What Role Does Political Environment Play? A Case Study of China

Transcription

1 What Role Does Political Environment Play? A Case Study of China Jingwen Yu (China Center for Economic Research, Peking University, ) Abstract: The uncertainty induced by the political environment affects the investment risk, and thus affects the investment decision which has a close relationship with the economic development. This paper investigates the economic cost of political environment, using the intension of cross-strait relation in China as a study case by synthetic control method. The intense situation of the cross-strait relation has great influence on the economic development of Fujian which is the closest province near Taiwan and also possesses the greatest preferential policies for Taiwan direct investment. The empirical result reveals that because of the intense situation, the loss of GDP per capita in Fujian has accumulated to yuan since 2001, which means that GDP per capita in Fujian declines about 12.1 percentage points annually relative to a synthetic control region without change of political environment. Keywords: Political Environment; Economic Development; Synthetic Control Method; Cross- Strait Relations JEL: E65; O12; P20 1. Introduction Political environment plays an important role in the economic development. Alesina and Perotte (1996) argue that the politico-economic environment reduces investment because of the huge investment risk. The investment is a primary engine of growth, and thus the political environment has an intimate relationship with economic growth. Svensson (1998), Devereux and Wen (1998) investigate the multinational differences in investment rate and establish a theoretical model explaining the mechanism of the influence of political instability on investment rate. Jong-A-Pin (2009) also examines multidimensionality of political instability by 25 indicators. The indicators are divided into four categories: politically motivated violence, mass civil protest, instability within the political regime and instability of the political regime. The generalized method of moments model is used to deal with the endogenous problems. Even though, it is quite difficult to estimate the economic cost of political factor because of the counterfactual essence in this situation. We cannot find the case of favorable political environment when the de-facto adverse political environment happens, and vice verse. Fortunately, the synthetic control method emerged recently in the comparative case study can overcome this difficulty 1. The data-driven property of synthetic control method guarantees that the synthetized control group can represent the treatment group. Abadie and Gardeazabal (2003), Abadie et al. (2010) come up with a new method, named synthetic control method to infer the causal effect in economics. Abadie and Gardeazabal (2003) study the terrorism led by Euskadi Ta Askatasuna (ETA), one terrorist organization in Basque of Spain. In their paper, they propose a new method in comparative case study, synthetic control method for the first time. The new method is applied to find the weights for the other two areas in Spain. These areas can be Electronic copy available at:

2 synthetized to a new control group for Basque by the obtained weights. This can overcome the identification problem. The estimated result illustrates that in the later period of 1960s, the GDP in Basque has been dropped by 10 percentage points due to the terrorist attack. Moreover, once ETA stops terrorist attack, the stock prices of enterprises located in Basque performs better than before, which confirms the economic costs generated by the conflicts. Abadie et al. (2010) use the synthetic control method to research on the effect of California s Proposition 99 implemented in 1988 on the tobacco consumption habits of residents in California. They find that by the year 2000, annual per capita cigarette sales in California are 26 packs lower than what California would have in the absence of Proposition 99. In this case, all the other states are synthetized as a control group for California. Because of the advantages of synthetic control method in the comparative case study, this paper will use it to study the impact of cross-strait intension in China after 2001 on economic development. As far as we know, Taiwan's democratic progressive party has elected to be the ruling party of Taiwan in 2000, which has a significantly negative influence on the cross-strait relations. Because Taiwan s democratic progressive party is willing to make Taiwan an independent country, in many circumstances, Taiwan s democratic progressive party advocates that Taiwan is an independent sovereign country and declares that it is necessary for referendum legislation to determine the future and the fate of Taiwan. This definitely breaks through the bottom line of China. During this period, the cross-strait relations come to a deadlock. We choose the beginning year 2001 of governing period by Taiwan s democratic progressive party as a turning point and year 2008 that Kuomintang (KMT) elected to be the ruling party as an ending point. We believe that the political environment deteriorates during this period and thus lead to a change in perceived levels of investment risk, which is quite damaging for economic development of Fujian. The object of this study focuses on Fujian province, which has the closest relationship with Taiwan not only in the geographical feature but also in the economic links. For example, Xiamen city of Fujian province has been set as special economic zone (SEZ) by central government in This special economic zone aims to attract the investment from Taiwan. The other three cities, Shenzhen, Shantou and Zhuhai, are also set up as special economic zones at the same time. Shenzhen and Shantou have the purpose of attracting the investment from Hong Kong and Zhuhai has the purpose of attracting the investment from Macao. Figure 1 reveals the fact that Fujian has closer relationship with Taiwan. The ratio of Taiwanese direct investment to FDI in Fujian is significantly greater than that in the nationwide since the 1990s, even there are some fluctuations. We believe that the intension of cross-strait relations will affect Fujian rather than others. As a matter of fact, in the middle period of 1990s, Denghui Li, president of Kuomintang also renounced the one-china policy, which leads to the intension of cross-strait relations. For example, the Chinese people's liberation army holds the military maneuver near Taiwan Strait in 1995 and When 1996 is set to be the turning point, we only find the gap between FDI of Fujian and synthetic FDI of Fujian. There is no gap between GDP per capita of Fujian and synthetic GDP per capita of Fujian. There are two reasons. Firstly, Kuomintang is in support of the one-china policy in principle. On the contrary, Taiwan's democratic progressive party is against the one-china policy. The Constitution of democratic progressive party contains the goal of Taiwan independence. As a result, the event that Taiwan s democratic progressive party is Electronic copy available at:

