International Journal of

Financial Studies

Article
Political Connections and Stock Price Crash Risk:
Empirical Evidence from the Fall of Suharto
Iman Harymawan 1, * , Brian Lam 2 , Mohammad Nasih 1 and Rumayya Rumayya 3
1 Department of Accountancy, Universitas Airlangga, Surabaya 60286, Indonesia; mohnasih@feb.unair.ac.id
2 Division of Business and Management, BNU-HKBU United International College, Zhuhai 519087, China;
brianmolam@uic.edu.hk
3 Department of Economics, Universitas Airlangga, Surabaya 60286, Indonesia; rumayya@feb.unair.ac.id
* Correspondence: harymawan.iman@feb.unair.ac.id; Tel.: +62-819851154




Received: 14 May 2019; Accepted: 6 September 2019; Published: 11 September 2019


Abstract: This study examines the relationship between firm-level political connections and stock price
crash risk in Indonesia. It employs the difference-in-difference design to deal with the self-selection
bias issue regarding the choice of the firms to become a politically connected firm. We use the sudden
resignation of the former President of Indonesia, Suharto, to show that politically connected firms are
associated with lower stock price crash risk and that the risk for these politically connected firms
increased after Suharto resigned. Furthermore, we found evidence that these negative associations
are more pronounced in firms with more complex firm structures.

Keywords: politically connected firms; stock price crash risk; complex firm structure

JEL Classification: D72; G10; L10

1. Introduction
Previous studies have found that firms with political connections receive preferential treatment to
finance, enjoy lower debt and equity costs, and have more procurement contracts (Claessens et al. 2008;
Boubakri et al. 2012; Houston et al. 2014; Goldman et al. 2013; Harymawan 2018). Other studies have
shown, however, that connected firms pay higher audit fees to auditors and charge higher prices for
debt to debtholders as a cost of political connections (Bliss and Gul 2012; Gul 2006). Grier et al. (1994)
have also examined the benefit–cost model across industries and the contributions of political action
committees in each industry, finding that industries with higher potential benefits from government
connections contribute more to their connections.
This study extends these prior findings in the political connections literature by investigating the
effect of political connections on firms’ stock price crash risk. Recent studies have provided evidence
that establishing political connections (i.e., hiring a politician as a director) can decrease the firm’s stock
price crash risk (Luo et al. 2016; Hu and Wang 2018). In contrast, Lee and Wang (2017) and Tee (2018)
have shown that connected firms tend to have a higher stock price crash risk. Meanwhile, Roberts (1990)
found a relationship between committee seniority in the US Congress and the distribution of federal
benefits. He also found that the stock market reacted significantly to the death of Senator Henry
Jackson, especially across his various constituent interests. Using the unexpected resignation of former
Indonesian President, Suharto, this study was able to investigate the stock price crash risk of connected
firms in the period before and after his resignation. This unique event provides a better setting than
a simple relationship model for answering the question of whether connected firms have a higher or
lower stock price crash risk than unconnected firms.

Int. J. Financial Stud. 2019, 7, 49; doi:10.3390/ijfs7030049 www.mdpi.com/journal/ijfs

Int. J. Financial Stud. 2019, 7, 49 2 of 16

Investigating the relationship between political connections and stock price crash risk in the pre-
and post-Suharto era provides us with a better research setting for several reasons. First, as Fisman (2001)
shows, during the Suharto era, political connections in Indonesia were centralized, which provides a
better proxy than a decentralized country for evaluating political connections; measuring the value of
political connections in a country with decentralized political decision-making is more complex due
to the variety of connection types. Second, the unexpected resignation of Suharto is an exogenous
event which allows us to make clearer estimates when evaluating the relationship between political
connections and stock price crash risk.
In this study, we first employed some univariate tests to describe our data and test the hypotheses.
The correlation matrix shows that firms connected to Suharto have a significantly higher stock price
crash risk. Then we compared the mean between a group of connected versus non-connected firms.
The results also show that firms with Suharto connections had a significantly higher mean of stock
price crash risk than firms with no connections.
In the multivariate analyses, we used the ordinary least square regression to test the hypotheses.
We found that firms with political connections had a negative and significant association with stock
price crash risk. To deal with the endogeneity issue, this study employed the difference-in-difference
model, using the unexpected resignation of Suharto as a natural experiment. We found that in the
period prior to the Suharto resignation, connected firms were more likely to have a lower stock price
crash risk. Interestingly, we found that after Suharto resigned, there was a positive and significant
association between political connections and stock price crash risk.
This study contributes to the existing literature of political connections in several ways.
Prior studies have examined the relationship between political connections to stock price crash
risk (citation). This study adds to this literature by being the first in providing evidence on how political
connections affect stock price crash risk by comparing when the political connections are in and out of
power. In addition, this study also extends the literature by examining the role of firm complexity
on the relationship between political connections and stock price crash risk. For practitioners, this
study provides useful information about the changes of stock price crash risk of politically connected
firms between pre and post political power. For a regulator, these findings provide an insight that
political activities disclosure might help the investor to reduce the information asymmetry between
firms and stakeholders to reduce the likelihood of stock price crash risk for politically connected
firms in the period post to political power resignation. The remainder of this article discusses the
institutional setting of political connections in Indonesia and presents the hypotheses, data, methods,
results, discussion, and conclusion.

2. Literature Review

2.1. Corporate Governance in Indonesia

Since the late 1970s, Indonesia has experienced economic transformation from state-owned
enterprise and a small and traditional business-dominated economy to modern private enterprise.
Unfortunately, rather than signaling the development of entrepreneurship and healthy business sectors,
the transformation marked the beginning of an era where political connections in business became
increasingly widespread, with Suharto’s cronies at the center. The success of these politically connected
firms heavily depended on their ability to acquire and maintain relationships with Suharto’s regime in
order to secure special privileges from the state.
As noted by McLeod (2000), these privileges included import protection, import licenses, no-bid
contracts from government and state-owned enterprises, cheap loans from the state and central banks,
rights to extract natural resources, tax relief, and even the right to collect taxes.
At first, the majority of these politically well-connected firms were dominated by the
Sino-Indonesian (of Chinese descent) business elites who had been close to Suharto since the 1950s.
Suharto’s administration granted them special rights to monopolize the import and distribution of sugar,
Int. J. Financial Stud. 2019, 7, 49 3 of 16

rice, forestry products, and automobiles. Over time, two other elite groups emerged: the well-connected
pribumi enterprises and businesses owned by Suharto’s children. These three groups formed a core
group of cronies in Suharto’s regime (Schwarz 1994).
Hill (2000) has shown that the 20 largest conglomerates in 1993, with a total turnover valued at
around 21% of GDP, were all connected to Suharto. Three of those twenty conglomerates were owned
by Suharto’s children. Claessens et al. (1999) has also shown that in 1996 Suharto’s cronies controlled
417 listed and non-listed companies on the Indonesian stock market. Together, these companies
contributed 16.6% of total stock market capitalization.

