economic policy uncertainty and information asymmetry€¦ · economic policy uncertainty and...
TRANSCRIPT
![Page 1: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/1.jpg)
Economic Policy Uncertainty and InformationAsymmetry∗
Venky NagarUniversity of Michigan
Jordan SchoenfeldUniversity of Utah
Laura WellmanUniversity of Utah
October 2017
Abstract
This study examines whether economic policy uncertainty exacerbates information
asymmetry among investors. We find that increased economic policy uncertainty is
associated with decreased stock liquidity, especially for firms more exposed to eco-
nomic policy uncertainty. Increased economic policy uncertainty also lowers investors’
reaction to earnings for firms with high liquidity risk. Management in turn increases
voluntary disclosure, which only partly reverses the liquidity drop. These results sug-
gest that information asymmetry is an important channel through which economic
policy uncertainty affects asset pricing.
Keywords: Corporate Disclosure; Information Asymmetry; Economic Policy Uncertainty;Stock LiquidityJEL Classification: D80; E61; E65; G12; G14; G18; L50
∗We appreciate the helpful suggestions from the Editor (John Core), an anonymous reviewer, Nick Bloom,Steve Davis, Beverly Walther, and workshop participants at BYU, Northwestern University, Temple Univer-sity, and the University of Utah.
![Page 2: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/2.jpg)
1 Introduction
Policymakers are motivated by a variety of factors, including economic, social, and polit-
ical forces. To the extent that investors cannot fully anticipate the effect of these competing
forces on policy outcomes, they face uncertainty over which policies the government will
implement, and ultimately over how these policies impact firm value. Accordingly, em-
pirical evidence suggests that economic policy uncertainty affects firms’ business prospects
and operating decisions, and thus equity prices and volatility (e.g., Baker et al., 2016, p.
1597). However, theory suggests that equity price levels and volatility patterns can change
both in a homogeneous information setting where economic policy uncertainty identically
changes every investor’s beliefs of the mean levels and the priced covariances of firms’ future
payoffs (e.g., the models of Kelly et al., 2016 and Pastor and Veronesi, 2012, 2013, where
all investors are identically informed), and in a heterogeneous information setting where
economic policy uncertainty changes price by changing the information asymmetry among
investors. We therefore empirically examine the effect of economic policy uncertainty on
investor information asymmetry.
Our main proxy for economic policy uncertainty is the economic policy uncertainty (EPU)
index of Baker et al. (2016) that is constructed from textual analysis of news sources. The
EPU index has found widespread acceptance in the literature in part due to the valida-
tion efforts of its creators and in part due to its continuous availability.1 Using a sample
of all U.S. public companies over 2003–2016, we first find that stocks, both on the aggre-
gate and individually, are less liquid in times of a higher EPU index, both at daily and
monthly intervals. This suggests that economic policy uncertainty exacerbates rather than
1Much of Baker et al. (2016) is devoted to validation efforts that show that the EPU index is correlatedwith key macroeconomic, individual firm-level, and political phenomena. We describe these features of theEPU index in detail in Section 2. Also see http://policyuncertainty.com/research.html.
1
![Page 3: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/3.jpg)
levels information asymmetry among investors.2 Since it is unlikely that the EPU index is
significantly influenced by any individual firm’s liquidity, we have assurance in attributing
causality from the EPU index to individual firm liquidity. To further establish that economic
policy uncertainty is indeed the factor behind this association, we follow Baker et al. (2016,
Section IV.B) and successfully find stronger (weaker) results for a subset of firms that have
high (low) exposure to economic policy uncertainty. We then show that a firm’s exposure to
illiquidity related to economic policy uncertainty also impacts how investors price earnings.
For the same earnings surprise in periods of a higher EPU index, earnings announcement
returns are lower for firms that have high exposure to liquidity risk. This result further
confirms the presence of increased information asymmetry among investors during periods
of increased economic policy uncertainty.
Given the above liquidity implications, we then delve more deeply into our reduced-
form estimation of the link between economic policy uncertainty and liquidity by following
Balakrishnan et al. (2014), who argue that managers shape liquidity and therefore respond
to exogenous liquidity reductions by increasing voluntary disclosure. We find this to be the
case in our setting as well. Managers increase disclosure after economic policy uncertainty
increases, and this disclosure reduces but does not eliminate the liquidity effect.3 These
findings buttress our main contention that economic policy uncertainty has a heterogeneous
impact on investors’ information sets.
We now give a sense of the economic magnitudes of our results. Using year-fixed effect
regressions, we find that a one standard deviation increase in the monthly EPU index is
associated with a 1.01% increase in monthly market-value-weighted quoted spreads and a
2Our proxy for information asymmetry is stock liquidity, as measured by quoted percent bid-ask spreads(from DTAQ) and Amihud illiquidity (from CRSP). Fong et al. (2017) and Holden and Jacobsen (2014) findthat these are the two best measures of stock liquidity. We conduct our tests at several levels of granularity,including monthly and daily data intervals, and at the aggregate market and firm levels. Monthly testsensure that the daily tests are not driven by spurious short-term correlations in the data, whereas firm-leveltests enable us to ascertain meaningful sources of cross-sectional variation in the aggregate tests. Figure 1shows that the EPU index has considerable time-series variation, so we do not examine liquidity effects fordurations longer than a month. We include appropriate fixed effects and error clustering.
3We also use the findings of Balakrishnan et al. (2014) to provide assurance that disclosure is likely drivingliquidity and not the other way around (see Section 3.4).
2
![Page 4: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/4.jpg)
1.93% increase in monthly market-value-weighted Amihud illiquidity.4 We then conduct
monthly tests at the firm level using firm-year-fixed-effect regressions. We find that a one
standard deviation increase in the monthly EPU index is associated with a 3.48% increase
in monthly firm-level quoted spreads and a 2.52% increase in monthly firm-level Amihud
illiquidity.
We next use daily data to link liquidity to the EPU index. At the aggregate market
level with year-fixed effects, we find that a one standard deviation increase in the daily EPU
index is associated with a 0.45% increase in daily market-value-weighted quoted spreads and
a 0.42% increase in daily market-value-weighted Amihud illiquidity. At the firm level with
firm-year-fixed effects, where we have more confidence in attributing causality, we find that
a one standard deviation increase in the daily EPU index is associated with a 1.8% increase
in daily quoted spreads and a 1.1% increase in daily Amihud illiquidity. By comparison,
Chordia et al. (2001, p. 509) find that the average daily change in a similar spread variable
is on the order of two percent. To corroborate these initial results, we perform several cross-
sectional analyses using both monthly and daily data. The link between the EPU index
and illiquidity is stronger for firms that are more exposed to economic policy uncertainty,
as measured by an EPU index-stock return sensitivity beta, for smaller firms, for firms with
more domestic exposure, and for firms with several additional characteristics. We also ensure
that our main findings generalize to other economic policy uncertainty settings such as being
in the healthcare business during the Affordable Care Act (ACA) policy debates.
To the extent that increased economic policy uncertainty decreases liquidity, this could
cause some investors to trade less aggressively on information or to withdraw from the market
altogether, which means that there would be fewer traders (or less trading) in the market
for important corporate information events such as earnings disclosures (e.g., Chordia et al.,
4Our use of value weights follows Chordia et al. (2005), though our results hold for equal weights aswell. Baker et al. (2016) find that the EPU index could be correlated with general market uncertainty, sowe control for the equity uncertainty (EU) index from Baker et al. (2016) in the above tests and, whenappropriate, in the analyses that follow. Baker et al. (2016, p. 1614) find that the EU index compares wellto other proxies of equity market uncertainty such as the VIX.
3
![Page 5: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/5.jpg)
2008; Kyle, 1985). Following the research design in Kelly and Ljungqvist (2012, Section
3.4), we compute the Pastor and Stambaugh (2003) liquidity risk exposure beta for each
firm and show that firms in the highest quintile have lower announcement returns for the
same earnings surprise during periods of high economic policy uncertainty, relative to firms
in the lower quintiles.5
We then look to managerial response following Balakrishnan et al. (2014), who argue that
management typically has an incentive to shape liquidity and therefore uses disclosure to
restore exogenous drops in liquidity. At the daily level, we find that a one standard deviation
increase in the average EPU index over the prior 30 days is associated with additional
guidance disclosures that lower spreads on the same day by 0.22%. This indirect effect of
0.22% amounts to 48.89% of the direct effect that the EPU index has on spreads on a daily
basis, suggesting that managers’ additional guidance disclosures offset the adverse liquidity
effects of economic policy uncertainty by about half. Managers attempt to reverse but do not
succeed in fully reversing the liquidity effects of economic policy uncertainty. This further
supports our inference that our reduced-form tests linking liquidity and economic policy
uncertainty reflect an increase in information asymmetry among investors.6
We lastly place our findings in the literature. Baker et al. (2016) argue that economic
policy uncertainty is an important uncertainty to investors in the market. Accordingly,
as stated in our opening paragraph, many studies examine the impact of economic policy
uncertainty on stock prices, but in a manner that does not explore whether investors have the
5Note that earnings disclosure can reduce information asymmetry among investors and thus offset investorpropensity to trade less. To complicate matters further, some investors may respond to economic policyuncertainty and earnings disclosures by increasing their information-acquisition activities and trading moreaggressively, which could also cause other less informed investors to withdraw (Kim and Verrecchia, 1994;Gao and Huang, 2016). The extent of this offset is an empirical issue. Our findings suggest that this offsetis not complete.
6Our findings also obtain when we control for guidance released on earnings announcement days. Ouruse of a daily setting is motivated by endogeneity considerations. If liquidity is low, management mayincrease disclosure, which would result in a positive association between illiquidity and disclosure, potentiallycanceling the liquidity improvement effect of disclosure (Balakrishnan et al., 2014, p. 2271). This is notwhat we find. To the extent that management responds to longer-term declines in liquidity, the impact ofthe endogeneity of disclosure is likely to be more pronounced over longer intervals such as a quarter, whichis why Balakrishnan et al. (2014, p. 2266) find a weak “naive” relationship between liquidity and disclosureover both the contemporaneous and the lagging quarter (see our Section 3.4).
4
![Page 6: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/6.jpg)
same information sets.7 To be sure, investor information heterogeneity has been examined
in other asset pricing settings such as brokerage closures (e.g., Kelly and Ljungqvist, 2012),
but not in the setting of economic policy uncertainty. Likewise, the impact of liquidity
on investor actions such as the pricing of financial reports and managerial actions such as
voluntary reporting has also been examined before, but in different settings (Balakrishnan
et al., 2014; Guay et al., 2016; Lang and Maffett, 2011; Sadka, 2011; Schoenfeld, 2017).
The remainder of this study is organized as follows. Section 2 describes our data. Section
3 provides our empirical results. Section 4 concludes.
2 Sample and Data
Our sample includes the entire NYSE daily TAQ (DTAQ) database from September 10,
2003 to December 31, 2016. We focus specifically on firms headquartered in the U.S., as
our primary measures for economic policy uncertainty pertain to the U.S. This reduces the
DTAQ universe by about 25%. We also link each DTAQ firm to CRSP data using the linking
table provided by DTAQ, which enables us to compute contemporaneous Amihud illiquidity.
We impose no other survival criteria (except where as noted). Our final sample comprises
nearly 7,000 firms and 15 million daily observations of liquidity, which suggests that our
regression coefficients will be precisely estimated.
To proxy for stock liquidity, we follow Holden and Jacobsen (2014) and use the daily
DTAQ database to compute daily percent quoted spreads for each firm-day over our sample
period. Fong et al. (2017) and Holden and Jacobsen (2014) provide convincing evidence that
this measure is the “first-best” proxy for liquidity. We compute percent quoted spread as
7One exception is Pasquariello and Zafeiridou (2014), who use Roll’s indirect measure of liquidity thatdoes not use bid-ask spreads but instead constructs the liquidity measure based on the covariance of pricemovements assuming an informationally efficient market with no information asymmetry (Fong et al., 2017).
5
![Page 7: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/7.jpg)
follows:
Percent Quoted Spreadsit = 100× Askit −Bidit(Askit +Bidit)/2
. (1)
Subscript i represents each firm, and subscript t represents each observation day. Askit is
the National Best Ask and Bidit is the National Best Bid, where both variables are time
weighted during trading hours for each day according to the procedure described in Holden
and Jacobsen (2014). To ensure that we compute this measure accurately, we vet and
implement the code provided by Holden and Jacobsen (2014).
Our second proxy for liquidity is Amihud illiquidity (Amihud, 2002). Fong et al. (2017)
and Holden et al. (2013) find that Amihud illiquidity is the second-best liquidity measure
behind quoted spreads, and the best measure that can be computed from CRSP. Following
Amihud (2002), we compute daily Amihud illiquidity as follows:
Amihudit = 106 ×∑ |Returnit|
Dollar Trade V olumeit. (2)
Returnit and Dollar Trade V olumeit are firm i’s stock return and dollar trading volume,
respectively, on day t from CRSP. While Chordia et al. (2008, p. 252) note that reporting
errors in the TAQ databases have declined considerably, we mitigate the influence of outliers
by winsorizing our liquidity proxies from the top at the 1% level (our results are qualitatively
similar without winsorizing). Furthermore, our analyses use the log of one plus the liquidity
measures as described in Appendix A, which further reduces outlier concerns and also makes
interpretations of the coefficients easier.8 All of our liquidity measures are available on a
daily basis and move inversely with the level of a firm’s stock liquidity. We conduct both
individual firm-level and aggregate market-level analyses. In our aggregate market analyses,
both of our liquidity proxies are market-value weighted on the day of observation. Our use
8Note that the number of observations in our regressions will differ slightly due to data availability ofour liquidity proxies. For example, on occasion CRSP does not provide trading volume, so for these days wecannot compute Amihud illiquidity.
