10 1111@acfi 12722

Accounting & Finance
Economic policy uncertainty exposure and earnings

management: evidence from China
Xin Cuia, Shouyu Yaoa, Zhenming Fanga, Hua Wangb

a
College of Management and Economics, Tianjin University, Tianjin, China
b
Zhongnan University of Economics and Law, Wuhan, China
Abstract
We investigate the impact of economic policy uncertainty (EPU) exposure on

the earnings management behaviour of Chinese firms. We find a significantly
positive relation between EPU exposure and firms’ earnings management. In
addition, the EPU exposure effect is more pronounced for firms with weaker
external monitoring mechanisms. We also find that the financial leverage and
growth rate of individual stocks have significant predictive ability for EPU
exposure. When we examine the potential mechanisms linking EPU exposure
to earnings management, we find that financial distress is the dominant
mechanism for firms with high leverage, while it is cash flow volatility for the
high-growth firms.
Key words: Cash flow volatility; Earnings management; Economic policy

uncertainty exposure; Financial distress
JEL classification: O31, G38
doi: 10.1111/acfi.12722
1. Introduction
Economic policy uncertainty (EPU) refers to the condition that economic

entities cannot predict exactly whether, when and how the government will
The authors would like to thank Tom Smith (Editor-in-Chief), Jing Shi (Editor),
anonymous referees, and the workshop of Tianjin University Financial Engineering
Research Center, for valuable comments and suggestions. We also acknowledge
financial support from the National Natural Science Foundation of China (Grant
numbers: 72073101; 71671122).
Please address correspondence to Hua Wang via email: [email protected]
© 2020 Accounting and Finance Association of Australia and New Zealand

2 X. Cui et al./Accounting & Finance
change its economic policies (Gulen and Ion, 2015). In recent years,
governmental changes to economic policies concerning regulatory, fiscal and
monetary issues are regarded as important sources of uncertainty and can
greatly impact financial markets. Studies investigate the effect of market EPU
on other economic and financial issues, such as commodity markets (Wang
et al., 2015), global stock market risk (Tsai, 2017), cash holdings (Demir and
Ersan, 2017), stock liquidity (Nagar et al., 2019), overseas investment (Wu
et al., 2020) and capital structure (Huang et al., 2020).
When linking external EPU with corporate decisions, most studies assume
that different stocks face the same market EPU (e.g., Demir and Ersan, 2017;
Huang et al., 2020; Wu et al., 2020), thus ignoring heterogeneity at the
individual stock level. Only a few studies consider the heterogeneity of EPU
exposure at the individual stock level when evaluating corporate innovation
(Xu, 2020) or firm value (Yang et al., 2019). In this paper, in line with Yang
et al. (2019) and Xu (2020), we also consider the heterogeneity of a stock’s
exposure to market EPU. Then, we explore the influence of idiosyncratic EPU
exposure on the micro-firm’s earnings management behaviour and add new
micro-level evidence to the findings on macro-level social and economic
impacts.
Our analysis is motivated by two opposing strands of research in corporate
finance concerning the effect of EPU on earnings management. In the first
strand, some studies find that firms may conduct more earnings management
when there is an increase in EPU. They report that changes in the external
economic environment may cause large fluctuations in corporate earnings,
resulting in a series of potentially negative consequences (Francis et al., 2004;
Graham et al., 2005; Borghesi et al., 2014). On the one hand, earnings volatility
may signal the unstable operating conditions of a firm (Francis et al., 2004),
which may have a negative impact on both managers’ reputation and
compensation (Graham et al., 2005; Borghesi et al., 2014). On the other hand,
only when earnings reach their target can firms preserve or obtain resources
and achieve their goals, such as preventing delisting (Campbell et al., 2015),
avoiding regulatory intervention (Pierk and Weil, 2016), reducing the possi-
bility of loan default (Ghosh and Moon, 2010), meeting analyst forecasts
(Burgstahler and Eames, 2006; McVay, 2006) and fulfilling the requirements of
initial public offerings (IPOs) and secondary equity offerings (SEOs) (Aharony
et al., 2010; He, 2016), among others. Therefore, when firms face high EPU
exposure, managers may have more incentives to engage in earnings manage-
ment to smooth earnings volatility, improve the persistence and predictability
of earnings, and convey a positive signal to stakeholders (Subramanyam, 1996;
Huang et al., 2013).
In the second strand of the literature, studies suggest that EPU possibly
reduces earnings management. Studies argue that external uncertainty (char-
acterised as possible changes in government policies and/or to political
leadership) may lead to uncertainty in a firm’s cash flow, resulting in higher

X. Cui et al./Accounting & Finance 3
capital costs and lower valuation (Pástor and Veronesi, 2012, 2013; Brogaard
and Detzel, 2015). However, firm managers understand their firm’s funda-
mentals and can predict the impact of external uncertainty on the firm. Thus,
the information asymmetry between managers and capital market participants
will increase when external uncertainty increases (Dai and Ngo, 2020). Facing
such information asymmetry, Nagar et al. (2019) suggest that to reduce the
negative impact of external EPU, such as higher capital costs (e.g., Lang and
Maffett, 2011; Balakrishnan et al., 2014), managers are strongly motivated to
increase their voluntary disclosures (Verrecchia, 1990; Baker et al., 2016).
Therefore, we expect that more voluntary information disclosure may weaken
the opportunity and motivation of managers who want to manipulate earnings.
However, other studies also point out that voluntary information disclosure
has high costs, and not all firms are willing to incur them (e.g., Bamber and
Cheon, 1998; Bova et al., 2015; Cao et al., 2018), so the higher information
asymmetry environment caused by EPU may still exist (Dai and Ngo, 2020). In
a high information asymmetry environment, Dai and Ngo (2020) argue that
conditional accounting conservatism is an efficient contracting technology,
which can alleviate information asymmetry between contracting parties in the
capital markets. They indicate that there will be an increase in the demand for
conditional conservation from less informed parties during periods of higher
uncertainty. Therefore, we infer that high economic policy uncertainty can also
lead to more accounting conservatism and reduce the managerial incentives to
manipulate earnings.
It can be seen from the above analysis that a high EPU exposure may lead to
an increase or decrease in earnings management. It is still an important
empirical issue to investigate whether and how EPU affects firms’ earnings
management. Using data from the Chinese stock market, we examine the
impact of EPU exposure on a firm’s earnings management. This investigation is
important not only for investors to evaluate firm performance and growth, but
also for regulators to improve the supervisory efficiency and enhance the
accuracy of accounting information.
We focus on the Chinese market for two reasons. First, China provides a
unique and ideal market to test the impact of EPU. Although China is the
second largest economy worldwide, it is still transitioning from a centrally
planned economy to a market-based economy. Therefore, government inter-
vention in the economy is the norm (Fan et al., 2013). During this transition,
the Chinese government has faced unprecedented challenges in economic
policy. The famous Chinese saying ‘crossing the river by feeling the stones’
describes the uncertainty surrounding policymaking in China (Chen et al.,
2017), and is an ideal approach to examining the role of EPU in financial
markets.
Second, as an emerging market, China’s institutional background is different
from that of developed countries like the United States (Kong et al., 2019; Liu
et al., 2019; Yao et al., 2020), and many findings for developed markets cannot

be directly applied in China. Due to the government’s interventions in

economic operations, Chinese firms are facing increasingly serious policy costs
(Fan et al., 2013; Xu et al., 2016). The financial activities of firms, especially
earnings management, are impacted by the costs associated with policy
uncertainty. Therefore, China provides a good setting for understanding how
policy costs impact earnings manipulation. In addition, similar to China, other
emerging market economies are also facing greater policy uncertainty, and the
ability of their firms to cope with such uncertainty is limited (Frot and Santiso,
2013; Carrière-Swallow and Céspedes, 2013; Winkler et al., 2015). Therefore,
exploring relevant issues in China will also help other emerging markets to
understand the impact of policy costs on firm earnings, and provide strong
policy guidance for the regulation of financial misconduct.
Differing from studies in which the EPU index is used to explore how EPU
influences corporate behaviour, following Brogaard and Detzel (2015) and Bali
et al. (2017), we combine the EPU index with the Fama-French model (Fama
and French, 2015). We use this approach to conduct a rolling-window
estimation to obtain the reactions of individual Chinese A-share firms to the
EPU index (hereafter, EPU exposure). Our evidence shows that EPU exposure
significantly aggravates firms’ earnings management behaviour. We find the
relation is robust in tests where we control for endogeneity. In addition, we
examine the relation between EPU exposure and earnings management from
the perspective of external monitoring mechanisms. We find the positive
association between EPU exposure and earnings management is more
pronounced for firms with weak external monitoring (e.g., low institutional
ownership, non-Big-4 auditors and low analyst coverage).
In addition, although we can estimate the EPU exposure of each stock, it is
not easy for more general market investors to observe the EPU exposure of a
stock. If investors and market regulators can determine the EPU exposure of a
firm, then they can mitigate potential risks. Therefore, we also link specific firm
characteristics to EPU exposure and explore which characteristic can predict a
firm’s EPU exposure. We find that rapidly growing firms and firms with high
financial leverage are more sensitive to external EPU. Finally, we further
examine the economic mechanisms between EPU exposure and earnings
management based on growth rate and financial leverage. We find that for
firms with high financial leverage, the EPU exposure may be more likely to
cause financial distress, thus triggering earnings management. However, in high
growth firms, the EPU exposure is typically related to the impact on corporate
cash flow volatility, which induces more earnings management.
Our study contributes the literature in three ways. First, this paper adds new
evidence to the research on the impact of EPU (Wang et al., 2015; Tsai, 2017;
Jin et al., 2019; Nagar et al., 2019; Wu et al., 2020; Xu, 2020), especially to the
relation between EPU and the behaviour of micro-firms (Demir and Ersan,
2017; Chen et al., 2017; Huang et al., 2020; Wu et al., 2020; Xu, 2020). More
specifically, we clarify the relation between EPU and earnings management

from the perspective of the heterogeneity of individual stocks’ EPU exposure.