3 elected to be the ruling party will have more negative effects on political environment. Secondly, during the middle period of 1990s, the scale of FDI is relative small, which has limited impact on the regional economic development. The positive effects of the preferential policy and its own potential economic growth counteract the negative effect of the decline in FDI. Therefore, we observe the decline in FDI but not in GDP per capita. The purpose of this paper is to utilize a quantitative method to estimate the economic cost of cross-strait tension. In current literature, there are few researches about the political environment of China and its influence on the economic development. The difficulty is how to distinguish the influences of the political environment on development of Fujian and the other influences of the essential heterogeneities owning by Fujian on economic development. The simple comparison of Fujian province and the other provinces may be misleading. On that account, this paper uses a new method in comparative case study, synthetic control method to analyze the impact of political environment on economic development of Fujian. Moreover, this paper tries to estimate the economic cost of such impact. Besides Abadie and Gardeazabal (2003), Enders and Sandler (1996) use Vector-Autoregression Model and Impulse Response Function to analyze the influence of terrorist activities on foreign direct investment in Spain and Greece. They found that the foreign direct investment decrease 13.5 percents since the happening of terrorist activities and the foreign direct investment decreases 11.9 percents due to the terrorist activities. Abadie and Dermisi (2008) employ the building-level data to study the vacancy rates in Chicago after September 11, The 9/11 attacks induces large increase in the perception of terrorism risk in Chicago Central Business District, including Sears Tower, Aon Center and the Hancock Center. Their results suggest that the economic activity in Central Business Districts can be affected negatively by the increase of terrorism risk. However, there is little evidence about China case. As far as I know, this is the first paper making an attempt to quantify the economic cost of political environment in China. The paper is organized as follows. Section 2 introduces the basic idea of synthetic control method. Section 3 uses the provincial data of China to analyze the influence of political environment on economic development. Section 4 tries to carry the placebo test for the preliminary results in the previously empirical study. Section 5 concludes. 2. Introduction of Synthetic Control Method During the period of cross-strait tension in China, Fujian is viewed as treatment group because of the intimate relationship with Taiwan. We need a control group substituting for Fujian after accepting treatment in order to estimate the influence of political environment on the economic development. However, the choice of suitable control group is also a difficulty in current research. Abadie and Gardeazabal (2003), Abadie et al. (2010) propose synthetic control method to deal with this difficulty. They use synthetic control method to study the economic cost of terrorism happened in Basque and the effect of the Proposition 99 in California respectively. The essential characteristic of synthetic control method is that the choice of control group depends on the existing data completely. By comparing the potential control groups and the treatment group before treatment, the weights of these potential control groups can be obtained under certain rules. Later on, these weights will be used to synthetize a new control group that can predict the performance of the treatment group in the case that it does not accept treatment. The next, we will introduce the basic idea of synthetic control method. Electronic copy available at:

