Political Connection, Government Patronage andFirm Performance: Evidence from Chinese

Manufacturing FirmsBei Qin1

IIES, Stockholm University

May, 2012

Abstract: The paper tests whether politically connected firms receive preferential favorfrom government, as measured by capital investment from central government and subsidy.Matching the working experience of top leaders from the State Council (Chinese centralgovernment) and Central Committee of Communist Party of China in power during 1998-2007 with the panel data of manufacturing firms in China, I got a set of firms and theirpolitical connection. Using firm fixed effects estimator, I exploit the variation for the samefirm over time to get clear identification of the political preference. My results show that firmsconnected with one more top leader from State Council tend to get more subsidy, amountingfor 9 percentage points more than the average. Firms connected with one more leader whoholds positions in both State Council and Central Committee tend to obtain 21 percentagepoints more state capital than the average, and then gain 2 percentage points higher markupthan the average. And when extra state capital brought by the connection kicks in, otherdomestic capitals are crowded out. Firms with more employees, but lower sales and lowerprofit are more likely to receive more state capital given being connected, while firms withhigher sales tend to obtain more subsidy. When extra state capital and subsidy enter the firm,they do not seem to improve firms performance, in terms of sales and profit.

JEL code: P16 P48

“From 1989 to 2002, China was led by a group of individuals imbued with heavy urban

biases in their views of economic development and with a strong industrial policy convic-

tion..... They followed a typical career path in a communist system-first serving as chief

technicians and engineers at large SOEs and then ascending through the bureaucracy.”

—– Yasheng Huang, 2008

1PhD student at Institute for International Economic Studies, IIES. Stockholm University. I thank JakobSvensson, David Strömberg, Masayuki Kudamatsu; Frederico Finan, Nancy Qian, Philippe Aghion, Tao Zhang,Zheng Michael Song, Maria Perrotta, Jinfeng Ge, Shuang Zhang, Shengxing Zhang, Konrad Burchardi, PamelaCampa, Nathaniel Lane, Abdulaziz Shifa for the great comments and suggestions.

1

1 Introduction

Substantial literature in political economy has shown that political leaders use their powerto grant economic favors to connected firms(Fisman ,2001; Johnson and Mitton, 2003;Fac-cio,2006). Most of the literature studies the political connection as the personal connectionbetween politicians and specific firms, either via cronyism or shareholding or managers, andoften suggests that the diverted resource are used inefficiently, Khwaja and Mian(2005) forexample. The political connection, however, besides with specific firms, could be extendedto a group of firms sharing similar characteristics.When top leaders start their influential po-sitions, not only firms but also the industries in the cities they arose from will gain moreattention from both the leader and the public. Political leaders may also show favor to thefirms in the same industry and city they ever worked for, which is the main question the paperwill answer.2

Existed studies of political connection usually deal with, at least institutionally defined as,democrat countries, and the trade off happens between politician accountability and private(pecuniary) benefit. Under the autocracy, without election system, accountability is thoughtto be hardly hold, which thereby fades the development. How much the political connectionwill affect the economy activity under an autocracy? Is it possible to discipline the politicianbehavior towards to what people expect? The paper explores the evidence from China, anautocrat developing country for answers to the above.

Taking advantage of a historical phenomenon in China and using the variation of top lead-ers in and out of power, the paper studies the impact of political connection on the resourcesfrom government to firms, and the second order effect, impact on firms’ market power. Thepaper defines that firms are connected with a top leader if they are in the industry and thecity the leader ever worked for and the leader is in power in the year, so the political con-nection for one specific firm will vary across years. Top leaders are defined as individualsholding position high as or above the minister level, while the resources are measured bystate capital invested from central government and subsidy, the market power is measured bythe product markup. Before economy reform(1978-) in China, firms are all publicly owned,de facto controlled by the central government (State-owned or controlled Enterprises, SOEs)or local governments (Collective Enterprises) . It was not until the adoption of the CompanyLaw in 1994 that a distinction is made clear between firms and government units under thebureaucracy. Before that, it was common to see people interchanging their jobs betweengovernment offices and firms at the time. Since industrial sector is the central part of thereform task, more and more individuals with working experience in industries are selected

2In the firm data used in the paper, firms top leader actually worked for only acocunt for 0.1% which is avery small group so I focus the discussion on a broader group with more general characteristics, firms in theindustry and city top leader ever worked for.

2

into governments when the economy reform initiated. This is how the situation described inthe beginning came into being, and the trend even continued after 2003. In the data of thepaper, 153 out of 295 top leaders from 1998-2002 term, 119 out of 280 from 2003-2007 termof central government and party organ indicate working experience in firms ever.

Once the individual enters the government, she/he will follow the Chinese officers promo-tion pattern: work in the local/city government for years, promoted into provincial govern-ment if perform well, and years later further promoted into central government if performedwell at provincial level. Then they finally get the chance to be promoted into minister orabove level positions. The promotion process from local to central government is usuallyvery long(more than 10 years) and the appointment of top Chinese leaders is actually a topside down process, suggesting that individual top leader’s in-and-out of power is exogenousto one specific firm. It provides the main identification for the paper.

Another endogeneity concern is the unobservable firm characteristics, which may en-dogenously determine their possibility of being connected with politicians and at the sametime obtain more resources from government. I collect a set of politician data which coverthe CV information of 350 top leaders in power during 1998-2007 from Central Committeeof Communist Party of China and economy related offices in State Council(the central gov-ernment in China), and then match with the panel data of manufacturing firms to get a setof firms and their political connection. By exploiting the panel data via firm fixed effectsestimator, I can reduce the unobservable firm characteristics concern, and use the variationfor same firm across years to get a clean identification.

Four mechanisms why leaders favor connected firms are suggested in the paper: socialnetworking, information view, reputation building and bribery view. When the politiciansarrive at the top level of the bureaucracy, connected firms might beat others in fight for moreresources from the government since they can access to the decision makers easier, givenother conditions equal. The mechanism is the so called social networking. By informationfew, firms with higher quality tend to receive more resource when connected since the leaderknows them well, if the leader aim to maximize the government investment return. The pub-lic would owe some credits to the leader when observing places they ever worked for developwell, and the public is also more likely to notice the good performance in places top lead-ers ever worked. To build up reputation, leaders would divert more resource to where theyever worked to promote their development. Social networking view is established once thepolitical connection is observed to bring more resource to firms. If the information view es-tablished, we would observe good firms(firms with higher sales, profit for example) tend toget more resources from governments; if the reputation building functions, we would observeresources going to where the public expect, for example, large firms with more employeesand higher sales. Besides checking the effect of political connection on resources from gov-ernment, the paper further checks what type of firm characteristics predict more resources

3

obtained given being connected, measured by the employment, sales, amount of capital andprofit. In lack of the bribery measurement, the paper is not able to test the bribery view.