6
![Page 8: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/8.jpg)
of value weights follows Chordia et al. (2005), though our results hold for equal weights as
well. We also conduct daily and monthly analyses. In our monthly analyses, we average
each liquidity measure (at both the market and firm level) over the observation month.
Our main proxy for economic policy uncertainty is the economic policy uncertainty (EPU)
index developed by Baker et al. (2016). The EPU index is constructed daily based on
the number of news articles that contain the terms economic or economy ; uncertain or
uncertainty ; and one or more of Congress, deficit, Federal Reserve, legislation, regulation,
and White House.9 Baker et al. (2016, p. 1593) note that several types of evidence—
including human readings of 12,000 newspaper articles—indicate that the EPU index proxies
for movements in economic policy uncertainty. The index spikes near tight presidential
elections, Gulf Wars I and II, the 9/11 attacks, the failure of Lehman Brothers, the 2011
debt ceiling dispute, and other major battles over economic policy. Using firm-level data,
Baker et al. (2016) find that the EPU index is associated with greater stock price volatility
and reduced investment and employment in policy-sensitive sectors like defense, health care,
finance, and infrastructure construction. At the macro level, the EPU index foreshadows
drops in investment, output, and employment in the United States.
We also address the finding in Baker et al. (2016) that the EPU index could be correlated
with the broader construct of general market uncertainty. Specifically, we control for the
equity uncertainty (EU) index when appropriate, as recommended by Baker et al. (2016).
The EU index is constructed based on the number of news articles that contain the terms
uncertainty or uncertain, the terms economic or economy, and one or more of the following
terms: equity market, equity price, stock market, or stock price.10 In our monthly analyses,
we compute the average of these measures over the observation month.
We conduct a cross-sectional analysis on the association between the EPU index and
liquidity based on several static, firm-level attributes. Our first attribute is the sensitivity of
9See http://www.policyuncertainty.com/us_daily.html.10See also http://www.policyuncertainty.com/equity_uncert.html. Our results also obtain when we
use the VIX and squared VIX in place of the EU index (Baker et al., 2016, p. 1614; Drechsler, 2013).
7
![Page 9: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/9.jpg)
the firm’s returns to the EPU index. We use a firm-level economic policy uncertainty returns
beta (Policy Beta), which we compute by first running firm-level regressions of monthly excess
returns on the daily EPU index over our sample period as follows:
rit = β0i + βEi EPUt + βMi MKTt + βSi SMBt + βHi HMLt + εit, (3)
where r is firm i’s excess return, MKT is the excess return on a market index, and SMB
and HML are long-short return spreads constructed on sorts of market capitalization and
book-to-market ratio (see Fama and French, 1993). Note that we run this regression at the
firm level, not the portfolio level, to ensure that we can directly link each firm’s βEi to its
respective liquidity.11 Our cross-sectional analysis also considers firm size, as measured by
log of market value of equity, as another factor driving the sensitivity of a firm to economic
policy uncertainty.
We also consider a firm’s international exposure, which might affect a firm’s sensitivity
to domestic U.S. economic policy uncertainty (e.g., Boutchkova et al., 2012). For example,
firms with less U.S. exposure might be insulated from economic policy uncertainty in the
U.S., which is our setting. Specifically, we use an indicator variable that equals 1 if a firm
had foreign-subsidiary revenue in any year during our sample period, derived from income
statement data provided by Compustat. The next attribute is the average number of foreign
trade words in a firm’s 10-Ks over our sample period, computed using the Baker et al.
(2016) foreign trade dictionary. This dictionary includes import tariffs, import duty, import
barrier, government subsidies, government subsidy, WTO, World Trade Organization, trade
treaty, trade agreement, trade policy, trade act, Doha Round, Uruguay Round, GATT, and
dumping. As with foreign revenue, foreign trade words in the 10-K might represent a firm’s
11Note that we are not interested in examining if the EPU index is a priced factor. We only want tomeasure how sensitive a firm’s returns are to the EPU index. In the cross-sectional analyses, we use theabsolute value of the beta because both positive and negative beta suggest that a firm’s return is sensitiveto economic policy uncertainty. Our use of absolute values is similar to the research design of Addoum andKumar (2016, p. 3473), who examine return sensitivity to political uncertainty.
8
![Page 10: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/10.jpg)
international exposure, which might affect its sensitivity to domestic U.S. economic policy
uncertainty.
We also consider two additional cross-sectional measures that could modulate the link
between economic policy uncertainty and the firm’s liquidity. Our first measure is a firm’s
lobbying expenditures, which could represent a firm’s effort to manage the economic policy
process through intervention (Gao and Huang, 2016; Hochberg et al., 2009; Wellman, 2017,
p. 225). We hand collect lobbying data from the Center for Responsive Politics (CRP) and
compute the average number of total lobbying dollars a firm spends per year over our sample
period.12 Our second measure is a firm-level economic policy uncertainty measure derived
from the 10-K (labeled EPU Words in the tables). We compute the average number of
economic policy uncertainty words in a firm’s 10-Ks over our sample period, relying on a set
of terms similar to that of Baker et al. (2016).13 An important caveat about using lobbying
dollars and 10-K policy terms is that both of these measures might proxy for increased
exposure to economic policy uncertainty, rather than effective management of economic
policy uncertainty (e.g., Adelino and Dinc, 2014; Hassan et al., 2017). We therefore do not
have predictions for these measures.
We compute all of the above firm-level measures as their averages for each firm over our
full sample period, because many of these measures do not change significantly over time or
can be measured only over relatively long time intervals. As a result, we cannot use these
measures to conduct meaningful within-firm-year analyses, and thus these measures vary
only in the cross-section of our firms. However, one advantage of this approach is that we
can still include firm-year-fixed effects in all of our cross-sectional tests.
Following the research design in Kelly and Ljungqvist (2012, Section 3.4), we next con-
duct tests to see if economic policy uncertainty changes the stock price reaction to earnings
12Lobbying reports are filed with the Secretary of the Senate’s Office of Public Records and are availableby calendar year beginning in 1998. The CRP maintains the lobbying data, which we manually match toCompustat by company name.
13These terms include economic, economy, uncertain, uncertainty, Congress, deficit, Federal Reserve, leg-islation, regulation, and White House.
9
![Page 11: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/11.jpg)
announcements based on the firm’s liquidity risk. We use a firm’s liquidity risk instead of
actual liquidity, because we want a measure of the firm’s steady-state return exposure to
liquidity and also because liquidity itself changes endogenously during earnings announce-
ments (Sadka, 2011). We measure a firm’s exposure to expected liquidity risk by the firm’s
βLi from Section 3 of Pastor and Stambaugh (2003). Following Chordia et al. (2014, footnote
7), we compute βLi by running the following regression for each firm using monthly data over
our entire sample period (we are not interested in constructing portfolios for trading):
rit = β0i + βLi Lt + βMi MKTt + βSi SMBt + βHi HMLt + εit, (4)
where L is the Pastor and Stambaugh (2003) Eq. (8) innovation in liquidity measure, r is
firm i’s excess return, MKT is the excess return on a market index, and SMB and HML
are long-short return spreads constructed on sorts of market capitalization and book-to-
market ratio.14 Note that we run this regression at the firm level, not the portfolio level,
to ensure that we can directly link each firm’s βLi to its respective earnings announcement
stock price reactions. Furthermore, note that we are not interested in time-sensitive higher
moments of returns (e.g., autocorrelation, cross-correlation, variance), so we do not need
to be concerned about market microstructure considerations such as bid-ask bounce and
asynchronous trading (Campbell et al., 1996, p. 129; Chordia et al., 2008).
Balakrishnan et al. (2014) argue that managers respond to illiquidity by increasing dis-
closure. Our main disclosure measure, following Balakrishnan et al. (2014), is the frequency
count of management guidance that incorporates capital expenditures, R&D, revenue, or
earnings. We focus on these guidance measures because prior studies have found that eco-
nomic policy uncertainty impacts capital expenditures, R&D, revenue, and earnings (e.g.,
Baker et al., 2016; Bloom et al., 2007; Gulen and Ion, 2016; Julio and Yook, 2012). We
obtain guidance data from I/B/E/S, which covers the universe of public U.S. companies.
14We obtain the innovation in liquidity measure from http://faculty.chicagobooth.edu/lubos.
pastor/research/.
10
![Page 12: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/12.jpg)
Similar to Balakrishnan et al. (2014), we find that the I/B/E/S guidance data do not have
the coverage gaps that were identified in Thomson Reuters’ First Call, which was discontin-
ued in 2012 (Chuk et al., 2013). I/B/E/S confirmed this in response to our inquiries. We
therefore follow Balakrishnan et al. (2014) and code no guidance in I/B/E/S as zero.15 We
also do not distinguish between good news and bad news guidance, as the type of the news is
more relevant when management is trying to raise or lower the stock price, which is not our
focus. Our focus is investor information asymmetry, which can be driven by the withholding
or release of any type of news. In our aggregate market analysis, each guidance disclosure
is weighted by a firm’s market-value weight (on the disclosure date) before aggregating the
guidance disclosures into one daily measure of guidance (our results are similar if we use
equal weights). This procedure ensures that our guidance disclosures are weighted in the
same way as our market-value-weighted liquidity measures. For our unexpected earnings
analysis, we obtain the consensus mean analyst earnings forecast and actual earnings from
I/B/E/S as well.16
We complement the EPU index with two additional means to measure economic pol-
icy uncertainty. First, we use the 2009–2010 Congressional healthcare debates as a quasi-
experimental difference-in-differences (D-in-D) setting that reduces economic policy uncer-
tainty for exposed healthcare firms. Second, we use U.S. presidential elections as a quasi-
experimental setting that increases economic policy uncertainty. We motivate these settings
and discuss them in detail in Section 3.5.
15Non-guiding firms rarely change their guidance practices, as noted in Balakrishnan et al. (2014, SectionIII.D), which could reduce the power of our tests and bias against finding results.
16We recognize that our quantitative metric of disclosure, while based on Balakrishnan et al. (2014), pre-cludes us from doing textual analyses that might measure the extent to which the disclosures explicitly referto economic policy uncertainty. However, in theory, managers’ disclosure response need not necessarily beincreased firm disclosure about economic policy uncertainty, but disclosure of any information that decreasesfirm-level uncertainty.
11
![Page 13: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/13.jpg)
3 Empirical Results
3.1 Univariate Statistics
During our sample period, the EPU index ranges from a low of 0.3821 during December,
2012, to a high of 2.5392 during August, 2011. As Figure 1 shows, other periods of rela-
tively high economic policy uncertainty include national election cycles. Table 1, Panel A
provides the distributions of our aggregate liquidity proxies, and Table 2, Panel A provides
the distributions of our firm-level liquidity proxies, which appear to be consistent with prior
research (Holden and Jacobsen, 2014). Table 1, Panel B and Table 2, Panel B indicate that,
as expected, percent quoted spreads and Amihud illiquidity are positively correlated. We
also find positive correlations between the EPU index and our liquidity proxies, providing
preliminary evidence that increased economic policy uncertainty might be associated with
decreased liquidity (recall that our stock liquidity proxies are inversely related to the level
of liquidity). Table 2, Panel C shows that our firm-level sample includes 6,897 firms, whose
median market value is about $325.2 million. About 29% of firms have foreign income. Table
2, Panel D indicates that market value is correlated with several of the firm-level variables,
underscoring the importance of using firm-year-fixed effects, which control for size and other
firm-specific factors that might vary by year.
3.2 EPU Index and Stock Liquidity
We study the association between the EPU index and stock liquidity using fixed-effect
regressions with clustered standard errors.17 We first study this association at the monthly
level using both aggregate market and firm-level measures of liquidity. For our aggregate
analyses, we weight both of our liquidity proxies by market value at the beginning of the
17For the aggregate regressions, we use time-fixed effects and cluster standard errors robustly or by timeas appropriate. For firm-level regressions, we use firm-time-fixed effects and cluster two way by firm andtime. Such clustering allows for error terms to be correlated both within a firm and across all firms in agiven period of time. The tables contain the exact time-period definitions for fixed effects and clustering.
12
![Page 14: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/14.jpg)
observation month. Our use of value weights follows Chordia et al. (2005), though our results
hold for equal weights as well. We run our monthly tests with year-fixed-effect regressions to
eliminate factors that might affect liquidity across time, including technological advances in
trading and market growth and attrition. Monthly tests mitigate concerns that the findings
of daily tests are driven by spurious short-term correlations in the data. In addition, the
Dickey-Fuller test rejects the null hypothesis of a unit root in all of our aggregate market
variables (at both the daily and monthly intervals), suggesting that our fixed-effect approach
with clustered standard errors is an appropriate specification.
