98% of observations.
10
most but by no means all donations go to ex post winners.4
To construct our outcome variable, we use daily stock returns from the Center for Research
in Security Prices. Following standard practice in the finance literature, we rely on the market
model to compute cumulative abnormal returns (CARs) for each firm (e.g., Campbell, Lo,
and MacKinlay 1996). Intuitively, CARs adjust stock returns for the performance of the
entire stock market over the same period. More precisely, the daily abnormal return of firm i on day t is defined as ARi,t = ri,t − ˆ αi − ˆ βimt, where ri,t denotes the realized return of the company’s stock on day t, and mt is the market return on the same day. ˆ βi and
ˆ αi, respectively, denote the sensitivity of the firm’s stock to overall market movements and
its usual risk-return performance.5 Residualizing stock returns in this fashion yields more
precise estimates by accounting for market-wide forces that are out of companies’ control
as well as the fact that some firms are more responsive to overall market conditions than
others. With the definition of ARi,t in hand, the cumulative abnormal return of a firm over a period of multiple consecutive days is given by CARi(t1,t2) = (Qt2 t=t1[1 + ARi,t])−1. For our main analyses, we calculate CARs from the day before the election to the day after, i.e., CAR(−1,1), but we also test for longer-term effects. Reassuringly, sensible alternative ways of constructing our outcome variable result in qualitatively equivalent conclusions. In
particular, we obtain nearly identical point estimates if we use simple, unadjusted returns
instead (see Appendix).
Our goal is to estimate the effect of electing a supported candidate on a firm’s value. To do
so, we would like to approximate the following hypothetical experiment. Suppose campaigns
proceed as usual, but election results are secretly determined by a coin flip.6 If who wins the
4For additional descriptive facts, see Appendix A. 5Following Acemoglu et al. (2016), we use a window from 230 to 30 trading days before the election to estimate ˆ βi and ˆ αi. Note, the efficient markets hypothesis implies that ˆ αi should be equal to zero. In our sample, enforcing this theoretical restriction produces a slightly tighter distribution of CARs and subsequently more precise estimates. The substantive difference, however, is negligible, which is why we have opted for the approach that more closely mirrors standard practice in the finance literature. In the Appendix, we also replicate our results using the Fama-French three-factor model instead of the CAPM to calculate abnormal returns. 6The coin flip must be secret because candidates and firms need to behave normally, believing that their campaign efforts influence election results.
11
election is random, then we can estimate the effect of a firm’s supported candidate (rather
than her opponent) rising to office by simply comparing the mean stock return of firms that
support winners with that of companies that support losers. Since such an experiment is not
feasible, we attempt to replicate it as closely as possible by focusing on close elections in a
regression discontinuity (RD) framework.
We implement our RD design by estimating the following equation:
(1) CAR(−1,1)i,j,t = βV ictoryi,j,t + γXi,j,t + λXi,j,tV ictoryi,j,t + i,j,t,
where V ictoryi,j,t is an indicator variable for whether firm i’s preferred candidate won election
j, and Xi,j,t denotes the candidate’s vote margin. The coefficient of interest is β. Because
many corporations given to multiple candidates in the same electoral cycle, β should be
interpreted as the return to the firm from one additional supported candidate rising to
office.
We cluster standard errors by election cycle to allow for almost arbitrary forms of correla
tion in the residuals across firms and elections. For our main results, we restrict attention to
elections in which the two-party vote share fell between .45 and .55, and we present robust
ness checks to demonstrate that our conclusions are insensitive to alternative bandwidths.7
Our primary specification uses CARs from the day before the election to the day after
in order to maximize statistical precision. A drawback of this strategy is that it relies on
the market to quickly internalize the effect of election results on the value of firms—even
before the newly elected officials take office. There is an enormous literature on the efficiency
of financial markets, which typically concludes that markets are close to efficient but not
perfectly so (see, e.g., Fama 1970; Shiller 1981). We are hesitant to take a strong stand on
7We settled on our RD specification after conducting ex ante power simulations. In each simulation, we randomly select a new date for each election year to serve as a placebo Election Day. Taking election outcomes, the pattern of corporate contributions, as well as actual stock returns around the placebo date as given, we add in a hypothetical treatment effect of a known magnitude and implement our empirical strategy. The simulations suggest that if the true effect is 0, then our empirical strategy will only reject the null about 5 percent of the time. If, however, the true effect is at least 0.25 percent, then we would reject the null about 93 percent of the time (see Appendix).