Our research supplements the literature on earnings management motivation
(Dhole et al., 2016; Mu et al., 2017; Na and Hong, 2017; Harris et al., 2019; Li,
2019; Thanh et al., 2020), and provides important evidence for market
regulators, as well as for investors to prevent potential investment risks.
Second, when exploring the relation between EPU and micro-firms’
behaviour, in most studies it is assumed that different firms face the same
EPU (e.g., Demir and Ersan, 2017; Huang et al., 2020; Wu et al., 2020).
However, in line with Xu (2020) and Yang et al. (2019), we take the
heterogeneity of individual stocks’ EPU exposure into account. More impor-
tantly, for the first time, we also use firm characteristics to reconfigure the
unobservable EPU exposure of ordinary investors into a relatively observable
factor. That is, we can judge the EPU exposure of individual stocks using their
financial leverage and growth rate. This practice will help investors to
distinguish firms with high EPU exposure, and assist regulators with super-
vising earnings management depending on the firm-specific characteristics.1
Third, we examine the cross-sectional variations in the relation between EPU
exposure and earnings management. Specifically, we find that the positive
association between EPU exposure and firms’ earnings management is more
pronounced in firms with weaker external governance (e.g., with low
institutional ownership, less analyst coverage and audited by non-Big 4
auditors). Our results indicate the influence of EPU is stronger when external
governance mechanisms are weaker. These results provide insight on the
impact of EPU, along with support for the necessity of external regulatory
mechanisms. Moreover, as a universal uncertainty factor, the impact of EPU
exposure on earnings management also has important implications for other
emerging markets.
The remainder of this paper is organised as follows. In Section 2, we
introduce the data sources, main variables and methodology. In Section 3, we
analyse the impact of EPU exposure on firm-specific earnings management. In
Section 4, we discuss the analyses, including the conditional analyses from the
perspective of external monitoring mechanisms, the relation between firm
characteristics and EPU exposure, and the economic linkages between EPU
1
Yung and Root (2019) explore the relation between macro policy uncertainty and
earnings management. Our paper is quite different from theirs in the following ways.
Unlike Yung and Root’s (2019) assumption that individual stocks face the same EPU,
we initially assume that firms have different exposures to the same market EPU, and
calculate the idiosyncratic EPU exposure of individual stocks. This approach provides a
more rigorous framework for the analysis of the environment and micro-firms’
behaviour. More importantly, using firm-specific characteristics, we turn the unobserv-
able EPU exposure of ordinary investors into a relatively observable factor. We find that
investors and market regulators can use the firms’ financial leverage and growth rate to
assess the extent of EPU exposure. Furthermore, we also analyse the mechanism
between EPU exposure and earnings management under different firm characteristics.

exposure and earnings management. Policy implications and conclusions are

drawn in Section 5.
2. Data, variables and methodology
2.1. Sample and data
Our sample includes all Chinese A-share listed firms on the Shanghai and
Shenzhen stock exchanges during the 2007–2018 period.2 Firm-specific data are
collected from the China Stock Market and Accounting Research (CSMAR)
database. Following Lin et al. (2011) and Li et al. (2017), we exclude (i)
financial services firms, (ii) special treatment (ST) firms, and (iii) firm-year
observations without sufficient financial data to construct control variables for
the regressions. The final sample includes 16,992 firm-year observations.
2.2. EPU index
The EPU index proposed by Baker et al. (2016) is the most widely used
indicator of EPU. Different from Baker et al. (2016), who use Hong Kong’s
South China Morning Post as the information carrier to calculate China’s EPU
index, Davis et al. (2019) use two official newspapers (the Renmin Daily and the
Guangming Daily)3 in mainland China to calculate the EPU index. Compared
to the newspaper published in Hong Kong, the Renmin Daily and the
Guangming Daily are the two most influential news outlets operated by the
Chinese government, and relay government policies in mainland China (Davis
et al., 2019). Therefore, following Davis et al. (2019), the EPU index we use is
based on the information on governmental policy published in the Renmin
Daily and the Guangming Daily.
2.3. The measurement of EPU exposure
Following Brogaard and Detzel (2015) and Bali et al. (2017), we combine the
EPU index4 with the Fama-French five-factor model (Fama and French, 2015),
2
China’s new accounting standards for listed firms were implemented before 2007, which
has made them more aligned with the International Financial Reporting Standards. The
accounting standards had an important impact on the financial information released by
listed firms on China’s stock markets, which helps us to analyse problems in a more
general system. Therefore, we choose 2007 as the first year in our sample period.
3
Relevant data and algorithms are obtainable at http://www.policyuncertainty.com/
china_monthly.html.
4
In order to eliminate the data dimension problem and smooth the data, following
Pástor and Veronesi (2013), we scale down the EPU index by 100.

as shown in Equation (1). Using the past 36 months (τ-36, τ-1) as the rolling
window, we regress Equation (1) to obtain βepu
i,τ , as follows:
Ri,τ r f,τ ¼ β0 þ βmkt

i,τ MKTτ þ βi,τ SMBτ þ βi,τ HMLτ þ βi,τ RMWτ
smb hml rmw
epu
(1)
þβcma
i,τ CMAτ þ β i,τ EPUτ þ ɛ i,τ ,
where Ri,τ r f,τ is the excess return on stock i in month τ; MKTτ, SMBτ, HMLτ
RMWτ and CMAτ represent Fama and French’s five factors; EPUτ is the EPU
index in month τ; and βepu i,τ is the sensitivity of stock i to EPUτ .
Pástor and Veronesi (2012) employ a theoretical model and find that stock
prices generally show a downward trend as EPU increases. If the market EPU
increases by one unit, a greater decrease in the price of an individual stock (a
greater decrease in a stock’s return) indicates it has a higher risk exposure to
EPU. Therefore, a stock with a more negative value of βepu i,τ (i.e., stock
experiencing a sharper return decline as EPU rises) indicates the stock has a
greater exposure to EPU. In addition, following Xu (2020), a stock with a more
negative βepu
i,τ can be regarded as the least optimal hedge against increase in
uncertainty. Thus, such a stock cannot hedge the fluctuation in EPU well,
suggesting that it is more exposed to EPU. Therefore, we use the negative value
of βepu
i,τ to proxy a firm’s exposure to EPU. A higher exposure to EPU (lower
excess stock return) also suggests that the firm is experiencing a more risky and
challenging business environment. Finally, we take the average monthly stock
value to obtain the annual measure of EPU exposure, CMff5 36i,t .
2.4. Earnings management measurement
Following Dechow et al. (1995) and Ghosh and Olsen (2009), we estimate the
modified Jones model as shown in Equation (2) for each year and industry to
calculate the absolute value of residuals (AbsAcc ) as the measurement of
earnings management:
TAi,t 1 ðΔREVi,t ΔRECi,t Þ PPEi,t

¼ αi þ βi þ γi þ ɛi,t , (2)
ATi,t1 ATi,t1 ATi,t1 ATi,t1
where TAi,t is total accruals scaled by the total assets of firm i in year t. ΔRECi,t
is the difference in net receivables between year t and year t − 1 scaled by total
assets in year t − 1 of firm i. PPEi,t is the gross property, plant and equipment
(PPE) of firm i in year t scaled by total assets in year t − 1;ΔREVi,t is the
difference in revenue between year t and year t − 1 scaled by total assets in year
t − 1 of firm i. ATi,t1 is the total assets in year t − 1 of firm i.
In addition, following Jones (1991), Kothari et al. (2005) and Ball and
Shivakumar (2006), we construct three alternative discretionary accruals
measures using the Jones model, nonlinear accruals model and ROA-matched

model. We then calculate three other earnings management measures: Jones,

Nonlinear and ROA matched . In addition, we follow Dechow and Dichev
(2002), Francis et al. (2008), and Ghosh and Moon (2010) and estimate
Equation (3) for each year and industry to determine earnings quality, EQ, as
follows:5
ΔWCi,t ¼ αi,0 þ αi,1 CFOi,t1 þ αi,2 CFOi,t þ αi,3 CFOi,tþ1 þ ɛi,t , (3)
where ΔWCi,t is the changes in working capital of firm i from year t − 1 to t,

which is calculated as (Δ accounts receivable + Δ inventory – Δ accounts
payable – Δ other assets) divided by the average total assets. CFOi,t1 , CFOi,t
and CFOi,tþ1 refer to the cash flows of firm i in period t − 1, period t, and
period t + 1, respectively. CFO is operational cash flow divided by average
total assets.
2.5. Methodology
To estimate the relation between EPU exposure and earnings management,

we use the following regression model:
EMi,tþ1 ¼ β0 þ β1 EPU exposurei,t þ ∑γ k Controlk,i,t þ ɛi,tþ1 , (4)

k
where EMi,tþ1 is the earnings management of firm i in year t + 1, measured by

AbsAcctþ1 , Jonestþ1 , Nonlineartþ1 and ROA matchedtþ1 . EPU exposurei,t is
EPU exposure, measured by CMff5 36i,t .
Following Thanh et al. (2020), we control for other firm-specific character-
istics, Controlk,i,t , that may influence firms’ earnings management, including
firm size (Size), book-to-market ratio (Bm), firm financial leverage (Lev), return
on assets (Roa), sales growth rate (Sales Growth), the listing age (Age), the net
loss (Loss), managerial ownership (Mgshare) and fixed assets ratio (Fixed). To
mitigate possible causality and other endogeneity problems, all independent
variables are lagged by one year in which the dependent variable is measured.
In addition, to control for sample heterogeneity caused by year and industry,
we adopt a two-way fixed effects model to estimate the regression. Detailed
descriptions of all variables are shown in the Appendix. All continuous
variables are winsorised at the 1 percent level at both tails.
5
Annual cross-sectional estimations of Equation (3) yield residuals. The standard
deviation of these residuals, obtained from the model for a five-year period from t − 4 to
t, is employed as the measure of earnings quality, where a lower standard deviation
denotes higher quality. Therefore, we use the negative value of standard deviation as the
proxy for earnings quality.

3. Results
3.1. Descriptive statistics and correlations
In Table 1, we present the descriptive statistics of the variables and their

average correlation coefficients. Adopting the Fama-MacBeth approach, we
estimate the cross-sectional statistics (including the mean, standard deviation,
first quartile (Q1), median, third quartile (Q3), maximum and minimum) of the
selected variables on a yearly basis. Then, we calculate and report the time
series averages of the cross-sectional statistics. As shown in Table 1, the mean
and median values of the representative indicators of earnings management
AbsAcc are 7.1 percent and 4.7 percent, respectively, which are in line with
prior studies.6 The mean of EQ is 0.001 and the standard deviation is 0.039,
showing that EQ varies greatly among sample firms. In addition, the maximum
and minimum of CMff5 36 are 2.115 and −2.017, respectively, while the
average value is −0.018, indicating that EPU exposure varies greatly among
sample firms.
Next, we again adopt the Fama-MacBeth approach and estimate the Pearson
and Spearman cross-sectional correlations between the applied variables. We
report the time series averages of these correlations in Table 2. We find that
AbsAcc is positively correlated with EPU exposure, proxied by CMff5 36. In
addition, EQ is negatively correlated with EPU exposure, and the negative
correlation between AbsAcc and EQ is significant at the 1 percent level.
3.2. Baseline results
Table 3 shows the baseline regression results of Equation (4). When we first
use AbsAcc as the proxy variable of earnings management, the results in
column (1) show that the coefficient between EPU exposure (CMff5 36) and
earnings management is significant and positive at the 1 percent level. The
results indicate that when EPU exposure is higher, there is more earnings
management. In addition, the coefficients for firm size (Size), book-to-market
ratio (Bm ), return on assets (Roa ) and total fixed assets (Fixed ) are all
significantly negative, while the coefficients for firm financial leverage (Lev),
sales growth rate (Sales Growth), listing age (Age) and managerial ownership
(Mgshare) are significantly positive, which is in line with the results in Harris
et al. (2019) and Thanh et al. (2020).
To verify the robustness of our results, we follow Jones (1991), Kothari et al.
(2005) and Ball and Shivakumar (2006) and construct alternative discretionary
accruals measures using the Jones model, nonlinear accruals model and ROA-
6
In order to be concise, only the representative earnings management indicators, AbsAcc
and EQ, are selected for analysis. The results of other earnings management proxies are
basically the same.