In addition, since the regressions that follow all include at least year-fixed effects, we
eliminate virtually all time trends in the data. Accordingly, we find Durbin-Watson test
statistics of over two for our fixed-effect regressions, which suggests that Newey-West stan-
dard errors are inappropriate for our analysis. Figure 1 also shows graphically that there
is no apparent time trend in the EPU index.18 Finally, note that our monthly and daily
windows are shorter than those of Balakrishnan et al. (2014), who examine the impact of
disclosure in the current quarter on liquidity in the next quarter. They are interested in the
long-run impact of permanent one-time events such as brokerage closures, whereas Figure 1
shows that the EPU index has considerable time-series variation; it is therefore unlikely that
today’s EPU index will have a long-run liquidity impact.
In Table 3, we regress monthly aggregate market averages of the value-weighted liquidity
proxies on the contemporaneous EPU index average for the same month. Table 3, Columns
1 and 2 show that the coefficient on the EPU index is positive and significant for both value-
weighted quoted spreads and value-weighted Amihud illiquidity (1% level). A one standard
deviation increase in the monthly EPU index is associated with a 1.0% and 1.9% increase
in value-weighted monthly spreads and Amihud illiquidity, respectively.19 These findings
obtain after we control for year-fixed effects and the equity uncertainty (EU) index, which
18Nonetheless, we find qualitatively similar results using Newey-West standard errors and Prais-Winstenregressions (untabulated).
19In these computations, we assume that for small x, x = ln(1 + x).
13
![Page 15: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/15.jpg)
is significantly positive for spreads but insignificant for Amihud illiquidity. We next perform
the contemporaneous monthly analysis at the firm level with firm-year-fixed effects.20 Table
3, Columns 3 and 4 show that the EPU index is positive and significant for both quoted
spreads and Amihud illiquidity. A one standard deviation increase in the monthly EPU
index is associated with a 3.4% and 2.5% increase in firm-level monthly spreads and Amihud
illiquidity, respectively. The EU index is positive and marginally significant for both quoted
spreads and Amihud illiquidity, but at smaller economic magnitudes relative to the EPU
index.21 We also find qualitatively similar results with no fixed effects, firm-fixed effects,
and firm-year-month-fixed effects. These findings suggest that increased economic policy
uncertainty decreases liquidity.
In Table 4, we use daily data of the EPU index and liquidity. Table 4, Columns 1 and
4 show that the coefficient on the contemporaneous EPU index is positive and significant
for both value-weighted quoted spreads and value-weighted Amihud illiquidity, respectively,
after we control for year-fixed effects and the EU index. A one standard deviation increase in
the daily EPU index is associated with a 0.45% and 0.42% increase in daily value-weighted
quoted spreads and Amihud illiquidity (1% level). At the daily firm level, Table 4, Columns
7 and 10 show that the coefficient on the EPU index is positive and significant for quoted
spreads and Amihud illiquidity, after we control for firm-year-fixed effects and the EU index.
A one standard deviation increase in the daily EPU index is associated with a 2.69% and
1.56% increase in daily quoted spreads and Amihud illiquidity (1% level). By comparison,
Chordia et al. (2001, p. 509) find the average daily change in their spread variables is on
the order of two percent. The EU index is also positive and significant in Columns 7 and
10, but at smaller economic magnitudes relative to the EPU index. These findings suggest
that economic policy uncertainty has a stronger economic effect on liquidity than general
equity market uncertainty, which is consistent with the conclusions of Baker et al. (2016),
20Chordia et al. (2014, p. 45) suggest supplementing portfolio tests with individual stock tests to avoidsensitivity to portfolio grouping procedures.
21We use the EPU and EU indices in their level forms, but all these findings are qualitatively similar whenwe log transform the indices.
14
![Page 16: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/16.jpg)
who argue that economic policy uncertainty is an important uncertainty to investors in the
market. It is also unlikely that the EPU index is significantly influenced by any individual
firm’s liquidity, so we have some assurance in attributing causality from the EPU index to
individual firm liquidity.
One advantage of our daily data is that we can study how long it takes economic policy
uncertainty to enter into the news (and thus the EPU index), as well as how long it takes
investors to process this information. In Table 4, Columns 2–3 and 5–6, we include five-day
lags and leads of the EPU index. Table 1, Panel A indicates that the daily autocorrelation
coefficient is 0.34 with year-fixed effects, so multicollinearity should not be a severe problem.
The two main results of note are that (1) all the significant EPU index coefficients are
positive, implying that increased economic policy uncertainty decreases liquidity, and (2) the
contemporaneous EPU index is largely significant in the firm-level regressions. Significant
positive lags suggest that asynchronous trading may cause news to be reflected later in
liquidity (Campbell et al., 1996, p. 84).22 Significant positive leads suggest that the EPU
index could be capturing news with some delay relative to investors. The contemporaneous
EPU index regressor is often insignificant in the aggregate regressions, suggesting that its
significance by itself in Columns 1, 4, 7, and 10 indicates that it represents economic policy
uncertainty not just on the day, but also a few days before and after (but it is often significant
in the firm-level regressions). In any event, we are not interested in the time-sensitive
autocorrelation structure, and interpreting the current day EPU index as an amalgamation
of a few days’ lags and leads does not impair our inferences.23
In the above analysis and throughout most of the analysis that follows, we use two-
way clustered errors by firm and an appropriate time dimension. However, since we cannot
directly identify the underlying covariance structure of the error terms, we have ensured
that all of our results obtain with alternative clustering methods, including clustering by
22Note that the EPU index spans from 1985 to current day, so we can include EPU index leads and lagswithout losing any observations.
23Recall that Durbin-Watson tests indicate no presence of autocorrelation in the regressions, suggestingthat EPU index autocorrelation is not driving significance.
15
![Page 17: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/17.jpg)
day, clustering by firm-year, and two-way clustering by firm and year. Also, the fixed
effects used above control for a number of alternative explanations, including persistent firm-
specific factors (such as industry membership) and time-varying firm-specific factors (such
as size and risk) that might affect liquidity. For example, our within-firm-year analyses
control for across-year variation in liquidity that might be due to time-varying firm-specific
or macroeconomic factors (e.g., Chordia et al., 2000). However, one limitation of these
regression specifications is that they restrict the coefficient on the EPU index to be the same
for all firms, a restriction we now relax by allowing the coefficient on the EPU index to change
with several firm-level factors. We add all the EPU index interaction effects at once and also
include the EPU and EU indices as main effects. We do not include the interacting factors
as main effects because they are firm-invariant and thus subsumed by the firm-year-fixed
effects.
We first check whether the variation in firm-specific exposure to economic policy uncer-
tainty changes the relation between economic policy uncertainty and liquidity. To do this,
we compute an EPU index-stock returns sensitivity beta for each firm. As noted in Section
2, to ensure that this beta does not proxy for risk, we use a specification that controls for
the risk factors identified in Fama and French (1993). We use the absolute value of the beta
coefficient because we do not have a clear theoretical expectation about whether a given firm
will be negatively or positively affected by economic policy uncertainty over time.24 By con-
struction, a higher absolute value of the policy beta suggests that a firm is more sensitive to
economic policy uncertainty. Table 5 indeed indicates that spreads and Amihud illiquidity
both increase more in the cross-section for firms with higher policy betas (1% level).
Our second cross-sectional effect relates to firm size. The asset pricing literature has long
found that larger firms, which potentially operate in a richer information environment, are
easier to value (e.g., Fama and French, 1995). This implies that the magnitude of the negative
relation between the EPU index and liquidity might decrease in firm size. Accordingly, in
24Our use of absolute values is similar to the research design of Addoum and Kumar (2016, p. 3473), whoexamine return sensitivity to political uncertainty.
16
![Page 18: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/18.jpg)
Table 5, we interact the EPU index with firm size, as proxied for by the log of market value
of equity, and find a significant negative coefficient on the interaction term for spreads and
Amihud illiquidity. This implies that liquidity declines less in larger firms for a given increase
in the EPU index (1% level). Firm size thus appears to partly mitigate the adverse selection
problems driven by economic policy uncertainty.
Our third cross-sectional effect is whether international exposure reduces a firm’s expo-
sure to U.S. economic policy uncertainty, which is what the EPU index represents. Firms
with international exposure are likely affected by the policy forces of all the countries in
which they operate. To the extent that this diversifies a firm’s exposure to U.S. economic
policy uncertainty, these firms might experience an attenuated relation between the EPU in-
dex and liquidity (e.g., Boutchkova et al., 2012; Desai et al., 2008). We use two measures to
proxy for international exposure: (1) an indicator variable for whether a firm had revenue in
a foreign income over our sample period, and (2) the average number of foreign trade words
in a firm’s 10-Ks over our sample period (using the foreign trade dictionary defined in Sec-
tion 2). Table 5 shows that the interaction effect of the EPU index and the foreign income
indicator is significantly negative for Amihud illiquidity, while the interaction effect with
10-K foreign trade words is uniformly negatively significant. A firm’s international exposure
thus appears to mitigate the effect of U.S. economic policy uncertainty on its liquidity.
Our fourth set of cross-sectional effects focus on lobbying expenditures and economic
policy uncertainty related disclosures in the 10-K (labeled EPU Words in the tables). Prior
literature is mixed on whether lobbying would actually mitigate economic policy uncertainty
or simply proxy for a firm’s exposure to it (e.g., Adelino and Dinc, 2014; Christensen et al.,
2017, p. 92). For example, firms may recognize their increased exposure to economic pol-
icy uncertainty but hire ineffective lobbyists. Economic policy uncertainty related 10-K
disclosures could likewise mitigate or proxy for economic policy uncertainty exposure. For
example, a 10-K disclosure about economic policy uncertainty could provide information to
investors about how economic policy uncertainty will impact the firm, or it could simply
17
![Page 19: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/19.jpg)
describe a firm’s exposure to economic policy uncertainty about which investors are already
informed. Thus, both of these measures could yield results in either direction. To proxy for
lobbying, we use a firm’s annual lobbying dollars averaged over our sample period. We also
compute the average number of economic policy uncertainty related words in a firm’s 10-Ks
over our sample period (see Section 2 for these words). Table 5 shows that the interaction
of the EPU index and lobbying is significantly negative for spreads and Amihud illiquidity,
and the interaction of the EPU index and 10-K economic policy uncertainty words is signifi-
cantly positive for spreads and Amihud illiquidity. Lobbying thus appears to partly mitigate
the adverse effect of economic policy uncertainty on liquidity, whereas 10-K economic policy
uncertainty words appear to proxy for the firm’s exposure to economic policy uncertainty.
3.3 EPU Index and Unexpected Earnings
To the extent that increased economic policy uncertainty decreases liquidity, this could
cause some investors to trade less aggressively on information or withdraw from the market
altogether, which would result in fewer traders (or less trading) in the market for important
corporate information events such as earnings disclosures (e.g., Chordia et al., 2008; Kyle,
1985). We therefore test whether economic policy uncertainty affects the stock price response
to unexpected earnings for firms with high liquidity risk exposure. For a given firm-quarter,
we follow Lys and Sohn (1990) and measure unexpected earnings using actual earnings minus
the analyst consensus mean earnings forecast at day −1 scaled by stock price at day −1,
where day 0 is the quarterly earnings announcement date. Our returns measure is [−1, +1
day] abnormal returns, computed as firm returns minus contemporaneous value-weighted
market returns. Table 6, Column 1 shows that, consistent with prior research, unexpected
earnings is significantly positively associated with [−1, +1 day] abnormal returns. Table 6,
Column 2 further shows that the interaction of unexpected earnings and the EPU index is
significantly negatively associated with [−1, +1 day] abnormal returns, after we control for
firm-year-fixed effects and the EU index. This result implies that increased economic policy
18
![Page 20: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/20.jpg)
uncertainty lowers the stock price response to and information content of earnings news (see
also footnote 5).
However, Table 6, Columns 1 and 2 do not indicate whether economic policy uncertainty
impacts the stock price reaction to unexpected earnings through liquidity. To estimate this
effect, we include a firm’s exposure to expected liquidity risk, as proxied for by βLi from
Section 3 of Pastor and Stambaugh (2003).25 Our focus on the firm’s exposure to liquidity
risk as opposed to its actual liquidity follows the research design in Kelly and Ljungqvist
(2012, Section 3.4).26 Since a firm’s exposure to expected liquidity risk is increasing in βLi ,
we set an indicator variable equal to 1 if a firm is in the top quintile of βLi and interact this
with the EPU index and unexpected earnings.27 We also include similar interaction terms
for a firm’s log of market value of equity and book-to-market ratio.
In Table 6, Column 3, we find a negative coefficient on the three-way interaction term
for the EPU index, unexpected earnings, and the βLi quintile indicator (1% level), which
suggests that for the same level of earnings surprise, investors of firms with high liquidity
exposure react less during periods of high economic policy uncertainty. Economic policy
uncertainty thus appears to have a liquidity channel for earnings. As expected, we also
find that the three-way interaction term with log of market value of equity is positive and
significant, which is consistent with there being fewer pricing frictions for larger firms. The
three-way interaction term with the book-to-market ratio is positive and significant, which
is consistent with there being fewer pricing frictions in relatively mature firms (Fama and
French, 1993).28
The triple-interaction results in Table 6, Column 3 also obtain in a sample that requires
25The exact equation for βLi is in Section 2.