12
how fast the market can internalize the impact of election results. Instead, we present a range
of specifications, allowing for longer lags until prices accurately reflect all effects of elections.
In particular, we present results for as much as 100 trading days after Election Day, when
the winners have taken office and begun enacting their policy agendas. Although increasing
standard errors make quantitative comparisons speculative, if anything, these specifications
suggest that firms benefit even less from their preferred candidate’s electoral triumph than
implied by our main results.
Our RD design serves several purposes. First, it directly addresses selection into who gives
to whom. If the results of very close races are, indeed, quasi-random, then firms supporting
the candidate who, ex post, barely won are, in expectation, identical to companies donating
to the candidate who barely lost. Second, our research design ensures that the market is
surprised by the outcomes that drive our inferences. For elections that are easily predictable,
stock prices will already reflect the value of any potential quid pro quo before Election
Day. Returns realized on or after Election Day would, therefore, be uninformative about the
benefits accruing to the firm. In very close elections, however, both candidates should have an
ex ante realistic chance of winning, which implies that the market receives new information
when the votes are tallied.8
Our RD results are, of course, local to close elections. That is, we estimate the monetary
value of the benefits accruing to a firm when its supported candidate narrowly wins rather
than narrowly loses. If we are interested in the mechanism whereby firms benefit from cam
paign contributions by changing the result of an election, then this is exactly the quantity
8One potential concern with our approach is that market participants may not have the opportunity to incorporate the effects of political contributions into their valuations of firms if they are unaware of who gave to whom. Fortunately for our purposes, PAC contributions are public record, and due to FEC reporting requirements, any contributions made by mid-October of the election year are publicly disclosed before Election Day. Traders can, therefore, know which firms gave to which candidates, and they can use this information if they deem it valuable. Another potential concern is that close elections are subject to recounts and court cases, meaning that there is lingering uncertainty about the outcome even after Election Day, which, in turn, would attenuate our estimates. Although recounts and court cases do occur, it is extremely rare for the initial vote tally to be reversed. Hence, the amount of residual uncertainty is likely small. Furthermore, this concern becomes less relevant as we examine longer time horizons, or when we conduct a donut RD design that ignores the closest of elections, which are the most likely to be affected by recounts and legal skirmishes.
13
of interest. After all, close races are the ones in which firms could potentially affect the out
come. If, however, we are more interested in the quid pro quo or informational mechanisms,
then our approach has both advantages and drawbacks. On one hand, the most powerful
politicians might be electorally safe, which could give them more leeway to hand out favors
relative to their counterparts who live under electoral threat. On the other hand, politicians
may be more willing to engage in quid pro quos precisely when their (re)election prospects
are uncertain—simply because campaign contributions are more valuable when the race is
expected to be close. Although we believe there are good theoretical reasons to be interested
in close races, we freely acknowledge that our RD results may not extrapolate to contri
butions to, say, members of the leadership, committee chairs, or other especially influential
incumbents, all of whom are unlikely to be involved in tight reelection battles.
RD Results
In the Appendix, we discuss and present several descriptive facts that help to motivate
our empirical approach. First, elections that were ex post close were not ex ante predictable,
meaning that the market received genuinely new information on Election Day. Second, almost
no corporations give to both candidates in a given race. This means that we can easily identify
a corporation’s supported candidate in virtually every election in which they contribute.
Third, although most firms give to winners, bigger firms especially give to winners. As a
result, a na¨ıve, cross-sectional study might significantly overestimate the effects of corporate
campaign contributions.
Also in the Appendix, we present several tests assessing the validity of the continuity
assumption necessary for our RD design. Even though corporations are overall more likely
to contribute to winners—as mentioned above—there is little to no evidence to suggest that
firms are systematically favoring bare winners over bare losers. Furthermore, we implement
several placebo tests using pretreatment covariates as outcome variables. There appears to
be no imbalance in cumulative abnormal returns leading up to Election Day, incumbency
status of the supported candidate, or overall firm performance. There is, however, a difference
14
in the number of other candidates that the firm supported, although this difference is not
statistically significant when we correct for multiple-hypothesis testing.