10
Table 1
Descriptive statistics
Variables Observations Mean SD 25th percentile Median 75th percentile Min Max
AbsAcc 16,992 0.071 0.078 0.021 0.047 0.091 0.001 0.452

Jones 16,992 0.070 0.078 0.021 0.046 0.089 0.001 0.457
Nonlinear 16,992 0.055 0.056 0.017 0.038 0.072 0.001 0.320
ROA_matched 16,992 0.064 0.071 0.019 0.042 0.083 0.001 0.417
EQ 16,992 0.001 0.039 −0.010 0.007 0.021 −0.133 0.071
CMff5_36 16,992 −0.018 0.678 −0.366 −0.024 0.318 −2.017 2.115
Size 16,992 15.044 1.092 14.278 15.004 15.734 12.715 18.036
Bm 16,992 0.600 0.238 0.418 0.601 0.783 0.103 1.099
Lev 16,992 0.443 0.224 0.265 0.434 0.607 0.048 1.065
Roa 16,992 0.043 0.068 0.014 0.041 0.076 −0.254 0.243
Sales_Growth 16,992 0.208 0.472 0.011 0.101 0.241 −0.381 3.394
Age 16,992 10.058 6.818 4.000 9.000 16.000 0.001 25.000
Loss 16,992 0.387 0.487 0.001 0.001 1.000 0.001 1.000
Mgshare 16,992 0.062 0.130 0.001 0.001 0.069 0.001 0.774
X. Cui et al./Accounting & Finance
Fixed 16,992 0.219 0.164 0.091 0.186 0.313 0.002 0.707
This table reports the descriptive statistics of the variables. The sample includes 16,992 firm-year observations listed on the SHSE and SZSE from
2007 to 2018. Adopting the Fama and MacBeth (1973) approach, we estimate the cross-sectional statistics (mean, standard deviation, Q1,
median, Q3, max and min) of the variables per year in the first step. In the second step, we calculate the time series average of the cross-sectional
statistics derived in the first step. The descriptions of all variables are shown in the Appendix.

matched model. The results in columns (2)–(4) of Table 3 indicate that the
coefficients of EPU exposure are significant, regardless of the measures of
earnings management. Moreover, with the rise in EPU, managers are more
inclined to manage earnings to whitewash their firms’ performance, which
causes a decline in earnings quality. To further explore the relation between
EPU exposure and earnings management, following Ghosh and Moon (2010),
we also use earnings quality (EQ) as an alternative variable to proxy earnings
management behaviour and investigate the effect of EPU exposure on earnings
quality. The results in column (5) show that there is still a significant negative
correlation between EPU exposure and earnings quality (EQ ). The results
further confirm that managers may manage earnings more frequently when
facing high EPU exposure, which verifies our previous results.
3.3. Robustness tests and endogeneity
In this subsection, we discuss six robustness tests that address possible biases:
(i) the sensitivity test of EPU exposure; (ii) an alternative model for EPU
exposure estimation; (iii) a longer rolling window for the estimation of EPU
exposure; (iv) an alternative measure of the EPU index (e.g., the EPU index
constructed by Baker et al. (2016)); (v) a multiple fixed effects model; and (vi)
an instrumental variable method to estimate the exogenous part of China’s
EPU exposure.
3.3.1. The sensitivity test of EPU exposure
EPU exposure is a beta type sensitivity measure of a firm’s stock price. We

conduct a robustness check to further test the sensitivity of EPU exposure to
the market EPU index. First, we construct the dummy variable High according
to the average of the market EPU index during the sample period. For the
sample period in which the market EPU index is higher than the mean, we
assign High a value of 1, and 0 otherwise. Then, we construct the interaction
term between High and EPU exposure, and add it to Equation (4). The results
in Panel A of Table 4 show that the coefficients of the interaction term
demonstrate a certain degree of statistical significance (at least at the 10 percent
level). The results show that our estimation method of EPU exposure is
reasonable, and shows a certain sensitivity to the original market EPU index:
when the EPU index is high, the firms with higher EPU exposure indeed
increase earnings management activities more.
3.3.2. An alternative measure of EPU exposure
In the above analysis, we use the more advanced Fama-French five-factor

model. However, the Fama-French three-factor model is a more classic and
widely applicable pricing model. Therefore, we also combine the EPU index

12
Table 2
Correlation coefficients
CMff5_
AbsAcc Jones Nonlinear Rmatched EQ 36 Size Bm Lev Roa Sales_Growth Age Loss Mgshare Fixed
AbsAcc 0.960 0.715 0.873 −0.110 0.024 −0.055 −0.109 0.130 −0.074 0.260 0.056 −0.025 −0.007 −0.167
Jones 0.955 0.692 0.836 −0.116 0.025 −0.058 −0.107 0.130 −0.068 0.266 0.060 −0.028 −0.014 −0.166
Nonlinear 0.723 0.702 0.547 −0.086 0.027 −0.055 −0.161 0.064 −0.088 0.185 0.027 −0.021 0.015 −0.054
ROA_matched 0.787 0.757 0.515 −0.113 0.015 −0.041 −0.096 0.106 0.038 0.297 0.048 −0.065 −0.007 −0.178
EQ −0.091 −0.095 −0.058 −0.086 −0.004 0.030 0.134 −0.050 0.002 0.011 0.043 0.043 0.029 0.047
CMff5_36 0.001 0.009 0.003 0.004 −0.004 0.012 −0.030 0.030 −0.048 0.003 0.050 −0.001 −0.028 −0.002
Size −0.055 −0.059 −0.041 −0.037 0.005 0.084 −0.114 0.179 0.101 0.040 0.386 −0.115 −0.237 0.015
Bm −0.084 −0.079 −0.139 −0.072 0.121 −0.013 −0.084 0.278 −0.138 0.044 0.063 0.106 0.022 0.095
Lev 0.087 0.088 0.010 0.070 −0.021 0.273 −0.013 0.399 −0.397 −0.005 0.384 0.054 −0.294 0.131
Roa 0.022 0.024 0.102 0.066 −0.031 −0.068 0.270 −0.219 −0.369 0.233 −0.257 −0.416 0.228 −0.150
Sales_Growth 0.106 0.107 0.050 0.133 −0.033 0.034 0.203 0.063 0.043 0.381 −0.076 −0.174 0.078 −0.141
Age 0.040 0.040 0.005 0.027 0.044 0.052 0.139 0.100 0.205 −0.133 −0.124 0.058 −0.431 0.069
Loss −0.023 −0.025 −0.039 −0.054 0.049 −0.011 −0.134 0.103 0.050 −0.433 −0.242 0.030 −0.065 0.072
Mgshare −0.021 −0.025 −0.006 −0.006 −0.010 −0.048 0.131 −0.060 −0.188 0.193 0.158 −0.326 −0.070 −0.157
Fixed −0.153 −0.157 −0.033 −0.162 0.085 −0.040 −0.100 0.073 −0.004 −0.137 −0.205 −0.105 0.082 −0.084
This table reports the time series averages of the cross-sectional Pearson (above diagonal) and Spearman (below diagonal) correlation estimates
from 2007 to 2018. Following the Fama and MacBeth (1973) approach, we first estimate the Pearson (above diagonal) and Spearman (below
diagonal) cross-sectional correlations between variables, and then we report the time series averages of these correlations. The correlations that

are significant at the 1% significance level are expressed in bold. The descriptions of all variables are shown in the Appendix.
Table 3
Results for the effect of EPU exposure on earnings management
Dependent variables
AbsAcct+1 Jonest+1 Nonlineart+1 ROA_matchedt+1 EQt+1
Variables (1) (2) (3) (4) (5)
CMff5_36t 3.019*** 3.811*** 2.309*** 3.367*** −1.110**

(3.43) (4.34) (2.86) (2.59) (−2.53)
Sizet −0.007*** −0.007*** −0.007*** −0.008*** 0.002***
(−8.26) (−8.50) (−8.81) (−7.08) (6.18)
Bmt −0.077*** −0.076*** −0.069*** −0.091*** 0.024***
(−22.59) (−22.68) (−22.23) (−19.99) (16.00)
Levt 0.037*** 0.038*** 0.035*** 0.050*** −0.016***
(9.27) (9.69) (9.68) (9.39) (−9.36)
Roat −0.044*** −0.042*** −0.019 −0.059*** −0.005
(−3.44) (−3.27) (−1.60) (−3.17) (−0.82)
Sales_Growtht 0.008*** 0.007*** 0.006*** 0.009*** −0.002***
(5.68) (5.02) (5.09) (4.47) (−2.74)

Aget 0.001*** 0.001*** 0.001*** 0.001*** −0.001***
(5.93) (5.89) (4.36) (4.16) (−4.01)
Losst 0.001 0.001 −0.001 0.001 −0.001
(0.41) (0.30) (−0.47) (0.63) (−1.06)

(continued)
13
14
Table 3 (continued)
Dependent variables
AbsAcct+1 Jonest+1 Nonlineart+1 ROA_matchedt+1 EQt+1
Variables (1) (2) (3) (4) (5)
Mgsharet 0.051*** 0.047*** 0.036*** 0.037*** −0.015***

(5.03) (4.63) (3.90) (2.66) (−3.28)
Fixedt −0.021*** −0.020*** −0.019*** −0.030*** 0.007***
(−4.20) (−4.14) (−4.26) (−4.78) (3.00)
Constant 0.189*** 0.191*** 0.179*** 0.193*** −0.093***
(13.26) (13.48) (13.88) (10.67) (−14.71)
Observations 16,992 16,992 16,992 16,992 16,992
Year fixed effects Yes Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes Yes
Adjusted R2 0.2675 0.2717 0.2612 0.2376 0.2004
This table shows the estimation results for the effect of EPU exposure on earnings management using the following regression model:
EMi,tþ1 ¼ β0 þ β1 EPU exposurei,t þ ∑γ k Controlk,i,t þ ɛ i,tþ1 ,
where EMi,tþ1 refers to the earningsk management of firm i in year t + 1, measured by AbsAcctþ1 , Jonestþ1 , Nonlineartþ1 , ROA matchedtþ1 and
EQtþ1 . EPUexposurei,t refers to EPU exposure, measured by CMff5 36i,t . The sample period is from 2007 to 2018. Controlsk,i,t refers to a set of
control variables. The variable descriptions are shown in the Appendix. All continuous variables are winsorised at the 1% level in each tail.
Industry and year fixed effects are controlled for and the standard errors are corrected using the double-clustering (firm and year) method, as
suggested by Petersen (2009). t-statistics are given in parentheses. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels,
respectively.