2.2. After the Fall of Suharto

In 1998, Indonesia was severely hit by a financial crisis. The crisis started in Thailand and
South Korea and then spread throughout Asia. During the crisis, the rupiah depreciated by 30%,
exports declined, the inflation rate escalated to almost 100%, the poverty rate doubled to over
27%, and economic growth contracted to almost 14% (Hofman et al. 2004). The financial crisis
then evolved into a political crisis which led to the resignation of Suharto after 32 years in power.
Pangestu and Habir (2002) have argued that the favorable treatment to politically connected firms
created moral hazard and caused the crisis as these firms became reckless in their business practices,
violating legal lending limits and generating non-performing loans.
Sato (2004) has found that the performance of Suharto-connected firms varied after the crisis.
The firms that survived were those established in the 1970s and ranked among the top 20 or top
30 richest in the 1980s and 1996. In contrast, the firms that suffered severely were younger firms
established in the 1980s and either ranked in the top 20 in 1996 or else outside the top 20 in the 1980s
but inside the top 10 in 1996. Carney and Hamilton-Hart (2015) also found that a significant number of
large firms shrank tremendously in 1996 and fell outside the top 200, including the firms of Suharto’s
children. Ultimately, Suharto’s old, well-established cronies seemed to successfully reconsolidate
their businesses and actively reconnect with current political actors and government officials in the
post-Suharto administration, who were possibly motivated to secure the benefits they had gained
under the Suharto government (Winter 2013).

2.3. Stock Price Crash Risk

Conceptually, stock price crash risk is premised on the tendency of managers to withhold bad
news, which leads to bad news being stockpiled in firms. At a certain point, it becomes too costly or
impossible for managers to withhold the bad news any longer. When such a point arrives, all the hidden
bad news is revealed to the market at once, resulting in a significant stock price drop, that is, a stock
price crash (Kim et al. 2011). Extant studies have investigated various determinants of stock price crash.
For instance, Kim et al. (2011) found that corporate tax avoidance enables managers to hoard bad news,
which leads to the risk of a stock price crash. He and Ren (2019) have provided evidence that financial
constraints are positively associated with future stock price crash risk, via both bad news hoarding
and default risk channels. Chen et al. (2016) have found that firms with a higher degree of earnings
smoothing are more prone to stock price crashes. For their part, Chauhan et al. (2017) have shown
that stock liquidity, which is identified through the threat of intervention and price informativeness,
decreases stock price crash risk as it works as a governance mechanism to discipline managers for
withholding bad news. In addition, prior research has shown that firms’ corporate social responsibility
significantly mitigates stock price crash risk (Lee 2016). Our research extends this stream of literature
by examining the relationship between stock price crash risk and political connections, using the
sudden resignation of Suharto to address the self-selection bias issue.

2.4. Hypothesis Development

There are two possible relationships between politically connected firms and stock price crash
risk. The first possibility is that firms with political connections are more likely to have higher stock
Int. J. Financial Stud. 2019, 7, 49 4 of 16

price crash risk. Prior literature has suggested that politically connected firms can derive significant
benefits in terms of financing, whereas non-politically connected firms are more likely to suffer from
financing constraints (Luo and Zhen 2008). To meet financing requirements, managers are encouraged
to whitewash financial statements and hide negative information. These actions may result in an
increase in information asymmetry, and the accumulation of bad news may eventually lead to a stock
price crash. He and Ren (2019) have also argued that if the institutional setting of the country is not
well developed (i.e., inefficient stock market, weak corporate governance, etc.), it will increase the
ability of firms to hoard bad news. This situation (hoarding bad news) thus increases the stock price
crash risk of connected firms. Based on this, we could say there is a positive association between
political connections and stock price crash risk, which is the first possible relationship.
On the other hand, Hu and Wang (2018) have also found that with the preferential access of
financing, politically connected firms can reduce bad news hoarding activities to avoid government
regulation and maintain access to government subsidies. In this way, the degree of information
asymmetry can be reduced, thus decreasing the likelihood of a stock price crash. Furthermore,
He and Ren (2019) have argued that if a country has a good institutional setting (i.e., efficient stock
market, good governance, pro-business policies, etc.) there is a lower probability of firms hoarding
bad news, therefore decreasing stock price crash risk. From this, we could argue that politically
connected firms actually have a lower stock price crash risk. Based on above discussion on the possible
relationships between politically connected firms and stock price crash risk, the formal hypothesis is
as follows:

Hypothesis 1 (H1). There is a relationship between firms with political connections and the level of
stock price crash risk.

Fisman (2001) has found that politically connected firms in Indonesia strongly relied on the
benefits from their connections under the Suharto regime and often had privileged access to financing.
Suharto’s resignation in 1998 represented a fundamental transition in Indonesia’s political economy
at the national level. The shift is evidenced by the decline in the number of people with military
and state backgrounds and the rise of those with private sector backgrounds in Indonesia’s political
elite (Poczter and Pepinsky 2016). Leuz and Oberholzer-Gee (2006) have also found that political
connections can lose their value through election losses. Connected firms thus need to be more
responsive to market pressures since they can no longer rely on the same benefits from their connections
(Harymawan and Nowland 2016). Suharto’s resignation and the subsequent regime change triggered
financial and operational shocks for politically connected firms. They then needed to face all kinds of
market pressures by themselves, resulting in a higher stock price crash risk. Therefore, we predict
that the resignation of Suharto increased the need for politically connected firms to respond to market
pressures, resulting in a higher stock price crash risk.

Hypothesis 2 (H2). After Suharto resigned, the risk of a stock price crash for firms with political
connections increased.

