Does political connectedness determine stock returns in Indonesia?

This study aims to investigate the influences of political connection on stock returns in Indonesia. We develop a comprehensive database of firm-level political connectedness among Indonesian firms from 2010 to 2017. Our sample is non-financial Indonesian listed firms that are selected in the Kompas 100 index for 16 consecutive periods, with a total of 448 firm-year observations. This study employs panel data regressions to estimate this relationship, then mitigate possible endogeneity issues using two-stage least square with fixed-effects. The finding of this study shows that political connectedness is associated with lower stock returns, more prominently in agriculture and consumer goods industries. Moreover, state-owned enterprises are more likely to earn lower stock returns. In summary, our result suggests that investing in politically connected firms could be a risky investment. The finding holds using alternative estimation methods


INTRODUCTION
The financial crisis has significant impacts on economic conditions. For example, the 2008 Global Financial Crisis affects most major economies in Asia and Europe, no exception in Indonesia. This crisis affects not only the housing market, but also economic productivity, unemployment, and asset prices. As explained by Reinhart & Rogoff (2009) in The Aftermath of Financial Crises, a sharp inclination of default and declined in output production growth occurred in emerging countries, for two consecutive years after crises.
Particularly in Indonesia, the economic growth drops around 4% during the crisis, as exports weakened and the downturn in security prices (Bank Indonesia, 2009). Due to lots of uncertainty, foreign capital outflows in the Indonesian stock market sharply increase as the investors reactively move out their capital to less risky places.
This sound effects of the financial crisis in real economy sectors and equity market stimulating the investors to find a way to predict the market so that they able to mitigate capital loss. This study aims to help the investors by investigating the determinant of stock returns in Indonesia, using the least explored capital, namely political connectedness. We investigate this issue because the political risk in Indonesia highly intervenes stock market returns (Amtiran & Indiastuti, 2017). Previous study suggests that having a good connection with politicians is a valuable capital for the firms (Ling, Zhou, Liang, Song, & Zeng, 2016).
We formulate the estimation models by developing testable political capital hypotheses used in Civilize & Young (2015) and Hahn & Lee (2009). We extend the study using more comprehensive measures of political connectedness, examining the connection of each board of director and board of commissioner. The evidence of this study is robust under different estimation methods and models.
We examine this linkage using Indonesian data. The reasons are; first, Indonesia has failed to impress Transparency International organization, due to the constant high level of corruption occurred in this country. Currently, Indonesia corruption index ranks number 90 out a total of 181 countries (Transparency International, 2018). Due to the fact that the sentiment is negative in the high level of corruption countries (Bathia & Goyal, 2013), the risk of investing in Indonesian stock markets is more pronounced. Therefore, we want to provide suggestions to investors in such a high-risk market setting. Second, according to Faccio (2006), political connection ubiquitously exploited in the country with high level of corruption, foreign investment restrictions and more transparent system. In fact, Indonesia fulfills those criteria, indicating the importance role of political connection in this country. Therefore, we attempt to address the influence of political ties on stock returns in Indonesia. As best to author's knowledge, this is the first study that examines this linkage using Indonesian data for a period of 2010 to 2017.
Research examining the determinant of stock returns in Indonesia suggest that macroeconomic conditions influence stock market performances. Inflation, exchange rate, interest rate, and bond yields affect stock returns (Defrizal, Sucherly, Wirasasmita, & Nidar, 2015). Furthermore, unique firm characteristics also play an important role in stock market performances. In the traditional model, stock returns are explained by these firm characteristics, i.e., firm size, book-to-market, debt ratio, and E/P ratio (Fama & French, 1992). Prior studies found that in Indonesia, firm size and debt ratio have a positive association with stock returns (Fauzi & Wahyudi, 2016). Also, liquidity ratio and market ratio have negative effects on stock returns in Indonesian markets (Fauzi & Wahyudi, 2016;Martani, Mulyono & Khairurizka, 2009). Therefore, we account for these firm characteristics in our estimation model.