Table 4
The robustness tests of the replacement of EPU exposure
Panel A: The sensitivity test of EPU exposure
Dependent variables
Variables AbsAcct+1 Jonest+1 Nonlineart+1 ROA_matchedt+1 EQt+1
CMff5_36t 2.564** 3.210** 1.892** 2.667** −0.953*

(2.20) (2.43) (2.12) (2.02) (−1.83)
CMff5_36t* High 1.267** 1.298* 0.806* 1.376** −0.306**
(2.29) (1.85) (1.81) (2.29) (−2.44)
High 0.015 0.016 0.008 0.021 −0.018
(0.45) (1.41) (0.69) (0.86) (−1.45)
Constant 0.076*** 0.074*** 0.063*** 0.065*** −0.022***
(10.89) (8.63) (10.68) (8.31) (−12.67)
Other controls Yes Yes Yes Yes Yes
Observations 16,992 16,992 16,992 16,992 16,992
Adjusted R2 0.2693 0.2746 0.2684 0.2453 0.2025
Panel B: An alternative model (FF3) for EPU exposure estimation

Dependent variables
CMff3_36t 2.573*** 3.455*** 2.062** 2.553*** −1.662***

(2.85) (3.83) (2.49) (2.92) (−3.74)
Constant 0.189*** 0.191*** 0.179*** 0.193*** −0.093***
(continued)
15
16
Table 4 (continued)
Panel B: An alternative model (FF3) for EPU exposure estimation
Dependent variables
(13.24) (13.47) (13.87) (10.65) (−14.70)

Other controls Yes Yes Yes Yes Yes
Observations 16,992 16,992 16,992 16,992 16,992
Adjusted R2 0.2664 0.2700 0.2609 0.2377 0.2003
Panel C: Using a longer horizon to estimate EPU exposure
Dependent variable = AbsAcct+1
Variables (1) (2) (3) (4)
CMff3_60t 4.958*** 2.721*

(3.42) (1.90)
CMff5_60t 4.295*** 2.863**
(2.95) (2.00)
Other controls No Yes No Yes
Constant 0.075*** 0.193*** 0.075*** 0.193***
(10.24) (12.68) (10.26) (12.68)

(continued)
Table 4 (continued)
Panel C: Using a longer horizon to estimate EPU exposure
Variables (1) (2) (3) (4)
Year fixed effects Yes Yes Yes Yes

Industry fixed effects Yes Yes Yes Yes
Observations 16,992 16,992 16,992 16,992
Adjusted R2 0.1240 0.2490 0.1247 0.2496
Panel D: An alternative measure of the EPU index
Variables (1) (2) (3) (4)
Baker_ff3_36t 8.132***
(3.99)
7.949***
Baker_ff5_36t
(4.03)
Baker_ff3_60t 11.120***
(3.28)
Baker_ff5_60t 9.589***
(2.89)

(continued)
17
18
Table 4 (continued)
Panel D: An alternative measure of the EPU index
Variables (1) (2) (3) (4)
Constant 0.190*** 0.190*** 0.193*** 0.193***

(13.28) (13.27) (12.70) (12.72)
Other controls Yes Yes Yes Yes
Observations 16,992 16,992 16,992 16,992
Adjusted R2 0.2692 0.2685 0.2501 0.2500
This table shows the robustness test results for the replacement of EPU exposure. Panel A shows the sensitivity test of EPU exposure. Panel B
shows the results with an alternative model (FF3) for EPU exposure estimation (CMff3 36t). Panel C shows the results under a longer horizon to
estimate EPU exposure (e.g., CMff3 60t and CMff5 60t ). Panel D shows the results under an alternative measure of the EPU index (the index
from Baker et al., 2016). The sample period is from 2007 to 2018. Controlsk,i,t refers to a set of control variables. The descriptions of all variables
are shown in the Appendix. All continuous variables are winsorised at the 1% level in each tail. Industry and year fixed effects are controlled for
and the standard errors are corrected using the double-clustering (firm and year) method, as suggested by Petersen (2009). t-statistics are given in
parentheses. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.

with the Fama-French three-factor model, as shown in Equation (5). Using the
past 36 months (τ-36, τ-1) as the rolling window, we use the following
equation to obtain the more traditional EPU exposure estimation – CMff3 36t:
epu
Ri,τ r f,τ ¼ β0 þ βmkt
i,τ MKTτ þ βi,τ SMBτ þ βi,τ HMLτ þ βi,τ EPUτ þ ɛ i,τ ,
smb hml
(5)
The results in Panel B of Table 4 show that there is a significant positive

correlation between CMff3 36t and the earnings management variables no
matter what kind of earnings management measure is adopted. The results
indicate that our findings are not affected by the estimation model.
In addition, the measurement of earnings management has little impact on
the results. Therefore, in the follow-up analysis, for the convenience of display
and analysis of the results, only the most representative variable, AbsAcc, is
used to measure earnings management.7
3.3.3. Using a longer horizon to estimate EPU exposure
In the previous subsection, we use a 36-month rolling window to estimate the

EPU exposure of individual stocks. In this subsection, we use a 60-month
rolling window and re-estimate Equations (1) and (5), which produces two
alternative measures of EPU exposure: CMff3 60t and CMff5 60t. We then re-
estimate Equation (4) using the new measures. As shown in Panel C of Table 4,
when adopting a longer horizon to estimate EPU exposure, its impact on firms’
earnings management does not change substantially.
3.3.4. An alternative measure of the EPU index
Studies on EPU generally use the EPU index proposed by Baker et al. (2016)
to measure EPU (e.g., Wang et al., 2015; Tsai, 2017; Nagar et al., 2019; Wu
et al., 2020; Xu, 2020). Compared with the EPU index reported in the
newspapers in mainland China, the EPU index of Baker et al. (2016) was
estimated using reports in the South China Morning Post. To avoid the
dependence of the estimated results on the EPU index, we combine the EPU
index proposed by Baker et al. (2016) with the Fama-French three-factor model
and the Fama-French five-factor model. Using 36 months and 60 months as
the rolling windows to re-estimate Equations (1) and (5), we get four new
proxies for EPU exposure: Baker ff3 36t , Baker ff3 60t ,Baker ff5 36t and
Baker ff5 60t . We then re-estimate Equation (4) using the four new measures.
The results in Panel D of Table 4 show that the coefficients for EPU exposure
are all significantly positive, which confirms that our baseline results are robust.
7
In untabulated results, when using other proxies of earnings management, the results
are not affected.

3.3.5. Multiple fixed effects model
To mitigate potential endogenous problems that may arise from omitting

firm-specific characteristics, time-varying industry-specific factors and pro-
vince-specific factors, we re-estimate Equation (4) using the multiple fixed
effects model (firm fixed, industry*year fixed and province*year fixed). Parsons
et al. (2018) show that a firm’s likelihood of engaging in financial misconduct is
related to the financial misconduct rates of firms in the same locale, and that
the relation is likely driven by social interactions among neighbouring firms. In
addition, other local factors such as local economic conditions and the political
environment may influence the incidence of financial fraud (Liu, 2016).
Following Liu (2016), in addition to controlling for firm-specific characteristics
(firm-fixed effect), we also control for time-varying industry-specific factors
such as industry growth opportunities and time-varying province-specific
factors, such as local cultural and economic conditions. The regression results
in Panel A of Table 5 indicate that our results do not change substantially.
3.3.6. Instrumental variable method
To further address potential endogeneity problems, following Wang et al.

(2017) and Liu and Zhang (2019), we use the EPU index of the US as an
instrumental variable to estimate the exogenous part of China’s EPU.8
Specifically, we employ China’s EPU index as the dependent variable and the
US’s EPU as the independent variable. Using the rolling window regression
method (the estimated range moves forward one month at a time) and the data
from t − 36 to t − 1, we obtain the predicted relation between these two
variables at time t. By substituting the US’s EPU value in time t, we obtain the
predicted value of China’s EPU, which is defined as the exogenous part of
China’s EPU. Using the predicted value of China’s EPU, we obtain the
individual stock’s exposure to the exogenous part of China’s EPU (i.e.,
EXO EPU ff3 36t , EXO EPU ff5 36t , EXO EPU ff3 60t , EXO EPU ff5 60t ).
We then re-estimate Equation (4) using EXO EPU ff3 36t , EXO EPU ff5 36t ,
EXO EPU ff3 60t and EXO EPU ff5 60t , respectively. The regression results,
shown in Panel B of Table 5, are qualitatively similar to the baseline regression
results.
8
As the largest economic entity worldwide, fluctuations in economic policies in the US
tend to spill over into other economies. However, compared to the US, China’s EPU
(especially the EPU exposure of individual stocks) has less impact on the EPU of the
US.

4. Further analysis
4.1. Do external monitoring mechanisms matter?
As a monitoring mechanism of agency problems, an independent auditor is

an institutional entity that can restrain the opportunistic behaviour of top
management (Jensen and Meckling, 1976), and firms using an international
Big-4 auditing firm are less likely to engage in earnings management (Chi and
Weng, 2014; Choi et al., 2018). Analyst coverage can also play an effective role
in reducing earnings management (Jensen and Meckling, 1976; Healy and
Palepu, 2001). In addition, institutional investors can often effectively monitor
manager behaviour, and thus restrain earnings manipulation (Chung et al.,
2005; Elyasiani et al., 2017). Therefore, we conjecture that faced with EPU,
managers’ earnings management behaviour is more serious in firms with non-
Big-4 auditors, low analyst coverage and low institutional ownership.
In this subsection, we examine whether the relation between EPU exposure
and firms’ earnings management is subject to the external monitoring
mechanisms discussed above and the institutional ownership of firms.
Following Yuan et al. (2016) and Cheng et al. (2020), we divide the sample
into three sub-groups: the Big-4/non-Big-4 sub-group, the high/low-analyst
coverage sub-group, and the high/low-institutional ownership sub-group.
Then, we re-estimate Equation (4) for each sub-group.9
Table 6 shows that the coefficients for CMff5 36t in columns (1)–(3) are all
insignificant, while those in columns (4)–(6) are all significantly positive at least
at the 5 percent level. The results suggest that the effect of EPU exposure on a
firm’s earnings management is more serious in firms with weak external
monitoring (e.g., non-Big-4 auditors, low analyst coverage and few institutional
shareholders). That is, the relation between EPU exposure and earnings
management is significantly dependent on the strength of external supervision
and governance. These findings are of great significance to the oversight of
earnings management.
4.2. Determinants of EPU exposure and potential economic channels
4.2.1. Determinants of the cross-sectional variation in EPU exposure
Our results suggest that even in the same market, firms have different EPU
exposure levels and outcomes (Yang et al., 2019; Xu, 2020), which conflicts
with previous studies on EPU and firm behaviour (e.g., Demir and Ersan, 2017;
Huang et al., 2020; Wu et al., 2020). Based on this logic, we combine the market
9
For the variables of analyst coverage and institutional ownership, if the value of the
corresponding variable of stock i is higher than the industry median in the same year, it
will be classified as a high group, and vice versa.