The complexity of firm structure may affect stock price crash risk. We measured the complexity
of firm structure according to the number of subsidiaries owned by the company. As the number of
subsidiaries increases, so too do the benefits derived from political connections.

Hypothesis 3 (H3). The negative relationship between stock price crash risk and political connections is more
pronounced for firms with a more complex firm structure.
Int. J. Financial Stud. 2019, 7, 49 5 of 16

3. Methodology

3.1. Data
Our sample covers the period from 1995 to 2001. We obtained political connection data from
the Indonesia Capital Market Directory (ICMD) and information on the total number of subsidiaries
from Osiris. Accounting and financial data were retrieved from Compustat Global. We deleted
observations with missing Compustat data, stock prices, returns, and trading volume data when
estimating our crash risk measures and control variables. We excluded the year of Suharto’s resignation
(1998) to achieve an unambiguous pre-event and post-event period for the main tests, especially for
the difference-in-difference test. The main sample includes 730 firm-year observations.

3.2. Measuring Stock Price Crash Risk

We followed Hutton et al. (2009) and Kim et al. (2016) to construct the three measures of stock price
crash risk. Our first measure was the firm-specific return crashes. The indicator variable CRASH equals
one for a firm-year that experiences one or more crash weeks during the fiscal year and zero otherwise.
To calculate the first measure of crash likelihood for each firm-year, we first estimated the weekly
returns by firm and by year with the following regression model:

rj,k = αj + β1jrm,k − 2 + β2jrm,k − 1 + β3jrm,k + β4jrm,k+1 + β5jrm,k + 2

(1)
+ β6jri,k − 2 + β7jri,k − 1 + β8jri,k + β9jri,k + 1 + β10jri,k + 2 + εjk,

where rj,k is the return on stock j in week k, rm,k is the return on the value-weighted market index in
week k provided by the Center for Research in Security Prices, and ri,k is the return on the Fama–French
value-weighted industry index in week k. As in Dimson 1979, the lead and lag terms for the market
and industry indexes were used to control for nonsynchronous trading. The firm-week return was
measured by the natural logarithm of one plus the residual return in Equation (1), with at least 26
weekly return observations available per firm-year. Crash weeks are firm-week returns which are
below 3.2 standard deviations of the mean firm-week return of the fiscal year.
For the second measure, we used the volatility of below-mean versus above-mean returns
(DUVOL) for each firm-year. The variable DUVOL was calculated by the natural logarithm of the ratio
of the standard deviation of weekly returns below the annual mean divided by the standard deviation
of weekly returns above the annual mean.
For the third measure, we used the negative skewness of firm-specific weekly returns (NCSKEW).
For each firm-year, we multiplied minus one to the third power of weekly returns and then divided it
by the third power of the standard deviation of weekly returns. Appendix A presents a summary of all
variables used in this research.

3.3. Main Model

To test our hypotheses, we performed multiple regression tests with a series of control variables
to ensure the results were not subject to some crucial factors which have systematic effects on the
cross-sectional variation in stock price crash risk.
To examine our first hypothesis (H1), we estimated the following linear regression, which regresses
stock crash risk (CRASH, DUVOL, or NCSKEW) on political connection (PCON) and the
control variables.
X X X
DepVart = β0 + β1 PCONt + βi Xt + βj YRDt + βk INDDt + et, (2)

where DepVar is CRASH, DUVOL, or NCSKEW; X is a set of the control variables, YRD is the year
dummy, and INDD is the industry dummy.
Int. J. Financial Stud. 2019, 7, 49 6 of 16

To examine our secondhypothesis (H2), we estimated the following linear regression,

which regresses stock crash risk (CRASH, DUVOL, or NCSKEW) on the difference-in-differences
dummy (DID), political connection (PCON), and control variables.
X X X
DepVart = β0 + β1 DIDt + β22 PCONt + βi Xt + βj YRDt + βk INDDt + et, (3)

where DepVar is CRASH, DUVOL, or NCSKEW. X is a set of the control variables, YRD is the year
dummy, and INDD is the industry dummy.
To examine our third hypothesis (H3), we estimated the linear regression using Equation (1),
which regresses stock crash risk (CRASH, DUVOL, or NCSKEW) on the political connection (PCON).
However, this time we divided the full sample into complex and less-complex subsamples. In the
complex subsample, the total number of subsidiaries was greater than the median. In the less-complex
subsample, the total number of subsidiaries was not greater than the median.

3.4. Sample Selection

Our sample consisted of listed companies from Indonesia, covering the period from 1995 to 2001.
We adopted the two-digit Standard Industrial Classification (SIC) system to classify the different
industrial sectors and removed observations that did not have firm-specific control variables. The final
sample for the main model contained 730 firm-year observations. Continuous variables were winsorized
at 1% or 99% to control for the effects of outliers.

4. Empirical Analysis
Table 1 presents the sample distribution for the period 1995–1997 and 1999–2001. We excluded the
year of Suharto’s resignation (1998) to achieve an unambiguous pre-event and post-event period for
the main tests, especially for the difference-in-difference test. Among the 730 firm-year observations,
the number of firms with political connections was 149 and those without political connections 581.

Table 1. Sample distribution.

Number of Firms with Number of Firms without

Year Total
Political Connections Political Connections
1995 13 29 42
1996 23 62 85
1997 29 106 135
1999 27 126 153
2000 28 125 153
2001 29 133 162
Total 149 581 730
Note: We excluded the year of Suharto’s resignation (1998) to achieve an unambiguous pre- and post-event period
for the difference-in-difference model.

Table 2 presents descriptive statistics for the period 1995–1997 and 1999–2001, totaling 730 firm-year
observations. All variables are defined in the appendix. The sample mean of PCON is 0.204, indicating
that about 20% of the sample was connected politically to Suharto and the remaining 80% was
not. The dependent variables (CRASH, DUVOL, and NCSKEW) and variable of interest (PCON) are
presented in current-year values. The control variables are presented in lag-one-year values because
lag-one-year values of these variables are used in the regression model.
Int. J. Financial Stud. 2019, 7, 49 7 of 16

Table 2. Summary statistics (n = 730).