Without neglecting these factors, the impact of political connectedness on stock returns has been tested using various methods (Chen, Ariff, Hassan, & Mohamad, 2014;Ferris, Houston, & Javakhadze, 2016;Fisman, 2001;Leuz & Oberholzer-Gee, 2006;Wu, Wu, & Rui, 2012 among others). Fisman (2001) uses an event study analysis to estimate the value of political connection in Indonesia. He found that the stock returns of politically connected firms highly depend on politicians' performances. Hahn & Lee (2009) find positive influences of political connection on stock returns because such firms are more likely to be assisted in bank loans access and regulation bureaucracies. However, Ling, Zhou, Liang, Song, & Zeng (2016) argue that the existence of politicians in the board members increase firm investment risks as they tend to overinvest the assets, which caused a sharp decline in firm performances. Chen, Li, Su, & Sun (2011) explain that politicians often utilize their power for rent-seeking, consistent with Shleifer & Vishny (1994) that suggest politicians focally point their best interests. Therefore, politically connected firms are less efficient, as government distorts firms' investment behaviour (S. Chen, Sun, Tang, & Wu, 2011).
Also, politically connected firms are associated with poorer corporate governance, due to lower quality information disclosure that caused by ineffectiveness of internal and external monitoring (Chaney, Faccio, & Parsley, 2011). With this regard, the markets punish political stocks. Using the sample of India, Ghosh (2011) shows that firms with political connection earn lower stock returns compared to non-connected firms. Moreover, the liquidity of stocks is also disrupted (Ding, 2014). Hence, we develop the hypothesis as follow.

H1: Political connection has a negative impact on stock returns
This study contributes to the discussion regarding stock returns determinants, by reporting a significant association between political connectedness and stock returns (Addoum & Kumar, 2016;Civilize & Young, 2015;Fama & French, 1992;Hahn & Lee, 2009 among others). Our study also accompanies the discussion regarding political connection influences on firm value and performances (Cao, Huang, Liu, & Tian, 2012;Fisman, 2001;Su & Fung, 2013;Wu, Wu, & Rui, 2012 among others). Also, our finding has practical contributions. By acknowledging the factors that determine stock returns, the investors able to formulate better investment decisions. Our study helps them in avoiding capital loss.
The structure of this paper is organized as follows. Section 1 is introduction, which describes the background of this research. Section 2 explains the data collection and research methods. Section 3 presents the result of the data analysis. Last, section 4 concludes the paper.

METHODS
The population of this research is all listed firms in Indonesian Stock Exchange from 2010 to 2017. We employ a purposive sampling method and use Indonesian listed firms that are selected in the Kompas 100 index for 16 consecutive periods, from 2010 to 2017 with total observations of 448 firm-year data. We exclude financial firms due to different behaviour and financial reporting. Stock returns and firm characteristics data are obtained from The Indonesian Capital Market Institute (TICMI) databases. For political connection, we hand-collect the data from boards' resumes disclosed in the annual reports and firms' websites. Last, we collect Indonesian 10-Year bond yields from investing.com databases. This study uses panel data regressions to investigate the effect of political connection on stock returns, adopting the employed model in Civilize & Young (2015) as follow: Where denotes the log return of firm i year t. denotes political connectedness of firm i year t, consists of polscoreit for political connection score of firm i year t, pSOEit for state-owned enterprises dummy of firm i year t, and pNSOEit for politically connected non-state owned enterprises of firm i year t. To calculate political connection score (polscore), we follow Boubakri, Cosset, & Saffar (2008) that calculates the ratio between connected board members to total board members on the firm. We define connected board members following Faccio (2006), if one of its top executives is a member/former of parliament, military, ministry, regent, or have served governmental organizations. White (1980) robust standard error is employed to account heteroskedasticity, serial correlation, and non-normality dispersions that may violate the estimation results.