26We want a measure of the firm’s steady-state return exposure to liquidity risk, which a firm’s liquidityby itself does not capture because it changes endogenously during earnings announcements (Sadka, 2011).
27We follow much of the literature by using βLi in a quantile as opposed to a raw form (e.g., Pastor and
Stambaugh, 2003). We also check and find that βLi in our sample is negatively correlated with the betas for
HML and MKT, as in Pastor and Stambaugh (2003, Table 3, Panel B, discussed on p. 668). However, ourcorrelation with the SMB beta is positive, perhaps because we use a later sample than Pastor and Stambaugh(2003).
28Note that we do not have two-way interactions for these risk factors; we assume that the firm-year-fixedeffects are adequate.
19
![Page 21: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/21.jpg)
each firm to have at least five years of monthly returns data in CRSP. This requirement
further ensures that each firm’s βLi is precisely estimated in Eq. (4) of Section 2. If this indeed
occurs, we should find a stronger triple-interaction effect. As expected, in this untabulated
sample we find that the coefficient on the three-way interaction term for the EPU index,
unexpected earnings, and the βLi quintile indicator increases in magnitude by about 21% (-
0.0635 to -0.0767 in levels; 1% level). This is despite the fact that the five-year requirement
decreases the sample by about 18%. This finding suggests that the liquidity channel through
economic policy uncertainty is strong in our setting.
3.4 EPU Index and Voluntary Disclosure
Following Balakrishnan et al. (2014), who argue that managers on average have an incen-
tive to shape liquidity and therefore respond to exogenous liquidity reductions by increasing
voluntary disclosures, we next link economic policy uncertainty to guidance disclosures and
guidance disclosures to liquidity.29
In Table 7, Column 1, we regress current day guidance disclosures on the average EPU
and EU indices from the 30 days prior. Balakrishnan et al. (2014, Section III.D) argue
that changing disclosure policy is costly, and therefore managers may look to longer-term
past EPU index trends rather than daily EPU index trends. To ensure that the market-
level regressions are consistent with and comparable to our liquidity regressions, we weight
each guidance disclosure by the market-value weight of the firm before aggregating all the
disclosures (similar to our liquidity measures). We find that a one standard deviation increase
in the daily EPU index is associated with a 15.87% increase in value-weighted guidance
disclosures, after controlling for year-fixed effects (1% level). In equal-weighted levels, this
translates to about 26 additional daily guidance disclosures. The EU index is insignificantly
associated with guidance disclosures.30 Following Balakrishnan et al. (2014, Table 5), in
29Balakrishnan et al. (2014, p. 2254) note that managers do not respond to liquidity drops due to marketmicrostructure shocks such as market-maker closures. We thus view economic policy uncertainty shocks asnon-market microstructure shocks.
30We also find similar results with equal-weighted guidance disclosures (untabulated).
20
![Page 22: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/22.jpg)
Table 7, Column 2, we validate this finding using a Poisson count model with year-fixed
effects and find that guidance frequency continues to be positively associated with the EPU
index.
The next challenge is linking management guidance to liquidity. This can be problematic
because management may issue guidance to shore up low liquidity. However, as noted above,
Balakrishnan et al. (2014, Section III.D) argue that managers are more likely to respond to
longer-term trends in liquidity, suggesting that at the daily level, liquidity responds to guid-
ance as opposed to the other way around. In fact, if management does respond in a day,
it would result in a positive association between daily illiquidity and disclosure, potentially
canceling the liquidity improvement effect of disclosure (Balakrishnan et al., 2014, p. 2271).
Indeed, Balakrishnan et al. (2014, p. 2266) find a weak “naive” relationship between liq-
uidity and disclosure over both the contemporaneous and the lagging quarter, while Healy
(2015) cites several studies that find a similar positive relation between guidance and mea-
sures of uncertainty (which could reflect both overall investor uncertainty and information
asymmetry among investors). Healy (2015) further notes that guidance is often issued along
with earnings announcements, and some studies show that liquidity is lower on these days
(e.g., Krinsky and Lee, 1996). We therefore show below that our results are not sensitive to
controlling for earnings announcement days. This means that if we find a positive relation
between liquidity and guidance at the daily level, we can be reasonably assured that the
relation is from disclosure to liquidity and not the other way around.31
In Table 7, Columns 3 and 4, we estimate daily aggregate liquidity models by regressing
quoted spreads and the Amihud illiquidity measure on daily guidance disclosures and the
daily EPU index. Most days, firms do not issue guidance and the guidance variable is zero,
which suggests that much of the variation in the guidance variable is coming from guidance
versus non-guidance days. To compute the indirect effect of economic policy uncertainty
(through disclosure) on liquidity, we multiply the guidance coefficients in Columns 3 and 4
31Following the findings of Balakrishnan et al. (2014), we are assuming that managers who do not wish toshape liquidity only reduce the power of our tests and work against our finding a result.
21
![Page 23: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/23.jpg)
by the EPU index coefficient in Column 1. The resulting negative product of −0.0034 for
spreads (0.2383×−0.0141; 1% level) signifies that the indirect effect is present and significant
in our sample.32 For a one standard deviation increase in the EPU index, the indirect effect
decreases spreads by 0.22%, which amounts to 48.89% of the direct effect that the EPU
index has on increasing spreads. This result implies that the additional guidance disclosures
associated with increased economic policy uncertainty increase liquidity, reversing by about
half the adverse liquidity consequences of increased economic policy uncertainty. There is
also a negative indirect effect for Amihud illiquidity, but this effect is statistically insignificant
at the aggregate level (the direct effect is still significant). The presence of year-fixed effects
implies that these magnitudes are to be interpreted as within-year.
We next conduct the same analysis at the firm level using firm-year-fixed effects. In Table
7, Columns 5 and 6, we find that the EPU index is again positively associated with guidance
disclosures (1% level). In Table 7, Columns 7 and 8, the indirect effect of additional guidance
on spreads and Amihud illiquidity is significantly negative (1% level), and the direct effect of
the EPU index on spreads and Amihud illiquidity is significantly positive (1% level). These
results are qualitatively consistent with those at the aggregate market level in Columns 3 and
4. However, the indirect effect reverses the direct effect by only 1.2%, which is smaller than
before. One reason for the difference could be that the aggregate market results are market-
value weighted, which means that those results are driven by proportionally larger firms that
tend to provide more guidance (Anilowski et al., 2007, p. 45). By contrast, the individual
firm analysis weights firms equally.33 Second and related, since large firms disclose guidance
more often, we have many small firms with guidance frequencies of zero that now have equal
weight in the regressions. Such firms rarely change their guidance practices, as noted in
Balakrishnan et al. (2014, Section III.D), which could reduce the power of the individual
firm tests and bias against finding our results. Note that the presence of firm-year-fixed
32We use the delta method (or the linear Taylor expansion) to compute standard errors (Krull and MacK-innon, 2001; Sobel, 1987).
33Consistent with this, when we limit the individual firm analysis to firms in the top market-value quintileof our sample, the ratio of the indirect effect to the direct effect increases.
22
![Page 24: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/24.jpg)
effects implies that these magnitudes are to be interpreted as within-firm, within-year.34
We also find that our firm-level results in Table 7, Columns 5–8 obtain when we include
a firm-level indicator variable for days when firms announce earnings.35 This indicator is
not significantly correlated with the EPU index regressor (the correlation is 0.02), which is
to be expected given that there is no reason for economic policy uncertainty to covary with
earnings announcement dates. Across Columns 5–8, this indicator is positive and significant,
consistent with studies that show that liquidity is lower on earnings announcement days
(e.g., Krinsky and Lee, 1996), and with studies that find that managers bundle guidance with
earnings announcements (e.g., Billings et al., 2015; Healy, 2015). However, the announcement
indicator does not quantitatively change the EPU index coefficient in Columns 5 and 6 or
the guidance and EPU index coefficients in Columns 7 and 8, which further affirms that the
relation we are capturing above is from disclosure to liquidity.36
Management of firms exposed to economic policy uncertainty might invest in lobbying in
part to gain access to additional information about pending policy (e.g., Christensen et al.,
2017; Wellman, 2017). Such managers may be more likely to make disclosures. To check for
such cross-sectional variation in disclosure response, we also run the regression in Table 7,
Column 5 with all the EPU index interaction terms from Table 5, except that we interact the
firm-level terms with the EPU index averaged over t(−30, 0). We find positive significance
for the policy beta and market-value interaction terms, which suggests that managers of
larger firms, who are more likely to provide guidance to begin with (Anilowski et al., 2007,
p. 45), and firms more exposed to economic policy uncertainty disclose more. However, we
do not find significance for the other interaction terms, including the log of lobbying dollars,
which suggests that management’s lobbying efforts (as we measure them) have no effect on
34In papers concurrently released with this study, Bird et al. (2017) and Boone et al. (2017) examinemanagement’s disclosure behavior during political elections for firms in battleground states. But theirinterest is political uncertainty, not economic policy uncertainty.
35We use an indicator instead of a continuous variable because most days are non-announcement days.36Management disclosures could also occur as a response to investors’ information-acquisition activities
as a result of economic policy uncertainty. We cannot directly measure these activities except throughfirm-year-fixed effects and the EPU index.
23
![Page 25: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/25.jpg)
their disclosure activities.
3.5 Additional Tests without the EPU Index
All our tests thus far have used the Baker et al. (2016) EPU index. This index is of high
quality, is widely used, and is available on a continuous basis, which allows us to perform
tests such as the stock price response to earnings in Table 6. However, it would be useful
to document our main results in other settings to ensure that our main results generalize.
We first exploit a recent and important setting that increases economic policy uncertainty
exposure for only a subset of firms. Specifically, we focus on the 2009–2010 Congressional
healthcare debates that occurred during the period leading up to the enactment of the
Affordable Care Act (ACA). This setting is unique in that, as we describe below, it potentially
reduces economic policy uncertainty for healthcare firms (our treatment sample), but not
for other types of firms (our control sample).37
The brief history that follows provides background on this setting. Discussions around
comprehensive healthcare reform emerged in great number over the course of the 2008 presi-
dential campaign, as virtually all of the Democratic platforms espoused significant healthcare
reform (Clinton, 2007; Edwards, 2007; Obama, 2007). Following Barack Obama’s inaugura-
tion in January, 2009, what began as his signature campaign promise became the focus of his
administration. Healthcare reform thus dominated the 2009 and early 2010 Congressional
meeting schedule (see Appendix B for the full schedule of these meetings). These meetings
culminated in the ACA, a 2,500-page bill that was signed into law on March 30, 2010.38 The
37To the extent that these debates increase economic policy uncertainty, this will work against our findinga result.
38The ACA was technically signed into law on March 23, 2010, but the Health Care and EducationReconciliation Act of 2010, which was a pre-planned amendment to the ACA, was signed into law on March30, 2010. This completed the ACA legislation process, and we therefore use the date of March 30, 2010. Wehand collect Congressional meeting dates from two sources. First, we review the legislative calendars fromCongress.gov pertaining to (1) The Affordable Health Choices Act (an early version of the ACA), (2) ThePatient Protection and Affordable Care Act (the bill that was ultimately passed), and (3) The Health Careand Education Reconciliation Act of 2010 (the bill that amended the ACA a week later). Second, we useSenate.gov and House.gov to obtain dates for healthcare hearings held prior to the introduction of the HouseTri-Committee Draft Proposal on June 23, 2009.
24
![Page 26: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/26.jpg)
ACA was the product of significant bargaining within and across party lines during formal
hearings and informal roundtable discussions with corporate interests, consumer groups, and
academic experts (Daschle and Nather, 2010).
One of the more fiercely contested issues throughout the healthcare legislative process was
the so-called “public option” for health insurance, which would allow Americans to select
a government-run healthcare plan. Supporters of the public option argued that it would
lower administrative costs, increase quality of care, and improve access to care (Congress,
2009). With strong support from The White House, House Speaker Nancy Pelosi pushed
to include the public option in the House bill that began circulating in June, 2009 (Bacon
and Kornblut, 2009). But by early July, 2009, as House Democrats were finalizing a version
of the bill, a faction of Democrats (the “Blue Dogs”) voiced concerns over including the
public option (Daschle and Nather, 2010). Meanwhile, Conservatives argued that a public
option would compromise the long-run viability of the private insurance market (Eckerly,
2009). Nonetheless, in the summer of 2009, House committees passed a bill that included
the public option, inciting a contentious debate among Senate Democrats. In October,
2009, Senator Joseph Lieberman, who normally caucused with the Democrats, threatened
to join a Republican filibuster to block any bill that contained a public option (Bash, 2009).
Consequently, the ACA does not include a public option.
The legislative debates surrounding healthcare reform were heated. But what one should
not overlook is that these debates clarified the parameters of healthcare reform and provided
a continuous flow of policy news to the market (e.g., Daschle and Nather, 2010). In other
words, the healthcare debates transformed the disparate proposals and campaign promises
into a fully defined piece of legislation whose design was influenced by healthcare corpora-
tions themselves and their corporate interest groups. As a result, we conjecture that the
Congressional healthcare meetings in 2009 and 2010 reduced economic policy uncertainty in
the healthcare industry.