Figure 1 presents our main result focusing on abnormal stock returns right around Election
Day. For illustrative purposes, we average CARs across all observations within one-quarter
percentage-point-wide bins of the vote share, and we display fitted values based on the regression model in equation (1). The size of the estimated discontinuity is −.0006, which implies that a firm’s value decreases by approximately 0.06 percent in response to a supported
politician winning office. Given that the 95% confidence interval associated with our point estimate ranges from −.0044 to .0032, we cannot reject the null hypothesis that the true effect equals zero. We can, however, statistically reject any purported impact greater than
about 0.3 percent.
To interpret this finding, consider the following back-of-the-envelope calculations. The
median firm in our data is valued at about $1 billion. Taking the upper end of the 95%
confidence interval at face value, this would mean that the median contributing firm is
about three million dollars more valuable as a result of its favored candidate winning. While
this may seem like an insignificant amount relative to the size of the company, it comes
at a low cost. The median firm spends only $4,000 on campaign donations per cycle and
supports three winners.9 Hence, there is a clear rationale for companies to contribute to
political candidates from their own treasuries.
This does not mean, however, that a company’s executives or shareholders have an eco
nomic incentive to give, and most PAC contributions are ultimately financed by individual
managers. According to the best estimates in the literature, CEOs personally receive about
$1 for every $100,000 of firm value that they create (see Murphy 1999). Thus, executives
within the company would value the same benefits at only $30.
9Calculating the exact return is challenging because corporations typically spend a lot more money on electioneering beyond their direct contributions to candidates. For example, companies make independent expenditures, contribute to party committees, and sometimes try to mobilize and persuade their employees (see, e.g., Bombardini and Trebbi 2011). Many companies also donate to lawmakers’ pet charities (Bertrand et al. 2018).
15
Based on these calculations, we cannot dismiss the possibility that small donations are
a good investment for firms. Publicly traded companies are so large and the absolute size
of most contributions is so small that it is difficult to envision a research design that could
statistically rule out this possibility. We can, however, statistically reject the possibility that,
on average, individual executives and managers benefit from their political donations through
stock-price effects on their compensation. More importantly, the evidence does not suggest
that, once elected, the average supported candidate engages in substantively meaningful quid
pro quos.
As discussed, for our main RD analysis we rely on a bandwidth of .05. In Figure 2, we
present estimates and confidence intervals for alternative choices ranging from .005 to .3. All
of our RD estimates are substantively small—some negative—and for larger bandwidths we
can statistically reject even medium-sized effects. In Table 1, we present robustness checks
for alternative regression models, including higher-order polynomials in the running variable,
and race fixed effects. Reassuringly, we obtain qualitatively and quantitatively similar results.
Relying on our preferred specification in equation (1), Figure 3 presents estimates for longer
time horizons. Even if financial markets are not perfectly efficient and cannot internalize the
effects of close elections immediately, at some point we would expect the impact of election
outcomes to be fully reflected in stock prices. We, therefore, present RD estimates for CARs
ranging from the day after the election up to 100 trading days afterward—when the electoral
winners have taken office and begun implementing their policy agendas. Our estimates are
never statistically distinguishable from zero, and after about 40 trading days they actually
become negative. Thus, regardless of the time horizon, there is no indication that firms
benefit from their preferred candidate being elected.
We also explore the possibility of heterogeneous effects. To this end, Table 2 presents RD
results for different subsamples. The first row shows our baseline estimate from Figure 1.
The second row estimates the same regression model on a donut sample, i.e., a sample from
which we removed all elections decided by less than 0.2 percentage points. These are the
16
races for which one might be worried about sorting, fraud, or legal challenges, all of which
may lead to nonrandom outcomes. Excluding these races has virtually no impact on the RD
estimate.
Next, we present results separately by office. A priori, one might have expected the largest
impact for governors. After all, governors are politically important and operate as inde
pendent executives rather than one of many members of a legislature. While actual point
estimate for governors is positive, it is substantively small and insignificant. In fact, and there
is no statistically significant evidence of a positive effect for any of the offices we consider.