Since stock risks are multidimensional, we adopt Fama & French (1992)'s multivariate analysis to determine the explanatory factors of stock returns. The following model is employed to test the effect of political connection on stock returns, with several control variables. = 0 + 1 + 2 + 3 + 4 + 5 + 6 + 7 + …………………….………… (2) Where denotes the log return of firm i at year t. Polscore denotes political connectedness of firm i at year t. Size is the natural logarithm of total assets of firm i at year t, to proxy firm size. BTM is book value of equity to market capitalizations of firm i at year t, to account for firm's growth opportunity. Solvency denotes total assets to total liabilities of firm i at year t, to account for solvency. LiqRatio denotes current assets to current liabilities of firm i at year t, to account firm's liquidity. BEP is operating profits to total assets of firm i at year t, to account basic earning power ratio. Last, BondYields denotes Indonesian 10-Year bond yields, to account for undiversifiable risks occurred in the market.
In the first part of our analysis, we use pooled regressions to estimate the models using industry and time controls to account each industry's unique feature and change in economic conditions. To check the robustness, in the second part of our analysis, we use panel data fixed effect regressions and two-stage least square regressions, to account endogeneity problem that may exist in our models. This work uses ratio of politically connected board of directors to total board of directors of firm i at year t (pBOD) as an alternative instrument of polscore in 2SLS regressions. Table 1 reports summary statistics for political connection variables. Panel A provides a proportion of politically connected firms by industry, from a sample of Indonesian listed firms that are selected in Kompas 100 index for 16 consecutive periods, from 2010 to 2017. Panel B reports total observations, mean, standard deviation, minimum value and maximum value of political connection variables used in this research from the total sample.  Table 1, we find 63% of Indonesian listed firms that are selected in Kompas 100 index for eight consecutive years from 2010 to 2017 are connected to politicians or government officials. Moreover, Basic Industry and Chemicals dominate political connection data set as 80% of the sample from this industry is politically connected. In fact, Basic Industry and Chemicals is the top contributor to Indonesian unemployment and GDP (Ministry of Industry Republic of Indonesia, 2012). Therefore, this industry is more controlled by politicians or government officials to ensure economic stability.

Summary Statistics
Panel B shows the summary statistics of all political connection variables from the total sample. The mean of political connection score (polscore) on our sample is 48.44%, meaning that, on average, the board members or a majority of the shareholders of the firms in our sample are connected to the politicians or government officials, with standard deviations of 0.6601. Also, in average, 2.04% of the board of directors of the firms in our sample are politically connected with maximum score of 28.57%.
Furthermore, we provide summary statistics of firm characteristic variables in Table  2. Panel A reports the total observations, mean, standard deviations, minimum value, and maximum value for total sample. To understand whether the characteristics of politically connected and non-connected firms are different, we provide the summary statistics for a sample of politically connected firms in Panel B and summary statistics of non-politically connected firms in Panel C.  Table 2, we see that the firm size of non-connected firms is higher than connected firms and total sample. The average value of Size for non-connected firms is 14.923 (deviation of 9.028), while Size of connected firms and total sample are 14.787 (deviation of 8.809) and 14.838 (deviation of 8.882), respectively. For BTM, we find that the ratio between book value of equity to market value of equity for connected firms are higher. The average value of BTM for connected firms is 74.6% (deviation of 0.907), while for non-connected and total sample are 62.3% (deviation of 0.57) and 69.9% (deviation of 0.797), respectively. It means that non-connected firms are relatively overpriced compared to connected firms, indicating lower market growth opportunity.
Furthermore, the ratio of total assets to total liabilities (Solvency) of non-connected firms is higher. The average value of Solvency for non-connected firms is 1.798 (deviation of 1.564), meaning that such firms are relatively more solvent due to lower debt ratio. On the other hand, connected firms own higher debt ratio because such firms have more access to finance, consequently it becomes less solvent (Claessens, Feijen, & Laeven, 2008;Yeh, Shu, & Chiu, 2013). Similarly, current ratio (LiqRatio) of non-connected firms is higher because the curent liabilities of non-connected firms is lower.
