International Review of Economics and Finance 42 (2016) 153–166

Contents lists available at ScienceDirect

International Review of Economics and Finance

journal homepage: www.elsevier.com/locate/iref

Political connection, government policy, and investor trading:

Evidence from an emerging market
Chih-Yung Lin b, Po-Hsin Ho a, Chung-Hua Shen c,⁎, Yu-Chun Wang d
a
Department of Business Administration, National Taipei University, Taiwan
b
College of Management & Innovation Center for Big Data and Digital Convergence, Yuan Ze University, Taiwan
c
School of Finance and Statistics, Hunan University, China
d
Department of Finance, Ming Chuan University, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: This study investigates whether political connection (PC), government policy (GP), or both affect
Received 22 January 2014 stock returns before and after the Taiwanese 2008 presidential election. We also examine whether
Received in revised form 8 July 2015 the two effects influence the five types of investor trading during the election. Past studies have
Accepted 30 September 2015
separately focused on either the PC or the GP effect, whereas this study considers both effects. Spe-
Available online 22 October 2015
cifically, we consider five types of investors, namely, foreign investors, security investment trust
companies (INVTRUSTs), security dealers, board of directors and supervisors, and individual in-
JEL classification: vestors. We find that firms connecting only with the winning party exhibit positive abnormal
G11
returns, and only INVTRUSTs increase the shareholdings of these firms (PC effect). Our results
G14
do not support the GP effect alone. Instead, we identify that firms with joint PC and GP effects
G18
G32 have the abnormal returns during election periods.
© 2015 Elsevier Inc. All rights reserved.
Keywords:
Political connection
Government policy
Stock returns
Investor trading
Presidential election

1. Introduction

An increasing number of studies have been devoted to the investigation of politics and finance. Political topics include political
events, such as elections, government policy (GP), political connection (PC), political intervention, and corruption. Among the studies
related to politics, Belo, Gala, and Li (2013) focus on the influence of the GP on government spending on the stock returns of firms,
whereas Goldman, Rocholl, and So (2009) and Cooper, Gulen, and Ovtchinnikov (2010) consider the PC effect on the stock market
during the presidential election. However, given that both GP and PC may affect stock returns during the presidential election, con-
sidering only one of them may cause inaccuracies in the findings of the effects of the two factors.
PC denotes that the chief executive officer (CEO) or board of directors of firms either serves in the central committee (e.g., Li, Meng,
Wang, & Zhou, 2008) or financially donates to a political party (e.g., Claessens, Feijen, & Laeven, 2008; Cooper et al., 2010). Firms con-
nected with politics are typically referred to as PC firms. Investors subsequently expect PC firms to obtain preferential treatment when
applying for loans, to be informed in advance of future policy directions, and to be the first to be bailed out, among many advantages.

⁎ Corresponding author.
E-mail addresses: phho@mail.ntpu.edu.tw (P.-H. Ho), d95723009@ntu.edu.tw (C.-Y. Lin), chshen01@ntu.edu.tw (C.-H. Shen), allan.ycwang@gmail.com
(Y.-C. Wang).

http://dx.doi.org/10.1016/j.iref.2015.09.008
1059-0560/© 2015 Elsevier Inc. All rights reserved.
154 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166

For example, using data on firm applications for capital under the Troubled Asset Relief Program (TARP),1 Duchin and Sosyura (2012)
show that PC firms are more likely to be funded than other types of firms. Accordingly, the “PC effect” denotes that stock returns and
shares trading of these firms tend to increase because investors purchase shares of firms with PCs with the winning party. This effect
has been confirmed by studies, such as those by Ferguson and Voth (2008) based on German data, Claessens et al. (2008) on Brazilian
data, and Goldman et al. (2009), Cooper et al. (2010), and Kim, Pantzalis, and Park (2012) on U.S. data.
GP denotes that given the political goals of each political party, investors are more likely to buy shares of firms whose businesses
may benefit from the policy, despite the lack of PCs of firms to the government (Pástor & Veronesi, 2012). For example, in the United
States, the Democratic Party focuses significantly on the minimum-wage policy, health-care reform, and income inequality, whereas
the Republican Party encourages laissez-faire economics, fiscal conservatism, and the elimination of government-run welfare pro-
grams. The business of a firm must match the policies to obtain the desired benefits. Belo et al. (2013) report that firms with high
level of industry exposure to government spending experience higher cash flows and stock returns during Democratic presidencies.
Therefore, the “GP effect” denotes that stock returns and shares trading of these firms tend to increase, because investors purchase
shares of firms whose businesses are consistent with the policies of the winning party in the election.
Both GP and PC affect the values of firms, but differ in their influence channels. The GP effect requires the business operations of
firms to be consistent with the policy direction without establishing a PC with the ruling party. By contrast, the PC effect focuses on
firms that connect with politicians, especially those from the winning party, without their businesses being necessarily integrated
with GP. An example is a company with a good relationship with politicians but whose business is not directed to favor established
policies. Firms may exhibit both GP and PC effects, indicating that their businesses are consistent with the policy direction and they
have PC with the winning party in the election. At the same time, firms may also exhibit neither of the two effects. GP and PC effects
have been examined separately in the past, but their effect may jointly influence the stock returns and trading shares. To the best of
our knowledge, no study has investigated the simultaneous GP and PC effects.
This study examines PC and GP effects during the 2008 presidential election in Taiwan. We examine these phenomena through the
following research questions: Does the influence of PC and GP effects generate abnormal stock returns during the election period?
Does the influence of PC and GP effects motivate investors to increase their stock holdings? What types of traders are most likely
to be affected by PC and GP effects? Do strategies of PC and GP bring profit?
These four issues differ from those examined in past studies in terms of two aspects. First, early studies did not distinguish PC and
GP effects.2 Therefore, the increasing stock returns or superior accounting performance of specific firms may be attributed to PC (GP)
effects when in fact they emerge from the PC (GP) effect during the election. Second, past studies implicitly assume that all types of
investor experiences have the same share-trading patterns (Goldman et al., 2009). The different types of investors may plausibly have
different priorities in terms of returns and risk, knowledge accumulation and information collection, trading experience, and the abil-
ity to use public information in evaluating the election outcomes and GPs. Thus, trading patterns may vary among different types of
traders in response to PC and GP effects. We consider the assumption of the same trading styles by classifying investors into five types,
namely, foreign investors (FIs), security investment trust companies (INVTRUSTs), security dealers (DEALERs), insiders, such as board
of directors and supervisors (BOARDs), and outsiders, such as individual investors (INDINVs).3 With no theoretical and empirical guid-
ance, we have limited knowledge on which investor type exhibits a trading pattern in line with PC and GP effects ex ante. Therefore,
the topic remains an empirical issue.
The use of the data on Taiwan is influenced by three unique features. First, the two major political parties in Taiwan, the
Kuomintang Party (KMT) and the Democratic Progressive Party (DPP) have significantly different economic policies, which aim to
provide an ideal environment for the identification of the consistency or inconsistency of business operations of firms with GP. Ac-
cording to Ross (2006) and Imai and Shelton (2011), KMT has adopted the trading policies of China and emphasizes the importance
of the Chinese market, whereas DPP prefers trading with other countries or regions and expresses that the Chinese market is less im-
portant than it appears to be. These two strikingly different policies easily allow us to identify the GP effect. Second, the exact and de-
tailed trading data by the five types of traders are rarely accessible in other developing countries. We collected the necessary data on
these investors from the Taiwan Stock Exchange (TWSE), which allows us to investigate this issue. Finally, Taiwan maintains data on
political donations since 2005 under the Sunshine Law. Thus, we can clearly distinguish PC from non-PC firms. Donation data are un-
usually available in other developing countries.
With the surveys indicating that 75% of voters believe that KMT will win the elections, the theoretical PC effect suggests that in-
vestors will purchase the shares of PC firms connected with KMT (KMT firms hereafter) and sell the shares of PC firms connected
with DPP (DPP firms) even prior to the elections. We examine the effects 40 days before and 5 days after the elections. By contrast,
the GP effect indicates that investors will buy the shares of firms with strong links to China. Hence, stock returns and shares trading
of PC firms connected with KMT (PC effect), as well as those of firms with businesses linked to China (GP effect) will increase.
Our empirical evidence modifies the traditional PC effect by finding that the joint effects of PC and GP are more significantly pos-
itive in the measure of cumulated abnormal results (CAR) during the 2008 presidential election in Taiwan. During the election, inves-
tors buy the shares of KMT firms to earn potential profit, and they sell the shares of DPP firms to avoid future losses. Hence, KMT and
DPP firms have positive and negative CAR, respectively, which are fully consistent with the expected PC effect. Second, we find that the