Few studies of political connection track down the usage of the extra resources gained bythe connection. Limiting sample to firms who do not have connection in the beginning ofsample period but get it later, I check years before and after they switch into the connectionto explore the heterogeneous effect of political connection as well as the resource utilizationafter more resources are rendered to firms.

The paper finds that one extra leader from State Council connected with the firm will bring9 percentage points more subsidy than the average level for the firm, while being connectedwith one more politician who holds positions in both State Council and Central Commit-tee will bring the firm 21 percentage points more state capital than the average level, andincrease the firm’s markup by 2 percentage points. The results show that firms with largeremployee size but lower yearly sales and lower profit tend to get more state capital, and butafter firms get more state capital, none of the firm characteristics above get statistically sig-nificant change. When similar analysis is applied to subsidy, firms with higher sales predictmore subsidy given being connected with top leader from State Council, and profit increaseonly in the year getting connected but decrease after that. The extra resources brought bypolitical connection has no long term effects on any firm characteristic The results lend moresupport to the reputation building view than the information view.

Another interesting question is, when governments invest more to a connected firm, howwill other investors react? By checking the capital structure change since connection year,the paper finds when extra state capital brought by political connection kicks in, collectivecapital, domestic private capital, legal person capital, and Hongkong & Macao capital arecrowded out, while foreign capital maintains. However, 2 years after the connection, statecapital begin the decreasing trend while domestic private capital start increasing. Further-more, the total amount of capital does not increase but actually decrease with the extra statecapital. It at least suggests that getting extra state capital does not signal as an attractiveinvestment for other investors.

The rest of the paper is organized as follows. Section 2 describes the background infor-mation of the government resources, enterprises and politicians, and summarizes the mecha-nisms through which the political connection works. Section 3 describes the data set buildingand the main econometric methodology used in the paper. Section 4 presents the basic resultsof political connection on the state capital, subsidy and markup, while section 5 exploresother firm characteristics change with the connection. Section 6 checks the heterogeneouseffect of the political connection and the firm performance after they get more resource fromgovernment via the connection. Section 7 concludes.

4

2 Background of politician and firms in China

2.1 Firms and government resources

As a developing country, China has been allocating plenty of resources to support firms, in-cluding direct investment, subsidies, special funds, tax breaks etc. To have a rough idea ofhow much the part of money is, figure 1 plots the total subsidy and additional working cap-ital given to enterprises as ratio of the yearly total government revenue. The yearly subsidyto enterprises, although decreasing across years, is still big, accounting for 0.5-3 percent-age points of yearly government expenditure; similarly, working capitals to enterprises alsodecreases over years, accounting for 0.05-0.5 percentage points of yearly government expen-diture; however, the innovation funds to enterprises still remains 2.4-3.7 percentage points ofyearly expenditure. Government investment and subsidy to firms (including the innovationfunds)3 are the main focus of the paper.Moreover, given the choice to look at top leaders,I will only analyze capital flows from the central government. Because there is no detailedinformation on subsidy source, subsidy studied in the paper is the total amount from all gov-ernments.

Capital from central government is named state capital, and falls into two types: capitalstowards the central enterprises and capital towards other regional enterprises4. The invest-ment to central enterprises are directly managed by the State-owned Assets Supervision andAdministration Commission(SASAC) in the State Council, whose core mission is to carryout the government’s functions as investor and owner of state assets. Another type of reg-ular investment planned by the State Council are named as central government within bud-getary investment, amounting to 2-3 hundred millions RMB(i.e. 30-45 million US dollars)per year, which is managed and decided by the National Development and Reform Com-mission(formerly State Planning Commission and State Development Planning Commissionbefore 2003), while Ministry of Finance issues the funds. The budgetary investment is kindof project investment funds, and there is no clear rules guiding which firms they should investbut principally it functions as industrial and fiscal policy for the macro-economy adjustment.There could also be irregular investment out from the offices list above.

There are many kinds of subsidies announced by various State council offices, and corre-sponding to each subsidy, there is a specific title for it. As discussed in Qin(2004), there are

3For individual firm, granted innovation funds is grouped into government subsidy.4The Chinese economic system classifies enterprises according to their level of administrative supervi-

sion. Central enterprise is the one with its control rights-managerial appointments, asset disposals, strategicdirections-of the firms and some or all of the income rights reside with the central government. Regional en-terprise is one where the same control and income rights belong to a regional government.(Huang, 2004). Forenterprises with more than one public share holder(central or local governments), the administrative supervisiongoes with the biggest investor. If the enterprises are completely private or foreign invested, the relationship ofadministrative supervision is decided by the level of government they registered with.

5

three types of subsidies toward SOEs : a) subsidies to help sustain and revive loss-makingSOEs b) subsidies to help privatize or restructure SOEs c) subsidies provided to foster keySOEs. In more detail, if a SOE wants to layoff extra employees (resulted by the old com-mand economy system), wants to reconstruct its production and business, plans to diversifyits ownership, or wants to purchase advance technology or replace the old machines, it canapply for corresponding titles of subsidy to realize the plan. For subsidies to laid off extraemployees, Ministry of Human Resources and Social Security(formerly Ministry of Labourand Social Security before 2003) is the office firms would interact with; subsidies for intro-ducing advance technology and production machines, or science creative production projectsare managed by Ministry of Science and Technology. Other offices that manage subsidiesincludes National Development and Reform Commission, State-Owned Assets Supervisionand Administration Commission, State Planning Commission, State Development PlanningCommission, State Commission for Economic Restructuring, Ministry of Finance as well asMinistry of Commerce. Actually, not only SOEs but also regional Chinese firms(collectivefirms) as well as foreign invested firms are qualified to apply for such subsidies. Figure2 shows the share of subsidies directed to foreign enterprises, increasing across years andamounting to 26% of the total subsidies to enterprises by central government. In the above-scale firm(with yearly sales larger than 5 million RMB, i.e. $0.9 million) data used in thepaper, all type of ownership firms report non-zero subsidy. Usually, firms that have admin-istrative supervision relationship with higher level of government (closer to central govern-ment) respond more to the subsidy announced by the State Council offices. At the sametime,there are also many subsidy titles from local governments firms can apply for. Given thevertical style of bureaucracy in China, we cannot deny that the top leaders from State Councilor Central Committee can also affect the local resource allocation decision.