An important feature of this setting is that we can conduct our analysis in a D-in-D
25
![Page 27: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/27.jpg)
manner: healthcare firms represent our treatment sample, and non-healthcare firms represent
our control sample. Accordingly, we conduct our analysis over the period of January 1, 2008
to March 30, 2010, and we set an ACA Hearing Period indicator equal to 1 from January
1, 2009 to March 30, 2010, and 0 otherwise. As Appendix B notes, these dates align with
the Congressional healthcare meetings and the enactment of the ACA. Our control period
is January 1, 2008 to December 31, 2008 (the healthcare meetings start in January, 2009);
we implement alternative control periods below. Following Christensen et al. (2017), we use
the following SIC codes to construct an indicator variable for healthcare firms: 2830, 2831,
2833–2836, 6300, 6310–6331, 6350, 6351, 6360, 6361, 6370–6379, 6390–6399, 6400–6411,
and 8000–8099. These SIC codes represent healthcare service companies, pharmaceutical
companies, and insurance companies. As before, we need not include the healthcare firm
indicator as a main effect in our regressions. This variable is constant within each firm,
which means that it is subsumed by the firm-year-fixed effects.
Table 8 shows that during the course of the 2009–2010 Congressional healthcare debates
from January 1, 2009 to March 30, 2010, liquidity decreases significantly for non-healthcare
firms relative to the control period of 2008. The coefficient on the ACA Hearing Period main
effect indicates that, for the control firms (i.e., the non-healthcare firms), quoted spreads in-
crease by 2.5% (5% level) and Amihud illiquidity increases by 4.2% (1% level). One potential
explanation for this is that the healthcare debates coincide with the financial crisis and a
recession. However, when we sum the ACA Hearing Period coefficient and the coefficient on
the interaction term of ACA Hearing Period and Healthcare Firm, we find that, for health-
care firms, spreads decrease by 4.7% and Amihud illiquidity decreases by 2.7% (1% level for
both) during the healthcare debate period.39 We also find qualitatively similar results using
data at monthly intervals. This D-in-D evidence suggests that a reduction in economic pol-
icy uncertainty increases stock liquidity. Importantly, our research design controls for other
39The coefficient for the ACA Hearing Period indicator can be interpreted in isolation only when the health-care firm indicator equals zero (i.e., this coefficient interpreted in isolation applies only to non-healthcarefirms). The result for healthcare firms is the sum of the ACA Hearing Period coefficient and the coefficientfor the interaction term, which is significant at the 1% level.
26
![Page 28: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/28.jpg)
macroeconomic phenomena that uniformly affect the liquidity of all firms, and we continue
to control for firm-fixed effects, which eliminate persistent firm-specific liquidity factors.
We also run alternative specifications for this analysis. First, we maintain the same ACA
Hearing Period of January 1, 2009 to March 30, 2010, but we tighten the pre period to smaller
windows ranging from three to nine months. Second, we move back the start date of the
ACA Hearing Indicator to June 23, 2009, the date when healthcare reform officially enters
Congress’s agenda. We find qualitatively similar results in both of these tests. In addition,
we find stronger results when we limit our definition of healthcare firms to health insurance
companies, for whom the healthcare debates likely reduce economic policy uncertainty the
most (SIC codes 6300, 6310–6329, 6370–6379, 6390–6399, and 6400–6411). These analyses
suggest that our selection of control windows and our specific definition of healthcare firms
are not driving our results.
We next focus on the timing of presidential elections. Several theories of election cycles
suggest that increased economic policy uncertainty leading up to U.S. presidential elections
stems from a variety of forces, including candidates’ preferences for inflation, labor policy,
trade, taxes, and government spending (e.g., Alesina, 1987, 1988). Furthermore, predicting
policy outcomes in presidential elections involves incorporating not only the probability that
a particular candidate will win, but also the probability that each piece of that candidate’s
policy agenda will actually be implemented. This issue is further compounded by potential
conflicts that the winner might encounter with administrative officials, judges, and members
of the House and Senate, some of whom will be newly elected as well (Brogaard and Detzel,
2015; Fowler, 2006; Goodell and Vahamaa, 2013; Jens, 2017). Accordingly, Baker et al.
(2016) note that the EPU index increases in the months leading up to presidential elections,
which means our previous EPU index analyses already test whether elections affect liquidity.
Nonetheless, we follow up with tests that explicitly use the election setting.
We define as presidential election periods the dates of July 1, 2004–November 2, 2004;
July 1, 2008–November 4, 2008; and July 1, 2012–November 6, 2012. We start each election
27
![Page 29: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/29.jpg)
period at July 1 because July is the month in which the Democratic and Republican parties
hold their national conventions; the ending November date for each election period represents
election day. To the extent that economic policy uncertainty emerges before July 1 of election
years or persists after election days, this would work against our finding a result (i.e., it might
cause liquidity to decline during the control dates). To appropriately control for time trends
in liquidity, we require that each election period have at least six months of data before and
after the election period. This precludes us from including the 2016 election, as we do not
have data for 2017.
In our first analysis, we regress liquidity on an indicator variable that equals 1 during
the dates of the election periods defined above, and 0 otherwise. We include firm-year-fixed
effects and use all non-election dates from 2003–2015 as our control dates. We exclude
observations from 2016 to mitigate any intervening effect of the 2016 election. Table 9 indi-
cates that our election periods are indeed strongly associated with decreased stock liquidity:
spreads increase by 2.6% and Amihud illiquidity increases by 2.3% (1% level). We also find
qualitatively similar results using data at monthly intervals.40
We then run separate regressions of liquidity for each election, using the same election-
period indicator defined above and using as control dates the non-election-period dates during
the election year, the one year before the election, and the one year after the election. For
example, for the 2008 election we include the years 2007, 2008, and 2009, and we set the
election period indicator to 1 from July 1, 2008–November 4, 2008. In Table 9, Columns 5
and 6, we find that the 2008 election period has the strongest effect (based on coefficient
magnitudes) on liquidity among the three elections that we analyze. In particular, spreads
increase by 5.8% and Amihud illiquidity increases by 3.3% over this period (1% level)—results
that are several times larger than those for the 2004 and 2012 elections. One explanation is
that term limits guaranteed that there would be a new president in 2008, which at the time
40Like Brogaard and Detzel (2015), Jens (2017), and Pasquariello and Zafeiridou (2014), we argue thatelections are exogenous events that unambiguously increase economic policy uncertainty, so we do not includethe EPU or EU index in the election regressions.
28
![Page 30: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/30.jpg)
was significant because both candidates were promising substantial policy changes from the
existing regime (including healthcare reform). By contrast, in the 2004 and 2012 elections,
there was a reasonable chance that the incumbent president would win and maintain (at least
partly) the status quo. To address the concern that the financial crisis and the recession
might be driving our results, we re-run our 2008 election analysis using as a control period
only the first half of 2009, which is when both the crisis and recession are near their peaks
(e.g., Foucault et al., 2013).41 We continue to find comparably strong results for the 2008
election. Collectively, these analyses provide additional evidence that increased economic
policy uncertainty is driving our election results.
Lastly, we test whether increased economic policy uncertainty due to the October, 2013
U.S. government shutdown affects liquidity. During the shutdown, the U.S. government
closed the majority of its agencies and cancelled the release of nearly all its regularly sched-
uled economic disclosures. Standard and Poor’s predicted that the shutdown “shaved at
least 0.6% off of annualized fourth-quarter 2013 GDP growth.” Macroeconomic Advisers
predicted “that a two-week shutdown would directly trim about 0.3 percentage point from
fourth quarter growth, mainly by interrupting the flow of services produced by federal em-
ployees.” Goldman Sachs predicted “that the shutdown would reduce GDP growth by 0.14
percentage points per week, even after most furloughed Department of Defense employees
returned to work.” Mark Zandi of Moody’s noted that “The 16-day Federal shutdown . . . hit
to fourth quarter real GDP is estimated at half a percentage point of growth.” In addition,
Baker et al. (2016) find that the EPU index spikes upward during the shutdown. The above
uncertainty over the impact of the shutdown implies that it can be viewed as a shock that
increases economic policy uncertainty.
Although the government shutdown period was October 1 to October 16, 2013, Baker
et al. (2016) report that the mention of “government shutdown” in news articles spikes lead-
ing up to and especially during the shutdown, and persists for a couple of weeks thereafter.
41Also see https://www.gpo.gov/fdsys/pkg/GPO-FCIC/pdf/GPO-FCIC.pdf.
29
![Page 31: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/31.jpg)
We therefore define the shutdown period as September 15, 2013 to October 30, 2013 and
construct a government shutdown indicator variable accordingly.42 Since we cannot analyze
liquidity during the shutdown days without an appropriate benchmark, we run our analysis
from August 15, 2013 to November 30, 2013, which provides a one-month control period
both before and after the shutdown period.
We run regressions of daily firm-level spreads and Amihud illiquidity on our government
shutdown indicator over the period of August 15, 2013 to November 30, 2013 with firm-
fixed effects. We find that the shutdown period is insignificantly negatively associated with
spreads and Amihud illiquidity. Our conclusions are similar if we limit the shutdown period
to October 1 to October 16, 2013, and if we expand the control period by a month or two in
both directions. These results imply that the shutdown either does not significantly affect
liquidity (relative to the control period), or our tests for this short period lack statistical
power.
4 Conclusion
Economic policy uncertainty is known to affect equity price levels and volatility (Baker
et al., 2016). However, volatility changes do not imply that information asymmetry among
investors has changed. For example, even in the CAPM setting that has no information asym-
metry among investors, price volatility can change if economic policy uncertainty changes the
covariance matrix of the firms’ payoffs. Furthermore, theory alone cannot predict whether
economic policy uncertainty increases some privately informed investors’ advantage over oth-
ers, or whether it is a common noise factor that levels the information playing field. This
study therefore attempts to understand empirically how economic policy uncertainty affects
information asymmetry among investors. Specifically, we show that increased economic pol-
icy uncertainty is associated with increases in firms’ bid-ask spreads and Amihud illiquidity,
42To the extent that economic policy uncertainty due to the shutdown does not fully materialize duringthe time leading up to and following the shutdown, this would work against our finding a result.
30
![Page 32: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/32.jpg)
with decreases to investors’ reactions to earnings in firms exposed to liquidity risk, and with
increases in management disclosures that partly reverse the liquidity drops. Collectively,
these findings suggest that economic policy uncertainty exacerbates information asymmetry
among investors.
Our study focuses exclusively on the stock market, but research has documented that
information asymmetry and liquidity in the stock market and the bond market are interlinked
(Chordia et al., 2005; Hu et al., 2013). Pasquariello and Vega (2009) further show how
public macroeconomic disclosures impact bond liquidity when investors have information
asymmetry. Future studies can therefore examine the effect of economic policy uncertainty
on stock market to bond market linkages.
31
![Page 33: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/33.jpg)
References
Addoum, J. M., Kumar, A., 2016. Political Sentiment and Predictable Returns. The Review
of Financial Studies 29, 3471–3518.
Adelino, M., Dinc, I. S., 2014. Corporate distress and lobbying: Evidence from the Stimulus
Act. Journal of Financial Economics 114, 256–272.
Alesina, A., 1987. Macroeconomic policy in a two-party system as a repeated game. The
Quarterly Journal of Economics 102, 651–678.
Alesina, A., 1988. Macroeconomics and politics. NBER Macroeconomics Annual 3, 13–52.
Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series effects. Journal
of Financial Markets 5, 31–56.
Anilowski, C., Feng, M., Skinner, D. J., 2007. Does earnings guidance affect market returns?
The nature and information content of aggregate earnings guidance. Journal of Accounting
and Economics 44, 36–63.
Bacon, Jr., P., Kornblut, A., 2009. Public Option Called Essential; Democratic Lawmakers
Express Concern. The Washington Post.
Baker, S., Bloom, N., Davis, S., 2016. Measuring Economic Policy Uncertainty. The Quar-
terly Journal of Economics 131, 1593–1636.
Balakrishnan, K., Billings, M., Kelly, B., Ljungqvist, A., 2014. Shaping Liquidity: On the
Causal Effects of Disclosure and Liquidity. The Journal of Finance 69, 2237–2278.
Bash, D., 2009. Public option would lead him to filibuster, key senator says. CNN Politics (on-
line) http://www.cnn.com/2009/POLITICS/10/27/health.care/index.html?iref=24hours.
Billings, M. B., Jennings, R., Lev, B., 2015. On guidance and volatility. Journal of Accounting
and Economics 60, 161–180.
Bird, A., Karolyi, S., Ruchti, T., 2017. Political uncertainty and corporate transparency.
Working Paper.
Bloom, N., Bond, S., Van Reenen, J., 2007. Uncertainty and Investment Dynamics. The
Review of Economic Studies 74, 391–415.
Boone, A., Kim, A., White, J., 2017. Political uncertainty and firm disclosure. Working
Paper.
32
![Page 34: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/34.jpg)
Boutchkova, M., Doshi, H., Durnev, A., Molchanov, A., 2012. Precarious Politics and Return
Volatility. The Review of Financial Studies 25, 1111–1154.
Brogaard, J., Detzel, A., 2015. The Asset-Pricing Implications of Government Economic
Policy Uncertainty. Management Science 61, 3–18.