1
TARP is a program of the U.S. government signed into law by President George W. Bush on October 3, 2008. The program purchases assets and equity from financial
institutions to strengthen the financial system.
2
Ferguson and Voth (2008), Claessens et al. (2008), Goldman et al. (2009), Cooper et al. (2010), and Kim et al. (2012) all confirm the PC effect during elections.
3
The Financial Supervisory Commission (FSC) is the official authority of the Banking, Securities, Derivatives, and Insurance of Taiwan; it requires all institutional in-
vestors to report daily firm-level trading information, providing us with rich trading data for this research.
 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166 155

GP effect alone is unsupported. Third, we also observe that the joint effects of PC and GP are strong and more significantly positive in
the measure of CAR. Fourth, our trading results reveal that only INVTRUSTs make significant purchases of the shares of KMT firms. Fi-
nally, we find that INVTRUSTs have significantly positive CAR based on their trading activities.
Our study contributes to the literature in three aspects. First, we provide a clear definition of the differences between PC and GP
effects and provide the operational definition for each effect. Failure to distinguish between the two effects may mix them together
and yield an incorrect inference. Second, we investigate the PC and GP effects not only on stock returns but also on trading behaviors
of investors. Few studies have examined the relationship between politics and the investment decisions of investors. For example,
using data on political contributions and stock holdings of U.S. investment managers, Hong and Kostovetsky (2012) suggest that mu-
tual fund managers who make campaign donations to Democrats hold less of their portfolios in firms deemed to be socially irrespon-
sible.4 In this study, we use daily share-trading data instead of mutual fund data used by Hong and Kostovetsky to examine this
relationship in this study. Third, we posit that different types of investors have different trading behaviors in response to the election,
which past studies have assumed to be the same.
The remainder of this paper is organized as follows: the next section introduces the background of political factors in Taiwan and
describes the data used in this study, Section 3 presents the empirical models and variables, Section 4 discusses the empirical results
and additional tests, and Section 5 concludes the paper.

2. Political factors and data in Taiwan

2.1. Political background in Taiwan

In 1996, Taiwan adopted the direct voting system in its presidential elections. Two major political parties, KMT and DPP, compete for
the presidency every four years. KMT was the ruling party until May 2000, when the party lost to DPP, which subsequently won the next
two elections and took charge of the government from 2000 to 2008. The presidency transferred from DPP to KMT in the 2008 election.
Prior to this election, major political analysts predicted the victory of KMT. The Gallup polls center even predicted that KMT would win by
landslide victory. As expected, KMT won by obtaining 16.9% (approximately two million) more votes than DPP. The assumed victory of
the KMT by a wide margin over DPP suggests that the market should respond to political factors prior to the day of the election.

2.2. Measurements of GP and PC

The first political factor is PC. Anecdotal evidence validates the finding that the effects of PC on firm value draw investor attention. For
example, Mr. Eric Chu, the vice president of the incumbent ruling party (KMT) in Taiwan, was appointed vice premier on September 7,
2009. The following day, the stock price of Twinhead Company immediately increased from NT$3.08 to NT$6.43 per share for 11 consec-
utive trading days (108% increase in stock returns).5 This stock price appreciation puzzled corporate finance analysts because the reve-
nues, strategies, corporate governance, research and development, credit ratings, and other facets of the company remained the same.
Such appreciation could have possibly resulted from the CEO of Twinhead Company being the father-in-law of the vice premier. If this
inference is true, then PC can thus be considered valuable to firms at least in the short term, such as during an election period.
Following the literature on PC, we define a firm as having PCs if it exhibits any one of the two measures of political activities. The
first measure is political party tendency, which is composed of three elements: firms whose chairman or CEO is a central committee
member of political parties (Li et al., 2008), firms whose chairman or CEO is well known as a close friend to a candidate,6 and firms
whose chairman or CEO shows public support for a specific candidate on newspapers, or the support is reported by news media.
The second measure of PCs is political donations. The active role of donations in the relationship between politicians and firms has
been modeled extensively (Grossman & Helpman, 1994; Kroszner & Stratmann, 2005). Thus, this measure is adopted as a proxy for
PCs in numerous studies, including those by Knight (2006), Claessens et al. (2008), and Cooper et al. (2010). Political party tendency
and political donations are important in establishing connections with political parties during elections in Taiwan. For instance, the
top officers in a firm who exhibit support for a candidate on newspapers, or firms that contribute to a candidate during the campaign,
would earn the gratitude of the candidate and would be considered as “allies” or “close friends.” Thus, upon winning the elections, the
candidate tends to reward these supporters either explicitly or implicitly (Cox & McCubbins, 1986).
The second political factor is GP. Investors also believe that policies would change considerably once KMT wins. As mentioned pre-
viously, KMT and DPP have opposite opinions on trading with Mainland China. Many investors believe the new policy from KMT
would benefit from the economic prospects, and as a result, the stock market positively reacted to such expectation because of the
following reasons. First, KMT suggested that in opposition to DPP, Taiwan should cooperate with Mainland China in terms of economic
activities. Given that Taiwan is an economy with high dependence on trade, a closer relationship with Mainland China would buoy
Taiwan's exports to that country and stimulate Taiwan's economic growth. A number of listed firms have also invested in China for
its lower labor costs and increasing market demand.7 An agreement to reduce commercial barriers would help the firms lower

4
Their benchmarks are non-donors or Republican donors. Socially irresponsible firms include tobacco producers, gun manufacturers, defense firms, and companies
with bad employee relations or diversity records.
5
The stock price of Twinhead Company increased again on February 22, 2010, when the nomination of Mr. Chu as county mayor of New Taipei City was announced.
6
Fisman (2001), Faccio (2006), Faccio and Parsley (2009), Su, Fung, Huang, and Shen (2014), and Chen, Shen, and Lin (2014) identify this close relationship in dif-
ferent countries to represent PCs.
7
For example, Hon Hai Precision Company, the parent company of Apple's supplier Foxconn, has established numerous factories in Guangdong, Shanxi, Sichuan, and
so on. Another big food conglomerate in Taiwan, Uni-President Company, has also received considerable attention in China's food market.
156 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166

their operation costs. Second, KMT has a longer history in governing the island and thus would have more experienced officers with
the sophistication to formulate economic policies. Accordingly, the changes of GP would occur if KMT wins the 2008 presidential elec-
tion, and then the stock returns of firms involving the GP effect will increase.

2.3. Data

Our sample consists of firms listed on the TWSE on the date of the 2008 presidential election. For the PC proxy, we hand-collect
information on the political party tendency of our sample firms from political party websites, the TWSE database, the Market Obser-
vation Post System database, and the database of two major news media in Taiwan, namely, the Central News Agency and the United
Daily News. Regarding data on political donations, the 2005 Sun Shine Law of Taiwan stipulates that political parties should provide
annual disclosure of the political donations they received.8 Hence, we successfully gathered the sheets with lists of firms that donated
to their respective presidential candidates in the 2008 presidential election from the Control Yuan (the top supervisory office in
Taiwan).
In contrast to most previous studies that employed only one of the two PC measures, our study creates a more reliable proxy for the
PC of listed firms. These data represent the first released donation data for the Taiwanese presidential elections, thereby restricting the
election event examined in this study. Considering that the comprehensive donation data contain all listed and unlisted firms, we also
identify listed donating firms manually to construct our PC proxy. Finally, we define firms as KMT or DPP firms when they have con-
nections with KMT or DPP.9 The definitions of all variables are listed in Table 1.
The GP may also change when the ruling party rotates. Following the concept of Belo et al. (2013) and considering the policy dif-
ferences in two competing parties, we use POLICY variable as the proxy of benefiting from the GP, which is the investment-to-total-
assets ratio (INV/Asset) in China. KMT proposes the policy of establishing a close economic relationship with China, whereas DPP does
not. Thus, firms with large amounts of investment in China are consistent with the KMT policy and therefore would possibly benefit
more from the policy favors, such as participating in the policy drafting. On the contrary, firms with small amounts of investment in
China are consistent with the DPP policy.
Table 2 presents the sample comprising 628 firms, but financial institutions are excluded because they belong to a highly regulated
industry. A total of 126 KMT firms, 74 DPP firms, and 15 firms are connected with both parties, and 413 firms are not politically con-
nected (non-PC firms) in our sample.10 Most of the firms are connected with KMT because the party served as the ruling party for
more than five decades (1949–2000 and 2008–2012).
Data on the five types of investors, namely, FIs, INVTRUSTs, DEALERs, BOARDs, and INDINVs, are accessible only after 2008.11
The first three types of investors are institutional investors, and the last two are individual investors. FIs consist of investment
banks, insurance companies, mutual funds, and pension funds mainly from the developed countries, such as the United States,
and European countries. FIs have a decisive influence on the Taiwan stock index because they hold more than 30% of the total
market value of the listed companies. They may even hold over 50% of the shares for some several listed companies.12 Next,
the trading activities of INVTRUSTs and DEALERs have increased in the last decade, reaching nearly 30% of the daily dollar trad-
ing volume; INVTRUSTs and DEALERs are often recognized as informed and influential traders in Taiwan (Chiang, Qian, &
Sherman, 2010). BOARDs have a natural advantage in corporate information, and thus, their trading activities responding to
the presidential elections might be an interesting topic for investigation. In particular, we separate FIs from BOARDs to clarify
the motives of shareholdings, because domestic directors, who aim to participate in corporate operations, may not change
their shareholdings for short-term capital gains. Finally, the trading activities of INDINVs account for over 60% of the daily dollar
trading volume of the entire stock market.13
We also collect the information on daily stock prices and financial reports submitted by listed firms to the TWSE, as well as data on
stock returns and the characteristics of a firm from the Taiwan Economic Journal.