From above description, we see that the state capital investment and subsidy allocationare the responsibility of the related State Council offices. However, China is lead by theunique party, Communist Party of China, and Central Committee is the top authority withinthe CPC, leading all the work of the Party and represents the CPC outside the Party. Althoughthere is no clear channel for the party to intervene the economy directly, it is reasonably be-lieved that Central Committee has the power to influence the decisions of State Council sincemost important social and economic policies are decided by CPC, and CPC holds the nomen-clature role for personnel appointments in State Council. According to the duty of CentralCommittee members, legally, we should not observe the impact of the political connectionwith a Central Committee member on the resources from central government to firms, butsince the Party influence is not clear, I leave the question open.

6

2.2 Top leaders arising from firms

By definition in this study, a firm is connected with a political leader in power if the firm liesin the city/county and industry where the politician ever worked for. The politicians the paperfocuses on include the heads of the State Council and the offices mentioned in section 2.1,and the politicians from Central Committee, in power between 1998 and 2007. In sum:

1. Members of Central Committee of the Communist Party of China except ones pro-moted from the army, since the administration system of the army is separated fromthe normal government in China.

2. Heads of departments in the State Council(the central government) which might beinvolved in the government decision of resources allocated to firms. Premiers, minis-ters, directors, secretary general or vice ones from National Development and ReformCommission; State-Owned Assets Supervision and Administration Commission; StatePlanning Commission; State Development Planning Commission; State Commissionfor Economic Restructuring; Ministry of Finance; Ministry of Commerce; Ministry ofLabour and Social Security; Ministry of Human Resources and Social Security; Min-istry of Science and Technology.

Political leaders of above two categories can be overlapping, that is, a leader can be a head ofthe chosen office in State Council and at the same time a member in the Central Committeeof CPC, and the leaders that hold position in both are usually more powerful. Among currenttop leaders, many of them have been working in SOEs in the past. It is worth notice thatmany of the SOEs they worked for have been re-constructed in both ownership and corporategovernance aspects today(Clarke, 2003). A typical career pattern of such leader is, firstlythey were re-assigned to local government from the firms, subsequently as a common case,they would serve in the local governments, possibly county to provincial level, for years andwere evaluated according to their performance there, and then if highly valued, they would bepromoted to the State Council and finally got the chance to be selected into the top positions,or be elected into the Central Committee.

To become a member of Central Committee, the politician would need to win the voting ofthe Party National Congresses composed by thousands of Party members across the country.The heads of departments in State Council are nominated by Premier each term, and passedby the Party, while the vices are elected in a more flexible fashion and not restricted by theterm timing. It is impossible for one individual firm to lobby the various persons who areinvolved in the assignment. Based on the complicated and long screening process before thepolitician gets to the top positions, I would argue that whether and when the politicians comeinto power is independent of the individual firms they ever worked for. Furthermore, for firms

7

with more than one directly connected or indirectly connected politicians, it is even harder topredict the total number of politicians they connected with in certain year.

Although when and whether individual politician are able to be in power are arguablyexogenous, industries in cities where the politicians ascend from may be important for the(local or national) economy in certain period of economy development, and thus the govern-ment want to enroll them in. The study will try to take the aspect into account by controllingfor interaction terms of industry and year dummies, of province and year dummies.

2.3 Why such political connection works?

Four mechanisms may make firms with political connection favored by the government.1) social networking. The previous colleagues or friends the leader got to know may

still work in the same industry and city where the leaders were years ago. Therefore, themaintained social capital make such firms easier to access to the leaders.

2) information view. Similar with the related lending discussed in La Porta, Lpeiz-de-silanes and Zamarripa(2003), leaders may know more about the firms in the industries in thecities they ever worked for, and thus better assess the cost-benefit of the investment projects.The information view will predict the resources are more likely diverted to higher qualityfirms and generate higher return, if leaders want to make an efficient investment for the gov-ernment.

3) reputation building. Once the leader became an important, well-known figure, indus-tries and places they ever worked for will also get attention from the public. If firms theredevelop well, people or superior decision makers will nevertheless owe some points to theleader. Such good image would do good to the leader for her/his future career. Similarly,given the public attention, if the top leader acts in these places, which will be quickly noticedby people. Therefore, the leader might want to divert more resource to the industry/city orcounty so as to push up their development, or more immediate effect, enlarge the employ-ment.

4) bribery view. Firms interchange private benefit with the top leaders for the resourcefrom governments.

When firms get connected with top leaders, more resources from government can beimmediately observed, and the market power of the firm will be the second order effect. Whengovernments show preferential treatment. to some firms, it may strengthen firms competitiveability comparing with exist firms, deter new entrance and thus increase their market power.So market power,measured by markup is also checked in the paper.

8

3 Data & Methodology

3.1 Data

Firm data is from the Annual Surveys of Industrial Production, collected by the Chinesegovernment’s National Bureau of Statistics every year since 1998. It is a census of all manu-facturing firms with more than 5 million RMB, approximately 0.9 million US dollars yearlysales revenue in mainland China, and firms with all types of ownership are included. Thepanel data start with over 160,000 firms in 1998, with new firms joining every year, and in2007 in total over 330,000 firms are included. The surveys cover the very detailed informationof the firms: names, addresses, age, ownership, capital amount and their sources, employees,wage, equity, tax , subsidy as well as other financial statistics. The paper limits the leadersunder study to the top leaders in central government and Central Committees, assuming theinfluencing power of top politicians decreasing along down the government cadre. There-fore, in the study, only firms with the administrative supervision relationship to the central,provincial and prefecture governments, at least 1 year during the data period, are chosen, i.e.86827 distinct firms with ten years panel from 1998 to 2007, 406774 obs. in total.