Variables Mean Std Min Q1 Median Q4 Max

CRASH 0.563 0.496 0.000 0.000 1.000 1.000 1.000
DUVOL −0.066 0.298 −1.026 −0.251 −0.079 0.100 2.092
NCSKEW −0.114 1.068 −3.861 −0.608 −0.139 0.268 17.407
PCON 0.204 0.403 0.000 0.000 0.000 0.000 1.000
SIZE(t−1) 4.800 1.446 1.328 3.763 4.666 5.743 8.865
ROA(t−1) −0.013 0.193 −1.175 −0.065 0.028 0.083 0.473
LEVERAGE(t−1) 0.477 0.317 0.000 0.242 0.468 0.646 1.613
MTB(t−1) 1.323 2.252 −8.558 0.385 0.908 1.711 23.928
NCSKEW (t−1) −0.089 1.257 −8.230 −0.545 −0.107 0.326 10.310
DTURN (t−1) 0.001 0.004 −0.022 −0.001 0.000 0.001 0.026
SIGMA(t−1) 0.063 0.038 0.015 0.036 0.055 0.078 0.421
RET (t−1) 0.554 2.551 −0.979 −0.465 −0.146 0.555 21.187
Observations
This table presents summary statistics for our sample for the period 1995–1997 and 1999–2001.

Table 3 presents the Pearson correlations. It displays the correlations between dependent variables,
the independent variable, and the control variables. The significance level is denoted at 10%, 5%, and 1%.
Table 4 presents the characteristics of politically connected and non-politically connected firms.
We observed that firms with political connections are in general larger in terms of total assets than
those without political connections. These politically connected firms also have higher leverage.
Table 5 presents the regression results for the relationship between political connection and stock
price crash risk. The coefficient of political connection (PCON) was negative and significant for all
three risk models (CRASH and DUVOL at the 1% level and NCSKEW at the 5% level) after controlling
the firm-specific control variables. This result was consistent with our prediction, which is that the
stock price crash risk is lower for politically connected firms. Therefore, hypothesis 1 is supported.
To use the difference-in-difference model, we needed to test whether the parallel trend assumption
holds. The visual inspection method was not appropriate for this study because the main dependent
variable (stock crash) is an indicator variable. Therefore, we conducted a formal test on the parallel
trend. Our sample contained three pre-treatment periods and three post-treatment periods. Following
Autor (2003), we specify the following model:

Yit = B−2 Dit + B−1 Dit + B1 Dit + B2 Dit + B3 Dit + Firm Fix Effects + Year Fixed Effects + eit , (4)

where Y is the outcome for firm i at time t. We included the interactions of the time dummies and the
political connection indicator for the first two pre-Suharto-resignation periods and dropped the one for
the year of Suharto’s resignation. Therefore, all the other interactions are shown with reference to this
omitted period. If the coefficients B−2 and B−1 were not significant, then the difference in differences
was not significantly different between the politically connected firms and non-politically connected
firms in the pre-Suharto-resignation periods, indicating that the parallel trend assumption holds.
As we included all the Suharto-connected firms as a treatment group and all the
non-Suharto-connected firms as a control group during the sample period, the random sampling
assumption for the treatment group and control group was not a concern in our study.
Table 6, panel A presents regression results testing the parallel trend. The coefficients for PCON1996
and PCON1997 were insignificant, indicating that the parallel trend assumption holds. Besides, it can
be observed from the coefficients for PCON1999, PCON2000, and PCON2001 that the increased stock
crash risk for Suharto-connected firms fades across the years in the post-Suharto period.
Int. J. Financial Stud. 2019, 7, 49 8 of 16

Table 3. Pearson correlations (n = 730).

Variables [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
[1] CRASH 1.000
[2] DUVOL 0.303 *** 1.000
[3] NCSKEW 0.296 *** 0.851 *** 1.000
[4] PCON 0.042 −0.096 *** −0.069 * 1.000
[5] SIZE(t−1) 0.168 *** −0.064 * 0.008 0.391 *** 1.000
[6] ROA(t−1) 0.053 −0.029 −0.022 −0.002 0.044 1.000
[7] LEVERAGE(t−1) −0.068 * 0.044 0.043 0.093 ** 0.132 *** −0.589 *** 1.000
[8] MTB(t−1) 0.070 * −0.013 0.006 0.063 * 0.141 *** 0.191 *** −0.115 *** 1.000
[9] NCSKEW (t−1) −0.059 0.060 0.064 * −0.082 ** −0.008 −0.061 0.029 −0.000 1.000
[10] DTURN (t−1) 0.077 ** −0.020 0.005 0.018 0.051 0.027 −0.054 0.058 −0.041 1.000
[11] SIGMA(t−1) −0.151 *** 0.113 *** 0.045 −0.065 * −0.328 *** −0.136 *** 0.170 *** −0.081 ** −0.029 −0.054 1.000
[12] RET (t−1) 0.007 −0.012 −0.049 0.051 −0.030 0.165 *** −0.035 0.209 *** −0.117 *** −0.007 0.239 *** 1.000
* p < 0.10, ** p < 0.05, *** p < 0.01.
Int. J. Financial Stud. 2019, 7, 49 9 of 16

Table 4. Characteristics of politically connected and non-politically connected firms.

Firms with Political Connections Firms without Political Connections

Variables Coef t-Value
N = 149 N = 581
CRASH 0.604 0.552 0.052 −1.131
DUVOL −0.123 −0.052 −0.071 *** 2.601
NCSKEW −0.258 −0.077 −0.182 * 1.854
SIZE(t−1) 5.916 4.514 1.402 *** −11.458
ROA(t−1) −0.014 −0.013 −0.001 0.065
LEVERAGE(t−1) 0.535 0.462 0.074 ** −2.533
MTB(t−1) 1.601 1.252 0.350 * −1.692
NCSKEW (t−1) −0.293 −0.037 −0.256 ** 2.220
DTURN (t−1) 0.001 0.001 0.000 −0.492
SIGMA(t−1) 0.058 0.064 −0.006 * 1.761
RET (t−1) 0.811 0.488 0.323 −1.381
The asterisks *, *** denote significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Table 5. Main results: results of regression of political connections on stock price crash risk.