4 Yit Suppose be the outcome that would be observed for region i at time t without treatment and I Yit be the outcome that would be observed for region i at time t with treatment, where i1,2, J 1, t 1,2, T. Region 1 receives the treatment at time T0 and ends at time T, where 1T0 T. The model can be specified as the following: Y Y D it it it it D it 1 if i 1 and t T 0 otherwise 0 When Dit equals 1, region i receives the treatment at time t. When Dit equals 0, region i does not receive the treatment at time t. Because only region 1 receives treatment, our purpose is to estimate the value of 1t when t T0. However, we cannot observe the value of Y1 t for t T0, I because we only observe the value of Y 1t, and thus Y1 t is a counterfactual variable here. Hence, the key point of comparative case study is to construct the counterfactual variable and estimate it. Let s suppose that Yit is given by the factor model: Y Z it t t i t i it Where t is an unknown common factor with constant factor loading across unit, Zi is a vector of control variables that are seldom affected by the treatment, t is a vector of parameters needed to be estimated, t is common factor with varying factor loadings, i across units, and it is the error term that standing for the unobserved transitory shocks at the region level with 0 mean for all i. W ( w The purpose here is to estimate a vector of weights 2,, wj 1)' wj 0 that for wj 1 j2,, J 1 and j1 W ( w. In fact, each value of vector 2,, wj 1)' can synthetize to a potential control group, that is, a particular weighted average of control regions. The value of the W ( w outcome variable for each synthetic control indexed by 2,, wj 1)' is as following: Suppose there is a vector of J 1 J 1 J 1 J 1 w Y w Z w w j jt t t j j t j j j jt j2 j2 j2 j2 (,, ) * * w2 wj 1 such that 2 : J1 J1 * * wjyj1 Y11,, wjyjt Y 0 1T0 j2 j2 (1) (2) (3) (4)

5 J 1 j2 w Z * j j Z 1 (5) If T0 t1 ' t t is nonsingular, then the following equation holds. J 1 J 1 T0 T0 J 1 * * ' ' * 1t j jt jt t t s js 1 s j js 1 s j2 j2 s1 n1 j2 Y w Y w ( ) ( ) w ( ) Abadie et al. (2010) has proven that under the general condition, the right hand side of equation J 1 * wy j jt j2 (6) will approach to zero. As a result, is the unbiased estimator of Y 1t. When t T0, J 1 * wy j jt Y1 t cannot be observed, j2 could be used to estimate Y 1t. The treatment effect in the time t can be estimated, where t { T0 1,, T) 3 : (6) J 1 * 1t Y1 t wjyjt j2 (7) 3. Empirical Studies on Economic Cost of Political Environment (1) Data Description This paper uses the provincial data in China from Comprehensive Statistical Data and Materials on 60 Years of ew China except Chongqing that is a new municipality after The dependent variable using to describe the economic development is GDP per capita adjusted by the price of The predicted variables include the ratio of fixed investment to GDP, the ratio of industrial output to GDP, the ratio of building industrial output to GDP, the ratio of farming forestry, animal husbandry and fishery output to GDP and the ratio of labor force to total population. Because the uncertainty of the economic reform, we choose the data from 1990 to Specifically, before 1990s, the economic reform of China is at the stage of destroy the old, the pilot projects are the dominating feature in this period. Moreover, the market economy system has not been set up yet. After 1990s, the economic reform of China is at the stage of establish the new, the goal of the reform is clear, and thus the economic performance is much better than before. Consequently, the predicted variables before 1990s and that after 1990s may have great discrepancy. Y The data from 1990 to 2000 will be used to predict the dependant variable it in equation (2). Because there are several critical events happened during this period, we choose the GDP per capita in 1992, the GDP per capita in 1994 and the GDP per capita in 1997 as predicted variables, which represent the events of Deng Xiaoping's southern tour, the integration of two-track exchange rate and the Asian financial crisis respectively. Table 1 illustrates that the differences between the predicted variables of Fujian and that of