3. Empirical model and variable construction

3.1. Regression of stock returns

The PC effect suggests that if firms have PCs with the winning party, then their stock returns will increase. Similarly, the GP effect
suggests that if firms are consistent with the policies of the winning party, then their stock returns will increase. We examine these

8
Congressional candidates also have to disclose all donations immediately after the elections. All these information can be found in the Control Yuan of Taiwan. How-
ever, the data are only accessible after a one-and-half-year lag because of the collection process involved.
9
Firms that donate to both parties are included in our sample. However, we also show the results in the Robustness check section by using the sample that excludes
firms that donate to both parties.
10
Firms with no PC effect in this study are defined as non-politically connected firms.
11
They are required to report their trading volumes to each firm to the Securities and Futures Bureau of the FSC.
12
Twelve firms in our sample are held by FIs for more than 50% before the election.
13
Although individuals are excluded from such disclosure rules, we use margin trading to proxy for the trading of INDINVs because only individuals can engage in such
activities. Institutional investors are not permitted to engage in margin trading, and board members rarely engage in it because the interest rates of over 6% would large-
ly shift the costs of long-term shareholdings. Board members in Taiwan often control and share the profits of a company through an investment company under the
consideration of tax rates.
 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166 157

Table 1
Variable definitions.

Variable Definition Source

CAR Cumulative abnormal return. Abnormal return estimates from market model and use 250 trading days before the testing TEJ & the current
period. In this study, CAR (−40, 5) refer to the cumulative abnormal returns from 40 days before the election to 5 days study
after the election. CAR (−40, −1), CAR (0), and CAR (1, 5) refer to the cumulative abnormal returns in the pre-election
period, election day, and post-election period, respectively.
ΔShare Change in the shareholding ratios of investors. In this study, five types of investors includes foreign investors (FIs), security TEJ & the current
investment trust companies (INVTRUSTs), security dealers (DEALERs), board of directors and supervisors (BOARD), and study
individual investors (INDINVs).
KMT Dummy variable that is equal to 1 if a firm is connected with KMT (the winning party), and 0 otherwise The current study
DPP Dummy variable that is equal to 1 if a firm is connected with DPP (the losing party), and 0 otherwise The current study
POLICY The investment to total assets ratio (INV/Asset) in China as the proxy of benefiting from the government policy. TEJ & The current
study
BETA Market beta estimated in the market model in the period (−290, −41) TEJ & the current
study
LNMKTCAP The logarithm of the last year-end market value of outstanding shares of each firm TEJ
BM Book-to-market ratio in the third quarter last year, calculated as book value of equity divided by market value of equity TEJ
MORET The raw return for each stock minus the market return one year before the testing period TEJ & the current
study
ROA Ratio of net income to asset TEJ
CASH Cash and cash equivalence divided by total assets TEJ
DIV The cash dividend payout divided by net income TEJ
SALESGR Growth ratio of firms' sales TEJ
LEV Debt-to-asset ratio TEJ

Table 2
Sample firms.

Political connection effect

Firm classification Number of firms From political party tendency measure From donation measure

(1) Firms that have connections with KMT (KMT firms) 126 90 36
(2) Firms that have connections with DPP (DPP firms) 74 35 39
(3) Firms that have connections with both parities 15 15
(4) Non-PC firms 413

This table presents the sample firms during the 2008 Taiwanese presidential election. The political connection proxy comprises two sub-measures, i.e., political party
tendency and political donation. They are defined in Section 2.2. There are a total of 628 firms in 2008.

two effects to investigate the CAR of firms before and after the voting day of the 2008 presidential election using the following
regression:

CARi;t ¼ β1 þ β2 KMTi;t−1 þ β 3 DPPi;t−1 þ β4 POLICYi;t−1 þ β 5 POLICYi;t−1  KMTi;t−1 þ β6 POLICYi;t−1  DPPi;t−1 þ λZi;t−1 þ ε i;t ; ð1Þ

where i denotes the ith firm, and t is the sample period, which is from 40 days before the election to 5 days after the election. CAR is the
cumulated abnormal returns, where AR is estimated from the market model.14 The estimation period covers 250 trading days before
the testing period. Any stock that traded for less than 200 days during the estimation period is excluded and thus, no initial public of-
fering effect can influence the returns. The AR sample period is from January 17, 2008 to March 31, 2008. The presidential election was
held on a Saturday (March 22, 2008), which was not a trading day in Taiwan. Thus, we define the first trading day (March 24, 2008)
after the election as the event day (t = 0). The POLICY variable is the INV/Asset in China. The dummy variables KMT and DPP are equal
to 1 if the firm is connected with KMT or DPP and 0 if they are non-PC firms. Z is the vector of the control variables and ε is the error
term. The industry-fixed effect in all equations is controlled using industry dummies. The t-values reported in all regressions are ad-
justed by heteroskedasticity and clustering at the industry level (Petersen, 2009; White, 1980).
Z is the vector of the control variables for the cross-sectional stock returns, which consist of BETA, LNMKTCAP, BM, MORET, ROA,
CASH, DIV, SALESGR, and LEV, following the studies of Faccio (2006), Goldman et al. (2009), Boubakri, El Ghoul, and Saffar (2013), and
Belo et al. (2013). Specifically, the control variables include market beta, market capitalization, book-to-market ratio, and momentum
returns, which control for the determinants of cross-sectional stock returns (Fama & French, 1992; Jegadeesh & Titman, 1993). The
term “BETA” for each stock is obtained by estimating the market model, as well as CAR.
LNMKTCAP is the logarithm of the last year-end market value of the outstanding shares of each firm and is expected to be nega-
tively associated with CAR because of small-size effects. BM refers to the book value of assets divided by the book value of debt plus the
market value of equity and is expected to be positively related with CAR for the effects of value stocks. MORET refers to the raw returns

14
Dasilas and Leventis (2011), Kollias, Papadamou, and Stagiannis (2011), Bae, Change, and Kim (2013), and Yao, Ma, and He (2014) use the market model to detect
abnormal returns.
158 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166

for each stock minus the market returns one year before the testing period; its expected correlation with CAR is mixed in different
market states (Cooper, Gutierrez, & Hameed, 2004). The following firm characteristics at the end of 2007 are also controlled for:
ROA, which refers to the earnings before interest and tax, divided by total assets; CASH, which refers to cash and cash equivalence di-
vided by total assets; DIV, which refers to the cash dividend payout divided by net income; SALESGR, which refers to the annual
growth rate of sales for each firm; and LEV, which refers to the leverage ratio measured as the total debt divided by the total assets.
If the coefficients of β2 and β3 are significantly positive and negative, respectively, it implies the CAR of KMT firms increase during
the election (the PC effect), whereas the CAR of DPP firms decrease. If the coefficient of β4 is significantly positive, then the GP effect
exists, that is, the CAR of firms with the GP effect will increase during the election. Meanwhile, if the coefficient of β5 is significantly
positive suggesting that the joint effects of GP and PC are larger than those of the PC effect alone.