(1) (2) (3)

Variables
CRASH DUVOL NCSKEW
PCON −0.366 *** −2.92 −0.087 *** −2.95 −0.276 ** −2.54
SIZE(t−1) 0.275 *** 4.47 0.010 0.68 0.053 1.18
ROA(t−1) −0.119 −0.12 0.058 0.91 0.256 1.16
LEVERAGE(t−1) −0.741 *** −2.65 0.010 0.19 0.090 0.49
MTB(t−1) 0.035 0.64 0.001 0.31 0.013 0.77
NCSKEW (t−1) −0.069 −1.16 0.007 0.80 0.031 1.07
DTURN (t−1) 6.830 0.39 −4.586 *** −2.97 −12.449 *** −2.98
SIGMA(t−1) −5.672 * −1.93 0.666 1.48 1.465 1.23
RET (t−1) 0.041 1.57 −0.000 −0.02 −0.004 −0.39
Constant 2.261 * 1.70 0.109 0.30 1.832 0.89
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Observations 704 730 730
R2 0.151 0.118
Adjusted R2 0.078 0.042
Pseudo R2 0.120
Notes: This table presents regression results testing the effect of political connections on stock price crash risk.
The second column displays t statistics. Standard errors were clustered by firm and year. The asterisks *, **, *** denote
significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Table 6. Panel A. Test on parallel trend; Panel B. Difference-in-difference results.

Panel A
(1) (2) (3)
Crash DUVol NCSkew
PCON1996 0.258 0.40 −0.073 −1.02 −0.271 −0.99
PCON1997 0.063 0.10 −0.050 −0.73 −0.118 −0.45
PCON1999 1.662 *** 2.81 0.222 *** 3.32 0.551 ** 2.15
PCON2000 0.976 * 1.72 0.196 *** 2.88 0.488 * 1.87
PCON2001 0.591 1.04 0.107 1.57 0.480 * 1.84
Constant −11.897 −0.02 −0.804 *** −2.86 −3.286 *** −3.06
Year FE Yes Yes Yes
Firm FE Yes Yes Yes
Observations 765 886 886
R2 0.317 0.248
Adjusted R2 0.129 0.041
Pseudo R2 0.138
Notes: this table presents regression results testing the parallel trend. The second column displays t
statistics. Standard errors are clustered by firm and year. The asterisks *, **, *** denote significance
at the 10%, 5%, and 1% levels (two-tailed), respectively.
Int. J. Financial Stud. 2019, 7, 49 10 of 16

Table 6. Cont.

Panel B
Variables (1) (2) (3)
CRASH DUVOL NCSKEW
DID 2.142 *** 2.74 0.284 *** 3.84 0.933 *** 2.64
PCON −1.576 * −1.79 −0.251 *** −2.61 −0.816 *** −2.92
SIZE(t−1) 0.281 *** 4.61 0.010 0.74 0.054 1.27
ROA(t−1) −0.203 −0.22 0.045 0.87 0.213 1.13
LEVERAGE(t−1) −0.835 *** −3.25 −0.004 −0.08 0.045 0.26
MTB(t−1) 0.041 0.70 0.002 0.42 0.015 0.82
NCSKEW (t−1) −0.051 −0.94 0.008 1.10 0.036 1.41
DTURN (t−1) −0.147 −0.01 −5.234 *** −3.97 −14.581 *** −2.71
SIGMA(t−1) −5.082 −1.61 0.713 * 1.65 1.620 1.40
RET (t−1) 0.043 * 1.73 0.000 0.18 −0.003 −0.30
Constant 2.637 ** 2.21 0.163 0.50 2.010 1.03
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Observations 704 730 730
R2 0.188 0.148
Adjusted R2 0.116 0.073
Pseudo R2 0.142
Notes: This table presents regression results testing the effect of Suharto’s resignation on stock price
crash risk. The second column displays t statistics. Standard errors are clustered by firm and year.
The asterisks *, **, *** denote significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Table 6, panel B presents the regression results for the effect of the fall of Suharto on the relationship
between political connection and stock price crash risk. The coefficient for differences in differences
(DID) is positive and significant at the 1% level for all three risk models after controlling the firm-specific
control variables and country-level control variables. This result is consistent with our prediction,
which is that the stock price crash risk of politically connected firms increased after the fall of Suharto.
Therefore, hypothesis 2 is supported.
Table 7 reports the results of the sensitivity test for Table 5 on the regression of stock price
crash risk on political connection because this test includes the year in which Suharto resigned, 1998.
The coefficient of political connection (PCON) for all three models was negative (CRASH at the 1%
level, and NCSKEW at the 10% level), indicating that political connection lowers stock price crash risk.
The significance level for NCSKEW and DUVOL in Table 7 was lower than in Table 5. This is because
the inclusion of observations for 1998, the year in which Suharto resigned, may have introduced some
noise into the regression model.
Table 8 reports the results of the sensitivity test for Table 5 on the regression of stock price crash
risk on DID because this test includes the year in which Suharto resigned, 1998. The coefficient of DID
for all three models was significantly positive, indicating that the stock price crash risk of politically
connected firms increased after the fall of Suharto.
To investigate whether a political connection exerts distinct impacts on stock price crash risk when
interacting with the differential complexity of firm structure (H3), the full sample was divided into two
subsamples according to the complexity of firm structure. Table 9 presents the regression results for
the effect of complexity of firm structure on the relationship between political connection and stock
price crash risk. The coefficient for political connection (PCON) was negative and significant at the 1%
level for all three risk models for complex firms after controlling the firm-specific control variables.
This result was consistent with our prediction, which is that the negative relationship between political
connection and stock price crash risk is more pronounced for firms with complex firm structures.
Therefore, hypothesis 3 is supported. In addition, we also tested (non-tabulated results) whether the
negative associations between political connection and stock price crash risk in complex firms are due
to the size effect (Dang et al. 2018). Hence, we run similar regression model except we exclude the
SIZE variable. We find that the coefficients of PCON are still negative.
Int. J. Financial Stud. 2019, 7, 49 11 of 16

Table 7. Results of regression of political connections on stock price crash risk (including year of
Suharto’s resignation, 1998).