6 synthetic Fujian are all smaller than the difference between the predicted variables of the Fujian province and that of the nationwide. This implies the weights from synthetic control method can match the characteristic of Fujian province much well, and thus can used to predict the economic development of Fujian province in the period of post cross-strait tension. (2) Empirical Result The preliminary result of the synthetic control method can be seen in graph 2. The potential control groups include 29 provinces in China. The weighs of each province are shown in table 2, among which, Jiangsu has the largest weight which is Shandong has the weight of and Tianjin has the weight of The others are 0. This implies that the weighted average of Jiangsu, Tianjin and Shandong can well fit the economic development of Fujian province. This combination can be used as a control group for Fujian without the change of political environment. Figure 2 shows that before 2001, the synthetic data fits quite well. However, after 2001, the gap between GDP per capita of Fujian and that of the synthetic Fujian emerges. The GDP per capita of synthetic Fujian is larger than that of Fujian. Furthermore, this gap shows an increasing trend, which means that the change of political environment due to cross-strait tension in China has an adverse effect on the economic development of Fujian. In order to observe the influence of cross-strait tension in China more intuitively, we calculate the gap between the GDP per capita and synthetic one. Figure 3 shows the gap changing with time. Before 2001, this gap changes little. However, after 2001, this gap drops dramatically. The gap between the GDP per capita and synthetic one drops from yuan in 2000 to yuan in After 2001, this gap keeps negative. In 2008, this gap becomes yuan and the average loss of GDP per capita is yuan, 12.1% lower than synthetic GDP per capita on average. The cross-strait tension increases the investment risk in Fujian province and lowers the predicted investment return, and thus goes against the economic development of Fujian. 4. Placebo Test (1) Change of Treatment Group The robust check in synthetic control method is also called placebo test. The previously empirical result reveals a gap between the GDP per capita of Fujian and that of synthetic Fujian. Is this gap really caused by the intension of cross-strait relations in China or just by chance? To some extent, the robust check will answer this question. The robust check in synthetic control method originates from Abadie and Gardeazabal (2003), Bertrand et al. (2004) and Abadie et al. (2010). The basic idea is that if the gap between GDP per capita of Fujian and that of synthetic Fujian indeed reveals the economic cost caused by the change of political environment, then gap between GDP per capita of the other provinces and the synthetic will not show the same pattern when they are viewed as treatment group. This is logical, because Fujian is affected most by the cross-strait tension. We will take three robust checks following the procedures of Abadie et al. (2010). Because in the previous analysis, Jiangsu has the largest weights, which implies Jiangsu province has similar characteristics with Fujian province, we will choose Jiangsu as treatment group and the other provinces as potential control groups. We anticipate that the GDP per capita of synthetic Jiangsu will has less change compared with that of synthetic Fujian. However, if the GDP per capita of Jiangsu have the same pattern as Fujian province, then the gap between GDP per capita of Fujian province and that of synthetic Fujian province is not caused by intension of