3.2. Investor trading regression

The PC and GP effects also suggest that investors will increase their shares for KMT and policy firms. Aside from examining the ag-
gregate shares trading, we also investigate the influence of the five types of investors in this study. Various types of investors have
different levels of investment sophistication and information processing; thus, their trading responses to the information during
the presidential elections are likely to differ (e.g., Barber & Odean, 2008; Barber, Lee, Liu, & Odean, 2009; Chen, Johnson, Lin, & Liu,
2009; Chiang et al., 2010; Grinblatt & Keloharju, 2000; Lin & Lin, 2014; Sias & Whidbee, 2010; Vo, 2008). Thus, we investigate the
PC and GP effects on trading using the following regression model:

ΔSharek;i;t ¼ β1 þ β2 KMTi;t−1 þ β3 DPPi;t−1 þ β 4 POLICYi;t−1 þ β 5 POLICYi;t−1  KMTi;t−1 þ β6 POLICYi;t−1  DPPi;t−1

ð2Þ
þλZi;t−1 þ ε i;t k ¼ FIs; INVTRUSTs; DEALERs; BOARDs; and INDINVs;

where subscripts i and k denote that stock i is held by the kth type of investor, which may be FIs, INVTRUSTs, DEALERs, BOARDs, and
INDINVs; ΔShare represents the shareholding changes during the election periods; and Z is the vector of control variables, which con-
sist of BETA, LNMKTCAP, BM, MORET, ROA, CASH, DIV, SALESGR, and LEV.
If the coefficients of β2 and β3 are significantly positive and negative, respectively, it confirms that investors increase their shares
for KMT firms and decrease their shares for DPP firms. If the coefficient of β4 is significantly positive, it suggests investors will increase
their shares for policy firms.

3.3. Regression of stock returns: evidence from investor trading

This section examines whether investors who adopt the PC and GP strategies can increase their CAR. We hypothesize that the pur-
chasing shares of KMT and policy firms will increase the CAR using the following equation:

CARi;t ¼ β1 þ β2 KMTi;t−1 þ β 3 DPPi;t−1 þ β4 POLICYi;t−1 þ β 7 KMTi;t−1  ΔSharek;i;t þ β8 POLICYi;t−1  ΔSharek;i;t ; þβ 9 KMTi;t−1

ð3Þ
POLICYi;t−1  ΔSharek;i;t þ λZi;t−1 þ εi;t k ¼ FIs; INVTRUSTs; DEALERs; BOARDs; and INDINVs;

where subscripts i and k denote stock i held by the kth type of investor, which may be FIs, INVTRUSTs, DEALERs, BOARDs, and INDINVs;
CAR is the cumulated abnormal returns during the election period; ΔShare represents the shareholding changes during the election
period; and Z is the vector of control variables, which consist of BETA, LNMKTCAP, BM, MORET, ROA, CASH, DIV, SALESGR, and LEV.
If the coefficients of β7 is significantly positive, respectively, it indicates that the increases in the shares of investors in KMT firms
are positively correlated with the abnormal stock returns. If the coefficient of β8 is significantly positive, it suggests that the increases
in the shares of policy firms by the investors are positively correlated with the abnormal stock returns. If the coefficients of β9 is sig-
nificantly positive, it indicates that the joint effects of PC and GP are strong and more significantly positive in the measure of CAR. Thus,
the increases in the shares of investors in KMT firms with policy benefits are more positively correlated with the abnormal stock
returns.

4. Empirical results and discussion

4.1. Basic results

Table 3 presents the basic statistics (mean, standard deviations, first quartile, median, and third quartile) of the study variables. The
mean of CAR (−40, 5) calculated across 46 trading days (i.e., 40 days before the election date and 5 days after the election date) is
3.89%, suggesting that the stock market has a slight upward trend during the entire period. We also divide our sample period into
three sub-periods, namely, pre-election period (t = − 40 to − 1), event day (t = 0), and post-election period (t = 1 to 5). The
mean values of CAR (−40, −1), CAR (0), and CAR (1, 5) are 0.03%, −0.14%, and 4.00%, respectively. On average, KMT firms comprise
22.45% and DPP firms make up 14.17% of the sample. The mean values of POLICY, BETA, LNMKTCAP, BM, MORET, ROA, CASH, DIV,
SALESGR, and LEV are 0.13, 0.89, 8.82, 0.85, 10.10, 0.12, 0.13, 0.48, 0.15, and 0.43, respectively.
The PC effect suggests that firms connected with the winning party (i.e., KMT) should have higher abnormal stock returns after the
election, when everything is certain. However, given that the polls suggested a landslide victory for KMT, the stock returns of KMT
 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166 159

Table 3
Basic statistics.

Variable Mean Stdev 25th Median 75th

CAR (−40,5) 3.8943 15.1536 −5.0683 3.2327 12.0778

CAR (−40, −1) 0.0334 12.9094 −7.8661 −0.2572 7.4126
CAR (0) −0.1406 2.4125 −1.7204 −0.1973 1.6450
CAR (1, 5) 4.0015 5.2929 1.2605 3.4350 6.4482
KMT 0.2245 0.4176 0.0000 0.0000 0.0000
DPP 0.1417 0.3490 0.0000 0.0000 0.0000
POLICY 0.1258 0.3316 0.0000 0.0000 0.0000
BETA 0.8868 0.2315 0.7413 0.9093 1.0487
LNMKTCAP 8.8172 1.4219 7.8500 8.6400 9.5825
BM 0.8477 0.2946 0.6575 0.8700 1.0600
MORET 10.0982 41.2395 −17.4131 2.8641 27.6912
ROA 0.1191 0.0881 0.0650 0.1070 0.1620
CASH 0.1288 0.1068 0.0510 0.1060 0.1715
DIV 0.4789 0.7341 0.0000 0.3915 0.6368
SALESGR 0.1462 0.3466 0.0180 0.1190 0.2460
LEV 0.4277 0.1707 0.2980 0.4350 0.5513

This table presents the summary statistics of variables used in this study. In this table, CAR (−40, 5) refers to the cumulative abnormal returns (%) from 40 days before
the election to 5 days after the election. CAR (−40, −1), CAR (0), and CAR (1, 5) refer to the cumulative abnormal returns in the pre-election period, election day, and
post-election period, respectively. Other variables are defined in Table 1. There are a total of 628 firms in 2008.

firms could also increase even prior to the election. Conversely, the stock returns of the losing party could drop (Goldman et al., 2009).
We compare the stock returns and the characteristic variables among KMT, DPP, and non-politically connected firms.
Table 4 shows the differences between KMT and DPP firms, KMT and non-PC firms, and DPP and non-PC firms in columns (4), (6),
and (8), respectively. The results in columns (4) and (6) show that KMT firms have larger CAR (−40, −1) than DPP and non-PC firms,
which confirms that KMT firms have higher stock returns in the pre-election period. KMT firms also have larger CAR (0) than non-PC
firms, indicating that KMT firms have higher stock returns on the event day. Thus, the results of the descriptive statistics confirm the
existence of the PC effect.
Fig. 1 plots the CAR of KMT, DPP, and non-PC firms, as well as the market index during the election. The CAR of KMT firms illustrate
a significantly upward trend after the 16th day before the election, reflecting the potentially politically connected profits in the future.
Similarly, the CAR of KMT firms are obviously higher than those of market index after the 16th day before the election. In contrast, the
CAR of DPP firms are lower than those of market index after the 12th day before the election, revealing a potential loss for DPP firms in
the future. The CAR of non-PC firms display a similar trend with those of market index. These outcomes confirm that the PC effect of
firms can bring them higher stock returns from 16 days before the election to 5 days after the election.
The significant difference in firm characteristics among the three types of firms might contribute to the problem of the missing
third variable; hence, multiple regression analysis is performed to examine the PC and GP effects.

4.2. GP and PC effects: stock returns

Table 5 presents the estimated results of Eq. (1) using data before and after the election, where CAR is the dependent variable. We
have four specifications that involve different sets of independent and control variables. First, the coefficients of KMT and DPP in

Table 4
Descriptive statistics: political connection effect.