The politician data is collected by the author: first, the name list of members of CPCCentral Committee, top leaders of State Council, and (vice) Ministers/ secretaries of StateCouncil departments that might interact with firms in terms of government resource allocationin power during 1998-2007 are found, in total 350 politicians; next, politician’s individualCV are collected from Dictionary of Central Committee Members of Communist Party ofChina 1921-2003, www.baidu.com, www.renwu360.cn/, www.xinhuanet.com/. Informationof individual characteristics, education, majors, working experience, province/city/countyever worked in are all collected.

3.2 Matching Politicians to Firms

Among the 350 politicians, around 200 of them indicate ever working in firms, but only 173politicians state clearly the firm names they ever worked for, and thus 277 distinct firm namesare got (not all of them are manufacturing firms). I then use the 173 politician records to buildup the political connection with firms from the firm data.

I check out the two digit industry code firms where the politician ever worked belongto and the corresponding city/county. For firms exist in the firm data, I use the industrycode given there(the matched firm can have more than one industry code across year, allare recorded); for firms not found in the sample, I first check their online profile to get theindustry they claim; for firms that do not state that clearly, I search in the manufacturingfirm data for the similar firms by products and take the top three frequently stated indus-try as the industry for them. Therefore, one firm could correspond to more than one industry.

9

Some politicians from Central Committee are at the same time in provincial governments, thecity/county associated with them are regarded as connected if it lies inside the same provincewhere the politician assumes the position; otherwise, regarded as not connected. The indi-rect connection is thus coded by the pair of city/county+ industry, which finally is shapedinto a table with number of politicians of each kind across year associated with each pairof city/county+ industry. It finally gives out 194 city/county+ industry connection pairs.When merging with the manufacturing firm data, 170 pairs associated with 131 politiciansare matched. In total 10,287 out of 86,827 distinct firms ever had political connection during1998-2007

3.3 Summary Statistics

Table 1 presents the summary statistics for the variables of interest for the firm data andmatching with politicians. State capital (capital from central government), and subsidy are themeasures of preferential treatment from government, and was obtained by taking logarithmof the amount of yearly state capital and subsidy received5.As we can see, that the statecapital for one single firm can be as large as 2000 billion RMB and yearly subsidy can belarge as 13.8 billion RMB. Markup is the measurement for firms’ market power, calculated bydividing the total manufacturing products profit by the total cost of manufacturing products(obtained by minus total profit from total sales). The range of the markup is huge, from -199.1to 16826.4. To give a picture of the difference across different type of firms, and also for laterchecks the paper will do, other firms’ characteristics, total amount of employee, yearly sales,profit, total number of actual received capital as well as the year the current firm establishedare also described in table 1.

Panel B and Panel C separate the sample into never connected and ever connected firms.Firms never and ever indirectly connected have closer mean value and standard deviation ofand subsidy, but firms ever connected have much lower mean of logarithm of state capital.Firms that ever had connection show much lower markup than the ones never had, while theyalso have much lower standard deviation, .39 against 28.9. In terms of the market power,firms ever connected are more similar with each other. For type of firm characteristics, firmsever connected and never show similar mean values and standard deviations for logarithm ofemployee amount, logarithm of yearly sales and total received capital. Firms that ever hadpolitical connection have mean profit three times more than ones never connected. To have abetter control group, I will limit the analysis to firms ever had connection during 1998-2007.

That a firm was ever connected with a politician does not mean it has political connectionevery year because the politician can be in and out of the top positions across years dueto various reasons. The number of connected politicians from both Central Committee and

5The two measures are logarithm of the number plus 1 so as to get rid of the negative infinite value.

10

State Council ranges from 0 to 2 for a given firm in a given year; the number of connectedpoliticians from Central Committee ranges from 0 to 9; the number of connected politiciansfrom State Council range from 0 to 3.

Table 2 shows all the politicians statistics across year: the in/out of power status of the131 top leaders across year. For the three type of politicians, connected politicians from bothCentral Committee and State Council are always the fewest, the number for ones from Centralcommittee is the highest, and ones from State Council only account for around one fourth tohalf of the ones from Central Committee. The data period covers two official terms for the twotype of politicians respectively, which are 1997-2002 and 2002-2007 for Central Committeeof CCP; 1998-2003 and 2003-2008 for State Council. Therefore, year 2002-2004 should bethe years when most changes were expected and observed. From table 2 we can see that year2002 and year 2003 are the most volatile years, and we can also tell less leaders from theState Council change compared to the ones from Central Committee. Departments in StateCouncil concerning the economy usually require professional knowledge and experience sothat the replacements of the leaders out of these offices are less frequent.

Politicians could also be placed in the positions out of the term switching years. Such ir-regular appointments are more likely to happened with State Council due to contingent needsand more likely to happen with the vice positions. The irregular appointments in CentralCommittee can happen only when some leaders are out of the board due to unpredicted rea-sons(sick, dead, checked/arrested due to malpractice, sick etc.) and in order to maintain thefixed number of CC members, new politicians are elected to fill out the absence. To explorethe change of different type of politicians across years, table 3 list the number of change eachyear for connected politicians, which are grouped by reason of change. Among the reasons,dismissed/arrested/fled away due to malpractice, term limit, retired can be regarded as totallyexogenous shock, which account for 8% of the total variance. Given the selection process ofthe top leaders in national government, plus the shock of being replaced out of offices, theamount of politicians firms connected each year is arguably exogenous to firms unobservablecharacteristics.

3.4 Methodology

The core difficulty in identifying the effect of political connection on preferential treatmentreceived from government is the endogeneity concern: some firms are important and influ-ential to the country so that politicians ever worked there are more likely to be promoted andat the same time more government resources are directed to the firms; or such firms have re-ceived government benefit for long time, and would further lobby the related decision makersto elect the politicians connected with them into the position. Given this concern, a convinc-ing estimation strategy is to exploit the variation for the same firm across year by using firm

11

fixed effects ( Khwaja and Mian (2005)):The main estimator in the analysis will be fixed effect estimator for panel data,

yit = ci +λt +Political_connectionitτ + Industryit ∗λt

+provincei ∗λt +Gi ∗λt +Ownershipit ∗λt + εit(1)

where yit are results of interest, state capital, subsidy and markup of firm i in year t.yit can also be other firm characteristics variables in auxiliary regressions and checks. Thereason I do not scale the two measure by yearly sales or total number of employee is thateither sales or employees can also be the result of the political connection. Similarly, sinceany individual firm characteristics related with its performance can affected by the politicalconnection the firm associated, I exclude them from the regressions. λt is the vector oftime period dummies, subscript t denotes year 1998 to 2007.Political_connectionit is thevector of political connection indexes for firm i in year t, the key independent variables ofinterest: # o f CCSC politicians, # o f CCSC politicians, # o f SC politicians. CC denotesCentral Committee, SC denotes State Council, and CCSC denotes Central Committee andState Council. Since one politician can be in and out of office across years, the number ofpoliticians connected changes across years. ci is the firm fixed effect representing unobservedfirm characteristics.