Table 5
22
The tests controlling for potential endogeneity
Panel A: Multiple fixed effects model
Variables (1) (2) (3) (4)
CMff3_36t 4.093*** 3.794***

(4.13) (3.75)
CMff5_36t 3.990*** 3.787***
(4.16) (3.87)
Constant 0.075*** 0.395*** 0.075*** 0.396***
(14.07) (16.78) (14.14) (16.29)
Firm fixed effects Yes Yes Yes Yes
Industry*Year fixed effects Yes Yes Yes Yes
Province*Year fixed effects No Yes No Yes
Observations 16,992 16,992 16,992 16,992
Adjusted R2 0.0045 0.0181 0.0076 0.0181
Panel B: Instrument variable method

Variables (1) (2) (3) (4)
EXO_EPU_ff3_36t 2.341***
(2.61)

EXO_EPU_ff5_36t 2.226***
(2.74)
(continued)
Table 5 (continued)
Panel B: Instrument variable method
Variables (1) (2) (3) (4)
EXO_EPU_ff3_60t 1.887**
(2.11)
EXO_EPU_ff5_60t 1.672**
(2.03)
Constant 0.223*** 0.220*** 0.231*** 0.229***
(14.15) (14.09) (14.22) (14.18)
Observations 16,992 16,992 16,992 16,992
Adjusted R2 0.2519 0.2521 0.2431 0.2422
This table shows the test results for controlling the potential endogeneity. First, to mitigate potential endogenous problems that may arise from
omitting firm-specific characteristics, and time-varying industry-specific and province-specific factors, we re-estimate Equation (4) using the
multiple fixed effects model. The results are shown in Panel A. Second, to further address potential endogenous problems, following Wang et al.
(2017) and Liu and Zhang (2019), we use the EPU index of the US as an instrumental variable to estimate the exogenous part of China’s EPU.
The results are shown in Panel B. The sample period is from 2007 to 2018.Controlsk,i,t refers to a set of control variables. The detailed descriptions
of all variables are shown in the Appendix. All continuous variables are winsorised at the 1% level in each tail. Industry and year fixed effects are
controlled for and the standard errors are corrected using the double-clustering (firm and year) method, as suggested by Petersen (2009). t-

statistics are given in parentheses. *, ** and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
23
Table 6
24
External monitoring and institutional ownership
Big-4 High-An High-Inst Non-Big4 Low-An Low-Inst
Dependent variables
AbsAcct+1 AbsAcct+1
Variables (1) (2) (3) (4) (5) (6)
CMff5_36t −3.672 −0.156 0.089 1.966** 2.543** 3.136**

(−1.14) (−0.12) (0.08) (2.17) (2.27) (2.25)
Sizet −0.006** −0.006*** −0.006*** −0.005*** −0.007*** −0.002
(−2.39) (−4.10) (−6.24) (−6.26) (−6.35) (−1.58)
Bmt −0.049*** −0.040*** −0.056*** −0.071*** −0.067*** −0.072***
(−4.28) (−6.45) (−14.41) (−22.16) (−18.78) (−13.82)
Levt 0.027* 0.046*** 0.039*** 0.045*** 0.043*** 0.049***
(1.74) (5.74) (7.84) (12.13) (11.09) (8.67)
Roat −0.057 0.140*** −0.086*** −0.160*** −0.231*** −0.198***
(−1.21) (5.94) (−5.40) (−13.76) (−16.98) (−11.86)
Sales_Growtht 0.052*** 0.047*** 0.039*** 0.051*** 0.053*** 0.049***
(9.38) (22.16) (23.25) (36.93) (31.07) (22.64)
Aget 0.001** 0.001* 0.001*** 0.001*** 0.001*** 0.001**
(2.02) (1.90) (4.86) (5.38) (5.13) (2.06)

Losst 0.004 0.002 −0.003* −0.003*** −0.004*** −0.001
(1.04) (1.23) (−1.86) (−2.61) (−2.81) (−0.54)
Mgsharet 0.006 0.009 0.033** 0.036*** 0.042*** 0.020*
(0.18) (0.77) (2.36) (4.15) (3.49) (1.66)
Fixedt −0.010 −0.024*** −0.027*** −0.039*** −0.043*** −0.041***

(−0.62) (−2.89) (−4.63) (−8.48) (−8.40) (−5.41)
Constant 0.150*** 0.147*** 0.181*** 0.168*** 0.194*** 0.118***
(continued)
Table 6 (continued)
Big-4 High-An High-Inst Non-Big4 Low-An Low-Inst
Dependent variables
AbsAcct+1 AbsAcct+1
Variables (1) (2) (3) (4) (5) (6)
(13.24) (15.89) (10.80) (11.98) (10.96) (14.74)

Observations 1,041 5,756 10,144 15,951 11,236 6,848
Year fixed effects Yes Yes Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes Yes Yes
Adjusted R2 0.2607 0.2503 0.2549 0.3063 0.2954 0.2429
We divide the full sample into the following subgroups: Big-4/Non-Big-4 auditors (Big-4), higher/lower analyst coverage (An), and higher/lower
institutional ownership (Inst). We then re-run Equation (4) using the subsamples for the period from 2007 to 2018, respectively. Controlsk,i,t refers
to a set of control variables. The detailed descriptions of all variables are shown in the Appendix. All continuous variables are winsorised at the
1% level in each tail. Industry and year fixed effects are controlled for and the standard errors are corrected using the double-clustering (firm and
year) method, as suggested by Petersen (2009). t-statistics are given in parentheses. *, ** and *** indicate significance at the 10%, 5% and 1%
levels, respectively.

25
EPU index with the Fama-French pricing model to estimate the heterogeneity
of firms’ EPU exposure. However, although we can estimate the EPU exposure
of each stock, the EPU exposure of individual stocks is not easily observed by
investors. If investors could evaluate EPU exposure, it would be vitally
important for both the investment judgement of investors and the mitigation of
potential risks by market regulators.
When exploring the cross-sectional pricing effect of the exposure of a stock to
economic uncertainty, Bali et al. (2017) use the firm’s economic uncertainty
exposure as the dependent variable, and explore the impact of individual firm
characteristics on the economic uncertainty exposure. Similar to Bali et al.
(2017), if we find that certain firm characteristics can significantly affect the
exposure of a stock to EPU, this can help investors and market regulators to
judge the exposure level of the stock to market EPU. Following Bali et al.
(2017), we select the following characteristics to explore the impact of firm-
specific factors on EPU exposure: size (Size), state-owned enterprise dummy
(Soe ), book-to-market ratio (Bm ), financial leverage (Lev ), return on assets
(Roa ), sales growth rate (Sales Growth ), listing age (Age ), net loss (Loss ),
managerial ownership (Mgshare) and fixed assets ratio (Fixed).
Columns (1)–(10) in Table 7 show the regressions results of EPU exposure
(CMff5 36t ) on each firm characteristic.10 Column (11) is the regression result
of mixing all firm factors together. Soe, Bm, Lev, Sales Growth and Fixed all
show certain prediction ability when the regression is conducted separately.
However, it is more meaningful to conduct the regression using all of the firm
characteristics we consider. The results in column (11) show that financial
leverage (Lev ) and sales growth rate (Sales Growth ) have the most obvious
prediction effect on EPU exposure, that is, firms with high financial leverage
and a high growth rate demonstrate higher EPU exposure. These findings
suggest that investors and market regulators should pay extra attention to the
potential risk caused by EPU exposure for firms with high leverage and high
growth.
4.2.2. Economic channels
In this subsection, we analyse the internal economic mechanisms. Frijns et al.

(2013) document that the uncertainty of the operational environment leads to
cash flow volatility, which has a negative impact on firm profitability, capital
structure, investment decision-making and ultimately reduces firm value
(Rountree et al., 2008; Oral and CenkAkkaya, 2015; Harris and Roark,
2019). Hirshleifer et al. (2012) also propose that cash flow volatility signals an
increase in operational risk and executives may engage in earnings management
to cover up the uncertainties. Therefore, cash flow volatility may be one of the
10
In untabulated results, when we use another alternative EPU exposure proxy and firm
fixed effects model, the results are basically the same.

potential economic channels through which EPU exposure affects earnings

management.
In addition, the increasing uncertainty of the external operational environ-
ment weakens the willingness of banks to approve loan applications and
increases the risk premiums they require (Pástor and Veronesi, 2013; Francis
et al., 2014). Bordo et al. (2016) and Hu and Gong (2019) also verify that EPU
significantly hinders the growth of bank credit. Moreover, external financing
constraints make firms more likely to fall into financial distress (Cheng et al.,
2020). He and Ren (2017) suggest that difficulties in raising funds induce
managers to use more discretionary accruals. Therefore, managers at distressed
firms may manage earnings more frequently than managers at healthy firms
(Burgstahler and Dichev, 1997; Habib et al., 2013).
As discussed, EPU may aggravate the cash flow volatility and financial
distress of firms, which can increase managers’ incentives to engage in earnings
management (Hirshleifer et al., 2012; Frijns et al., 2013; Habib et al., 2013;
Cheng et al., 2020). In addition, from the discussion in Subsection 4.2.1, the
financial leverage and growth rate of a stock are the significant factors that can
influence EPU exposure. In this subsection, we combine these to explore the
role of cash flow volatility (Cashvola) and financial distress (Z Score).
First, we analyse the power of the above two channels. Using the median of
the financial leverage of each industry in each year, we divide the samples into a
high financial leverage subsample and a low financial leverage subsample.
Following Wen et al. (2004) and Bentley-Goode et al. (2019), we use the
following mediating model to explore the role of cash flow volatility and
financial distress in the high financial leverage subsample (this subsample
means stronger EPU exposure, which can help us to study the potential
mechanism of the impact of EPU exposure):
AbsAcci,tþ1 ¼ β0 þ β1 EPU exposurei,t þ ∑βk Controlk,i,t þ ɛ i,tþ1 : (Step A)

k
Economic channeli,t ¼ β0 þ β1 EPU exposurei,t þ ∑βk Controlsk,i,t þ ɛi,t : (Step B)

k
AbsAcci,tþ1 ¼ β0 þ β1 EPU exposurei,t þ β2 Economic channeli,t

þ∑βk Controlsk,i,t þ ɛi,tþ1 : (Step C)
k
As shown in Panel A of Table 8, in the high financial leverage subsample, the

financial distress variable (Z Score ) exhibits a significant mediating effect
while the cash flow volatility variable (Cashvola) does not. The results show
that in the high financial leverage subsample, the external EPU may lead to
financial distress (e.g., limited future financing), forcing managers to use
earnings management to mitigate negative effects. As shown in Panel B, in the

28
Table 7
Firm characteristics and the prediction of EPU exposure
Dependent variable = EPU exposure (CMff5_36t)
Firm characteristics (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Sizet 0.001 −0.002