(1) (2) (3)

Variables
CRASH DUVOL NCSKEW
PCON −0.373 *** −3.32 −0.059 −1.54 −0.200 * −1.65
SIZE(t−1) 0.255 *** 4.85 0.002 0.11 0.016 0.33
ROA(t−1) −0.283 −0.32 0.049 0.85 0.230 1.28
LEVERAGE(t−1) −0.591 ** −2.22 0.011 0.28 0.045 0.34
MTB(t−1) 0.050 1.10 0.001 0.24 0.010 0.98
NCSKEW (t−1) −0.087 −1.50 0.009 0.95 0.038 1.15
DTURN (t−1) 5.944 0.42 −1.445 −0.52 −2.049 −0.22
SIGMA(t−1) −4.252 * −1.80 0.616 * 1.73 1.681 * 1.91
RET (t−1) 0.040 1.63 −0.001 −0.37 −0.003 −0.31
Constant 1.737 1.25 0.113 0.38 1.616 1.00
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Observations 879 886 886
R2 0.125 0.088
Adjusted R2 0.062 0.022
Pseudo R2 0.111
Notes: This table presents regression results testing the effect of political connections on stock price crash risk. The
second column displays t statistics. Standard errors are clustered by firm and year. The asterisks *, **, *** denote
significance at the 10%, 5%, and 1% levels (two-tailed), respectively.

Table 8. Difference-in-difference results (including year of Suharto’s resignation, 1998).

(1) (2) (3)

Variables
CRASH DUVOL NCSKEW
DID 2.441 *** 3.46 0.274 *** 4.18 0.961 *** 2.80
PCON −1.496 *** −2.66 −0.190 ** −2.38 −0.662 *** −2.78
SIZE(t−1) 0.262 *** 5.23 0.002 0.15 0.018 0.38
ROA(t−1) −0.361 −0.44 0.040 0.85 0.197 1.35
LEVERAGE(t−1) −0.657 ** −2.50 0.002 0.07 0.017 0.14
MTB(t−1) 0.064 1.22 0.002 0.67 0.016 1.26
NCSKEW (t−1) −0.073 −1.25 0.010 1.14 0.041 1.40
DTURN (t−1) 1.527 0.10 −1.871 −0.62 −3.540 −0.33
SIGMA(t−1) −3.713 −1.54 0.644 * 1.89 1.777 ** 2.07
RET (t−1) 0.034 1.38 −0.001 −0.48 −0.005 −0.48
Constant 2.007 1.49 0.148 0.54 1.739 1.14
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Observations 879 886 886
R2 0.161 0.121
Adjusted R2 0.100 0.057
Pseudo R2 0.142
Notes: This table presents regression results testing the effect of Suharto’s resignation on stock price crash risk.
The second column displays t statistics. Standard errors are clustered by firm and year. The asterisks *, **, *** denote
significance at the 10%, 5%, and 1% levels (two-tailed), respectively.
Int. J. Financial Stud. 2019, 7, 49 12 of 16

Table 9. Results of regression for complex and less-complex firms.

CRASH DUVOL NCSKEW

Variables (1) (2) (3) (4) (5) (6)
Complex Firms Less-Complex Firms Complex Firms Less-Complex Firms Complex Firms Less-Complex Firms
PCON −0.610 *** −9.48 −0.028 −0.07 −0.154 *** −3.09 −0.103 * −1.73 −0.601 *** −3.14 −0.237 −1.19
SIZE(t−1) 0.296 * 1.68 0.166 1.52 −0.005 −0.18 0.030 1.46 0.055 0.68 0.075 1.25
ROA(t−1) 0.312 0.24 −0.067 −0.05 0.046 0.40 0.066 0.78 0.251 0.60 0.222 0.99
LEVERAGE(t−1) −2.047 ** −2.43 −0.110 −0.18 −0.042 −0.36 0.006 0.11 0.015 0.04 0.071 0.53
MTB(t−1) 0.077 1.21 0.062 0.76 0.006 1.31 0.004 0.54 0.029 1.03 0.021 1.20
NCSKEW (t−1) −0.071 −0.50 −0.053 −0.44 −0.002 −0.14 0.011 0.53 0.017 0.28 0.046 1.02
DTURN (t−1) −21.782 −0.53 30.546 0.92 −5.211 ** −2.50 −1.083 −0.24 −19.726 *** −2.72 2.715 0.17
SIGMA(t−1) −5.253 −1.07 −6.479 *** −2.87 0.569 0.67 0.553 1.33 1.053 0.35 1.182 1.26
RET (t−1) 0.037 0.83 0.067 ** 2.33 −0.007 −1.27 0.008 1.50 −0.019 −0.70 0.017 1.24
Constant 3.415 ** 2.24 2.929 * 1.69 0.343 0.92 −0.197 −0.88 2.270 1.07 −0.308 −0.62
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Observations 299 372 315 397 315 397
R2 0.278 0.165 0.167 0.151
Adjusted R2 0.148 0.058 0.016 0.042
Pseudo R2 0.197 0.131
Notes: this table presents regression results testing the effect of political connections on stock price crash risk in complex and less-complex subsamples. Complex is a subsample where the
total number of subsidiaries is greater than the median. Less-complex is a subsample where the total number of subsidiaries is not greater than the median. The second columns displays t
statistics. Standard errors are clustered by firm and year. The asterisks *, **, *** denote significance at the 10%, 5%, and 1% levels (two-tailed), respectively.
Int. J. Financial Stud. 2019, 7, 49 13 of 16

To ensure that the assignment of observations into the treatment group and control group was
random, we adopted the coarsened exact matching (CEM) method. We set each covariate into five equal
bins, or strata. Eight covariates were input into the CEM model. Table 10, panel A presents the matching
CEM summary. Out of a total of 171 strata generated by the CEM model, 36 strata contained both
connected and unconnected observations. A total of 109 out of 149 connected observations were
matched with 403 out of 581 unconnected observations.

Table 10. Panel A. Coarsened exact matching method—matching summary. Number of strata: 171;
Number of matched strata: 36. Panel B. Coarsened exact matching method—diagnosis of matching
quality. Panel C. Coarsened exact matching method—difference-in-difference results.