Industryit ∗ λt , provincei ∗ λt , are interaction terms between industries(2 digit industrycode, 40 categories), provinces(24 provinces for ever connected firms) and years respectively,representing the provincial trend and industry trend that may affect the resources allocationdecision and the promotion of politicians ascending from that industry and region. The factthat the politicians were selected from original SOEs to governments can be because of theindustry they worked in become important for the economy at the time so that governmentwant some experts on that. Although the years they were promoted from firms to the gov-ernment are years before the sample years, it is possible that some industries have been andare still important for the national economy during the sample years. Similar concerns applyto provinces.The industry trend and provincial trend in the regressions will take care of theseconcerns.

Although the sample is limited to firms ever connected, the connection patterns differacross firms. There are four categories of firms: firms switch at least twice between connectedand not connected during the sample period-type 1; firms have no connection in the beginningof the sample period but gain it later on-type 2; firms have the connection in the beginningbut lose it later-type 3; firms have connection all the sample years-type 4 firms. It is possiblethat firms in same connection pattern share some unobservable characteristics differing fromothers that will vary across years. So I will try to control for the different connection patterntrend by the interaction term between four types of firms and year dummy, Gi ∗λt . Gi is the

12

vector of the connection pattern dummies.According to Huang(2004), Chinese government has political pecking order: government

holds certain political preferences towards different ownership type of firms, and they aretranslated into economic policies, regulatory practices, financial support decisions etc. In theorder, state-owned or state controlled enterprises are ranked first, and then collective firmsclass, foreign invested firms are next, while the domestic wholly private firms are the last6

. I will control for the ownership type trend. In all the regressions, standard errors arecorrected to account for correlation of the error term across observations in different yearsthat correspond to the same firm and thus are clustered at firm level.

4 Main Results: Impact of Political Connection on StateCapital , Subsidy and Market Power

Table 4 presents the results of the estimation of equation (1). One more leader holding posi-tions in both Central Committee and State Council connected with the firm in one year willbring 21 percentage points more state capital compared with the average level of firms, andincrease the markup by 2 percentage points. One more leader from State Council connectedwith the firm tend to bring 9 percentage points subsidy more than the average. Estimates forconnection with Central Committee member on all three measurements, show little effect, interms of both economic and statistic significance. The connection with leaders from StateCouncil here do not help with the state capital, small negative point estimate and too largestandard errors. The connection with leaders from both seem not to help with the subsidy.

Usually the leader holding positions in both Central Committee and State Council rankson the top of the hierarchy. Hence, it is not surprising to see they influence the big amount(state capital investment) rather than the small amount(subsidy). The targeted number offirms for state capital investment is much smaller than the one for subsidy. And comparedwith the subsidy, state capital is much larger amount for an individual firm. The chance toget more investments from central government is much lower but the amount is much biggerthan subsidy. Therefore, if a top leader plans to favor preferred firms, she/he would notbother to check the small amount of subsidy but rather they will consider the investment. Ifthe decision for state capital is influential and noticeable, when the firm in question is noteligible to obtain, a lower leader, compared with the leader holding position in both StateCouncil and Central Committee, would not risk to divert the capital to firms they preferred.

However, will extra capital brought by the connection come with more or better projects,and will it boost the firm production and improve their performance in the subsequent years?We will look into the former question in section 5 and the latter question in section 7.

6In total, 23 ownership types exist in the data.

13

5 Is the state capital accompanied by more projects?

Bertrand, Kramarz, Schoar and Thesmar (2007) find politically connected CEOs hire moreemployees during the election periods but such employees increase does not go with moreprojects or contracts. Similarly, I would ask what is the goal of more investment from centralgovernment to the enterprises. Is it because the top leaders think the firms from the industryand city/county they once worked for are promising by the information view, so that theywould like to build more projects there and gather the best return for the government invest-ment? If so, the state capital increase should be associated with more projects planned bythe government. In this case, we would observe bigger production size and more profit whenthey are connected with at least a leader who hold position in both State Council and Centralcommittee. It is also possible that the extra state capital will come with or be invested tosome long run projects and the return will be able to see only years after. For this possibility,I will discuss in the part how firms use the extra resources they obtained in section 7.

Replacing the dependent variable with number of employees, yearly sales, profits andtotal amount of capital, I re-estimate equation (1). Table 5 presents the estimates of politicalconnection on the above four firm practice measures. From column (1), (2) and (4), wecannot reject the zero effect of the indirect connection with a leader holding positions in bothtop authorities, which suggests the connection does not necessary come with more and betterprojects that will function immediately. The estimate of top leaders from Central Committeeand State Council do have positive sign on sales and profit, which help explain the positiveeffect on markup(show in section 5). It is not obvious and significant that such connectionincrease sales and profit, but the combined effect is salient.

However, no statistically significant effect on firm employee, sales and profit can also bebecause of the heterogeneity of the connected firms. Firms differ too much and then drive themean estimation down. Section 6.1 will further discuss the heterogeneous effects of politicalconnection on more resources, and section 6.2 limit the sample to only firms benefit from thepolitical connection and study the dynamic effects of connection on firm performance.

Column (3), surprisingly, presents no effect of connection with leaders from Central Com-mittee and State Council on the total capital, negative and statistically insignificant. Whyextra state capital invested into firms do not increase the total capital? Given the negativesign of state capital on the actual received capital, an explanation would be: when the statecapital is increased with the connection with top leaders, it at the same time crowds out othersources of capitals, by at least equivalent or even more amount.

14

6 What characteristics predict the favor and what happenslater on?

The chance to obtain more state capital investment is small in the whole sample, and theconnection variation occurs at industry and city level, it would not be possible that all firmswithin the connected industry in the city gained the preferential treatment. Certain type offirms are more likely targeted by the leaders. By information view, firms with higher sales orprofit should be targeted; by reputation building, firms with larger employee size should betargeted. To explore what type of characteristics helps given being connected, I use the firmcharacteristic one year before getting the connection to interact with the first connected yeardummies, and firm characteristics one year before losing the connection to interact with thefirst disconnected year dummies to investigate the heterogeneous effects.