Variables KMT [A] DPP [B] Non-PC [C] Diff. [A] − [B] t-Value Diff. [A] − [C] t-Value Diff. [B] − [C] t-Value

(1) (2) (3) (4) (5) (6) (7) (8) (9)

CAR (−40,5) 7.5751 1.9958 2.9708 5.5793** (2.38) 4.6043*** (3.15) −0.9750 (−0.52)
CAR (−40, −1) 2.8759 −2.4422 −0.5076 5.3182*** (2.72) 3.3835*** (2.72) −1.9346 (−1.20)
CAR (0) 0.4999 0.3145 −0.4353 0.1854 (0.56) 0.9353*** (4.03) 0.7499** (2.41)
CAR (1, 5) 4.1993 4.1235 3.9137 0.0758 (0.09) 0.2856 (0.54) 0.2098 (0.33)
POLICY 0.0633 0.0450 0.0655 0.0183** (2.13) −0.0022 (−0.31) −0.0205** (−2.21)
BETA 0.9415 0.8456 0.8753 0.0960*** (2.96) 0.0662*** (2.97) −0.0297 (−0.99)
LNMKTCAP 9.5072 8.6137 8.6180 0.8936*** (4.41) 0.8892*** (6.47) −0.0043 (−0.03)
BM 0.8752 0.8783 0.8332 −0.0031 (−0.08) 0.0421 (1.45) 0.0451 (1.17)
MORET 12.1721 12.4024 9.0020 −0.2303 (−0.04) 3.1700 (0.78) 3.4004 (0.65)
ROA 0.1063 0.1052 0.1258 0.0011 (0.10) −0.0195** (−2.24) −0.0207* (−1.78)
CASH 0.1143 0.0937 0.1397 0.0207 (1.42) −0.0253** (−2.42) −0.0460*** (−3.38)
DIV 0.6279 0.4138 0.4395 0.2141* (1.78) 0.1884** (2.53) −0.0257 (−0.31)
SALESGR 0.1315 0.1777 0.1458 −0.0463 (−1.03) −0.0143 (−0.41) 0.0320 (0.71)
LEV 0.4612 0.4481 0.4129 0.0131 (0.56) 0.0482*** (2.89) 0.0352 (1.59)

This table presents the differences in descriptive statistics between KMT, DPP, and non-politically connected firms. The sample includes 628 firms at 2008. For the def-
initions of all variables, see Table 1. Superscripts *, **, and *** denote significance at 10%, 5%, and 1%, respectively.
160 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166

8.00

6.00

4.00

CAR(%) 2.00 MARKET

KMT
DPP
0.00 NPC

-2.00

-4.00

-6.00

Fig. 1. Cumulative abnormal returns in political connection effect.

Models (1) and (3) are both significantly positive and negative, respectively, suggesting that the firms connected to the winning and
losing parties have positive and negative CAR. The results confirm that KMT firms have positive CAR, whereas that DPP firms have neg-
ative CAR, which are fully consistent with the expected PC effect. Second, the coefficients of POLICY in Models (2) and (3) are insig-
nificant. The results do not support the GP effect, that is, firms consistent with the policies of the winning party will not increase their
stock returns during the election.

Table 5
Regression analysis of stock returns.

Variable (1) (2) (3) (4)

Constant 18.7197*** 16.6967*** 20.3308*** 20.9256***

(3.06) (2.73) (3.22) (3.38)
KMT 3.3028*** 3.4328*** −0.0093
(2.57) (2.68) (−0.01)
DPP −2.6622** −2.7419** −2.2324
(−2.14) (−2.21) (−1.43)
POLICY −5.9149 −8.6293 −18.4065**
(−0.64) (−0.98) (−2.29)
POLICY × KMT 54.7457***
(2.73)
POLICY × DPP −10.2602
(−0.54)
BETA −1.0497** −0.7874* −1.1609** −1.1582**
(−2.15) (−1.66) (−2.35) (−2.38)
LNMKTCAP 1.7902 2.2923 1.7417 2.4195
(0.66) (0.85) (0.65) (0.94)
BM 0.6911 1.5152 0.7029 0.5410
(0.26) (0.57) (0.26) (0.20)
MORET −0.1717*** −0.1718*** −0.1724*** −0.1700***
(−10.41) (−10.42) (−10.52) (−10.51)
ROA 4.7983 2.7077 4.2299 3.9088
(0.51) (0.29) (0.45) (0.42)
CASH −4.9819 −3.9502 −4.9730 −4.6336
(−0.85) (−0.67) (−0.85) (−0.79)
DIV −0.8427 −0.6343 −0.8379 −0.9468
(−1.38) (−1.09) (−1.38) (−1.41)
SALESGR 3.3293* 3.0976 3.1987 3.0574
(1.65) (1.54) (1.58) (1.59)
LEV −9.9388*** −9.7078*** −10.1906*** −9.7212***
(−2.93) (−2.93) (−3.01) (−2.94)
Control for
Industry fixed effect Yes Yes Yes Yes
Adj-R2 0.4330 0.4236 0.4334 0.4420
N 628 628 628 628

This table presents the OLS regression results of CAR before and after the election. The dependent variable CAR (−40, 5) refers to the cumulative abnormal returns from
40 days before the election to 5 days after the election. Dummy variables KMT and DPP are equal to 1 if the firm is connected with KMT and DPP, respectively, and 0 if
otherwise. POLICY is the investment to total assets (INV/Asset) ratio in China as the proxy of benefiting from the government policy. For the definitions of all variables,
see Table 1. Through all equations, we also control for industry fixed effect by using industry dummies. The t-values reported in all regressions are adjusted by the
heteroskedasticity and clustering at the industry level (Petersen, 2009; White, 1980). Superscripts *, **, and *** denote significance at 10%, 5%, and 1%, respectively.
Table 6
Regression analysis of stock returns: evidence from three sub-periods.

Panel A. Pre-election: CAR (−40, −1) Panel B. Election day: CAR (0) Panel C. Post-election: CAR (1, 5)

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Constant 10.2590* 7.8218 10.8496* 11.1720** 2.2736** 1.9310* 2.0096* 2.0405** 6.1871*** 6.9439*** 7.4716*** 7.7131***
(1.84) (1.42) (1.91) (1.98) (2.29) (1.93) (1.96) (2.00) (3.16) (3.70) (3.92) (4.03)
KMT 2.4555** 2.5032** 0.3635 0.1117 0.0904 0.0252 0.7356 0.8392 −0.3979

C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166
(2.18) (2.26) (0.25) (0.58) (0.47) (0.11) (1.22) (1.41) (−0.57)
DPP −3.3635*** −3.3927*** −2.5978* −0.0221 −0.0090 −0.1982 0.7234 0.6598 0.5636
(−3.07) (−3.11) (−1.93) (−0.10) (−0.04) (−0.79) (1.50) (1.37) (0.93)
POLICY −0.7656 −3.1634 −8.4380 1.4664 1.4139 0.8946 −6.6158 −6.8797** −10.8631***
(−0.09) (−0.39) (−1.15) (1.32) (1.27) (0.73) (−1.91) (−2.05) (−3.47)
POLICY × KMT 34.3163* 0.9191 19.5102*
(1.72) (0.40) (1.93)
POLICY × DPP −15.8264 3.7345 1.8317
(−0.88) (1.35) (0.22)
BETA −0.2359 −0.0015 −0.2767 −0.2729 0.1175* 0.1454** 0.1357** 0.1349* −0.9312*** −0.9314*** −1.0199*** −1.0202***
(−0.54) (−0.01) (−0.62) (−0.62) (1.73) (2.21) (1.97) (1.96) (−4.82) (−5.34) (−5.50) (−5.53)
LNMKTCAP 0.2804 0.7149 0.2627 0.7125 −2.5917*** −2.5715*** −2.5837*** −2.5827*** 4.1015*** 4.1489*** 4.0628*** 4.2897***
(0.11) (0.29) (0.11) (0.30) (−5.68) (−5.64) (−5.66) (−5.66) (4.56) (4.66) (4.55) (4.89)
BM −0.8917 −0.1722 −0.8874 −1.0679 0.3956 0.4095 0.3937 0.4238 1.1872 1.2780 1.1966 1.1851
(−0.35) (−0.07) (−0.35) (−0.43) (0.89) (0.93) (0.89) (0.96) (1.34) (1.43) (1.35) (1.35)
MORET −0.1559*** −0.1561*** −0.1562*** −0.1546*** −0.0048** −0.0047** −0.0047** −0.0047** −0.0109* −0.0109* −0.0114** −0.0106*
(−10.43) (−10.38) (−10.49) (−10.51) (−2.21) (−2.16) (−2.15) (−2.14) (−1.86) (−1.88) (−1.97) (−1.84)
ROA 0.4353 −0.6200 0.2270 −0.1072 −1.9299 −1.8888 −1.8368 −1.7870 6.2929** 5.2165* 5.8397** 5.8029**
(0.05) (−0.07) (0.03) (−0.01) (−1.35) (−1.34) (−1.28) (−1.25) (2.09) (1.78) (1.97) (1.97)
CASH −6.0626 −5.1190 −6.0593 −5.8871 −1.7064** −1.6898** −1.7079** −1.6853** 2.7871 2.8586 2.7942 2.9388
(−1.13) (−0.95) (−1.12) (−1.09) (−2.18) (−2.16) (−2.17) (−2.15) (1.26) (1.30) (1.29) (1.33)
DIV −1.1611* −1.0100 −1.1594* −1.2430* 0.1078 0.1123 0.1070 0.1116 0.2106 0.2634 0.2145 0.1846
(−1.79) (−1.58) (−1.80) (−1.82) (1.24) (1.30) (1.22) (1.26) (0.88) (1.09) (0.87) (0.71)
SALESGR 3.8325* 3.7078* 3.7846* 3.6774* 0.3624** 0.3813** 0.3838** 0.3892** −0.8656 −0.9915* −0.9697* −1.0091*
(1.82) (1.76) (1.79) (1.78) (2.12) (2.21) (2.22) (2.28) (−1.55) (−1.74) (−1.73) (−1.75)
LEV −6.8064** −6.5452** −6.8988** −6.5382** 0.0208 0.0747 0.0621 0.0424 −3.1531** −3.2373** −3.3539** −3.2254**
(−2.27) (−2.21) (−2.31) (−2.21) (0.04) (0.15) (0.13) (0.09) (−2.29) (−2.38) (−2.45) (−2.39)
Control for
Industry fixed effect Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Adj-R2 0.3761 0.3633 0.3753 0.3791 0.3845 0.3867 0.3848 0.3840 0.2278 0.2307 0.2332 0.2420
N 628 628 628 628 628 628 628 628 628 628 628 628