7 cross-strait relations but by the other unobserved variables. (2)Change of Potential Control Groups The next, we use different potential control groups by the consideration that the previous potential control groups may not capture the main feature of Fujian. Since Fujian is an eastern province, we only choose eastern provinces as potential control groups which have similar features with Fujian. Figure 5 is the estimated result which is quite similar to the preliminary estimated result. Before 2001, the GDP per capita of synthetic Fujian fit the GDP per capita of Fujian quite well. After 2001, the gap emerges implying the negative influence of cross-strait tension on economic development of Fujian province. In this case, Jiangsu province also has the largest weight Shangdong province has the weight of and Hebei province has the smallest weight (3) Permutation Test Finally, we take a permutation test which is similar to rank test in statistics. The aim here is to test the statistical significance of our result. Are there any other provinces shows a gap between GDP per capital and synthetic GDP per capita when these provinces are viewed as treatment group respectively? What is the probability of appearing the same large gap as Fujian province does? To assess the significance of our estimates, we iteratively apply the synthetic control method to estimate the effect of cross-strait tension on every other province. Before doing our test, we need exclude the provinces that do not fit the original data before 2001 well. For example, the province with the worst fit is Guangdong which has the largest mean squared prediction error (MSPE) If the fit is not well enough, then the gap after 2001 may reflect not only the effect of the cross-strait tension but also the unfitted feature. Therefore, we exclude the provinces whose MSPE before 2011 are larger than 100. As a result, we obtain 22 provinces as potential control groups here 4. Figure 6 shows the results of permutation test. Each province is viewed as treatment group each time and the gap between GDP per capita and synthetic GDP per capita is drawn respectively. We find that the largest gap is also Fujian province, which means the probability of estimating a gap of the magnitude of the gap for Fujian under a random permutation of the treatment in our data is 1/22. In other words, the previous result is statistical significant at 4.5% or so. Furthermore, we calculate the ratio of MSPE after 2001 to MSPE before The main advantage here is that it avoids the subjective choice of cutoff for the exclusion of ill-fitting provinces. If the final result is convincing, then the MSPE before 2001 should be small and the MSPE after 2001 should be large, which means the ratio will be larger. Figure 7 shows the result of permutation test. We find that the ratio of MSPE after 2001 to MSPE before 2001 of Fujian stands out in the figure. The MSPE of post cross-strait tension period is about 16 times the MSPE of the pre cross-strait tension period. o control province achieves such a large ratio. If the treatment is assigned at random in the data, the probability of obtaining a ratio of MSPE after 2001 to MSPE before 2001 as large as Fujian is 1/30. This implies that we can refuse the null hypothesis that cross-strait tension has no negative effect on economic development of Fujian at 3.3% statistical significance. Summarily, we can conclude that compared with the potential level of economic development of Fujian, deterioration of political environment decreases the level of economic development of Fujian. This means that Fujian should have a better economic performance in the case of no

8 change of political environment. Furthermore, we find that the negative effect of deterioration of political environment becomes larger from 2001 to Conclusion Comparative case study is widely used in social science. However, there is a difficult about how to choose a suitable control group and how to catch the causal effect of among the variables. Fortunately, to some extent, the synthetic control method can solve these problems. It is a kind of data driven method fully utilizing the information of the data, and thus obtain the weights of the potential control groups. These weights can be sythetize to be a control group that matches the treatment group before the treatment well. In addition, Abadie et al. (2010) propose an inference methods about the result obtained from the synthetic control method. In light of these advantages, this paper uses the synthetic method to study the influence of political environment on the economic development through the case of China. Taiwanese independence activists try to divide the country, and thus make the political environment deteriorative, which will increase the investment risk of both the domestic investors and foreign investors in Fujian province. Taiwan's democratic progressive party coming into power in 2000 would be a turning point for the political environment, and thus the investment environment of Fujian province. Our research finds that even if there are lots of preferential policies, GDP per capita of Fujian is found to be lesser than GDP per capita of synthetic Fujian province after the election of Taiwan in This means the political environment deterioration has a negative effect on the economic development of Fujian province. Furthermore, the accumulated loss of GDP per capita of Fujian province is estimated to be yuan from 2001 to 2008 and the average loss rate is about 12.1%. In sum, the paper provides a new case of China to illustrate the influence of political factor on economic development and the estimated economic cost of political environment. Reference: Abadie, A. and S. Dermisi, 2008, Is Terrorism Eroding Agglomeration Economies in Central Business Districts? Lessons from the Office Real Estate Market in Downtown Chicago, Journal of Urban Economics, 64(2), pp Abadie, A., A. Diamond and J. Hainmueller, 2010, Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California s Tobacco Control Program, Journal of the American Statistical Association, 105(409), pp Abadie, A. and J. Gardeazabal, 2003, The Economic Costs of Conflict: A Case Study of the Basque Country, American Economic Review, 93(1), pp Alesina, A. and R. Perotti, 1996, Income Distribution, Political Instability, and Investment, European Economic Review, 40(6), pp Bertrand, M., E. Duflo, and S. Mullainathan, 2004, How Much Should We Trust Differences- in- Differences Estimates, Quarterly Journal of Economics, 119(1), pp Card, D., 1990, The Impact of the Mariel Boatlift on the Miami Labor Market, Industrial and Labor Relations Review, 43(2), pp