To explore further what happens with the extra resources brought by political connection,I limit the analysis to only firms who actually obtain (lose) more resources in the year or oneyear after they gain (lose) political connection. By standardizing the year into ith years beforeor after getting(losing) connection years and exploring the variation from different timings, Icould draw a more clear graph indicating what is going on in the timing axis. When analyzethe state capital, I construct the getting (losing) connection year timing according connectionwith leaders from Central Committee and State Council; when it apply to subsidy, I constructaccording to connection with leaders from State Council.

6.1 State Capital

6.1.1 Heterogeneous effects of connection with leaders from CCSC on obtaining morestate capital

Before trying to find out the firm characteristics predicting de facto more state capital whenconnected, I need to identify when the increase(decrease) will happen. After check, I find theextra resources obtaining(losing) are more likely happened in the same year getting(losing)connection, and a much smaller change in one year after.7 So additional to equation (1)specification, I add the interaction terms between the firm characteristics one year beforeconnection (disconnection) and the getting connected (disconnected) year dummies. Run theequation (2) as the following:

yit = ci +λt +Political_connectionitτ +# o f CCSCit∗α +# o f CCSCit∗×Charactit∗−1Φ

+CCSCit0β +CCSCit0×Charactit0−1Ω+ Industryit ∗λt

+provincei ∗λt +Gi ∗λt +Ownershipit ∗λt + εit

(2)

7Results are provided by request.

15

where subscription t∗denotes the year firms gain connection, and t0 denotes the year firmslose connection. # o f CCSCit∗ denote the number of connected leaders that holds positionsin both Central Committee and State Council. CCSCit0 is the dummy for losing the connec-tion with leaders holding positions in both CC and SC. Charactit∗−1 is the vector of firmcharacteristics(employees, profit, sales, and total other type of capital)one year before theconnection, and similarly Charactit0−1 is the vector of firm characteristics before the yearlosing the connection. We would expect to observe exactly opposite sign for the interactionsof # o f CCSCit∗×Charactit∗−1 and CCSCit0×Charactit0−1, if these characteristics sensitiveto every change of the political connection.

Results are reported in table 6 column (1). Firms with more employees are more likelyto gain more state capital in the connection year. However, yearly sales and profit predictthe state capital on the opposite way, firms with lower yearly sales and profit are more likelyto get the state capital in the the connection year. Assume one firm gets 1 leader connectedin year t∗, if the firm has 1 percentage points higher employee than the average, it willobtain 0.12 percentage points more state capital; if the firm has 10,000 RMB lower profit,it will obtain e− 05 percentage points more state capital than the average; if the firm has 1percentage points lower sales than the average, it will obtain 0.08 percentage points more statecapital. The effect of the employee size is quite considerable and suggest the big weight ofthe employee size in the state capital investment consideration. The level of total other capitaldoes not seem to help attract more investment from the central government (no statistic powerfor the estimate).

When checking the interaction terms between firm characteristics and losing connectionyear, except for employees, interaction terms with yearly sales and profit both show oppositesigns. It suggest that firms with lower profit or lower sales lose state capital in the year theylose the connection. Firms with more employees still maintain the state capital when theylose the connection. The estimates for the interaction term with the year losing connection,except for the profit, do not have the statistic power.

The results suggest that the state capital diverted by the top leaders does not go to thefirms with obvious better performance, but the larger firms in terms of employee number.The purpose that the top leader disposes the state capital to firms in the industry and citythey once worked for is not likely for maximizing the profit as an investor. The reputationbuilding view is supported here, which predicts that the leaders would want to build up agood image for themselves by boosting the employment of the industries and cities they everworked. More employment and more firms exist in the place, no mater efficiency or not, willdefinitely be a good image in front of the public. In fact, if the central government care moreabout big firms (with more employees), such preference of the leaders is consistent with thisaim. If it is not true, the national economy will pay for the individual leaders’ reputation.

No matter what the true aim of the central planner is, state capital do go for employment

16

enlarging under the functioning of the political connection. Leave the efficiency aside, higheremployment is always the public would like to see. Hence, in an autocracy and developingcountry like China, officials’ career concern will discipline the leaders behave somewhat asthe public expect.

6.1.2 Capital structure change across years around connection (disconnection)

From section 5, I do not find any increase with the total capital when state capital increase. Iwill thus check what happens to the capital structure when firms obtain extra state capital dueto political connection. To check this, I limit the study sample to firms who obtain (lose) statecapital in the year or 1 year after they gain (lose) connection with at least a leader holdingposition in both Central Committee and State Council, and run the following estimate.

Kit = ci +λt +CCSCit ′−1α1 +CCSCit∗α2 +CCSCit∗+1α3 +CCSCit∗+2α4

+CCSCit0β1 +CCSCit0+1β2 +CCSCit0+2β3 + Industryit ∗λt

+provincei ∗λt +Gi ∗λt +Ownershipit ∗λt + εit

(3)

Kitdenotes one of the certain types of capital: state capital, collective capital(from localgovernment), domestic private capital, legal person capital (capital invested from domesticcompanies), Hongkong or Macao capital, foreign capital, total actual received capital.Thesubscription t∗ denotes the year getting connection and t0 denotes the year losing connection.Results are reported in table 7. From table 7, except for foreign capital, all other types ofcapital show decreasing in the year or one to two years after getting connection, and thedecrease maintained even to the year losing connection except for the collective capital. Inone to two years after losing the connection, other types of capital show increasing. To seethe pattern clearly, I plot the coefficient estimates α1-α4, and β1-β3 for each type of capitalregression in figure 3.

From figure 3, we can see that the state capital level increase in the beginning of theconnection but almost fall back to the level as before the connection in two years after gettingthe connection. When extra state capital kicks in, all other domestic capitals are crowdedout and capital from local government (collective capital) and domestic legal person capitaldecrease all the time even till the two years after the connection year, but domestic privatecapital finally go back and beyond the level before getting connection in two years after theconnection year. Hongkong & Macao capital and foreign capital has the most flat pattern,slightly affected by the extra state capital. The most crowded out capital are legal personcapital and collective capital when extra state capital kicks in, while legal person capital canhardly go back to the original level.