Interestingly, connected firms have less basic earning power because the ratio between operating profits to total assets (BEP) is lower. The average value of BEP on connected firms is 10.7% (deviation of 0.125), while BEP on non-connected firms is 12.3% (deviation of 0.130). It means that connected firms are less efficient in utilising the total assets compared to non-connected firms. This inefficiency may occur due to rentseeking activities from government officials that often occurred on connected firms Chen, Li, Su & Sun 2011). After the data distribution of variables used in this study is identified, we conduct a Pairwise Correlation test to ensure that our estimation models examining the influence of political connection on stock returns contain no multicollinearity problems. To provide unbiased estimation results, we use independent variables that correlated below 60% (Huang & Zhao, 2016). The result of Pairwise Correlation tests presented in Table 3 shows that our models contain no multicollinearity problems because the correlation matrix does not exceed 60%. After that, we conduct Breusch-Pagan Lagrange Multiplier, and Haussmann tests to decide the estimator of our models. The result shows that pooled OLS regressions fit our data set because it meets the best linear unbiased estimator (BLUE) conditions.

Estimation results
We provide the results of our estimation models in Table 4, using pooled OLS regressions. Panel A shows the estimated coefficient of political connection score (polscore) to stock returns. In Panel B, following Lin, Tan, Zhao, & Karim (2015), we separate the political connectedness under different types of ownership, i.e. state-owned enterprises (pSOE) and non-state-owned enterprises (pNSOE) to understand where the value of political connectedness occurred. We also provide the result of two sample meancomparison test to ensure that the independent variables used in the model are statistically different, in order to avoid multicollinearity issue.  Table 4, stock returns for politically connected firms are significantly lower than non-connected firms. The polscore coefficient suggests that when firms tying closer connection to the politicians or government official, the average stock returns of the firms decrease 2.3% per year (statistically significant at 10%). Another significant variable is state-owned enterprises dummy (pSOE) that shows such firms earn lower average stock return of 4.1% per year (statistically significant at 10%) compared to average stock returns of non-SOEs. Previous study shows that the officials intervene connected firms for rent-seeking, resulting in lower firm performances (Acemoglu & Johnson, 2014). This exploitation is cutting the profit pie that initially distributed to the investors. Indeed, our result confirms the evidence in China that shows political rent-seeking is associated with negative stock returns (Fan, Rui, & Zhao, 2008).
To ensure that the evidence is not caused by other factors that also explain stock returns in Indonesia, we further run the regressions with control variables. We use pooled OLS regressions to estimate the relationship, using the robust standard error to account heterokesticity and non-normality dispersions (White, 1980). We run three models to estimate this relationship; model 1 does not account industry and time effects. Model 2 accounts time effect, and model 3 accounts industry effect. Other variables used are similar, as presented in Table 5. After controlling for firm characteristics, the results confirm the negative relationship between political connection and stock returns in Indonesia. Table 5 shows political connectedness (polscore), firm size (Size), book-to-market (BTM), basic earning power (BEP), and bond yields (BondYields) determine stock returns in Indonesian market. In model 1, we see that one-unit standard deviation of political connection score decrease 2.8% average stock returns (statistically significant at 10%). Furthermore, when the variance of time effects is controlled, the magnitude of political connectedness influence on stock returns decrease 1% to 1.8%. However, the estimated coefficient of political connectedness (polscore) in model 3 become insignificant when we control for industry effect. Therefore, we further estimate this association within each industry to see where this effect is mediated.
For the influence of firm characteristics on stock returns, Table 5 shows larger firms tend to gain higher stock returns per year. We find one-unit increase in firm size increase 0.7% stock returns per year. We also find negative relationship between book-to-market (BTM) to stock returns, where one-unit increase in BTM decreases 4.8% stock returns. The result consistents with Hahn & Lee (2009) and Fama & French (1992) that found financially unconstrained firms (proxied by Size and BTM) have been predicted earning higher stock returns.
Similarly, the ratio between operating income to total assets (BEP) have positive and significant association with stock returns, consistent with Civilize & Young (2015). We find one-unit change in basic earning power increases 22.3% of stock returns per year. Moreover, Indonesian 10-Year Bond Yields is negatively associated with stock returns since bond is naturally an alternative investment choice for stock market. We see oneunit change in BondYields decreases 10.9% stock returns, and this finding is consistent with Glascock, Lu, & So (2000).