Campbell, J., Lo, A., MacKinlay, A. C., 1996. The Econometrics of Financial Markets.
Princeton University Press.
Chordia, T., Roll, R., Subrahmanyam, A., 2000. Commonality in liquidity. Journal of Fi-
nancial Economics 56, 3–28.
Chordia, T., Roll, R., Subrahmanyam, A., 2001. Market Liquidity and Trading Activity.
The Journal of Finance 56, 501–530.
Chordia, T., Roll, R., Subrahmanyam, A., 2008. Liquidity and market efficiency. Journal of
Financial Economics 87, 249–268.
Chordia, T., Sarkar, A., Subrahmanyam, A., 2005. An empirical analysis of stock and bond
market liquidity. The Review of Financial Studies 18, 85–129.
Chordia, T., Subrahmanyam, A., Tong, Q., 2014. Have capital market anomalies attenu-
ated in the recent era of high liquidity and trading activity? Journal of Accounting and
Economics 58, 41–58.
Christensen, D., Mikhail, M., Walther, B., Wellman, L., 2017. From K Street to Wall Street:
Political Connections and Stock Recommendations. The Accounting Review 92, 87–112.
Chuk, E., Matsumoto, D., Miller, G., 2013. Assessing methods of identifying management
forecasts: CIG vs. researcher collected. Journal of Accounting and Economics 55, 23–42.
Clinton, H., 2007. American Health Choices Plan: Quality, Affordable Health Care for Every
American, Hillary Clinton for President.
Congress, U. S., 2009. H.R.3200 - America’s Affordable Health Choices Act of 2009,
https://www.congress.gov/bill/111th-congress/house-bill/3200/text.
Daschle, T., Nather, D., 2010. Getting It Done: How Obama and Congress Finally Broke
the Stalemate to Make Way for Health Care Reform. Macmillan.
Desai, M., Foley, C. F., Hines, Jr., J., 2008. Capital structure with risky foreign investment.
Journal of Financial Economics 88, 534–553.
33
![Page 35: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/35.jpg)
Drechsler, I., 2013. Uncertainty, Time-Varying Fear, and Asset Prices. The Journal of Fi-
nance 68, 1843–1889.
Eckerly, S., 2009. Letter Opposing America’s Affordable Health Choices Act of 2009
(H.R.3200), national Federation of Independent Business.
Edwards, J., 2007. Universal health care through shared responsibility, John Edwards for
President.
Fama, E., French, K., 1995. Size and Book-to-Market Factors in Earnings and Returns. The
Journal of Finance 50, 131–155.
Fama, E. F., French, K. R., 1993. Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics 33, 3–56.
Fong, K. Y. L., Holden, C. W., Trzcinka, C. A., 2017. What Are the Best Liquidity Proxies
for Global Research? Review of Finance 21, 1355–1401.
Foucault, T., Pagano, M., Roell, A., 2013. Market Liquidity: Theory, Evidence, and Policy.
Oxford University Press.
Fowler, J. H., 2006. Elections and Markets: The Effect of Partisanship, Policy Risk, and
Electoral Margins on the Economy. The Journal of Politics 68, 89–103.
Gao, M., Huang, J., 2016. Capitalizing on Capitol Hill: Informed trading by hedge fund
managers. Journal of Financial Economics 121, 521–545.
Goodell, J., Vahamaa, S., 2013. US presidential elections and implied volatility: The role of
political uncertainty. Journal of Banking & Finance 37, 1108–1117.
Guay, W., Samuels, D., Taylor, D., 2016. Guiding through the fog: Financial statement
complexity and voluntary disclosure. Journal of Accounting and Economics 62, 234–269.
Gulen, H., Ion, M., 2016. Policy Uncertainty and Corporate Investment. Review of Financial
Studies 29, 523–564.
Hassan, T., Hollander, S., van Lent, L., Tahoun, A., 2017. Aggregate and idiosyncratic
political risk: Measurement and effects. Working Paper.
Healy, P. M., 2015. Discussion of “On Guidance and Volatility”. Journal of Accounting and
Economics 60, 136–140.
34
![Page 36: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/36.jpg)
Hochberg, Y., Sapienza, P., Vissing-Jorgensen, A., 2009. A Lobbying Approach to Evaluating
the Sarbanes-Oxley Act of 2002. Journal of Accounting Research 47, 519–583.
Holden, C., Jacobsen, S., 2014. Liquidity Measurement Problems in Fast, Competitive Mar-
kets: Expensive and Cheap Solutions. The Journal of Finance 69, 1747–1785.
Holden, C., Jacobsen, S., Subrahmanyam, A., 2013. The Empirical Analysis of Liquidity.
Foundations and Trends in Finance 8, 263–365.
Hu, G., Pan, J., Wang, J., 2013. Noise as Information for Illiquidity. The Journal of Finance
68, 2341–2382.
Jens, C. E., 2017. Political uncertainty and investment: Causal evidence from U.S. guber-
natorial elections. Journal of Financial Economics 124, 563–579.
Julio, B., Yook, Y., 2012. Political Uncertainty and Corporate Investment Cycles. The Jour-
nal of Finance 67, 45–83.
Kelly, B., Ljungqvist, A., 2012. Testing Asymmetric-Information Asset Pricing Models. The
Review of Financial Studies 25, 1366–1413.
Kelly, B., Pastor, L., Veronesi, P., 2016. The Price of Political Uncertainty: Theory and
Evidence from the Option Market. The Journal of Finance 71, 2417–2480.
Kim, O., Verrecchia, R., 1994. Market Liquidity and Volume around Earnings Announce-
ments. Journal of Accounting and Economics 17, 41–67.
Krinsky, I., Lee, J., 1996. Earnings announcements and the components of the bid-ask spread.
The Journal of Finance 51, 1523–1535.
Krull, J., MacKinnon, D., 2001. Multilevel Modeling of Individual and Group Level Mediated
Effects. Multivariate Behavioral Research 36, 249–277.
Kyle, A., 1985. Continuous Auctions and Insider Trading. Econometrica 53, 1315–1336.
Lang, M., Maffett, M., 2011. Transparency and liquidity uncertainty in crisis periods. Journal
of Accounting and Economics 52, 101–125.
Lys, T., Sohn, S., 1990. The Association Between Revisions of Financial Analysts’ Earnings
Forecasts and Security-Price Changes. Journal of Accounting and Economics 13, 341–363.
Obama, B., 2007. Barack Obama’s Plan for a Healthy America: Lowering health care costs
and ensuring affordable, high-quality health care for all, Barack Obama for President.
35
![Page 37: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/37.jpg)
Pasquariello, P., Vega, C., 2009. The on-the-run liquidity phenomenon. Journal of Financial
Economics 92, 1–24.
Pasquariello, P., Zafeiridou, C., 2014. Political Uncertainty and Financial Market Quality.
Working Paper.
Pastor, L., Stambaugh, R., 2003. Liquidity risk and expected stock returns. Journal of
Political Economy 111, 642–685.
Pastor, L., Veronesi, P., 2012. Uncertainty about Government Policy and Stock Prices. The
Journal of Finance 67, 1219–1264.
Pastor, L., Veronesi, P., 2013. Political uncertainty and risk premia. Journal of Financial
Economics 110, 520–545.
Sadka, R., 2011. Liquidity risk and accounting information. Journal of Accounting and Eco-
nomics 52, 144–152.
Schoenfeld, J., 2017. The effect of voluntary disclosure on stock liquidity: New evidence from
index funds. Journal of Accounting and Economics 63, 51–74.
Sobel, M., 1987. Direct and Indirect Effects in Linear Structural Equation Models. Sociolog-
ical Methods & Research 16, 155–176.
Wellman, L., 2017. Mitigating political uncertainty. Review of Accounting Studies 22, 217–
250.
36
![Page 38: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/38.jpg)
Appendix A: Variable Definitions for DTAQ Firms from 2003–2016Index i represents each firm, and index t represents the observation day/month. CRP stands for the Center for Responsive Politics. For our monthly tests, we average the EPUand EU indices and the liquidity proxies over the observation month. For our daily tests, we use the daily values of the EPU and EU indices and the liquidity proxies. For theaggregate market tests, we value weight the liquidity proxies and each guidance disclosure based on firm market capitalization on the observation day or at the beginning of theobservation month (depending on daily/monthly tests). Each firm-level static is constant for each firm. See Section 2 for more details on all the variables.
Variable Definition Source
Uncertainty IndicesEconomic Policy Uncertainty (EPU) indext Daily index based on the number of news articles that contain the terms defined in Section 2 Baker et al. (2016)
Equity Uncertainty (EU) indext Daily index based on the number of news articles that contain the terms defined in Section 2 Baker et al. (2016)
Stock Liquidity Proxies Computed Daily
Log of Percent Quoted Spreadsit ln[1 +
[100× Askit−Bidit
(Askit+Bidit)/2
]], time weighted as in Holden and Jacobsen (2014) DTAQ
Log of Amihud Illiquidityit ln[1 +
[106 ×
∑ |Returnit|Dollar Trade V olumeit
]], as in Amihud (2002) CRSP
Disclosure ProxiesUnexpected Earningsit Actual earnings minus the analyst consensus mean earnings forecast at t = −1 scaled by
stock price at t = −1, where t = 0 is the quarterly earnings announcement date I/B/E/S
Log of Guidance Disclosuresit ln[1+ Frequency of guidance that incorporates capital expenditures, R&D, revenue,
or earnings], 0 in the absence of disclosure I/B/E/S
Firm-Level Statics|Policy Beta|i |βE
i | from the regression rit = β0i + βE
i EPUt + βMi MKTt + βS
i SMBt + βHi HMLt + εit, CRSP,
which we run monthly by firm over our sample period Baker et al. (2016)
Log of Market Valuei ln[1 +
[Price× Shares Outstanding
]], averaged over our sample period CRSP
Log of Foreign Incomei 1 if a firm had revenue in a foreign income over our sample period, 0 otherwise Compustat
Log of Trade Wordsi ln[1+The average number of foreign trade words in a firm’s 10-Ks over our sample period
using the foreign trade dictionary defined in Section 2]
SEC EDGAR
Log Lobbying Dollarsi ln[1+ A firm’s annual lobbying dollars averaged over our sample period
]CRP
Log of EPU Wordsi ln[1+The average number of economic policy uncertainty words in a firm’s 10-Ks over our
sample period using the policy uncertainty dictionary defined in Section 2]
SEC EDGAR
Liquidity Risk Exposure (βLi ) βL
i of the regression rit = β0i + βL
i Lt + βMi MKTt + βS
i SMBt + βHi HMLt + εit, CRSP,
which we run monthly by firm over our sample period Pastor and Stambaugh (2003)
37
![Page 39: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/39.jpg)
Appendix B: Affordable Care Act (ACA) Meeting Schedule
Date Chamber Title
1/29/2009 Senate Crossing the Quality Chasm in Health Reform
2/23/2009 Senate Principles of Integrative Health: A Path to Health Care Reform
2/24/2009 Senate Addressing Underinsurance in National Health Reform
2/25/2009 Senate Scoring Health Care Reform: CBO’s Budget Options
3/10/2009 Senate The President’s Fiscal Year 2010 Health Care Proposals
3/11/2009 House Hearing on Health Reform in the 21st Century: Expanding Coverage, Improving Quality, Controlling Costs
3/12/2009 Senate Workforce Issues in Health Care Reform: Assessing the Present and Preparing for the Future
3/18/2009 Senate What is Health Care Quality and Who Decides?
