How Do Firms Respond to Political Uncertainty? Evidence from U.S. Gubernatorial Elections

• Published date: 01 September 2023
• Author: ANDREW BIRD,STEPHEN A. KAROLYI,THOMAS G. RUCHTI
• Date: 01 September 2023
• DOI: http://doi.org/10.1111/1475-679X.12482

DOI: 10.1111/1475-679X.12482

Journal of Accounting Research

Vol. 61 No. 4 September 2023

Printed in U.S.A.

How Do Firms Respond to Political

Uncertainty? Evidence from U.S.

Gubernatorial Elections

ANDREW BIRD,∗STEPHEN A. KAROLYI ,†

AND THOMAS G. RUCHTI ‡

Received 10 July 2017; accepted 5 December 2021

ABSTRACT

We examine the joint response to political uncertainty along two margins:

changes in real activity and voluntary disclosure. We focus on within-firm vari-

ation in exposure to ex ante competitive U.S. gubernatorial elections using

data on preelection poll margins and firms’ state exposures. Despite real ac-

tivity falling in the years leading up to a close election, we find that voluntary

disclosure increases both in frequency and content, including mentions of

risk in filings that reference states holding elections. Our tests use a decom-

position of 8-K filings into real activity and voluntary disclosure to address

the endogenous complementarity between these two responses. These results

hold when using alternative ex ante measures of political uncertainty based

on term-limited incumbents, historically competitive offices, or state legisla-

ture gridlock. Both effects of political uncertainty are stronger for firms in

∗Chapman University George L Argyros School of Business and Economics; †Office of the

Comptroller of the Currency; ‡Office of Financial Research, US Department of the Treasury,

Washington, DC, United States

Accepted by Christian Leuz. The views stated herein are those of the authors and are not

necessarily the views of the Office of the Comptroller of the Currency, the Office of Finan-

cial Research, the U.S. Department of the Treasury,or any federal agency, and do not estab-

lish supervisory policy, requirements, or expectations. We wish to thank Vikas Agarwal, Pat

Akey, Mike Alvarez, Aytekin Ertan, Candace Jens, Andy Koch, Stefan Lewellen, Matt Shum,

Andy Sinclair, and Bobby Stoumbos for valuable discussions and feedback, and Qijia Yuan

for diligent research assistance. An online appendix to this paper can be downloaded at

https://www.chicagobooth.edu/jar-online-supplements.

1025

© 2023 The Chookaszian Accounting Research Center at the University of Chicago Booth School of

Business.

1026 a. bird, s. a. karolyi, and t. g. ruchti

highly regulated industries and weaker for those least exposed to the local

market, linking the real activity and disclosure responses to uncertainty.

JEL codes: D72, G14, G18, G34

Keywords: political uncertainty; real activity; voluntary disclosure

1. Introduction

The question of how uncertainty affects corporate activity is central to our

understanding of organizational decision making and financial markets.

Responding to fluctuating uncertainty is a vital strategic challenge for firms

today.1Because uncertainty alters both opportunities and investor percep-

tions, firms may respond both by changing real activities and by changing

disclosure. Theory disagrees about the direction of uncertainty’s impact

on the marginal profitability of investment (Bernanke [1983], Caballero

[1991]). Likewise, theory highlights competing mechanisms that relate un-

certainty to disclosure (Verrecchia [1983], Dye [1985], Lambert, Leuz, and

Verrecchia [2007], Heinle and Smith [2017]). Furthermore, any changes

in real activity may themselves be related to the manager’s preference for

voluntary disclosure. In this paper, we use political uncertainty to study

whether and how firms jointly respond along these two margins. We focus

on political uncertainty because surveys of top executives reveal that gov-

ernment policies and the political process that shapes them rank in the top

five concerns for U.S. businesses, higher than rising input prices, employee

productivity and morale, data security, or macroeconomic uncertainty.2

We face both measurement and identification challenges in this en-

deavor. First, we use U.S. gubernatorial elections to generate measurable

cross-sectional and time-series variation in firms’ exposure to political un-

certainty (Levitt and Wolfram [1997], Jens [2017]). Peltzman [1987] and

Ang and Longstaff [2013] show that political uncertainty arising from gu-

bernatorial elections is critical for businesses that operate in the state, and

Snowberg, Wolfers, and Zitzewitz [2007] argue that cross-election variation

in competitiveness is a credible exogenous source of variation in political

uncertainty.