In addition, we study cases in which corporations donated relatively large sums. Even in
races where a firm gave more than $2,500—roughly the 90th percentile of donations—we
obtain a small negative and statistically insignificant point estimate. If there were political
quid pro quos, say in the form of a single grant or procurement project, we would expect to
find a greater effect on firm value for small rather than large, diversified companies. We test
this prediction by splitting our data according to firms’ market capitalization. Again, it is
not possible to reject the null of no quid pro quos. The same holds true when we separately
analyze different economic sectors and when we consider elections before and after Citizens
United. Although the sample size for the latter period is small, there is no evidence of an
effect in either era.
Table 2 further examines heterogeneity across the number of other winners that a firm
supported in the same election cycle. If there are diminishing marginal returns to politi
cal connections, an electoral victory should be most valuable when the company did not
contribute to other elected officials. However, even in these cases, the RD estimate is small
and statistically insignificant. Moreover, our results do not depend on the overall number of
candidates that the firm supported, suggesting that the sample imbalance with respect to
this variable is inconsequential.
Since one might suspect that corporations giving to candidates on both sides of the aisle
are especially access-oriented, we separately analyze firms that donated primarily to Repub
17
licans, Democrats, or both. Our results, however, yield no evidence of effects for any of these
subgroups. One may also expect that quid pro quos are more likely to arise when only a few
firms contribute to a candidate’s campaign. Yet, we detect no meaningful variation across
the number of firms supporting the winner. Lastly, for the legislative settings in our sample,
we might expect greater effects when a company’s favored legislator belongs to the majority
party and thus stands a greater chance of influencing policy. But again, neither majority
nor minority-party winners have a meaningful impact on firm value.
Broadly summarizing, our RD results imply that, on average, firms do not derive signifi
cant benefits from the electoral victory of a supported candidate. In addition, we find little
evidence of heterogeneity in effect size. Even in settings that are a priori most likely to yield
evidence of political quid pro quos, there appear to be none.
Alternative Research Design
To ameliorate potential concerns with our RD analysis, we implement an alternative research
design that relies on a different source of variation and, therefore, on a different set of iden
tifying assumptions. Since both empirical approaches produce similar results, we conclude
that our substantive results are neither driven by the assumptions underlying our RD design
nor the unrepresentativeness of very close elections.
Our alternative approach uses within-campaign variation in market beliefs—as measured
by betting odds—about the outcomes of U.S. Senate elections from 2004 to 2010. Instead
of comparing returns across firms that contributed to different candidates, this approach
holds firm–candidate pairs fixed. Identification comes from high-frequency changes in the
probability that a corporation’s preferred candidate ends up winning the race.
The sample for this analysis consists of 3,371 firm–candidate pairs across 120 Senate races
that were listed on Intrade. As explained in the Appendix, the betting price provides the
market’s implied belief about the probability that a particular candidate will win (Wolfers
and Zitzewitz 2004). For each firm–election, we focus on betting prices and stock returns in
the 40 trading days leading up to Election Day. Restricting attention to a short period before
18
the election ensures that the vast majority of corporations have already distributed their
contributions. Furthermore, this is a period of intense campaigning, with often-significant
swings in polls.
The within-campaign approach complements our RD design in a number ways. First, since
it conditions on “who gave to whom,” the resulting estimates are not subject to the concern
that firms contributing to winners may be systematically different from those supporting
losers. Second, our within-campaign design leverages additional, high-frequency variation,
resulting in more statistical power for any given election. Thus, if one is especially interested
in recent Senate races, then this alternative approach yields more informative results than the
RD estimates. Third, this design draws on all elections for which market beliefs fluctuated
over the final weeks of the campaign. Since this is even the case for races involving prominent
party figures, powerful committee members, and other influential incumbents, the evidence
is not limited to candidates who are electorally vulnerable. In sum, a within-campaign design
helps to address reservations about the internal and external validity of our RD results.