Panel A
PCON = 0 PCON = 1
All 581 149
Matched 403 109
Unmatched 178 40
Panel B
Pre-match L1 statistics Post-match L1 statistics
SIZE 0.391 0.230
ROA 0.156 0.144
LEVERAGE 0.195 0.163
MTB 0.107 0.138
NCSKEW 0.153 0.149
DTURN 0.151 0.098
SIGMA 0.203 0.123
RET 0.069 0.083
Panel C
(1) (2) (3)
Variables
Crash DUVol NCSkew
DID 1.077 ** 2.10 0.312 *** 3.13 1.158 *** 2.67
PCON (t−1) 0.591 *** 3.11 −0.257 ** −2.05 −0.871 ** −2.37
SIZE(t−1) 0.294 *** 5.91 0.009 0.82 0.061 * 1.69
ROA(t−1) −1.337 ** −2.18 0.047 0.60 0.146 0.68
LEVERAGE(t−1) −0.823 *** −3.60 0.008 0.09 −0.047 −0.24
MTB(t−1) 0.083 0.85 0.007 1.08 0.050 ** 2.23
NCSKEW (t−1) −0.099 −0.56 0.032 * 1.84 0.080 1.49
DTURN (t−1) −2.268 −0.06 −4.079 −1.06 −17.753 −1.16
SIGMA(t−1) −6.599 −1.35 0.549 0.76 2.651 1.20
RET (t−1) 0.211 1.55 0.035 *** 2.77 0.087 *** 3.43
Constant 3.019 * 1.81 0.292 0.95 2.591 1.31
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Observations 475 512 512
R2 0.242 0.197
Adjusted R2 0.143 0.093
Pseudo R2 0.124
Notes: this table presents regression results testing the effect of Suharto’s resignation on stock price crash risk.
Standard errors are clustered by firm and year. The asterisks *, **, *** denote significance at the 10%, 5%, and 1%
levels (two-tailed), respectively.

We used L1 statistics to measure the matching quality (Iacus et al. 2012; He et al. 2019). L1 statistics
are the absolute difference in the value of covariates between the connected and unconnected firms.
Table 10, panel B presents the diagnosis of the matching quality of the CEM method. The results show
that the post-match L1 statistics were generally lower than the pre-match L1 statistics, indicating there
is a significant improvement in the matching quality with CEM.
Int. J. Financial Stud. 2019, 7, 49 14 of 16

Table 10, panel C presents the result of the replication of the baseline model with the
difference-in-difference model by the CEM method. The table reveals a consistent result with that
in Table 6 Panel B, further supporting our hypothesis.

5. Conclusions
Establishing political connections can generate benefits and costs for firms. On the one hand,
if politicians are involved in a firm, they can use their network and power to support that firm.
However, this can also increase the firm’s risk due to potential conflicts of interest. This study has
focused on providing some evidence on the relationship between political connections and stock
price crash risk in a country with centralized political decision-making. In general, we have shown
that firms with political connections have a lower stock price crash risk. This finding indicates that
political connections can reduce the probability of a firm experiencing a stock price crash. Using the
difference-in-difference method, we found that firms with political connections have a higher stock
price crash risk after their connections fall from power. This result implies that the cost of having
political connections emerges only after these connections lose power. In addition, we also found that
negative associations between political connections and stock price crash risk are more pronounced in
firms with more complex firm structures. Despite the issue of the sample period, this study provides a
better research setting which produces better information on how political connections affect the stock
price crash risk of firms.

Author Contributions: Conceptualization, I.H., M.N., R.R., and B.L.; data curation, B.L.; formal analysis, I.H.,
M.N., R.R., and B.L.; methodology, I.H. and B.L.; software, I.H.; supervision, I.H., M.N., and R.R.; validation,
I.H., M.N., and R.R.; visualization, I.H. and B.L.; writing—original draft, R.R.; writing—review and editing,
I.H. and B.L.
Funding: This research was partially supported by the Ministry of Research Technology and Higher Education of
Indonesia under the 2018 SIMLITABMAS Research Grant (PDUPT Scheme).
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A

Table A1. Variable definitions.

Variable Definition Data Source

1 for a firm–year that experiences one or more firm-specific weekly returns
falling 3.2 standard deviations below the mean firm-specific weekly returns
CRASH Compustat
over the fiscal year, with 3.2 chosen to generate frequencies of 0.1% in the
normal distribution during the fiscal year period, otherwise 0
DUVOL Down-to-up volatility of firm-specific weekly returns Compustat
NCSKEW Negative skewness of firm-specific weekly returns over the fiscal year period. Compustat
PCON 1 for a firm that is politically connected with Suharto, otherwise 0 ICMD
Average monthly share turnover over the current fiscal year period minus the
average monthly share turnover over the previous fiscal year period, where
DTURN Compustat
monthly share turnover is calculated as the monthly trading volume divided
by the total number of shares outstanding during the month
SIGMA Standard deviation of firm-specific weekly returns over the fiscal year period. Compustat
RET Mean of firm-specific weekly returns over the fiscal year period Compustat
SIZE Natural logarithm of the market value of equity Compustat
MTB Market value of equity divided by the book value of equity Compustat
LEVERAGE Total long-term debts divided by total assets Compustat
ROA Income before extraordinary items divided by lagged total assets Compustat
Int. J. Financial Stud. 2019, 7, 49 15 of 16