17

6.1.3 Usage of the extra state capital.

Another question the study tries to answer is how firms will deal with the extra state capitalbrought by the political connection; will it be used to increase the efficiency of the produc-tion? To answer the question, also by limiting samples to firms who actually obtain morestate capital in the year or 1 year after getting connection, I replaced Kit with employees,yearly sales, profit in equation (3),and re-estimate the model. Results are reported in table 8.In the connection year and one year after, all the three index show decrease. The sales andprofit show increase in the two years after the connection year. All the three characteristicsshow mixing signs in years around the year losing connection. However, all of these estimatehave very large standard errors to be statistically significant, and we cannot deny the zeroeffects for all of them.

In sum, I did not find any statistically change on firms employee, yearly sales and profit inthe year and till two years after the firm obtaining connection with leaders holding positionsin both Central Committee and State Council, when they obtain more state capital becauseof the connection. Combing this results with the result from section 5, I can conclude thatthe extra state capital really does not come with more or better projects. Extra state capital atleast does not improve production size, nor yearly sales and profit.

6.2 Subsidy

6.2.1 Heterogeneous effects of connection with leaders from SC on obtaining moresubsidy

As described in section 2, many purposes are associated with subsidy, laid off subsidy, pro-duction reconstructing and innovation subsidy etc. It is difficult to predict what type of firmsare more favored before looking into the data. Revising the equation (2) a bit and I run thefollowing equation (4).

yit = ci +λt +Political_connectionitτ +# o f SCit∗α +# o f SCit∗×Charactit∗−1Φ

+SCit0β +SCit0×Charactit0−1Ω+ Industryit ∗λt

+provincei ∗λt +Gi ∗λt +Ownershipit ∗λt + εit

(4)

The key timing in equation (4) is defined as the year t∗ (t0) when the firm get connectedwith a leader holding position in State Council, because we only observe such connectionhelp firm obtain more subsidy in section 4. Results are reported in table 6 column (2). Fromthat, we can observe only yearly sales get the statistically significant estimate: firms withhigher yearly sales are more likely to receive the subsidy. It suggests that leaders in StateCouncil seem to care more about business active firms. Other firm characteristics, however,seem to be trivial in the subsidy decision, both the economic size and statistic significance

18

are small.

6.2.2 Usage of extra subsidy

Compared with state capital, subsidy is more like cash flow and in small amount, which issupposed to be spent all out right after they obtain it. So the effect of extra subsidy, if thereis any, should be observe immediately after the firm obtain (lose) the connection. Revise theequation (3), I run the estimate for the equation (5) as following:

Charactit = ci +λt +SCit ′−1α1 +SCit∗α2 +SCit∗+1α3 +SCit∗+2α4

+SCit0β1 +SCit0+1β2 +SCit0+2β3 + Industryit ∗λt

+provincei ∗λt +Gi ∗λt +Ownershipit ∗λt + εit

(5)

Charactit denotes the firm characteristics of interest, employee, yearly sales and profit.Results are reported in table 9. From table 9, we can see that in the year when the firm get theconnection with leader from State Council, the firm increase their profit by 23.572 millionRMB more. Given the average level of yearly profit is 28.473 million RMB, this number isreally huge, accounting for almost 83% of the average level. The yearly sales in the sametiming at most will increase by 9.6 percentage points than the average, and the employeeat most increase by 5.4 percentage points than the average(if we assume the statistic poweris enough, though it is not), it is hard to believe the increased profit is generated by theenlarged production and sales caused by connection with SC leaders. And the number dropsdramatically in one year later and even turns to negative in two years after the connection.Therefore, the increased profit is probably driven by the simply increased income source -subsidy. Still, no sustainable effects on firm performance is found for extra subsidy obtained.

7 Conclusion

A special historical fact of the leader component of Chinese government this decade is thatmany persons with working experience in enterprises are enrolled and promoted into the gov-ernment offices. (Huang, 2008). Taking advantage of it, I collect the individual informationof the top leaders from top authorities of government and party in poer during 1998-2007 andthen match it with the above scale manufacturing firm survey to get a set of firms and polit-ical connection. I investigate the effects of the political connection -defined as a top leaderfrom the offices I choose ever worked in the same industry and city/county the firm lies- onthe preferential favor of government, measured by state capital and subsidy, as well as themarket power measured by markup.

My results suggest that firms connected, with a leader from State Council are more likelyto obtain more subsidy from governments, but it doesn’t seem to help firm get more state

19

capital. A connection with leaders from Central Committee does not show help on the twomeasure of government favor nor the firm markup, somehow indicating the party do not crossthe line and intervene the government decision on allocating resource to firms. Connectionwith a politician who holds position in both State Council and Central committee wouldbring more state capital to the firm. Given the extra investment, I further check whetherthis increased capital comes with more projects or contracts planned by the government.However, the result does not support it.

Checking firm characteristics right before (after) the connection (disconnection), I foundthat larger (with more employee) and worse performed firms(lower yearly sales and profit)are more likely to obtain the state capital brought by the connection with top leader hold-ing positions in both Central Committee and State Council. The fact support the reputationbuilding view discussed. For firms who did get more state capital, no statistically significantfirm characteristics change is found in the 1st year till 3rd year into the connection, nor in the1st year till 3rd year out of the connection.

Firms with larger sales size are more likely to obtain more subsidy when they get connec-tion with leaders from State Council. In the first year of connection, profit show incredibleincrease but dramatically decrease after that without obviously sales or employee increase.It suggests the extra subsidy brought by the connection actually does not bring sustainableimprovement to firms.

However, corruption channel is not discussed given the data I have cannot test the story,but it would be an interesting topic for future study.