An important limitation of the within-campaign approach is that rich betting market data
are only available for a subsample of elections. Additionally, this design relies heavily on
market efficiency. For our inferences to be valid, financial markets must accurately respond
to high-frequency changes in candidates’ electoral prospects, and betting markets have to
be efficient enough for these changes to be incorporated into odds. Since betting markets
are thinner than financial markets, the latter assumption may be problematic. If variation in
betting prices is due to noise rather than genuine information, then our subsequent estimates
would be attenuated. To speak to this issue, the Appendix presents a case study of the 2006
Senate race in Virginia. At least within this particular setting, bettors are quite responsive to
new information. In particular, most of the meaningful changes in betting odds are explained
by gaffes, campaign events, and new polls. We also note that restricting attention to the most
liquid and, therefore, least noisy contracts on Intrade has virtually no impact on the results
below.
19
To implement the within-campaign design, we estimate the following equation:
(2) ARi,∆t = α + β∆Pr(FavoredCandidate)i,j,∆t + i,j,∆t,
where ∆Pr(FavoredCandidate)i,j,∆t is the change in the perceived winning probability of
company i’s preferred candidate in election j over time period ∆t, and ARi,∆t denotes the
firm’s abnormal return over the same time frame. The parameter of interest is β. It measures
the increase in market value that would result from an electoral victory of the firm’s favored
candidate, relative to a counterfactual loss.
By regressing returns (i.e., changes in stock prices) on changes in the electoral prospects
of candidates, the regression model in equation (2) is akin to a first-differences design and
therefore holds all firm- and election-specific factors constant.10 Since we work with abnormal
rather than unadjusted returns, our results also control for overall market conditions.
The crucial assumption for estimates based on (2) to be unbiased is that changes in beliefs
about the electoral prospects of a particular candidate are uncorrelated with changes in other,
unobserved factors determining the value of the firms that contributed to her campaign. This
assumption would be violated if, for instance, corporate scandals had spillover effects on the
supported politicians, or if financially troubled companies could withdraw earlier donations.
Table 3 presents the results from estimating equation (2) by ordinary least squares. Columns
(1)–(3) use daily observations, while columns (4)–(6) rely on weekly data. For the latter anal
ysis, the dependent variable is the CAR from Friday to Friday, and the independent variable
is the change in the betting market probability over the same period. If one is concerned that
10To see why the model in (2) conditions on who gives to whom, consider the data generating process
log(stockprice)i,t = µi,t + βPr(FavoredCandidate)i,j,t + i,j,t,
where µi,t is a firm–candidate specific factor that is priced into the company’s stock. Taking the difference between time t and t0 gives
Ri,∆t = β∆Pr(FavoredCandidate)i,j,∆t + ∆i,j,∆t,
with Ri,∆t denoting the stock’s return. Above, we rely on abnormal rather than simple returns to also account for overall market conditions.
20
daily fluctuations in betting odds are noisy and only weakly related to changes in election
fundamentals, then the weekly analysis will be more informative.
Our most inclusive regression models in columns (3) and (6) also include firm–election fixed
effects. We, therefore, not only condition on “who gives to whom,” but we also implicitly
account for linear time trends in the electoral prospects of candidates and the performance
of individual stocks. Identification in these specifications comes from temporary fluctuations
around the respective trends. Arguably, the models in columns (3) and (6) rely on weaker
identifying assumptions, but they require more faith in market efficiency.
Regardless of specification, all point estimates in Table 3 are statistically indistinguishable
from zero. The coefficients in columns (4)–(6) even have the “wrong” sign. In sum, the
evidence from this alternative research design is fully consistent with our RD estimates.
Discussion
In this paper, we provide systematic evidence on the impact of money in politics by studying
corporate campaign contributions in over 18,000 elections for governor, Congress, and state
legislature across three decades. Our research designs isolate the present value of the benefits
a company derives if its favored candidate wins rather than loses the race. Surprisingly, we
find no evidence that campaign contributions produce significant benefits for the firm, at
least not on average. Our estimates are precise enough to statistically reject meaningfully
large effect sizes as well as the possibility that campaign contributions are a good investment
for individual executives.
We should emphasize that our results only speak to the impact of one additional supported
candidate winning office. Previous research has convincingly shown that partisan majorities
in Congress impact firm values through their (anticipated) effects on policy (see, e.g., Jay
achandran 2006; Snowberg, Wolfers and Zitzewitz 2007b). Our results are not at odds with
this finding. The policy-making process is complicated, and electing one additional favored
candidate may not be enough to shift the overall balance of power in a democracy with many
checks and balances. In this sense, legislative systems with diffuse powers protect against
21
the influence of special interests.