References
Autor, David H. 2003. Outsourcing at will: The contribution of unjust dismissal doctrine to the growth of
employment outsourcing. Journal of Labor Economics 21: 1–42. [CrossRef]
Bliss, Mark A., and Ferdinand A. Gul. 2012. Political connection and cost of debt: Some Malaysian evidence.
Journal of Banking & Finance 36: 1520–27. [CrossRef]
Boubakri, Narjess, Omrane Guedhami, Dev Mishra, and Walid Saffar. 2012. Political connections and the cost of
equity capital. Journal of Corporate Finance 18: 541–59. [CrossRef]
Carney, Richard W., and Natasha Hamilton-Hart. 2015. What Do Changes in Corporate Ownership in Indonesia
Tell Us? Bulletin of Indonesian Economic Studies 51: 123–45. [CrossRef]
Chauhan, Yogesh, Satish Kumar, and Rajesh Pathak. 2017. Stock liquidity and stock prices crash-risk: Evidence
from India. The North American Journal of Economics and Finance 41: 70–81. [CrossRef]
Chen, Changling, Jeong-Bon Kim, and Li Yao. 2016. Earnings Smoothing: Does It Exacerbate or Constrain Stock
Price Crash Risk? Journal of Corporate Finance. [CrossRef]
Claessens, Stijn, Simeon Djankov, and Larry HP Lang. 1999. Who Controls East Asian Corporations, and the
Implications for Legal Reform. World Bank Policy Research, 2054. [CrossRef]
Claessens, Stijn, Erik Feijen, and Luc Laeven. 2008. Political connections and preferential access to finance: The role
of campaign contributions. Journal of Financial Economics 88: 554–80. [CrossRef]
Dang, Chongyu, Zhichuan Frank Li, and Chen Yang. 2018. Measuring firm size in empirical corporate finance.
Journal of Banking & Finance 86: 159–76.
Fisman, Raymond. 2001. Estimating the Value of Political Connections. The Amerian Economic Review 91: 1095–102.
[CrossRef]
Goldman, Eitan, Jörg Rocholl, and Jongil So. 2013. Politically connected boards of directors and the allocation of
procurement contracts. Review of Finance 17: 1617–48. [CrossRef]
Grier, Kevin B., Michael C. Munger, and Brian E. Roberts. 1994. The determinants of industry political activity,
1978–86. American Political Science Review 88: 911–26. [CrossRef]
Gul, Ferdinand A. 2006. Auditors’ response to political connections and cronyism in Malaysia. Journal of Accounting
Research 44: 931–63. [CrossRef]
Harymawan, Iman. 2018. Why do firms appoint former military personnel as directors? Evidence of loan interest
rate in militarily connected firms in Indonesia. Asian Review of Accounting 26: 2–18. [CrossRef]
Harymawan, Iman, and John Nowland. 2016. Political connections and earnings quality: How do connected firms
respond to changes in political stability and government effectiveness? International Journal of Accounting &
Information Management 24: 339–56. [CrossRef]
He, Guanming, and Helen Ren. 2019. Are Financially Constrained Firms Susceptible to a Stock Price Crash?
Working Paper. Durham University, Durham/Stockton-on-Tees, UK. Available online: https://ssrn.com/
abstract=3021307 (accessed on 21 June 2019).
He, Guanming, David Marginson, and Xixi Dai. 2019. Do voluntary disclosures of product and business expansion
plans impact analyst coverage and forecasts? Accounting and Business Research, 1–33. [CrossRef]
Hill, Hal. 2000. The Indonesian Economy. Cambridge: Cambridge University Press.
Hofman, Bert, Ella Rodrick-Jones, and Kian Wie Thee. 2004. Indonesia: Rapid growth, weak institutions.
Paper presented at World Bank Shanghai Conference, Shanghai, China, May 25–27. Available online:
http://www.worldbank.org/wbi/reducingpoverty/case-Indonesia-PovertyReduction.html (accessed on 15
April 2019).
Houston, Joel F., Liangliang Jiang, Chen Lin, and Yue Ma. 2014. Political connections and the cost of bank loans.
Journal of Accounting Research 52: 193–243. [CrossRef]
Hu, Guoliu, and Yu Wang. 2018. Political connections and stock price crash risk: The role of intermediary
information disclosure. China Finance Review International 8: 140–57. [CrossRef]
Hutton, Amy P., Alan J. Marcus, and Hassan Tehranian. 2009. Opaque financial reports, R2, and crash risk. Journal
of Financial Economics 94: 67–86. [CrossRef]
Iacus, Stefano M., Gary King, and Giuseppe Porro. 2012. Causal inference without balance checking: Coarsened
exact matching. Political analysis 20: 1–24. [CrossRef]
Kim, Jeong-Bon, Yinghua Li, and Liandong Zhang. 2011. Corporate tax avoidance and stock price crash risk:
Firm-level analysis. Journal of Financial Economics 100: 639–62. [CrossRef]
Int. J. Financial Stud. 2019, 7, 49 16 of 16

Kim, Jeong-Bon, Zheng Wang, and Liandong Zhang. 2016. CEO overconfidence and stock price crash risk.
Contemporary Accounting Research 33: 1720–49. [CrossRef]
Lee, Ming-Te. 2016. Corporate Social Responsibility and Stock Price Crash Risk: Evidence from an Asian Emerging
Market. Managerial Finance 42. [CrossRef]
Lee, Wei, and Lihong Wang. 2017. Do political connections affect stock price crash risk? Firm-level evidence from
China. Review of Quantitative Finance and Accounting 48: 643–76. [CrossRef]
Leuz, Christian, and Felix Oberholzer-Gee. 2006. Political relationships, global financing, and corporate
transparency: Evidence from Indonesia. Journal of Financial Economics 81: 411–39. [CrossRef]
Luo, Donglun, and Liming Zhen. 2008. Private control, political relationship and financing constrain of private
listed enterprises. Journal of Financial Research 12: 164–78.
Luo, Jin-hui, Manning Gong, Yilong Lin, and Qifeng Fang. 2016. Political connections and stock price crash risk:
Evidence from China. Economics Letters 147: 90–92. [CrossRef]
McLeod, Ross H. 2000. Soeharto’s Indonesia: A Better Class of Corruption. Agenda: A Journal of Policy Analysis and
Reform, 99–112. [CrossRef]
Pangestu, Mari, and Manggi Habir. 2002. The Boom, Bust and Restructuring of Indonesian Banks. IMF Working Paper
WP/02/66. Washington: International Monetary Fund.
Poczter, Sharon, and Thomas B. Pepinsky. 2016. Authoritarian Legacies in Post–New Order Indonesia: Evidence
from a New Dataset. Journal Bulletin of Indonesian Economic Studies 52: 77–100. [CrossRef]
Roberts, Brian E. 1990. A dead senator tells no lies: Seniority and the distribution of federal benefits. American
Journal of Political Science, 31–58. [CrossRef]
Sato, Yuri. 2004. Corporate Ownership and Management in Business in Indonesia: New Challenges, Old Problems.
Singapore: Institute of South East Asian Studies.
Schwarz, A. 1994. A Nation in Waiting: Indonesia in the 1990s. St. Leonards: Allen & Unwin.
Tee, Chwee Ming. 2018. Political connections and stock price crash risk: Evidence of institutional investors’
heterogeneous monitoring. Spanish Journal of Finance and Accounting/Revista Española de Financiación y
Contabilidad, 1–18. [CrossRef]
Winter, Jeffrey A. 2013. Oligarchy and Demoracy in Indonesia. Indonesia 96: 11–33. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).