(0.05) (−0.09)
Soet −0.072** −0.041
(−2.24) (−1.21)
Bmt −0.428** −0.528
(−2.13) (−1.56)
Levt 0.132** 0.262***
(2.09) (3.56)
Roat −0.058 −0.168
(−0.36) (−0.79)
Sales_Growtht 0.046** 0.100***
(2.08) (4.00)
Aget 0.004 0.004

(1.40) (1.26)
Losst −0.012 0.012
(continued)

Table 7 (continued)
Dependent variable = EPU exposure (CMff5_36t)
Firm characteristics (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
(−0.54) (0.46)
Mgsharet −0.002 −0.005
(−0.02) (−0.03)
Fixedt −0.171* −0.071
(−1.88) (−0.74)
Constant −7.572*** −7.521*** −7.389*** −7.629*** −7.562*** −7.571*** −7.585*** −7.558*** −7.507*** −7.511*** −7.386***
(−9.34) (−8.23) (−7.15) (−8.07) (−9.36) (−9.20) (−8.75) (−8.98) (−6.98) (−7.23) (−4.78)
Observations 16,992 16,992 16,992 16,992 16,992 16,992 16,992 16,992 16,992 16,992 16,992
Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Adjusted R2 0.2024 0.2032 0.2131 0.2039 0.2031 0.2037 0.2033 0.2024 0.2018 0.2029 0.2273
This table reports the slope coefficients from the regressions of the EPU exposure on the firm-level characteristics, including size (Size), state-
owned enterprise dummy (Soe), book-to-market ratio (Bm), financial leverage (Lev), return on assets (Roa), sales growth rate (Sale Growth), the
listing age (Age), the net loss (Loss), managerial ownership (Mgshare) and fixed assets ratio (Fixed). The sample period is from 2007 to 2018. All
continuous variables are winsorised at the 1% level in each tail. Industry and year fixed effects are controlled for and the standard errors are
corrected using the double-clustering (firm and year) method, as suggested by Petersen (2009). t-statistics are given in parentheses. *, ** and ***
indicate statistical significance at the 10%, 5% and 1% levels, respectively.

29
Table 8
30
Economic channels between EPU exposure and earnings management
Panel A: High-leverage sub-sample
Dependent variables
AbsAcct+1 Cashvolat AbsAcct+1 AbsAcct+1 −Z_Scoret AbsAcct+1

Variables Step A Step B Step C Step A Step B Step C
CMff5_36t 3.234*** 7.202** 2.743*** 3.234*** 1.923** 2.562**

(3.57) (2.36) (2.84) (3.57) (2.41) (2.43)
Channel variable 0.212 0.001***
(1.53) (3.17)
Sizet −0.009*** 4.913*** −0.011*** −0.009*** 0.129 −0.008***
(−7.53) (7.22) (−9.17) (−7.53) (0.09) (−7.46)
Bmt −0.079*** 1.642*** −0.085*** −0.079*** 1.195 −0.080***
(−17.43) (6.28) (−8.45) (−17.43) (0.23) (−17.54)
Levt 0.037*** 1.447*** 0.032*** 0.037*** 2.786*** 0.038***
(4.77) (9.19) (4.11) (4.77) (3.33) (4.93)
Roat −0.067*** 1.247 −0.066*** −0.067*** −8.044*** −0.074***
(−3.80) (0.04) (−3.72) (−3.80) (−5.26) (−4.16)
Sales_Growtht 0.009*** −5.630* 0.009*** 0.009*** −0.165 0.010***
(5.27) (−1.87) (5.40) (5.27) (−0.10) (5.51)
Aget 0.001*** −0.971 0.001*** 0.001*** 0.013 0.001***

(3.01) (−1.51) (3.27) (3.01) (0.04) (2.99)
Losst 0.000 −6.176** 0.000 0.000 1.156 −0.000
(0.01) (−2.00) (0.15) (0.01) (0.71) (−0.08)
Mgsharet 0.048*** −5.863 0.051*** 0.048*** −5.040 0.048***
(2.74) (−1.42) (2.90) (2.74) (−0.29) (2.75)

Fixedt −0.018*** −2.161 −0.017*** −0.018*** 1.332 −0.019***
(−2.82) (−1.47) (−2.70) (−2.82) (1.30) (−2.92)
(continued)
Table 8 (continued)
Panel A: High-leverage sub-sample
Dependent variables

Constant 0.216*** −0.442*** 0.254*** 0.216*** 0.195*** 0.214***

(10.78) (−16.36) (12.25) (10.78) (9.47) (10.71)
Observations 8,878 8,878 8,878 8,878 8,878 8,878
Adjusted R2 0.2555 0.2456 0.2600 0.2555 0.2421 0.2596
Panel B: High-growth sub-sample
Dependent variables

CMff5_36t 3.388*** 5.020** 2.651*** 3.388*** 3.595* 2.835***

(3.63) (2.05) (2.71) (3.63) (1.70) (2.91)

Channel variable 0.235** 0.001
(2.44) (1.43)
Sizet −0.007*** 5.906*** −0.008*** −0.007*** −0.108 −0.010***
(−7.86) (4.47) (−9.94) (−7.86) (−0.06) (−7.14)
Bmt −0.070*** 1.755*** −0.134*** −0.070*** 1.160 −0.066***

(−20.80) (5.35) (−8.44) (−20.80) (0.66) (−9.26)
Levt 0.036*** 1.459*** 0.041*** 0.036*** 2.802*** 0.057***
(continued)
31
32
Table 8 (continued)
Dependent variables

(9.03) (7.50) (7.37) (9.03) (3.41) (7.21)

Roat −0.030** 1.244 −0.054*** −0.030** −7.772*** −0.078***
(−2.39) (1.29) (−4.52) (−2.39) (−4.35) (−6.12)
Sales_Growtht 0.007*** 0.333 0.005*** 0.007*** −0.256 0.009***
(5.52) (0.11) (4.02) (5.52) (−0.11) (4.89)
Aget 0.001*** −0.289 0.001*** 0.001*** 0.004 0.001***
(5.79) (−0.70) (5.29) (5.79) (0.11) (2.63)
Losst 0.002 −5.371 0.001 0.002 1.045 0.003
(1.34) (−1.09) (1.00) (1.34) (0.81) (0.93)
Mgsharet 0.044*** −3.739 0.030*** 0.044*** −5.597 0.042***

(4.41) (−1.41) (3.24) (4.41) (−0.91) (2.70)
Fixedt −0.018*** −2.591 −0.014*** −0.018*** 1.612 −0.015***
(−3.63) (−1.33) (−3.31) (−3.63) (0.92) (−4.54)
(continued)

Table 8 (continued)
Dependent variables

Constant 0.176*** −0.557*** 0.484*** 0.176*** 0.209*** 0.225***

(12.46) (−13.45) (12.65) (12.46) (8.18) (9.21)
Observations 9,137 9,137 9,137 9,137 9,137 9,137
Adjusted R2 0.2287 0.3230 0.2364 0.2287 0.0358 0.2283
This table shows the power of two potential channels (cash flow volatility and financial distress) in two subsamples: high financial leverage and
high growth rate. Using the median of financial leverage/sales growth rate of each industry in each year, we divide the samples into the high (low)
financial leverage/sales growth rate sub-sample. Following Wen et al. (2004) and Bentley-Goode et al. (2019), we use the following intermediation
model to explore the role of cash flow volatility and financial distress in the more meaningful subsample with high financial leverage and high
growth rate, with the sample period from 2007 to 2018. Panel A shows the results for the high financial leverage sample. Panel B shows the results
for the high sales growth rate sample. Controlsk,i,t refers to a set of control variables. The detailed descriptions of all variables are shown in the
Appendix. All continuous variables are winsorised at the 1% level in each tail. Industry and year fixed effects are controlled for and the standard
errors are corrected using the double-clustering (firm and year) method, as suggested by Petersen (2009). t-statistics are given in parentheses. *, **
and *** indicate statistical significance at the 10%, 5% and 1% levels, respectively.
AbsAcci,tþ1 ¼ β0 þ β1 EPU exposurei,t þ ∑βk Controlk,i,t þ ɛ i,tþ1 ðStep AÞ
k
Economicchanneli,t ¼ β0 þ β1 EPU exposure i,t þ ∑β k Controlsk,i,t þ ɛ i,t ðStep BÞ
k

AbsAcci,tþ1 ¼ β0 þ β1 EPU exposurei,t þ β2 Economic channeli,t þ ∑βk Controlsk,i,t þ ɛ i,tþ1 ðStep CÞ
k
33
high growth rate subsample, the cash flow volatility variable (Cashvola )
demonstrates a significant mediating effect when compared to the financial
distress variable (Z Score). The results show that in the high growth sub-
sample, the external EPU may lead to a huge fluctuation in corporate cash
flow, which will likely induce managers to use earnings management to deal
with the variations in firm performance.
5. Conclusion
We investigate the impact of EPU exposure on Chinese listed firms’ earnings

management. First, our primary finding is that EPU exposure is positively and
significantly related to firms’ earnings management. This result still holds after
we conduct several robustness tests to address endogeneity. Second, the results
of conditional analyses reveal that the effect of EPU exposure on earnings
management is more pronounced for firms with weaker external monitoring
mechanisms (e.g., non-Big-4 auditors, low analyst coverage and fewer
institutional shareholders). Moreover, for the first time, we link firm charac-
teristics to EPU exposure. We find that the financial leverage and growth rate
of a firm have the most significant predictive ability for EPU exposure. Finally,
we also examine the internal mechanism between EPU exposure and corporate
earnings management. We find that for firms with high financial leverage, EPU
exposure may be more likely to cause them to become financially distressed,
thus triggering earnings management. However, in firms with high growth,
EPU exposure affects the fluctuations of corporate cash flows, resulting in
earnings management behaviour.
Overall, this study adds new evidence on the impact of EPU exposure on
corporate decision-making. We support the proposal that EPU exposure
appears to have a significant impact on corporate finance. In addition, our
paper also has important policy implications. On the one hand, policymakers
should focus on smoothing adjustments when formulating and changing
economic policies to mitigate the negative influences of EPU. In this way, they
can mitigate firms’ incentives to engage in earnings management. On the other
hand, we also find that effective external monitoring mechanisms, such as Big-4
auditors and analyst coverage, can help alleviate the impact of EPU exposure
on earnings management. Therefore, China’s market as well as other emerging
markets should further strengthen the supervisory mechanisms in the market.
References
Aharony, J., J. Wang, and H. Yuan, 2010, Tunneling as an incentive for earnings
management during the IPO process in China, Journal of Accounting and Public Policy
29, 1–26.
Baker, S. R., N. Bloom, and S. J. Davis, 2016, Measuring economic policy uncertainty,
Quarterly Journal of Economics 131, 1593–1636.