3/24/2009 Senate Addressing Insurance Market Reform in National Health Reform
3/25/2009 Senate The Role of Long-Term Care in Health Reform
4/1/2009 House Health Reform in the 21st Century: Reforming the Healthcare Delivery System
4/22/2009 House Health Reform in the 21st Century: Insurance Market Reforms
4/28/2009 Senate Learning from the States: Individual State Experiences with Health Care Reform Coverage Initiatives in the Context of National Reform
4/29/2009 House Health Reform in the 21st Century: Employer Sponsored Insurance
5/6/2009 House Health Reform in the 21st Century: A Conversation with Health and Human Services Secretary Kathleen Sebelius
6/11/2009 Senate Healthcare Reform
6/23/2009 House The Tri-Committee Draft Proposal for Health Care Reform
6/23/2009 House Comprehensive Health Reform Discussion Draft (Day 1)
6/24/2009 House Health Reform in the 21st Century: Proposals to Reform the Health System
6/24/2009 House Comprehensive Health Reform Discussion Draft (Day 2)
6/25/2009 House Comprehensive Health Reform Discussion Draft (Day 3)
7/14/2009 House America’s Affordable Healthcare Choices Act
7/15/2009 House America’s Affordable Healthcare Choices Act
7/16/2009 House America’s Affordable Healthcare Choices Act
7/17/2009 House America’s Affordable Healthcare Choices Act
7/20/2009 House America’s Affordable Healthcare Choices Act
7/21/2009 House Markup of H.R. America’s Affordable Health Choices Act
7/31/2009 House America’s Affordable Healthcare Choices Act
9/17/2009 Senate Affordable Health Choices Act
9/17/2009 House Patient Protection and Affordable Care Act
9/22/2009 Senate Open Executive Session to Consider an Original Bill Providing for Health Care Reform
9/23/2009 Senate Continuation of the Open Executive Session to Consider an Original Bill Providing for Health Care Reform
9/24/2009 Senate Continuation of the Open Executive Session to Consider an Original Bill Providing for Health Care Reform
9/25/2009 Senate Continuation of the Open Executive Session to Consider an Original Bill Providing for Health Care Reform
9/29/2009 Senate Continuation of the Open Executive Session to Consider an Original Bill Providing for Health Care Reform
9/30/2009 Senate Continuation of the Open Executive Session to Consider an Original Bill Providing for Health Care Reform
(continued on next page)
38
![Page 40: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/40.jpg)
Appendix B: Affordable Care Act (ACA) Meeting Schedule (continued)
Date Chamber Title
10/1/2009 Senate Continuation of the Open Executive Session to Consider an Original Bill Providing for Health Care Reform
10/7/2009 House Patient Protection and Affordable Care Act
10/8/2009 Senate Patient Protection and Affordable Care Act
10/8/2009 House Patient Protection and Affordable Care Act
10/13/2009 Senate Continuation of the Open Executive Session to Consider an Original Bill Providing for Health Care Reform
10/13/2009 Senate Patient Protection and Affordable Care Act
10/14/2009 House America’s Affordable Healthcare Choices Act
11/19/2009 Senate Patient Protection and Affordable Care Act
11/20/2009 Senate Patient Protection and Affordable Care Act
11/21/2009 Senate Patient Protection and Affordable Care Act
11/30/2009 Senate Patient Protection and Affordable Care Act
12/1/2009 Senate Patient Protection and Affordable Care Act
12/2/2009 Senate Patient Protection and Affordable Care Act
12/3/2009 Senate Patient Protection and Affordable Care Act
12/4/2009 Senate Patient Protection and Affordable Care Act
12/5/2009 Senate Patient Protection and Affordable Care Act
12/6/2009 Senate Patient Protection and Affordable Care Act
12/7/2009 Senate Patient Protection and Affordable Care Act
12/8/2009 Senate Patient Protection and Affordable Care Act
12/9/2009 Senate Patient Protection and Affordable Care Act
12/10/2009 Senate Patient Protection and Affordable Care Act
12/13/2009 Senate Patient Protection and Affordable Care Act
12/14/2009 Senate Patient Protection and Affordable Care Act
12/15/2009 Senate Patient Protection and Affordable Care Act
12/16/2009 Senate Patient Protection and Affordable Care Act
12/17/2009 Senate Patient Protection and Affordable Care Act
12/19/2009 Senate Patient Protection and Affordable Care Act
12/20/2009 Senate Patient Protection and Affordable Care Act
12/21/2009 Senate Patient Protection and Affordable Care Act
12/22/2009 Senate Patient Protection and Affordable Care Act
12/23/2009 Senate Patient Protection and Affordable Care Act
12/24/2009 Senate Patient Protection and Affordable Care Act
12/29/2009 Senate Patient Protection and Affordable Care Act
3/17/2010 House Health Care and Education Reconciliation Act of 2010
3/21/2010 House Health Care and Education Reconciliation Act of 2010
3/21/2010 House Patient Protection and Affordable Care Act
(continued on next page)
39
![Page 41: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/41.jpg)
Appendix B: Affordable Care Act (ACA) Meeting Schedule (continued)
Date Chamber Title
3/22/2010 House Patient Protection and Affordable Care Act
3/23/2010 Senate Health Care and Education Reconciliation Act of 2010
3/23/2010 House Patient Protection and Affordable Care Act
3/23/2010 President Patient Protection and Affordable Care Act
3/24/2010 Senate Health Care and Education Reconciliation Act of 2010
3/25/2010 Senate Health Care and Education Reconciliation Act of 2010
3/30/2010 President Health Care and Education Reconciliation Act of 2010
40
![Page 42: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/42.jpg)
Figure 1: The Baker et al. (2016) Economic Policy Index Plotted from 2003–2016
This figure charts the Baker et al. (2016) economic policy uncertainty (EPU) index averaged by year-month for display purposes (we use the dailyvalue in our regressions). The index is provided on a daily basis on http://www.policyuncertainty.com/. The source for news articles is Newsbank,which covers U.S.-based media ranging from large national media like USA Today to small local media. See Baker et al. (2016) for more details.
0
50
100
150
200
250
300
2003
01
2003
05
2003
09
2004
01
2004
05
2004
09
2005
01
2005
05
2005
09
2006
01
2006
05
2006
09
2007
01
2007
05
2007
09
2008
01
2008
05
2008
09
2009
01
2009
05
2009
09
2010
01
2010
05
2010
09
2011
01
2011
05
2011
09
2012
01
2012
05
2012
09
2013
01
2013
05
2013
09
2014
01
2014
05
2014
09
2015
01
2015
05
2015
09
2016
01
2016
05
2016
09
Eco
nom
ic P
olicy
Unce
rtain
ty I
ndex
Year-Month
41
![Page 43: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/43.jpg)
Table 1: Variable Descriptive Statistics for Daily Aggregate Measures from 2003–2016Value-weighted (VW) measures are weighted by market capitalization at day t and logged for all analyses. The column AR(1) ρ w/ yr. F.E. inPanel A is the autocorrelation coefficient with year-fixed effects (i.e., within year). For display purposes, we multiply the value-weighted aggregatedisclosures by 100. Index t represents the observation day. All variables are defined in Appendix A.
Panel A: Descriptive Statistics of the Uncertainty Variables and the Aggregate Liquidity and Guidance Variables
AR(1) ρVariable N Mean σ P25 P50 P75 AR(1) ρ w/ yr. F.E.
EPU Indext 3,351 0.978 0.666 0.510 0.803 1.262 0.55 0.34EU Indext 3,351 0.469 0.679 0.121 0.259 0.542 0.45 0.40Log VW Quoted Spreadt 3,351 0.086 0.059 0.069 0.077 0.089 0.07 0.04Log VW Amihud Illiquidityt 3,351 0.025 0.043 0.009 0.013 0.026 0.18 0.12Log VW Aggregate Guidance Disclosurest × 100 3,351 1.258 0.834 0.564 1.188 1.863 0.49 0.38
Panel B: Correlation Matrix of the Uncertainty Variables and the Aggregate Liquidity and Guidance Variables
(1) (2) (3) (4) (5)
EPU Indext 1.00EU Indext 0.40∗∗∗ 1.00Log VW Quoted Spreadt 0.07∗∗∗ 0.11∗∗∗ 1.00Log VW Amihud Illiquidityt 0.21∗∗∗ 0.13∗∗∗ 0.18∗∗∗ 1.00Log VW Aggregate Guidance Disclosurest -0.02 -0.01 -0.23∗∗∗ -0.04∗ 1.00
42
![Page 44: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/44.jpg)
Table 2: Variable Descriptive Statistics for Daily Firm-Level Measures from 2003–2016The number of observations in Panel A and in our regressions vary according to data availability in DTAQ and CRSP. We multiply |Policy Beta|i by1,000 for display purposes. Index i represents each firm, and index t represents the observation day. All variables are defined in Appendix A.
Panel A: Descriptive Statistics of the Daily Economic Policy Uncertainty Index and Daily Firm-Level Liquidity Variables
Variable N Mean σ P25 P50 P75
Log Quoted Spreadit 15,127,411 0.455 0.520 0.115 0.243 0.579Log Amihud Illiquidityit 14,775,942 0.146 0.446 0.001 0.005 0.042Log Guidance Disclosuresit 15,129,038 0.013 0.129 0.000 0.000 0.000
Panel B: Correlation Matrix of the Daily Economic Policy Uncertainty Index and Daily Firm-Level Liquidity Variables
EPU Indext EU Indext Log Quoted Spreadit Log Amihud Illiquidityit Log Guidance Disc.it
EPU Indext 1.00EU Indext 0.40∗∗∗ 1.00Log Quoted Spreadit 0.08∗∗∗ 0.06∗∗∗ 1.00Log Amihud Illiquidityit 0.07∗∗∗ 0.04∗∗∗ 0.68∗∗∗ 1.00Log Guidance Disclosuresit 0.01∗ -0.01∗∗∗ -0.04∗∗∗ -0.03∗∗∗ 1.00
Panel C: Descriptive Statistics of Static Firm-Level Variables
Variable N Mean σ P25 P50 P75
|Policy Beta|i×1, 000 6,897 0.011 0.023 0.003 0.007 0.014Log of Market Valuei 6,897 19.820 1.811 18.522 19.682 21.000Foreign Incomei 6,897 0.287 0.453 0.000 0.000 1.000Log of Trade Wordsi 6,897 0.074 0.283 0.000 0.000 0.000Log Lobbying Dollarsi 6,897 2.552 5.051 0.000 0.000 0.000Log of EPU Wordsi 6,897 2.120 2.005 0.000 3.1381 3.9576
Panel D: Correlation Matrix of Static Firm-Level Variables
|Policy Beta|i Log of Market Valuei Foreign Incomeit Trade Wordsit Log Lobbying Dollarsit Log of EPU Wordsit
|Policy Beta|i 1.00Log of Market Valuei -0.15∗∗∗ 1.00Foreign Incomei -0.07∗∗∗ 0.21∗∗∗ 1.00Log of Trade Wordsi -0.00 0.07∗∗∗ 0.07∗∗∗ 1.00Log Lobbying Dollarsi -0.10∗∗∗ 0.37∗∗∗ 0.12∗∗∗ 0.07∗∗∗ 1.00Log of EPU Wordsi 0.00 0.25∗∗∗ -0.02∗ 0.07∗∗∗ 0.07∗∗∗ 1.00
43
![Page 45: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/45.jpg)
Table 3: Regressions of Stock Liquidity on the Economic Policy Uncertainty Index using Monthly Data from 2003–2016This table uses the monthly average value of all variables. The EPU Index stands for the economic policy uncertainty (EPU) index, and the EU Indexstands for the equity uncertainty (EU) index. This analysis spans our entire sample of 2003 to 2016. Value-weighted (VW) measures are weightedby market capitalization at the beginning of observation month t. We find Durbin-Watson test statistics of over two for all of our regressions, whichindicates that standard errors clustered across time are appropriate for our analysis (see Section 3.2). When two parameters are separated by acomma in the S.E. Clustering row, this implies two-way clustering. YM stands for year-month. Index i represents each firm, and index t representsthe observation month. All variables are defined in Appendix A. T-statistics are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
Monthly Data
Aggregate Market Level Firm Level
Log VW Quoted Spreadt Log VW Amihud Illiquidityt Log Quoted Spreadit Log Amihud Illiquidityit
(1) (2) (3) (4)
EPU Indext 0.0151*** 0.0290*** 0.0523*** 0.0379***(3.18) (3.17) (3.16) (3.43)
EU Indext 0.0217*** -0.0029 0.0639** 0.0263**(2.97) (-0.35) (2.51) (1.69)
Fixed Effects Year Year Firm-Year Firm-YearS.E. Clustering Robust Robust Firm, YM Firm, YMObservations 160 160 725,842 720,960R2 0.47 0.57 0.94 0.88
44
![Page 46: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/46.jpg)
Table 4: Regressions of Stock Liquidity on the Economic Policy Uncertainty Index using Daily Data from 2003–2016This table uses the daily value of all variables. The EPU Index stands for the economic policy uncertainty (EPU) index, and the EU Index stands forthe (EU) equity uncertainty index. This analysis spans our entire sample of 2003 to 2016. Value-weighted (VW) measures are weighted by marketcapitalization at day t. We find Durbin-Watson test statistics of over two for all of our regressions, which indicates that standard errors clusteredacross time are appropriate for our analysis (see Section 3.2). When two parameters are separated by a comma in the S.E. Clustering row, this impliestwo-way clustering. Index i represents each firm, and index t represents the observation day. All variables are defined in Appendix A. T-statisticsare in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
Daily Data
Aggregate Market Level Firm Level
Log VW Quoted Spreadt Log VW Amihud Illiquidityt Log Quoted Spreadit Log Amihud Illiquidityit