We should also emphasize that our findings complement rather than contradict the extant
literature on the value of political connections. Previous work, for instance, demonstrates
that political connections are quite valuable for firms in less-developed countries (e.g., Fisman
2001; Faccio 2006; Boas and Hidalgo 2014). Even in a mature democracy like the U.S.,
ties between corporations and important members of the executive branch appear to yield
nontrivial benefits (Goldman, Rocholl, and So 2009; Do, Lee, and Nguyen 2015; Brown and
Huang 2017), especially in times of crisis (Acemoglu et al. 2016). Our results differ, at least
in part, because we study connections between firms and a much broader set of politicians,
who are, on average, less powerful. Notably, our results are silent on whether companies
profit from ingratiating themselves with members of the federal executive.
Taken together, the extant evidence suggests that context matters a great deal—not all
connections between firms and candidates are created equal. Taking both our findings as
well as those in the literature at face value, corporate campaign contributions may be cause
for concern in (only) a limited number of circumstances. Delineating the conditions under
which corporate electioneering is and is not damaging to the democratic process remains an
important question for future work.11 For example, based on our RD results it appears that
special interests are not particularly effective at influencing electorally constrained politicians
through PAC donations. In any case, we need more information on the entire distribution
of effects—be they good, bad, or null—in order to properly evaluate the overall impact of
money in politics.
In addition, our null findings raise an interesting puzzle. Scholars have argued that special
interests give strategically (Barber 2016; Fouirnaies and Hall 2014, 2018; Powell and Grim
mer 2016), even buying access (Kalla and Broockman 2016). Yet, we find no evidence that
contributions actually benefit the average firm. Is access not valuable? In order for contri
11Fouirnaies and Fowler (2018) take a first step in this direction by studying the influence of the insurance industry in U.S. state politics. Although this industry is a big player in state politics, and despite the fact that it is heavily regulated on the state level, Fouirnaies and Fowler find no evidence that the ability to make corporate campaign contributions benefits insurance companies.
22
butions to consistently affect firm value through this channel, campaign contributions need
to buy access, access has to change the behavior of elected officials, that behavior has to
translate into policy, which in turn must help firms. Determining why, or at which point,
this causal chain breaks down is also a fruitful question for future research.
If corporations do not meaningfully profit from one additional favored candidate winning
office, why do they nonetheless contribute to so many political campaigns? Our results help
to narrow down the set of possibilities. First, as in the signaling theories of Gordon and Hafer
(2005) and Schnakenberg and Turner (2016), the benefits that accrue to the company may
not depend on who wins the election. Second, companies might “give a little and get a little”
(Ansolabehere, de Figueiredo, and Snyder 2003; p. 126). That is, the true benefits to the
firm may be so small that they are not statistically detectable. Third, individual managers
may derive consumption value from donating, and agency problems within the company may
prevent shareholders from effectively monitoring the use of these funds.
The last explanation is consistent with Bonica’s (2016) argument that even CEOs and
other corporate elites donate largely for ideological reasons. Still, even spending driven by
consumption motives may alter election results in favor of the preferences of a small class
of individuals; and our analysis is silent on the number of races that might have seen a
different winner in the absence of such contributions. If the political leanings of high-ranking
executives are not representative of the electorate as a whole, this might be reason enough
to worry about giving by corporations and other wealthy donors.
23
References
Akey, Pat. 2015. “Valuing Changes in Political Networks: Evidence from Campaign Contributions to Close Congressional Elections.” Review of Financial Studies, 28(11): 3188–3223.
Acemoglu, Daron, Simon Johnson, Amir Kermani, James Kwak, and Todd Mitton. 2016. “The Value of Connections in Turbulent Times: Evidence from the United States.” Journal of Financial Economics, 121(2): 368–391.
Ansolabehere, Stephen, John M. de Figueiredo, and James M. Snyder, Jr. 2003. “Why is there so Little Money in U.S. Politics?” Journal of Economic Perspectives, 17(1): 105–130.