Balakrishnan, K., M. B. Billings, B. Kelly, and A. Ljungqvist, 2014, Shaping liquidity:

on the causal effects of voluntary disclosure, The Journal of Finance 69, 2237–2278.
Bali, T. G., S. J. Brown, and Y. Tang, 2017, Is economic uncertainty priced in the cross-
section of stock returns?, Journal of Financial Economics 126, 471–489.
Ball, R., and L. Shivakumar, 2006, The role of accruals in asymmetrically timely gain
and loss recognition, Journal of Accounting Research 44, 207–242.
Bamber, L. S., and Y. S. Cheon, 1998, Discretionary management earnings forecast
disclosures: antecedents and outcomes associated with forecast venue and forecast
specificity choices, Journal of Accounting Research 36, 167–190.
Bentley-Goode, K. A., T. C. Omer, and B. J. Twedt, 2019, Does business strategy
impact a firm’s information environment?, Journal of Accounting, Auditing and
Finance 34, 563–587.
Bordo, M. D., J. V. Duca, and C. Koch, 2016, Economic policy uncertainty and the
credit channel: aggregate and bank level US evidence over several decades, Journal of
Financial Stability 26, 90–106.
Borghesi, R., J. F. Houston, and A. Naranjo, 2014, Corporate socially responsible
investments: CEO altruism, reputation, and shareholder interests, Journal of
Corporate Finance 26, 164–181.
Bova, F., Y. Dou, and O. K. Hope, 2015, Employee ownership and firm disclosure,
Contemporary Accounting Research 32, 639–673.
Brogaard, J., and A. Detzel, 2015, The asset-pricing implications of government
economic policy uncertainty, Management Science 61, 3–18.
Burgstahler, D., and I. Dichev, 1997, Earnings management to avoid earnings decreases
and losses, Journal of Accounting and Economics 24, 99–126.
Burgstahler, D., and M. Eames, 2006, Management of earnings and analysts’ forecasts
to achieve zero and small positive earnings surprises, Journal of Business Finance and
Accounting 33, 633–652.
Campbell, L., L. Liu, and K. Martin, 2015, Accrual earnings management prior to
delisting, Journal of Accounting and Finance 15, 53–63.
Cao, S. S., G. Ma, J. W. Tucker, and C. Wan, 2018, Technological peer pressure and
product disclosure, The Accounting Review 93, 95–126.
Carrière-Swallow, Y., and L. F. Céspedes, 2013, The impact of uncertainty shocks in
emerging economies, Journal of International Economics 90, 316–325.
Chen, J., F. Jiang, and G. Tong, 2017, Economic policy uncertainty in China and stock
market expected returns, Accounting and Finance 57, 1265–1286.
Cheng, F., C. Chiao, Z. Fang, C. Wang, and S. Yao, 2020, Raising short-term debt for
long-term investment and stock price crash risk: evidence from China, Finance
Research Letters 33, 101200.
Chi, H. Y., and T. C. Weng, 2014, Managerial legal liability and Big 4 auditor choice,
Journal of Business Research 67, 1857–1869.
Choi, A., J. H. Choi, and B. C. Sohn, 2018, The joint effect of audit quality and legal
regimes on the use of real earnings management: international evidence, Contempo-
rary Accounting Research 35, 2225–2257.
Chung, R., M. Firth, and J. B. Kim, 2005, Earnings management, surplus free cash flow,
and external monitoring, Journal of Business Research 58, 766–776.
Dai, L., and P. Ngo, 2020, Political uncertainty and accounting conservatism, European
Accounting Review, https://doi.org/10.1080/09638180.2020.1760117.
Davis, S. J., D. Liu, and X. S. Sheng, 2019, Economic policy uncertainty in China since
1946: the view from mainland newspapers. Working paper. https://site.stanford.edu/
sites/g/files/sbiybj8706/f/5331-economic_policy_uncertainty_in_chia_since_1949-the_
view_from_mainland_newspapers.pdf

Dechow, P. M., and I. D. Dichev, 2002, The quality of accruals and earnings: the role of
accrual estimation errors, The Accounting Review 77, 35–59.
Dechow, P. M., R. G. Sloan, and A. P. Sweeney, 1995, Detecting earnings management,
The Accounting Review 70, 193–225.
Demir, E., and O. Ersan, 2017, Economic policy uncertainty and cash holdings: evidence
from BRIC countries, Emerging Markets Review 33, 189–200.
Dhole, S., H. Manchiraju, and I. Suk, 2016, CEO inside debt and earnings management,
Journal of Accounting, Auditing and Finance 31, 515–550.
Elyasiani, E., Y. Wen, and R. Zhang, 2017, Institutional ownership and earning
management by bank holding companies, Journal of Financial Research 40, 147–178.
Fama, E. F., and K. R. French, 2015, A five-factor asset pricing model, Journal of
Financial Economics 116, 1–22.
Fama, E. F., and J. D. MacBeth, 1973, Risk, return, and equilibrium: empirical tests,
Journal of Political Economy 81, 607–636.
Fan, J. P., J. Huang, and N. Zhu, 2013, Institutions, ownership structures, and distress
resolution in China, Journal of Corporate Finance 23, 71–87.
Francis, J., R. LaFond, P. M. Olsson, and K. Schipper, 2004, Costs of equity and
earnings attributes, The Accounting Review 79, 967–1010.
Francis, J., P. Olsson, and K. Schipper, 2008, Earnings quality, Foundations and Trends
in Accounting 1, 259–340.
Francis, B. B., I. Hasan, and Y. Zhu, 2014, Political uncertainty and bank loan
contracting, Journal of Empirical Finance 29, 281–286.
Frijns, B., A. Gilbert, T. Lehnert, and A. Tourani-Rad, 2013, Uncertainty avoidance,
risk tolerance and corporate takeover decisions, Journal of Banking and Finance 37,
2457–2471.
Frot, E., and J. Santiso, 2013, Political uncertainty and portfolio managers in emerging
economies, Review of International Political Economy 20, 26–51.
Ghosh, A., and D. Moon, 2010, Corporate debt financing and earnings quality, Journal
of Business Finance and Accounting 37, 538–559.
Ghosh, D., and L. Olsen, 2009, Environmental uncertainty and managers’ use of
discretionary accruals, Accounting, Organizations and Society 34, 188–205.
Graham, J. R., C. R. Harvey, and S. Rajgopal, 2005, The economic implications of
corporate financial reporting, Journal of Accounting and Economics 40, 3–73.
Gulen, H., and M. Ion, 2015, Policy uncertainty and corporate investment, The Review
of Financial Studies 29, 523–564.
Habib, A., B. Uddin Bhuiyan, and A. Islam, 2013, Financial distress, earnings
management and market pricing of accruals during the global financial crisis,
Managerial Finance 39, 155–180.
Harris, C., and S. Roark, 2019, Cash flow risk and capital structure decisions, Finance
Research Letters 29, 393–397.
Harris, O., J. B. Karl, and E. Lawrence, 2019, CEO compensation and earnings
management: does gender really matter?, Journal of Business Research 98, 1–14.
He, G., 2016, Fiscal support and earnings management, International Journal of
Accounting 51, 57–84.
He, G., and H. Ren, 2017, Are financially constrained firms susceptible to a stock price
crash?, WBS Finance Group Research Paper No. 239. Available at: http://dx.doi.org/
10.2139/ssrn.3021307.
Healy, P. M., and K. G. Palepu, 2001, Information asymmetry, corporate disclosure,
and the capital markets: a review of the empirical disclosure literature, Journal of
Accounting and Economics 31, 405–440.
Hirshleifer, D., K. Hou, and S. H. Teoh, 2012, The accrual anomaly: risk or mispricing?,
Management Science 58, 320–335.

Hu, S., and D. Gong, 2019, Economic policy uncertainty, prudential regulation and
bank lending, Finance Research Letters 29, 373–378.
Huang, X., S. H. Teoh, and Y. Zhang, 2013, Tone management, The Accounting Review
89, 1083–1113.
Huang, Y., Y. Ji, X. Ni, Y. Tan, and X. Wang, 2020, Economic policy uncertainty, state
ownership, and credit allocation, Available at: http://dx.doi.org/10.2139/ssrn.3177982.
Jensen, M. C., and W. H. Meckling, 1976, Theory of the firm: managerial behavior,
agency costs and ownership structure, Journal of Financial Economics 3, 305–360.
Jin, X., Z. Chen, and X. Yang, 2019, Economic policy uncertainty and stock price crash
risk, Accounting and Finance 58, 1291–1318.
Jones, J. J., 1991, Earnings management during import relief investigations, Journal of
Accounting Research 29, 193–228.
Kong, D., J. Xiang, J. Zhang, and Y. Lu, 2019, Politically connected independent
directors and corporate fraud in China, Accounting and Finance 58, 1347–1383.
Kothari, S. P., A. J. Leone, and C. E. Wasley, 2005, Performance matched discretionary
accrual measures, Journal of Accounting and Economics 39, 163–197.
Lang, M., and M. Maffett, 2011, Transparency and liquidity uncertainty in crisis
periods, Journal of Accounting and Economics 52, 101–125.
Lee, E., M. Walker, and C. Zeng, 2014, Do Chinese government subsidies affect firm
value?, Accounting, Organizations and Society 39, 149–169.
Li, L., 2019, Is there a trade-off between accrual-based and real earnings management?
Evidence from equity compensation and market pricing, Finance Research Letters 28,
191–197.
Li, D., M. Zheng, C. Cao, X. Chen, S. Ren, and M. Huang, 2017, The impact of
legitimacy pressure and corporate profitability on green innovation: evidence from
China top 100, Journal of Cleaner Production 141, 41–49.
Lin, C., P. Lin, F. M. Song, and C. Li, 2011, Managerial incentives, CEO characteristics
and corporate innovation in China’s private sector, Journal of Comparative Economics
39, 176–190.
Liu, X., 2016, Corruption culture and corporate misconduct, Journal of Financial
Economics 122, 307–327.
Liu, G., and C. Zhang, 2019, Economic policy uncertainty and firms’ investment and
financing decisions in China, China Economic Review 63, 101279.
Liu, Q., T. Luo, and G. G. Tian, 2019, How do political connections cause SOEs and
non-SOEs to make different M&A decisions/performance? Evidence from China,
Accounting and Finance 59, 2579–2619.
McVay, S. E., 2006, Earnings management using classification shifting: an examination
of core earnings and special items, The Accounting Review 81, 501–531.
Mu, C., A. Wang, and J. Yang, 2017, Optimal capital structure with moral hazard,
International Review of Economics and Finance 48, 326–338.
Na, K., and J. Hong, 2017, CEO gender and earnings management, Journal of Applied
Business Research 33, 297–308.
Nagar, V., J. Schoenfeld, and L. Wellman, 2019, The effect of economic policy
uncertainty on investor information asymmetry and management disclosures, Journal
of Accounting and Economics 67, 36–57.
Oral, C., and G. CenkAkkaya, 2015, Cash flow at risk: a tool for financial planning,
Procedia Economics and Finance 23, 262–266.
Parsons, C. A., J. Sulaeman, and S. Titman, 2018, The geography of financial
misconduct, The Journal of Finance 73, 2087–2137.
Pástor, Ľ., and P. Veronesi, 2012, Uncertainty about government policy and stock
prices, The Journal of Finance 67, 1219–1264.