This table presents the OLS regression results of CAR in three sub-periods. In Panels A, B, and C, the dependent variables are CAR (−40, −1), CAR (0), and CAR (1, 5) which are the cumulative abnormal returns in the three sub-
periods. Dummy variables KMT and DPP are equal to 1 if the firm is connected with KMT and DPP, respectively, and 0 if otherwise. POLICY is investment to total assets (INV/Asset) ratio in China as the proxy of benefiting from the
government policy. For the definitions of all variables, see Table 1. Through all equations, we also control for industry fixed effect by using industry dummies. The t-values reported in all regressions are adjusted by the
heteroskedasticity and clustering at the industry level (Petersen, 2009; White, 1980). Superscripts *, **, and *** denote significance at 10%, 5%, and 1%, respectively.

161
162 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166

Third, we investigate whether the combined effects of GP and PC are larger than the PC effect alone. The coefficient of
POLICY × KMT in Model (4) is significantly positive, suggesting that the joint effects of PC and GP are larger than the PC
effect alone. Thus, the evidence demonstrates that the increases in the shares of investors in KMT firms with larger invest-
ments in China (policy benefits) are more positively correlated with the abnormal stock returns.
Moreover, BETA, MORET, and LEV negatively influence share prices. These factors all indicate their expected influence on stock
returns based on an analysis of the effects of market beta, momentum returns, leverage ratio, and other factors.
We divide our sample period into three sub-periods (pre-election period, event day, and post-election period) to investigate
whether the effects of PC and GP vary over the election period. Table 6 presents the estimated results of Eq. (1) using data from
three sub-periods, where CAR is the dependent variable. Panel A shows that the coefficients of KMT and DPP are significantly positive
and negative, respectively. Hence, KMT firms have positive CAR whereas DPP firms have negative CAR during the pre-election period.
However, Panels B and C show that the coefficients of KMT and DPP become insignificantly positive and negative, respectively, imply-
ing that the PC effect weakens after the election.15 The three panels also show that the coefficients of POLICY are all insignificant, in-
dicating that the expectation of the existence of the GP effect in the three sub-periods is rejected.
Finally, we investigate whether the combined effects of PC and GP are larger than the PC effect alone in the three sub-periods. In
Models (4) and (12), the coefficients of POLICY × KMT are both significantly positive, indicating that the joint effects of GP and PC are
larger than the PC effect alone. Therefore, the evidence confirms the expectation that the joint effects of PC and GP have the largest
effects, even when our sample is divided into three sub-periods.

4.3. GP and PC effects: shares trading of investors

Table 7 presents the estimated results of Eq. (2), in which the shareholding change (ΔShare) of each investor is used as dependent
variable for the entire period. For brevity, we only present the estimated results for the entire period. The estimated results of the three
sub-periods are available upon request.
First, we discuss whether the PC effect influences the share trading of the five types of investors. In the regression for INVTRUSTs,
the coefficient of KMT is significantly positive after controlling for other firm characteristics. Thus, this result confirms the PC effect,
i.e., INVTRUSTs increase more shareholdings in KMT firms connected with the winning party. The results also confirm the arguments
of existing literature that domestic institution investors can understand the typical nuances in political contexts, and possess first-
hand experience in rationalizing and using soft information (Bae, Stulz, & Tan, 2008; Choe, Kho, & Stulz, 2005), such as potential po-
litical profit in the future. However, none of the coefficients of KMT is significant when shareholding changes of FIs, DEALERs, BOARDs,
and INDINVs are used as dependent variables, thereby suggesting that they do not increase their shareholdings for the PC effect
around the time of the elections. Thus, except for security investment trust companies, no evidence to confirm the PC effect is found.
Second, we discuss whether the GP effect influences the share trading of the five types of investors. Except for DEALERs, the coef-
ficients of POLICY are insignificant when the shareholding changes of FIs, INVTRUSTs, BOARDs, and INDINVs are used as dependent
variables. Thus, this result suggests that those investors do not increase their shareholdings for the GP effect around the time of the
elections.
We caution that a few common factors adequately explain investor trading, especially in the short term. Despite our best attempts
to account for as many variables as we can, we find that most variables are insignificant. For example, the coefficients of LNMKTCAP
and BM in the INVTRUSTs regression are significantly negative and positive. We assume that when processing trading in the short
term, INVTRUSTs may focus on size and book-to-market effects. Thus, during a short period of a specific event, INVTRUSTs may pur-
chase more shares in firms with smaller sizes and higher BMs.

4.4. Do strategies of PC and GP bring profit?

An examination of whether investors gain profit by adopting strategies of PC and GP is of significant interest among researchers. If
profits are positive, then this scenario would indicate that investment strategies of investors are “effective” investment decisions.
Table 8 presents the estimated results of Eq. (3) on shareholding changes and CAR. In Eq. (3), we use CAR as the dependent variable
and shareholding changes (ΔShare) of each investor as the independent variable in the entire period.16 Among the five specifications,
the coefficients of KMT are significantly positive, suggesting that firms connected to the winning party have positive CAR. The coeffi-
cients of POLICY are significantly and insignificantly negative, which rejects the GP effect again. Model (2) shows that the coefficients
of KMT × POLICY × ΔSHARE are significantly positive, suggesting that INVTRUSTs will acquire higher CAR if they increase the share-
holding of KMT firms with larger investments in China. Therefore, the joint effects of PC and GP having the largest effects are con-
firmed again.
Combined with the results in Tables 7 and 8, we find that only INVTRUSTs significantly purchase shares of KMT firms, and that they
have significantly positive CAR based on their trading activities.

15
A possible explanation is that investors expect that the KMT will win the election. Hence, the PC effect was realized before the election.
16
For brevity, we only present the estimated results of the entire period. The estimated results of the three sub-periods are available upon request.
 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166 163

Table 7
Regression results of investor trading.

FIs INVTRUSTs DEALERs BOARDs INDINVs

Variable (1) (2) (3) (4) (5)

Constant 1.6584** 0.2126 −0.3609*** −0.1138 −0.5990

(2.20) (0.46) (−3.84) (−0.25) (−0.52)
KMT 0.1100 0.3005** −0.0016 0.1090 −0.4563
(0.44) (1.97) (−0.04) (1.62) (−1.33)
DPP 0.0454 −0.0893 0.0385 0.0581 −0.0636
(0.17) (−0.54) (1.21) (0.95) (−0.20)
POLICY 0.5343 −0.4017 0.2230* −0.4772 0.2568
(0.69) (−0.58) (1.90) (−0.53) (0.18)
KMT × POLICY 1.0402 −0.2693 −0.0264 −0.5980 −0.2716
(0.44) (−0.18) (−0.08) (−0.79) (−0.08)
DPP × POLICY −4.5277 1.8576 −0.2571 1.1972 1.2056
(−1.47) (1.14) (−0.57) (1.12) (0.31)
BETA −0.1479** −0.0010 0.0319*** 0.0067 0.0467
(−2.03) (−0.02) (3.73) (0.14) (0.46)
LNMKTCAP 0.2625 −0.3621* 0.1123*** 0.0723 −0.0159
(0.80) (−1.88) (3.24) (0.32) (−0.03)
BM 0.0665 0.4567** 0.0546 0.0878 −0.7049
(0.19) (2.20) (1.33) (0.37) (−1.29)
MORET 0.0009 −0.0025* 0.0005 −0.0042 0.0092**
(0.47) (−1.67) (1.45) (−1.62) (2.05)
ROA −2.5576** 0.8110 −0.0120 0.8634 −0.1284
(−2.38) (1.00) (−0.08) (0.94) (−0.06)
CASH −0.3972 0.1772 0.1146 −0.2751 −0.5873
(−0.58) (0.34) (1.05) (−0.49) (−0.47)
DIV 0.0689 −0.0490 0.0086 −0.0404 0.0459
(0.97) (−1.54) (1.00) (−0.99) (0.49)
SALESGR 0.0754 −0.1122 0.0301 0.0395 0.3116
(0.48) (−0.98) (1.15) (0.59) (0.86)
LEV −0.0148 −0.0942 −0.0199 −0.3594 0.0306
(−0.03) (−0.36) (−0.39) (−1.39) (0.05)
Control for
Industry fixed effect Yes Yes Yes Yes Yes
Adj-R2 0.0362 0.0470 0.0809 0.0079 0.0745
N 628 628 628 628 628