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
EPU Indext−5 0.0035* 0.0051*** 0.0104*** 0.0079***(1.68) (3.63) (3.39) (4.50)
EPU Indext−4 0.0098*** 0.0019 0.0111*** 0.0064***(4.39) (1.25) (3.11) (3.34)
EPU Indext−3 0.0008 0.0017 0.0155** 0.0068***(0.37) (1.15) (2.18) (3.24)
EPU Indext−2 0.0008 0.0048*** 0.0082* 0.0048***(0.37) (3.21) (1.93) (2.69)
EPU Indext−1 -0.0008 0.0019 0.0052 0.0048***(-0.37) (1.28) (1.60) (2.68)
EPU Indext 0.0067*** 0.0020 0.0006 0.0063*** 0.0009 0.0021 0.0269*** 0.0054 0.0097*** 0.0156*** 0.0050*** 0.0069***(3.47) (0.92) (0.27) (4.87) (0.61) (1.27) (8.28) (1.46) (2.82) (9.78) (2.92) (3.90)
EPU Indext+1 0.0031 0.0017 0.0089** 0.0041**(1.32) (0.75) (2.27) (2.39)
EPU Indext+2 0.0032 0.0052 0.0117*** 0.0069***(1.40) (1.72) (3.24) (3.65)
EPU Indext+3 0.0020 0.0038 0.0082** 0.0047**(0.89) (1.69) (2.06) (2.51)
EPU Indext+4 0.0029 -0.0002 0.0081** 0.0041**(1.26) (-0.16) (2.23) (2.42)
EPU Indext+5 0.0072** 0.0024 0.0154*** 0.0068***(2.58) (1.43) (4.39) (3.94)
EU Indext 0.0047*** 0.0032* 0.0024 0.0037*** 0.0083*** 0.0019 0.0183*** 0.0098*** 0.0113*** 0.0078*** 0.0046*** 0.0043***(2.88) (1.96) (1.23) (3.37) (4.76) (1.29) (6.74) (3.99) (4.43) (5.08) (3.56) (3.03)
Fixed Effects Year Year Year Year Year Year Firm-Year Firm-Year Firm-Year Firm-Year Firm-Year Firm-YearS.E. Clustering Year-Qtr. Year-Qtr. Year-Qtr. Year-Qtr. Year-Qtr. Year-Qtr. Firm, Day Firm, Day Firm, Day Firm, Day Firm, Day Firm, DayObservations 3,351 3,351 3,351 3,351 3,351 3,351 15,127,411 15,127,411 15,127,411 14,775,942 14,775,942 14,775,942R2 0.06 0.08 0.08 0.20 0.21 0.21 0.84 0.84 0.84 0.57 0.57 0.57
45
![Page 47: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/47.jpg)
Table 5: Regressions of Stock Liquidity on Firm-Level Statics Interacted with the Economic Policy Uncertainty Index from2003–2016This analysis spans our entire sample of 2003 to 2016. The interacted variables with subscript i are constant for each firm and vary only in the cross-section, which means that they are collinear with the fixed effects. Our interaction tests should thus be interpreted as across-firm, not within-firm,results. We include the main effect for the EPU and the EU indices. When two parameters are separated by a comma in the S.E. Clustering row,this implies two-way clustering. YM stands for year-month. Index i represents each firm, and index t represents the observation day/month. Allvariables are defined in Appendix A. T-statistics are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
Log Quoted Spreadit Log Amihud Illiquidityit
Monthly Daily Monthly Daily(1) (2) (3) (4)
EPU Indext 0.0533824*** 0.226018*** 0.039977*** 0.233178***(3.17) (13.88) (3.09) (12.84)
EPU Indext× |Policy Beta|i 5.630217*** 1.515130*** 7.620135*** 3.098189***(4.39) (3.85) (3.97) (6.12)
EPU Indext× Log of Market Valuei -0.000300** -0.010349*** -0.000786*** -0.011545***(-2.60) (-14.27) (-5.05) (-13.02)
EPU Indext× Foreign Incomei -0.000022 -0.000002 -0.000021** -0.000014***(-1.56) (-0.51) (-2.48) (-5.66)
EPU Indext× Log of Trade Wordsi -0.000057*** -0.000028*** -0.000069*** -0.000029***(-4.57) (-8.08) (-4.42) (-7.82)
EPU Indext× Log of Lobbying Dollarsi -0.000012*** -0.000001*** -0.000001*** -0.000003***(-4.56) (-2.74) (-4.30) (-12.26)
EPU Indext× Log of EPU Wordsi 0.000015** 0.000020*** 0.000035*** 0.000041***(2.18) (6.18) (5.12) (11.78)
EU Indext 0.063782** 0.018273*** 0.026221* 0.007768***(2.52) (6.76) (1.70) (5.07)
Fixed Effects Firm-Year Firm-Year Firm-Year Firm-YearTwo-Way S.E. Clustering Firm, YM Firm, Day Firm, YM Firm, DayObservations 725,842 15,127,411 720,960 14,775,942R2 0.94 0.84 0.88 0.57
46
![Page 48: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/48.jpg)
Table 6: Stock Price Reaction to Unexpected Earnings from 2003–2016For a given firm i quarter, we measure unexpected earnings as actual earnings minus the analyst consensus mean earnings forecast at t = −1 scaled bystock price at t = −1, where t = 0 is the earnings announcement date (Lys and Sohn, 1990). The EPU and EU indices are averaged over year-quarterq prior to the earnings announcement date. Returns are firm returns minus the value-weighted market return from CRSP. To estimate the effect ofeconomic policy uncertainty on the stock price reaction to unexpected earnings through liquidity, we first follow Section 3 of Pastor and Stambaugh(2003) and run the following regression for each firm using monthly data over our sample period: rit = β0
i +βLi Lt+β
Mi MKTt+β
Si SMBt+β
Hi HMLt+
εit, where L is the Pastor and Stambaugh (2003) innovation in liquidity measure (which we obtain from their website), r is firm i’s excess return,MKT is the excess return on a market index, and SMB and HML are long-short return spreads constructed on sorts of market capitalization andbook-to-market ratio (see Fama and French, 1993). Note that we run this regression at the firm level, not the portfolio level, to ensure that we candirectly link each firm’s βL
i to its respective stock price reaction to earnings. Since a firm’s exposure to expected liquidity risk is increasing in βLi ,
we then set an indicator variable equal to 1 if a firm is in the top quintile of βLi (0 otherwise) and interact this with the EPU index and unexpected
earnings. We interpret this three-way interaction term as the effect of economic policy uncertainty on the stock price reaction to unexpected earningsthrough liquidity (see Section 3.3). Note that we do not have two-way interactions for these risk factors; we assume that the firm-year-fixed effectsare adequate. Index i represents each firm, and index t represents the observation day. All variables are defined in Appendix A. T-statistics are inparentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
(1) (2) (3)Returns [−1, +1 day]it Returns [−1, +1 day]it Returns [−1, +1 day]it
Unexpected Earningsit 1.0078*** 1.2239*** 1.2124***(25.50) (13.09) (14.72)
EPU Indexq -0.0019 -0.0022(-0.54) (-0.64)
EPU Indexq × Unexpected Earningsit -0.1776*** -1.3110***(-3.37) (-4.53)
EU Indexq -0.0046 -0.0047(-1.61) (-1.59)
EPU Indexq × UEit × Top βLi Quintileit -0.0635***
(-2.70)
EPU Indexq × UEit × Log of Market Valueit 0.1010***(4.11)
EPU Indexq × UEit × BTMit 0.0003***(2.97)
Fixed Effects Firm-Year Firm-Year Firm-YearS.E. Clustering Year-Quarter Year-Quarter Year-QuarterObservations 198,414 198,414 198,414R2 0.32 0.33 0.33
47
![Page 49: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/49.jpg)
Table 7: Daily Stock Liquidity, Voluntary Disclosure, and the Economic Policy Uncertainty Index from 2003–2016We estimate the effect of economic policy uncertainty induced disclosure on liquidity by first regressing daily guidance on the daily EPU index inColumn 1. In Columns 3 and 4, we then regress daily quoted spreads and Amihud illiquidity on daily guidance and the EPU index. We estimatethe effect of aggregate guidance disclosures on aggregate quoted spreads and Amihud illiquidity by multiplying the guidance disclosure coefficient inColumns 3 and 4 by the EPU index coefficient in Column 1. We likewise estimate the effect of firm-level guidance disclosures on firm-level quotedspreads and Amihud illiquidity by multiplying the guidance disclosure coefficient in Columns 7 and 8 by the EPU index coefficient in Column 5. Wedescribe the model further and interpret the results in Section 3.4. Z-statistics are provided in parentheses for the Poisson count regressions; otherwiset-statistics are in parentheses. We do not log guidance frequency for the Poisson count regressions. Value-weighted (VW) measures are weighted bymarket capitalization at day t. Index i represents each firm, and index t represents the observation day. Variables with only subscript t representday t = 0. Variables with subscript t(−30, 0) are averaged over days t = −30 to t = 0. When two parameters are separated by a comma in the S.E.Clustering row, this implies two-way clustering. All variables are defined in Appendix A. * p < 0.10, ** p < 0.05, *** p < 0.01.
Daily Data
Aggregate Market Level Firm Level
(1) (2) (3) (4) (5) (6) (7) (8)Log VW Guidancet
∑iGuidanceit Log VW Spreadst Log VW Amihudt Log Guidanceit Guidanceit Log Spreadsit Log Amihudit
EPU Indext(−30,0) 0.2383*** 0.0392*** 0.0225*** 0.0124***(3.39) (9.21) (2.83) (6.65)
EU Indext(−30,0) -0.1443 -0.0149** -0.0015 -0.0112*(-1.19) (-2.44) (-1.25) (-1.96)
Log VW Guidancet -0.0141*** -0.0005(-6.96) (-1.35)
Log Guidanceit -0.0143*** -0.0026***(-4.04) (-8.53)
EPU Indext 0.0068*** 0.0063*** 0.0269*** 0.0156***(3.60) (4.86) (8.28) (9.78)
EU Indext 0.0044*** 0.0036*** 0.0183*** 0.0078***(2.77) (3.35) (6.74) (5.08)
Estimation OLS Poisson OLS OLS OLS Poisson OLS OLSFixed Effects Year Year Year Year Firm-Year Firm-Year Firm-Year Firm-YearS.E. Clustering Robust Robust Robust Robust Firm, Day Firm, Day Firm, Day Firm, DayObservations 3,351 3,351 3,351 3,351 15,129,038 15,129,038 15,127,411 14,775,942R2 0.02 0.01 0.12 0.20 0.01 0.01 0.86 0.58
48
![Page 50: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/50.jpg)
Table 8: Difference-in-Differences Regressions of Stock Liquidity for the 2009–2010 Affordable Care Act Congressional DebatesThe sample period is January 1, 2008 to March 30, 2010, the date when the ACA was signed into law. The ACA Hearing Period indicator equals 1from January 1, 2009 to March 30, 2010 (0 otherwise). Following Christensen et al. (2017), the healthcare indicator variable represents firms withSIC codes of 2830, 2831, 2833–2836, 6300, 6310–6331, 6350, 6351, 6360, 6361, 6370–6379, 6390–6399, 6400–6411, and 8000–8099. These SIC codesinclude healthcare service companies, pharmaceutical companies, and insurance companies. The remainder of the sample is included as non-healthcarecontrol firms. As in Table 5, we do not include the healthcare firm indicator as a main effect in our regressions. This variable is constant within eachfirm and varies only in the cross-section, which means it is collinear with the firm-fixed effects. As Section 3.5 notes, we also run several alternativespecifications. When two parameters are separated by a comma in the S.E. Clustering row, this implies two-way clustering. Index i represents eachfirm, and index t represents the observation day. All variables are defined in Appendix A. T-statistics are in parentheses. * p < 0.10, ** p < 0.05,*** p < 0.01.
(1) (2)Log Quoted Spreadit Log Amihud Illiquidityit
ACA Hearing Periodt 0.024783** 0.041764***(1.99) (4.84)
ACA Hearing Periodt× Healthcare Firmi -0.071629*** -0.068840**(-3.17) (-2.46)
Fixed Effects Firm FirmS.E. Clustering Firm, Day Firm, DayObservations 2,769,827 2,694,356R2 0.80 0.56
49
![Page 51: Economic Policy Uncertainty and Information Asymmetry€¦ · Economic Policy Uncertainty and Information Asymmetry Venky Nagar University of Michigan Jordan Schoenfeld ... we control](https://reader035.vdocument.in/reader035/viewer/2022062916/5ebff7f302ee571eef6ae677/html5/thumbnails/51.jpg)
Table 9: Regressions of Stock Liquidity for the U.S. Presidential Elections of 2004, 2008, and 2012The U.S. presidential election period indicator equals 1 for the following dates: July 1, 2004–November 2, 2004 (2004 election); July 1, 2008–November4, 2008 (2008 election); and July 1, 2012–November 6, 2012 (2012 election); and 0 otherwise. We start each election period at July 1 because Julyis the month in which the Democratic and Republican parties hold their national conventions; the ending November date for each election periodrepresents election day. As Section 3.5 notes, we also run several alternative specification windows. When two parameters are separated by a commain the S.E. Clustering row, this implies two-way clustering. Index i represents each firm, and index t represents the observation day. All variablesare defined in Appendix A. T-statistics are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01.
2004, 2008, 2012 Elections 2004 Election 2008 Election 2012 Election
Log of Log of Log of Log of(1) (2) (3) (4) (5) (6) (7) (8)
Spreadsit Amihudit Spreadsit Amihudit Spreadsit Amihudit Spreadsit Amihudit
Presidential Election Periodt 0.0262*** 0.0232*** 0.0105*** 0.0287*** 0.0581*** 0.0333*** 0.0039** 0.0072***(2.97) (6.88) (3.13) (12.05) (2.91) (3.84) (2.30) (4.07)
Fixed Effects Firm-Year Firm-Year Firm-Year Firm-Year Firm-Year Firm-Year Firm-Year Firm-YearS.E. Clustering Firm, Day Firm, Day Firm, Day Firm, Day Firm, Day Firm, Day Firm, Day Firm, DayObservations 14,015,922 13,677,327 3,314,059 3,222,753 3,661,403 3,557,507 3,421,711 3,357,980R2 0.84 0.57 0.83 0.52 0.83 0.59 0.84 0.58
50