Pástor, Ľ., and P. Veronesi, 2013, Political uncertainty and risk premia, Journal of
Financial Economics 110, 520–545.
Petersen, M. A., 2009, Estimating standard errors in finance panel data sets: comparing
approaches, The Review of Financial Studies 22, 435–480.
Pierk, J., and M. Weil, 2016, Price regulation and accounting choice, Journal of
Accounting and Public Policy 35, 256–275.
Rountree, B., J. P. Weston, and G. Allayannis, 2008, Do investors value smooth
performance?, Journal of Financial Economics 90, 237–251.
Subramanyam, K. R., 1996, The pricing of discretionary accruals, Journal of Accounting
and Economics 22, 249–281.
Thanh, S. D., N. P. Canh, and N. T. T. Ha, 2020, Debt structure and earnings
management: a non-linear analysis from an emerging economy, Finance Research
Letters 35, 101283.
Tsai, I. C., 2017, The source of global stock market risk: a viewpoint of economic policy
uncertainty, Economic Modelling 60, 122–131.
Verrecchia, R. E., 1990, Information quality and discretionary disclosure, Journal of
Accounting and Economics 12, 365–380.
Wang, Y., B. Zhang, X. Diao, and C. Wu, 2015, Commodity price changes and the
predictability of economic policy uncertainty, Economics Letters 127, 39–42.
Wang, Y., Y. Wei, and F. M. Song, 2017, Uncertainty and corporate R&D investment:
evidence from Chinese listed firms, International Review of Economics and Finance 47,
176–200.
Wen, Z., L. Chang, K. T. Hau, and H. Liu, 2004, Testing and application of the
mediating effects, Acta Psychologica Sinica 36, 614–620.
Winkler, J., C. P. J. W. Kuklinski, and R. Moser, 2015, Decision making in emerging
markets: the Delphi approach’s contribution to coping with uncertainty and
equivocality, Journal of Business Research 68, 1118–1126.
Wu, J., J. Zhang, Y. Wu, and D. Kong, 2020, When to go abroad: economic policy
uncertainty and Chinese firms’ overseas investment, Accounting and Finance 60,
1435–1470.
Xu, N., Q. Chen, Y. Xu, and K. C. Chan, 2016, Political uncertainty and cash holdings:
evidence from China, Journal of Corporate Finance 40, 276–295.
Xu, Z., 2020, Economic policy uncertainty, cost of capital, and corporate innovation,
Journal of Banking and Finance 111, 105698.
Yang, Z., Y. Yu, Y. Zhang, and S. Zhou, 2019, Policy uncertainty exposure and market
value: evidence from China, Pacific-Basin Finance Journal 57, 101178.
Yao, S., Z. Wang, M. Sun, J. Liao, and F. Cheng, 2020, Top executives’ early-life
experience and financial disclosure quality: impact from the Great Chinese Famine,
Accounting and Finance, https://doi.org/10.1111/acfi.12659.
Yuan, R., J. Sun, and F. Cao, 2016, Directors’ and officers’ liability insurance and stock
price crash risk, Journal of Corporate Finance 37, 173–192.
Yung, K., and A. Root, 2019, Policy uncertainty and earnings management:
international evidence, Journal of Business Research 100, 255–267.
Zhang, L., E. I. Altman, and J. Yen, 2010, Corporate financial distress diagnosis model
and application in credit rating for listing firms in China, Frontiers of Computer
Science in China 4, 220–236.

Appendix
Variables Definitions
Economic policy uncertainty exposure

CMff3_36 (CMff5_36) or The economic policy uncertainty exposure, measured by the negative value of βepu i,τ , which is obtained by regressing
CMff3_60 (CMff5_60) stock returns on the EPU index (following Davis et al. (2019), which we scale down by 100), controlling for the
Fama-French three factors (or five factors), using the past 36 months (τ-36, τ-1), or the past 60 months (τ-60, τ-1),
as the rolling window. We take the average of monthly values to get the annual measure. See Equations (1) and (5)
for details.
Dependent variables
AbsAcc The proxy for earnings management, see Equation (2) for details.
Jones Following Jones (1991), we construct the alternative discretionary accruals measures using the Jones model for the
proxy of earnings management.
Nonlinear Following Ball and Shivakumar (2006), we construct the alternative discretionary accruals measures using the nonlinear
accruals model for the proxy of earnings management.
ROA_matched Following Kothari et al. (2005), we construct the alternative discretionary accruals measures using the ROA-matched
model for the proxy of earnings management.
EQ The proxy for earnings quality. See Equation (3) for details.
Control variables
Size Firm size, the natural logarithm of the market value of equity.
Bm Book-to-market ratio, calculated as the book value of equity divided by the market value of equity.
Lev Firm financial leverage, defined as total debt divided by total assets.
Roa Return on assets, calculated as net profit divided by total assets.
Sales_Growth Sales growth rate, the percentage change in sales compared to prior year’s sales.

Age Listing age, which is calculated as the natural logarithm of the number of years since a firm’s IPO.
(continued)
39
40
Appendix 1 (continued)
(continued)
Variables Definitions
Loss Net loss, which is an indicator variable that equals one if the firm reports a net loss for the year, and zero otherwise.
Mgshare Managerial ownership, which is measured by the ratio of the
market value of managerial shareholdings to the total market value of the firm.
Fixed Fixed assets ratio, gross value of property, plant and equipment to total assets.
Other variables of interest
Cashvola The volatility of cash flows from operations, measuring by the standard deviation of cash flows from firm’s operating
activities from t − 1 to t + 1.
−Z_Score Financial distress. Following Zhang et al. (2010) and Lee et al. (2014), we calculate the Z-Score. We use the negative
value of Z-Score to proxy for financial distress.
Soe A dummy variable that equals one if the firm is an SOE (state-owned enterprise), and zero otherwise.
Big4 A dummy variable that equals one if firm i in year t is audited by one of the Big-4, which refer to the biggest four
international auditors, Deloitte, PwC, EY, and KPMG, and zero otherwise.
High-An A dummy variable that equals one if the number of analyst coverage of firm i in year t is above the average analyst
coverage of all sample firms in the same industry, and zero otherwise.
High-Inst A dummy variable that equals one if the percentage of stocks held by institutional investors of firm i in year t is
above the average institutional shareholding of all sample firms in the same industry, and zero otherwise.

10 1111@acfi 12722

Uploaded by

Copyright:

Available Formats

10 1111@acfi 12722

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

10 1111@acfi 12722

Uploaded by

Copyright:

Available Formats

Accounting & Finance

Economic policy uncertainty exposure and earnings

Xin Cuia, Shouyu Yaoa, Zhenming Fanga, Hua Wangb

We investigate the impact of economic policy uncertainty (EPU) exposure on

Key words: Cash ﬂow volatility; Earnings management; Economic policy

JEL classiﬁcation: O31, G38

Economic policy uncertainty (EPU) refers to the condition that economic

Please address correspondence to Hua Wang via email: [email protected]

© 2020 Accounting and Finance Association of Australia and New Zealand

© 2020 Accounting and Finance Association of Australia and New Zealand

© 2020 Accounting and Finance Association of Australia and New Zealand

be directly applied in China. Due to the government’s interventions in

© 2020 Accounting and Finance Association of Australia and New Zealand

from the perspective of the heterogeneity of individual stocks’ EPU exposure.

© 2020 Accounting and Finance Association of Australia and New Zealand

exposure and earnings management. Policy implications and conclusions are

2. Data, variables and methodology

2.1. Sample and data

2.2. EPU index

2.3. The measurement of EPU exposure

© 2020 Accounting and Finance Association of Australia and New Zealand

Ri,τ r f,τ ¼ β0 þ βmkt

2.4. Earnings management measurement

TAi,t 1 ðΔREVi,t ΔRECi,t Þ PPEi,t

© 2020 Accounting and Finance Association of Australia and New Zealand

model. We then calculate three other earnings management measures: Jones,

where ΔWCi,t is the changes in working capital of ﬁrm i from year t − 1 to t,

To estimate the relation between EPU exposure and earnings management,

EMi,tþ1 ¼ β0 þ β1 EPU exposurei,t þ ∑γ k Controlk,i,t þ ɛi,tþ1 , (4)

where EMi,tþ1 is the earnings management of ﬁrm i in year t + 1, measured by

© 2020 Accounting and Finance Association of Australia and New Zealand

3.1. Descriptive statistics and correlations

In Table 1, we present the descriptive statistics of the variables and their

3.2. Baseline results

© 2020 Accounting and Finance Association of Australia and New Zealand

AbsAcc 16,992 0.071 0.078 0.021 0.047 0.091 0.001 0.452

Fixed 16,992 0.219 0.164 0.091 0.186 0.313 0.002 0.707

© 2020 Accounting and Finance Association of Australia and New Zealand

3.3. Robustness tests and endogeneity

3.3.1. The sensitivity test of EPU exposure

EPU exposure is a beta type sensitivity measure of a ﬁrm’s stock price. We

3.3.2. An alternative measure of EPU exposure

In the above analysis, we use the more advanced Fama-French ﬁve-factor

© 2020 Accounting and Finance Association of Australia and New Zealand

© 2020 Accounting and Finance Association of Australia and New Zealand

AbsAcct+1 Jonest+1 Nonlineart+1 ROA_matchedt+1 EQt+1

Variables (1) (2) (3) (4) (5)

CMﬀ5_36t 3.019*** 3.811*** 2.309*** 3.367*** −1.110**

(5.68) (5.02) (5.09) (4.47) (−2.74)

© 2020 Accounting and Finance Association of Australia and New Zealand

AbsAcct+1 Jonest+1 Nonlineart+1 ROA_matchedt+1 EQt+1

Variables (1) (2) (3) (4) (5)

Mgsharet 0.051*** 0.047*** 0.036*** 0.037*** −0.015***

© 2020 Accounting and Finance Association of Australia and New Zealand

Panel A: The sensitivity test of EPU exposure

Variables AbsAcct+1 Jonest+1 Nonlineart+1 ROA_matchedt+1 EQt+1

CMﬀ5_36t 2.564** 3.210** 1.892** 2.667** −0.953*

Panel B: An alternative model (FF3) for EPU exposure estimation

Variables AbsAcct+1 Jonest+1 Nonlineart+1 ROA_matchedt+1 EQt+1

CMﬀ5_36t 3.019* 3.811* 2.309* 3.367* −1.110**

Mgsharet 0.051* 0.047* 0.036* 0.037* −0.015***

CMﬀ5_36t 2.564 3.210 1.892 2.667 −0.953*

CMﬀ3_36t 2.573* 3.455* 2.062 2.553* −1.662***

Constant 0.190* 0.190* 0.193* 0.193*

CMﬀ3_36t 4.093* 3.794*

CMﬀ5_36t −3.672 −0.156 0.089 1.966 2.543 3.136**