This table presents the OLS regression results regarding the shareholding changes of investors using the sample before and after the election. The dependent variable is
the shareholding changes of five types of investors from 40 days before the election to 5 days after the election. In this table, the five types of investors include foreign
investors (FIs), security investment trust companies (INVTRUSTs), security dealers (DEALERs), board of directors and supervisors (BOARD), and individual investors
(INDINVs). Dummy variables KMT and DPP are equal to 1 if the firm is connected with KMT and DPP, respectively, and 0 if otherwise. POLICY is investment to total assets
(INV/Asset) ratio in China as the proxy of benefiting from the government policy. For the definitions of all variables, see Table 1. Through all equations, we also control
for industry fixed effect by using industry dummies. The t-values reported in all regressions are adjusted by the heteroskedasticity and clustering at the industry level
(Petersen, 2009; White, 1980). Superscripts *, **, and *** denote significance at 10%, 5%, and 1%, respectively.

4.5. Robustness check

In Table 5, the evidence demonstrates that the PC effect is more important than the GP effect during the 2008 presidential election
in Taiwan. The results indicate that KMT firms have positive CAR whereas DPP firms have negative CAR. This finding is fully consistent
with the PC effect. However, this result may vary if we adopt the different definitions of PC. Hence, we use three additional settings to
determine the association of the PC and GP effects on CAR and establish the robustness of our findings.

4.5.1. Different categories of PC

In this study, we define firms to be politically connected when they satisfy one of the following conditions: (1) firms have political
party tendency and (2) firms make political donations. We then classify our PC firms into two types to examine whether one of them
has more influence on the CAR during the election. Eq. (1) is re-estimated by using measures of political party tendency and political
donations. The results of political party tendency remain supportive of the joint effects of PC and GP, that is, the stock returns of firms
will increase during the elections if they are consistent with KMT and the policies of KMT. For example, in Model (1) of Table 9, the
coefficients of POLICY × KMT is significantly positive. However, the joint effects of PC and GP weaken when we use political donations
to measure PC effect in Model (2).

4.5.2. Excluding firms connected to both parties

The relationship between PCs and stock returns may be driven by the sample of firms connected to both parties. Therefore, to re-
examine the issue, we exclude the firms connected to both parties from our sample; these firms account for 15 observations. After
164 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166

Table 8
Stock returns and investor trading.

k = FIs k = INVTRUSTs k = DEALERs k = BOARDs k = INDINVs

Variable (1) (2) (3) (4) (5)

Constant 28.4622*** 27.7447*** 30.5589*** 29.9884*** 29.4178***

(4.87) (4.83) (5.22) (5.22) (5.40)
KMT 2.9189** 2.3549* 3.1548** 3.3461*** 1.9641*
(2.33) (1.90) (2.37) (2.60) (1.69)
DPP −1.8518 −2.2852* −2.0963 −2.2917* −1.8555
(−1.48) (−1.82) (−1.60) (−1.78) (−1.50)
POLICY −15.3883 −14.7514* −13.586 −8.3837 −15.9463**
(−1.72) (−1.96) (−1.55) (−0.91) (−2.21)
KMT × ΔSHAREk 0.7171 1.1496 2.1145 8.3764 −1.5172***
(0.85) (1.58) (0.40) (0.66) (−3.08)
POLICY × ΔSHAREk 10.4237** 20.1873*** 51.4942 13.4862*** −13.6389***
(2.36) (3.28) (0.86) (3.27) (−6.70)
KMT × POLICY × ΔSHAREk 8.3769 37.2569*** 52.9203 −71.778 −6.4817
(1.17) (3.30) (0.59) (−0.62) (−0.96)
Control for
Firm characteristic Yes Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes Yes
Adj-R2 0.4200 0.4496 0.4117 0.4133 0.4802
N 628 628 628 628 628

This table presents the OLS regression results of CAR before and after the election. The dependent variable CAR (−40, 5) refers to the cumulative abnormal returns from
40 days before the election to 5 days after the election. Dummy variables KMT and DPP are equal to 1 if the firm is connected with KMT and DPP, respectively, and 0 if
otherwise. POLICY is investment to total assets (INV/Asset) ratio in China as the proxy of benefiting from the government policy. ΔSHAREk is the share holding change of
five types of investors from 40 days before to 5 days after the election. In this table, the five types of investors include foreign investors (FIs), security investment trust
companies (INVTRUSTs), security dealers (DEALERs), board of directors and supervisors (BOARD), and individual investors (INDINVs). For the definitions of all vari-
ables, see Table 1. Through all equations, we also control for industry fixed effect by using industry dummies. The t-values reported in all regressions are adjusted by
the heteroskedasticity and clustering at the industry level (Petersen, 2009; White, 1980). Superscripts *, **, and *** denote significance at 10%, 5%, and 1%, respectively.

Table 9
Robustness check.

Political party tendency Political donation Exclude both connection firms Heckman's two-stage regression

Variable (1) (2) (3) (4)

Constant 19.8884*** 18.1144*** 21.1165*** 9.4478

(3.27) (2.89) (3.37) (0.84)
KMT −0.3042 1.7293 −1.2390 −13.0455
(−0.15) (0.85) (−0.67) (−1.27)
DPP −2.1933 −3.1652* −4.0028** −2.3875
(−1.20) (−1.80) (−2.26) (−1.53)
POLICY −16.7283** −10.9403 −20.2390** −18.0548**
(−2.16) (−1.21) (−2.51) (−2.27)
KMT × POLICY 80.8451*** −24.9507 64.9770*** 53.9953***
(3.24) (−1.49) (2.95) (2.65)
DPP × POLICY −27.4552 25.0775 10.9336 −10.9369
(−1.00) (1.34) (0.50) (−0.58)
IMR 7.5592
(1.24)
Control for
Firm characteristic Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes
Adj-R2 0.4506 0.4240 0.4401 0.4426
N 628 628 613 628

This table presents the OLS regression results of CAR before and after the election. The dependent variable CAR refers to the cumulated abnormal returns from
40 days before the election to 5 days after the election. For the definitions of all variables, see Table 1. In Models (1) and (2), we separately employ two types of
PC, namely, political party tendency and political donation, respectively, to re-examine the issue. In Model (3), we exclude firms connected to both parties from
our sample to test the robustness of our results. In Model (4), we use the method of Heckman's two-stage regression to address the endogeneity problem.
Inverse Mills ratio (IMR) is obtained from Heckman's first step of estimating a determinant equation through the probit model, which is used to avoid the
self-select problem created when firms decide to hire a politically connected CEO. In the probit model, we include firm characteristic variables defined in
Table 1 as independent variables to assess the possible motives of top managers in building PCs. Dummy variables KMT and DPP are equal to 1 if the firm is
connected with KMT and DPP, respectively, and 0 if otherwise. POLICY is investment to total assets (INV/Asset) ratio in China as the proxy of benefiting
from the government policy. For the definitions of all variables, see Table 1. Through all equations, we also control for industry fixed effect by using industry
dummies. The t-values reported in all regressions are adjusted by the heteroskedasticity and clustering at the industry level (Petersen, 2009; White, 1980).
Superscripts *, **, and *** denote significance at 10%, 5%, and 1%, respectively.
 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166 165

excluding these firms, the sample size is reduced to 613. In Model (3) of Table 9, the new results are very similar to those of Model
(4) in Table 5. The result also confirms the joint effects of PC and GP.