Ansolabehere, Stephen and James M. Snyder, Jr. 2002. “The Incumbency Advantage in U.S. Elections: An Analysis of State and Federal Offices, 1942-2000.” Election Law Journal, 1(3): 315–338.
Baron, David P. 1989. “Service-Induced Campaign Contributions and the Electoral Equilibrium.” Quarterly Journal of Economics, 104(1): 45–72.
Barber, Michael J. 2016. “Donation Motivations: Testing Theories of Access and Ideology.” Political Research Quarterly, 69(1): 148–159.
Barber, Michael J., Brandice Canes-Wrone, and Sharece Thrower. 2017. “Ideologically Sophisticated Donors: Which Candidates Do Individual Contributors Finance?” American Journal of Political Science, 61(2): 271–288.
Bertrand, Marianne, Matilde Bombardini, Raymond Fisman, and Francesco Trebbi. 2018. “Tax-Exempt Lobbying: Corporate Philanthropy as a Tool for Political Influence.” NBER Working Paper No. 24451.
Boas, Taylor C., F. Daniel Hidalgo and Neal P. Richardson. 2014. “The Spoils of Victory: Campaign Donations and Government Contracts in Brazil.” Journal of Politics, 76(2): 415–429.
Bombardini, Matilde and Francesco Trebbi. 2011. “Votes or Money? Theory and Evidence from the U.S. Congress.” Journal of Public Economics, 95(7–8): 587–611.
Bonica, Adam. 2013. Database on Ideology, Money in Politics, and Elections. Stanford University Libraries, Stanford, CA <data.stanford.edu/dime>.
Bonica, Adam. 2016. “Avenues of Influence: On the Political Expenditures of Corporations and their Directors and Executives.” Business and Politics, 18(4): 367–394.
Brown, Jeffrey R. and Jiekun Huang. 2017. “All the President’s Friends: Political Access and Firm Value.” NBER Working Paper No. 23356.
24
Campbell, John Y., Andrew W. Lo, and A. Craig MacKinlay. 1996. The Econometrics of Financial Markets. Princeton University Press, Princeton, NJ.
Denzau, Arthur T. and Michael C. Munger. 1986. “Legislators and Interest Groups: How Unorganized Interests Get Represented.” American Political Science Review, 80(1): 89– 106.
Do, Quoc-Anh, Yen Teik Lee, and Bang Dang Nguyen. 2015. “Political Connections and Firm Value: Evidence from the Regression Discontinuity Design of Close Elections.” mimeographed <bit.ly/2lOlF0u>.
Eggers, Andrew C., Anthony Fowler, Jens Hainmueller, Andrew B. Hall, and James M. Snyder, Jr. 2015. “On the Validity of the Regression Discontinuity Design for Estimating Electoral Effects: New Evidence from Over 40,000 Close Races.” American Journal of Political Science, 59(1): 259–274.
Faccio, Mara. 2006. “Politically Connected Firms.” American Economic Review, 96(1): 369– 386.
Fama, Eugene F. 1970. “Efficient Capital Markets: A Review of Theory and Empirical Work.” Journal of Finance, 25(2): 383–417.
Fisman, Raymond. 2001. “Estimating the Value of Political Connections.” American Economic Review, 91(4): 1095–1102.
Fouirnaies, Alexander, and Anthony Fowler. 2018. “Do Campaign Contributions Buy Favorable Policies? Evidence from the Insurance Industry.” mimeographed <bit.ly/2AjfNmU>.
Fouirnaies, Alexander, and Andrew B. Hall. 2014. “The Financial Incumbency Advantage: Causes and Consequences.” Journal of Politics, 76(3): 711–724.
Fouirnaies, Alexander, and Andrew B. Hall. 2018. “How Do Interest Groups Seek Access to Committees?” American Journal of Political Science, 62(1): 132–147.
Goldman, Eitan, J¨org Rocholl, and Jongil So. 2009. Do Politically Connected Boards Affect Firm Value? Review of Financial Studies, 22(6): 2331–2360.
Gordon, Sanford C. and Catherine Hafer. 2005. “Flexing Muscle: Corporate Political Expenditures as Signals to the Bureaucracy.” American Political Science Review, 99(2): 245–261.