Second, we introduce data and measures that better capture uncertainty

shocks, firm-level exposure to these uncertainty shocks, and potential real

activity and disclosure responses by exposed firms. We collect 2,982 preelec-

tion polls from gubernatorial elections between 1995 and 2014 to build

an election-level measure of uncertainty that captures the environment

in each state preceding the election. Because firms may be exposed to

1For example, Manela and Moreira [2017] find that NVIX, a news-based historical coun-

terpart to VIX, has been fluctuating significantly since the 1890s and, if anything, these fluc-

tuations have accelerated.

2See, for example, surveys from Duke CFO Global Business Outlook (www.cfosurvey.org)

and Deloitte CFO Signals: 2017 Q2 (https://www2.deloitte.com/us/en/pages/finance/

articles/cfos-signals-survey- growing-concerns-about- political-and-policy- uncertainty.html#1).

how do firms respond to political uncertainty? 1027

elections in multiple states simultaneously and to varying degrees, we col-

lect a data set of corporate mentions of U.S. states in 10-K filings (Garcia

and Norli [2012], Jens [2017]) and use these mentions to construct a time-

varying index of a firm’s exposure to each state.

To construct measures of the corporate response to uncertainty, we col-

lect data from 8-K filings and use the 8-K classification system to decompose

these filings into real activity and voluntary disclosure. Our measure of real

activity is based on the frequency of mandatory 8-K items, such as signing a

contract (Item 1.01) or acquiring or disposing of an asset (Item 2.01). The

frequency of the remaining, voluntary, items (7.01 and 8.01) constitutes

our primary measure of voluntary disclosure (He and Plumlee [2020]).3

Of course, even filings concerning real activity have an element of volun-

tary disclosure because firms can decide how much information to disclose

about these events and so we also study the content of 8-K filings as a sup-

plemental measure of voluntary disclosure.

Starting from 8-K filings and elections data, we construct a firm-election-

month panel data set, starting two years before each election. Like Jens

[2017], we explore differences in behavior between firms that are exposed

and firms that are not exposed to close elections. To ensure that we com-

pare firms at similar points in an election cycle, a firm in a given month

enters our sample if that firm is exposed to a state election in the next

24 months. We then implement a weighted least squares regression using

firm exposures to different states as weights. Our empirical strategy thus

allows us to compare outcomes for firms with varying degrees of exposure

to close gubernatorial elections in the two years preceding elections.

We find evidence that firms respond to increased political uncertainty, as

measured by the closeness of preelection polls, by both decreasing their real

activities and increasing their disclosures. Specifically, our log-linear specifi-

cation implies that firms exposed to close elections decrease their 8-K filings

related to real activities by 12%–33% and increase their 8-K filings specifi-

cally related to voluntary disclosure by 8%–29% in the two years preceding

the elections.4These results suggest that firms respond to uncertainty by

disclosing more but delaying decisions that have a real impact on corpo-

rate operations (Baker, Bloom, and Davis [2016], Hassan et al. [2019]). We

find results that are similar in sign and magnitude when we employ alterna-

tive definitions of close elections, different measures of political uncertainty

3We provide a more thorough discussion of the details of our measure in the data sec-

tion and in the online appendix. In particular, in the online appendix, we provide empirical

evidence that the frequency of mandatory 8-K item types is a measure of real activity.

4These ranges reflect the set of results presented in table 2 and associated robustness tests

included in the online appendix. Our preferred estimates are 28% and 23%, respectively,

and are presented in table 2. To calculate these magnitudes from log-linear specifications,

we exponentiate the coefficients and subtract one. In results that are available upon request,

we also investigate sensitivity to functional form of our log-linear specification with an alterna-

tive dependent variable: the fraction of filings that are voluntary. Estimates with this alternative

dependent variable are consistent with those in our main voluntary disclosure specifications.