4.5.3. Endogeneity of PCs

The decision of top managers to establish PC may result from motives to overcome certain business limitations (Claessens et al.,
2008; Cooper et al., 2010). Therefore, the motivation to build PCs may be determined endogenously, leading to inconsistent estimates
and spurious interpretation. Hence, we use the conventional method of Heckman's two-stage regression to address this problem. In
the first stage, we perform a probit regression with PC as dependent variable. Firm characteristics at the firm level (BETA, LNMKTCAP,
BM, MORET, ROA, CASH, DIV, SALESGR, and LEV) are included as independent variables to assess the possible motives of top managers
in building PCs. The resulting inverse Mills ratio (IMR) is inserted in the second-stage regressions to correct any potential bias. Model
(4) in Table 9 presents the estimated results of the second-stage regressions based on Eq. (1). The new results remain similar to those
reported in Table 5, and the new results still support the joint effects of PC and GP.

5. Conclusion

This study posits that presidential elections create two effects on share trading and stock returns, namely, PC and GP effects. The
former suggests that investors are interested in firms connected with the winning party, while the latter stresses that investors are
interested in the policies of the winning party. Previous studies have usually focused on either the PC or the GP effect separately,
whereas our study examines both effects. This approach motivates us to use Taiwanese firm data because the policies of the two po-
litical parties in Taiwan differ dramatically in their views on investments in China. Hence, we can distinguish the GP effect successfully
by checking whether firms are within the beneficiary industries.
Different types of investors have different experiences in handling diverse information and knowledge; hence, their portfolio al-
locations before and after the presidential election may also vary. We consider five types of investors, namely, FIs, INVTRUSTs,
DEALERs, BOARDs, and INDINVs. The first three types of investors are institutional investors whereas the last two are individual
investors.
Our results indicate that KMT firms have positive CAR whereas DPP firms have negative CAR, which are fully consistent with the
expected PC effect. We also find that the GP effect alone is not supported. We further observe that the joint effects of PC and GP are
strong and more significantly positive in the measure of CAR. Our trading results also reveal that only INVTRUSTs significantly pur-
chase the shares of firms with KMT. Finally, based on trading activities, INVTRUSTs are found to have significantly positive CAR.
In summary, discussing the influence of the presidential elections without considering both the PC and GP effects may lead to the
loss of important information, leading to an inaccurate interpretation of stock returns and share trading. Although our results are spe-
cific to Taiwan and to a certain election, the results can still serve as a reference for other countries and elections.

References

Bae, K. H., Stulz, R. M., & Tan, H. (2008). Do local analysts know more? A cross-country study of the performance of local analysts and foreign analysts. Journal of
Financial Economics, 88, 581–606.
Bae, S., Change, K., & Kim, D. (2013). Determinants of target selection and acquirer returns: evidence from cross-border acquisitions. International Review of Economics
& Finance, 27, 552–565.
Barber, B. M., & Odean, T. (2008). All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors. Review of Financial
Studies, 21, 785–818.
Barber, B. M., Lee, Y. T., Liu, Y. J., & Odean, T. (2009). Just how much do individual investors lose by trading? Review of Financial Studies, 22, 609–632.
Belo, F., Gala, V. D., & Li, J. (2013). Government spending, political cycles, and the cross section of stock returns. Journal of Financial Economics, 107, 305–324.
Boubakri, N., El Ghoul, S., & Saffar, W. (2013). Cash holdings of politically connected firms. Journal of Multinational Financial Management, 23, 338–355.
Chen, L. W., Johnson, S. A., Lin, J. C., & Liu, Y. J. (2009). Information, sophistication, and foreign versus domestic investors' performance. Journal of Banking & Finance, 33,
1636–1651.
Chen, Y. S., Shen, C. H., & Lin, C. Y. (2014). The benefits of political connection: Evidence from individual bank loan contracts. Journal of Financial Services Research, 45,
287–305.
Chiang, Y. -M., Qian, Y., & Sherman, A. E. (2010). Endogenous entry and partial adjustment in IPO auctions: Are institutional investors better informed? Review of
Financial Studies, 23, 1200–1230.
Choe, H., Kho, B. C., & Stulz, R. M. (2005). Do domestic investors have an edge? The trading experience of foreign investors in Korea. Review of Financial Studies, 18,
795–829.
Claessens, S., Feijen, E., & Laeven, L. (2008). Political connections and preferential access to finance: The role of campaign contributions. Journal of Financial Economics,
88, 554–580.
Cooper, M. J., Gulen, H., & Ovtchinnikov, A. V. (2010). Corporate political contributions and stock returns. Journal of Finance, 65, 687–724.
Cooper, M. J., Gutierrez, R. C., & Hameed, A. (2004). Market states and momentum. Journal of Finance, 59, 1345–1365.
Cox, G. W., & McCubbins, M. D. (1986). Electoral politics as a redistributive game. Journal of Politics, 48, 370–389.
Dasilas, A., & Leventis, S. (2011). Stock market reaction to dividend announcements: Evidence from the Greek stock market. International Review of Economics &
Finance, 20, 302–311.
Duchin, R., & Sosyura, D. (2012). The politics of government investing. Journal of Financial Economics, 106, 24–48.
Faccio, M. (2006). Politically connected firms. American Economic Review, 96, 369–386.
Faccio, M., & Parsley, D. C. (2009). Sudden deaths: Taking stock of geographic ties. Journal of Financial and Quantitative Analysis, 44, 683–718.
Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. Journal of Finance, 47, 427–465.
Ferguson, T., & Voth, H. J. (2008). Betting on Hitler—The value of political connections in Nazi Germany. Quarterly Journal of Economics, 123, 101–137.
Fisman, R. (2001). Estimating the value of political connections. American Economic Review, 91, 1095–1102.
Goldman, E., Rocholl, J., & So, J. (2009). Do politically connected boards affect firm value? Review of Financial Studies, 22, 2331–2360.
Grinblatt, M., & Keloharju, M. (2000). The investment behavior and performance of various investor types: A study of Finland's unique data set. Journal of Financial
Economics, 55, 43–67.
Grossman, G., & Helpman, E. (1994). Protection for sale. American Economic Review, 84, 833–850.
166 C.-Y. Lin et al. / International Review of Economics and Finance 42 (2016) 153–166

Hong, H., & Kostovetsky, L. (2012). Red and blue investing: Values and finance. Journal of Financial Economics, 103, 1–19.
Imai, M., & Shelton, C. A. (2011). Elections and political risk: New evidence from the 2008 Taiwanese Presidential Election. Journal of Public Economics, 95, 837–849.
Jegadeesh, N., & Titman, S. (1993). Returns to buying winners and selling losers: Implications for stock market efficiency. Journal of Finance, 48, 65–91.
Kim, C., Pantzalis, C., & Park, J. C. (2012). Political geography and stock returns: The value and risk implications of proximity to political power. Journal of Financial
Economics, 106, 196–228.
Knight, B. (2006). Are policy platforms capitalized into equity prices? Evidence from the Bush/Gore 2000 presidential election. Journal of Public Economics, 90, 751–773.
Kollias, C., Papadamou, S., & Stagiannis, A. (2011). Terrorism and capital markets: The effects of the Madrid and London bomb attacks. International Review of Economics
& Finance, 20, 532–541.
Kroszner, R., & Stratmann, T. (2005). Corporate campaign contributions, repeat giving, and the rewards to legislator reputation. Journal of Law and Economics, 48,
41–71.
Li, H., Meng, L., Wang, Q., & Zhou, L. A. (2008). Political connections, financing and firm performance: Evidence from Chinese private firms. Journal of Development
Economics, 87, 283–299.
Lin, A. Y., & Lin, Y. N. (2014). Herding of institutional investors and margin traders on extreme market movements. International Review of Economics & Finance, 33,
186–198.
Pástor, L. U., & Veronesi, P. (2012). Uncertainty about government policy and stock prices. Journal of Finance, 67, 1219–1264.
Petersen, M. (2009). Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies, 22, 435–480.
Ross, R. S. (2006). Taiwan's fading independence movement. Foreign Affairs, 85, 141–148.
Sias, R. W., & Whidbee, D. A. (2010). Insider trades and demand by institutional and individual investors. Review of Financial Studies, 23, 1544–1595.
Su, Z. Q., Fung, H. G., Huang, D. S., & Shen, C. H. (2014). Cash dividends, expropriation, and political connections: Evidence from China. International Review of Economics
& Finance, 29, 260–272.
Vo, M. T. (2008). Strategic trading when some investors receive information before others. International Review of Economics & Finance, 17, 319–332.
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817–838.
Yao, J., Ma, C., & He, W. P. (2014). Investor herding behaviour of Chinese stock market. International Review of Economics & Finance, 29, 12–29.