Gordon, Sanford C., Catherine Hafer, and Dimitri Landa. 2007. “Consumption or Investment? On Motivations for Political Giving.” Journal of Politics, 69(4): 1057–1072.
Grossman, Gene M. and Elhanan Helpman. 2001. Special Interest Politics. MIT Press, Cambridge, MA.
25
Green, Donald P. and Alan S. Gerber. 2015. Get Out the Vote: How to Increase Voter Turnout (3rd ed.). Brookings Institution Press, Washington, DC.
Hall, Richard L. and Frank W. Wayman. 1990. “ Buying Time: Moneyed Interests and the Mobilization of Bias in Congressional Committees.” American Political Science Review, 84(3): 797–820.
Jayachandran, Seema. 2006. “The Jeffords Effect.” Journal of Law and Economics 49(2): 397–425.
Kalla, Joshua L. and David E. Broockman. 2016. “Campaign Contributions Facilitate Access to Congressional Officials: A Randomized Field Experiment.” American Journal of Political Science, 60(3): 545–558.
Knight, Brian. 2007. “Are Policy Platforms Capitalized into Equity Prices? Evidence from the Bush/Gore 2000 Presidential Election.” Journal of Public Economics 91(1–2): 389– 409.
Lee, Chisun, Brent Ferguson and David Earley. 2014. After Citizens United: The Story in the States. Report of the Brennan Center for Justice at NYU School of Law.
Lee, David, and Thomas Lemieux. 2010. “Regression Discontinuity Designs in Economics.” Journal of Economic Literature 48: 281–355.
Milyo, Jeffrey, David Primo and Timothy Groseclose. 2000. “Corporate PAC Campaign Contributions in Perspective.” Business and Politics, 2(1): 75–88.
Murphy, Kevin J. 1999. “Executive Compensation,” (pp. 2485–2563) in Orley Ashenfelter and David Card (eds.), Handbook of Labor Economics, Vol. 3B. Elsevier, Amsterdam.
Powell, Eleanor N. and Justin Grimmer. 2016. “Money in Exile: Campaign Contributions and Committee Access.” Journal of Politics, 78(4): 974–988.
Roscoe, Douglas D., and Shannon Jenkins. 2005. “A Meta-Analysis of Campaign Contributions’ Impact on Roll Call Voting.” Social Science Quarterly, 86(1): 52–68.
Schnakenberg, Keith E. and Ian R. Turner. 2016. “Helping Friends or Influencing Foes: Electoral and Policy Effects of Campaign Finance Contributions.” mimeographed, <bit.ly/2lT0mvk>.
Shiller, Robert J. 1981. “Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?” American Economic Review, 71(3): 421–436.
Snowberg, Eric, Justin Wolfers, and Eric Zitzewitz. 2007a. “Partisan Impacts on the Economy: Evidence from Prediction Markets and Close Elections.” Quarterly Journal of Economics, 122(2): 807–829.
26
Snowberg, Eric, Justin Wolfers, and Eric Zitzewitz. 2007b. “Party Influence in Congress and the Economy.” Quarterly Journal of Political Science, 2(3): 277–286.
Stevens, John Paul. 2010. Citizens United v. Federal Election Commission 588 U.S. 08-205 (concurring in part and dissenting in part).
Stratmann, Thomas. 2005. “Some Talk: Money in Politics. A (Partial) Review of the Literature.” Public Choice, 124: 135–156.
Tullock, Gordon. 1972. “The Purchase of Politicians.” Western Economic Journal, 10(3): 354–355.
Wawro, Gregory. 2001. “A Panel Probit Analysis of Campaign Contributions and Roll-Call Votes.” American Journal of Political Science, 45(3): 563–579.
Welch, W. P. 1982. “Campaign Contributions and Legislative Voting: Milk Money and Dairy Price Supports.” Western Political Quarterly, 35(4): 478–495.
Wolfers, Justin and Eric Zitzewitz. 2004. “Prediction Markets.” Journal of Economic Perspectives, 18(2): 107–126.
Wolfers, Justin and Eric Zitzewitz. 2016. “What Do Financial Markets Think of the 2016 Election?” mimeographed <brook.gs/2njI8U3>.