9 Card, D. and A.B. Krueger, 1994, Minimum Wages and Employment: A Case Study of the Fast- Food Industry in ew Jersey and Pennsylvania, American Economic Review, 84(5), pp Devereux, M.B. and J.F. Wen, 1998, Political Instability, Capital Taxation, and Growth, European Economic Review, 42(9), pp Enders, W. and T. Sandler, 1996, Terrorism and Foreign Direct Investment in Spain and Greece, Kyklos, 49(3), pp Jong-A-Pin, R., 2009, On the Measurement of Political Instability and its Impact on Economic Growth, European Journal of Political Economy, 25(1), pp Svensson, J.,1998, Investment, Property Rights and Political Instability: Theory and Evidence, European Economic Review, 42(7), pp

10 % Fujian Province 0ationwide Year Data Source: China s Economy and Trade Yearbook and China Business Yearbook Figure 1. Ratio of Taiwanese Direct Investment to FDI in China Yuan Synthetic Fujian Fujian Year Figure 2. GDP per capita: Fujian VS Synthetic Fujian

11 Yuan Yuan Year Figure 3. Gap between GDP per capita of Fujian and That of Synthetic Fujian Jiangsu Synthetic Jiangsu Year Figure 4. GDP per capita: Jiangsu VS Synthetic Jiangsu

12 Yuan Synthetic Group Fujian Year Figure 5. GDP per capita: Fujian VS Synthetic Fujian (Different Potential Control Groups) 500 Yuan Fujian Year Figure 6. Gap between GDP per capita and Synthetic GDP per capita (MPSE<100)

13 Figure 7. Distribution of Ratio of MPSE after 2001 to MPSE before 2001

14 Table 1. Mean Value of Predicted Variables Fujian (1) China (2) Synthetic Fujian (3) Fixed Investment/GDP (%) farming forestry, animal husbandry and fishery /GDP(%) Industry/GDP (%) Building Industy/GDP (%) Labor/Population (%) GDP per capita in 1992 (Yuan) GDP per capita in 1994 (Yuan) GDP per capita in 1997 (Yuan) Data Source: Comprehensive Statistical Data and Materials on 60 Years of ew China Table 2. Weights of Control Group Province Weight Province Wight Province Wight Beijing 0 Zhejiang 0 Chongqing Tianjin Anhui 0 Sichuan 0 Hebei 0 Jiangxi 0 Guizhou 0 Shanxi 0 Shandong Yunnan 0 Inner Mongolia 0 Henan 0 Tibet 0 Liaoning 0 Hubei 0 Shanxi 0 Jilin 0 Hunan 0 Gansu 0 Heilongjiang 0 Shandong 0 Qinghai 0 Shanghai 0 Guangxi 0 ingxia 0 Jiangsu Hainan 0 Xinjiang 0 Table 3. Weights of ew Control Groups Province Weight Province Wight Province Wight Beijing 0 Zhejiang 0 Chongqing Tianjin 0 Anhui 0 Sichuan 0 Hebei Jiangxi 0 Guizhou 0 Shanxi 0 Shandong Yunnan 0 Inner Mongolia 0 Henan 0 Tibet 0 Liaoning 0 Hubei 0 Shanxi 0 Jilin 0 Hunan 0 Gansu 0 Heilongjiang 0 Shandong 0 Qinghai 0 Shanghai 0 Guangxi 0 ingxia 0 Jiangsu Hainan 0 Xinjiang 0 1 The details about comparative case study can be found in Card (1990), Card and Krueger (1994). Card (1990) investigates Mariel Boatlift in Cuba and its influence on Miami economy. Card and Krueger (1994) study the case of minimum wage increase in ew Jersey. 2 ( w,, w ) X can be obtained from minimization of 1 X0W V ( X1 X 0W )' V ( X1 X 0W ) * * 2 J 1 where X represents the predicted variables and V is the symmetric positive semi-definite matrix. The detail is discussed in the appendix B of Abadie and Gardeazabal(2003). 3 The technical detail about the synthetic control method is discussed in the appendix B of Abadie et al (2010). 4 The dropped provinces include Tianjin, Inner Mongolia, Heilongjiang, Shanghai, Hainan, Guizhou, ingxia and Tibet.,