After the effects of political connectedness and firm characteristics are examined, we further analyze this association within each industry to understand the magnitude of this impact within each sector. We use fixed effect panel data regressions to estimate the models, with time effects to account for the variance of economic change conditions across time. The estimation results are reported in Table 6. From table 6, we see that political connection influences occurred on specific industries. This study finds political connectedness in Indonesia matters in Agriculture and Consumer Goods sectors. The estimated coefficient of political connectedness (polscore) is negative and significant in Model 1, meaning that one-unit change of political connection score of firms in Agriculture industry decrease 23.3% its average stock returns. Similarly, the estimated coefficient of political connection score (polscore) in model 3 is also negative and significant. For one-unit change in political connection score of firms in Consumer Goods industry, decreases 144.6% stock returns per year. To deliver robust evidence for the link between political connection and stock returns, we estimate the models using different methods. First, using fixed effect panel data regression to account for inconsistency of our data set. Second, using two-stage least square with fixed effects, the models account for endogeneity issue that may violate our regressions. We use political connection score of board of directors (pBOD) as an alternative measure of polscore. The result is presented in Table 7. Table 7 shows consistent results for the link between political connectedness and stock returns. In model 1, we see that one-unit change of political connection score (polscore) decreases 25.2% stock returns per year. Similarly, using instrument variable of political connectedness of firm's board of directors, we find 56.6% average stock returns decrease as if board of directors become more connected to political party. In summary, it can be concluded that investors in Indonesia see political connectedness as a negative value.
For control variables, the result consistents with our main models. Larger firm (Size), low book-to-market (BTM), and high earning ratio (BEP) are positively influence stock returns. Furthermore, using fixed effect panel data regression and two-stage least square regression, we find other characteristics determine stock returns in Indonesia. Firm with higher assets to liabilities ratio (Solvency) is associated with positive returns because such firms are less likely to be financially distressed. In summary, this work exploits an alternative determinant of stock returns in Indonesia using the least discussed factor, namely political connectedness. We find political connectedness determine stock returns, more prominently in Agriculture and Consumer Goods industries. The association between political connectedness and stock returns is negative due to rent-seeking behavior that is perceived negatively by the investors (Chen, Li, Su & Sun, 2011). Rent-seeking behavior that commonly occurred in connected firms decrease firm performances and wealth distribution, hence the investors in Indonesia prefer to avoid these firms.
The finding of this study suggests that the presence of politicians in the board members cause negative sentiment in Indonesian stock markets. The politicians may exploit the company's profit for 2014 general election campaign and political party operations. Therefore, when politicians or government officials enter the company as the board member, the investors perceive that such agents will confiscate their wealth. With this regard, the stock price of such firms will befall. Our finding explains this phenomenon by demonstrating political connectedness statistically significant decreasing firm's stock returns using various methods.

Conclusions
This study reports political connectedness determines stock returns in Indonesia. We find the existence of a politically connected board of directors or board of commissioners influence negatively to stock returns, more prominently in Agriculture and Consumer Goods sectors. Also, state-owned enterprises are more likely to gain lower returns compared to privately-owned enterprises. Prior literature explains that politicians often exploit the profit earned for rent-seeking that cause investors distrust. This study also find larger firms, low book-to-market, more solvent, more earning power, and liquid firms are more likely to gain high stock returns. The results hold under different alternative methods.

Recommendations
This study has several limitations. First, political connection data is compiled using all available information disclosed in boards' resume, and neglect unobserved aspect of connection, such as personal relationship with the political party that might also important on explaining political connection power. Second, this study does not examine the link between political connection and rent-seeking directly. Therefore, we suggest further research to test the direct relationship between political connection and rent-seeking in Indonesia, to deliver empirical explanations regarding negative effects of political connection on stock returns.
Nevertheless, the finding of this study is important for investors. By understanding the determinants of stock returns presented in this research, we help the investors on optimizing their investment decisions. Our results suggest the investors consider political connectedness of the board members, prominently in Agriculture and Consumer Goods sectors.