References:Bertrand Marianne, Kramarz Francis, Schoar Antoinette, and Thesmar, 2007, “Politi-

cians, Firms and the Political Business Cycle: Evidence from France” , working paper ver-sion, http://www.crest.fr/ckfinder/userfiles/files/Pageperso/kramarz/politics_060207_v4.pdf

Faccio Mara, 2006, “Politically Connected Firms”, The American Economic Review, Vol96, No.1

Fisman Raymond, 2001, “Estimating the Value of Political Connections”, American Eco-nomic Reviews, Vol. 91, 1095-1102,

Fisman Raymond and Wang Yongxiang, 2011, “Evidence on the Existence and Impact ofCorruption in State Asset Sales in China ”, working paper,

http://www2.gsb.columbia.edu/faculty/rfisman/papers/new/transfers%20paper%20-%20March272011.pdfHuang yasheng, 2004, Selling China-Foreign Direct Investment During the Reform Era,

Cambridge University PressHuang yasheng, 2008, Capitalism with Chinese Characteristics: Entrepreneurship and the

State, Cambridge University Press

20

Johnson Simon, and Todd Mitton, 2003, “Cronyism and Capital Controls: Evidence fromMalaysia”, NBER working paper, w8521

Khwaja Asim Ijaz and Mian Atif, 2005, “Do Leaders Favor Politically Connected Firms?Rent Provision in an Emerging Financial Market”, Quarterly Journal of Economics, Vol 120,Issue 4

La Porta Rafael, Lopez-De-Silanes Florencio and Zamarripa Guillermo, 2003, “RelatedLending”, Quarterly Journal of Economics, February 2008

Qin Julia Ya , 2004, “WTO regulation of Subsidies to State-Owned Enterprises(SOEs)-ACritical Appraisal of the China Accession Protocol”, Journal of international Economic Law,Vol 7, No.4

0.05.1

.5

1

0

1

2

3

shar

e of

yea

rly g

over

nmen

t exp

endi

ture

, %

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007year

subsidies innovation fundsworking capital

Source: calculated from government finance part, Premium China Database, CEIC data.

Figure 1: Yearly Subsidies, Innovation funds, Working Capital to Enterprises

21

10

15

20

25sh

are

of to

tal y

early

sub

sidi

es to

ent

erpr

ises

, %

1998 2000 2002 2004 2006year

Source: calculated from government finance part, Premium China Database, CEIC data.

figure 2, Yearly subidies to foreign enterprises, by central government

Figure 3 Capital structure change during years since connectin

-10

12

Coe

ffici

ent

t*-2 t*-1 t* t*+1 t*+2 t0 t0+1 t0+2Timing

state capital collective capitalprivate capital legal person capitalforeign capital Hongkong or Macao capital

t * : year into connection. t0 : year off connection

22

Table 1 Summary Statistics of Firms VARIABLES N mean sd min max

Panel A: Full Sample State Capital (log 10,000 RMB) 406,629 3.987 4.612 0 19.07 Subsidy (log 10,000 RMB) 406,109 1.094 2.506 -2.671 14.02 Markup 391728 .0177757 26.90143 -199.1 16826.4 Employee (log) 406,773 5.159 1.645 0 12.18 Yearly Sales (log 10,000 RMB) 406,747 9.557 2.839 0 18.98 Total actual received capital (log 10,000 RMB) 389,365 9.069 1.941 -0.119 19.07 Yearly profit (10,000 RMB) 406,773 11,444 405,358 -5.194e+06 1.108e+08 # of CCSC connected 406,774 0.0364 0.213 0 2 # of CC connected 406,774 0.0967 0.520 0 9 # of SC connected 406,774 0.0392 0.227 0 3 Having connection with CCSC 406774 0.031531 0.1747481 0 1 Having connection with CC 406774 0.0525279 0.2230894 0 1 Having connection with SC 406774 0.0307689 0.1726913 0 1 Year the firm established 401,345 1,982 19.36 1,600 2,007 Number of firms 86,827

Panel B: Never had Political Connection State Capital (log 10,000 RMB) 352,280 4.119 4.626 0 18.22 Subsidy (log 10,000 RMB) 351,832 1.097 2.510 0 14.02 Markup 338709 .0220536 28.92998 -199.1 16826.4 Employee (log) 352,420 5.175 1.641 0 12.02 Yearly Sales (log 10,000 RMB) 352,398 9.509 2.864 0 18.89 Total actual received capital (log 10,000 RMB) 336,542 9.049 1.921 -0.119 18.22 Yearly profit (10,000 RMB) 352,420 8,818 174,529 -5.194e+06 2.742e+07 Year the firm established 347,367 1,982 19.40 1,600 2,007 Number of firms 76,540

Panel C Ever had Political Connection State Capital (log 10,000 RMB) 54,349 3.131 4.425 0 19.07 Subsidy (log 10,000 RMB) 54,277 1.077 2.478 -2.671 13.64 Markup 53019 -.0095539 .3886813 -53.66667 10 Employee (log) 54,353 5.051 1.667 0 12.18 Yearly Sales (log 10,000 RMB) 54,349 9.865 2.657 0 18.98 Total actual received capital (log 10,000 RMB) 52,823 9.196 2.059 -0.119 19.07 Yearly profit (10,000 RMB) 54,353 28,473 1.016e+06 -4.436e+06 1.108e+08 # of CCSC connected 54,354 0.273 0.526 0 2 # of CC connected 54,354 0.724 1.252 0 9 # of SC connected 54,354 0.294 0.556 0 3 Year the firm established 53,978 1,983 19.08 1,837 2,007 Number of firms 10,287

Note: CCSC, Central Committee & State Council; CC, Central Committee; SC, State Council, same in all tables. Amount is deflated by CPI, all in 1998 RMB currency value.

Table 2 Type of Politicians Connected with Firms across Years 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Not in power 57 56 51 49 26 53 58 58 60 61 Central Committee 49 49 53 54 81 44 46 47 46 48 State Council 19 20 21 22 18 23 19 18 16 13 Both CC & SC 6 6 6 6 6 11 8 8 9 9 Total 131 131 131 131 131 131 131 131 131 131

23

Table 3 Reasons for Connected Leaders Varying across years Type of change 1999 2000 2001 2002 2003 2004 2005 2006 2007 Total

placed in CC 1 25 1 27 placed in SC 3 2 2 1 7 1 2 18

leave due to position change 2 1 1 2 2 1 2 3 3 17 change from Central Committee to

Both 5 1 1 7 change from State Council to

Central Committee 1 1 change from State Council to Both 2 2

change from Both to Central Committee 2 1 2 5

change from Both to State Council 1 1 dismissed due to abusing power 1 1 flee away due to corruption case 1 1

arrested due to corruption 1 1 retire 4 1 5

Leave after term terminated 28 2 30 term limit 1 1

Total 5 4 3 33 50 8 4 5 5 117 Note: the number of politicians between year n and n-1 corresponding to perspective reason are list in the table.

Table 4 Impact of political connection on resources obtained and market power (1) (2) (3) VARIABLES State Capital Subsidy Markup # of CCSC politicians connected 0.207*** 0.0250 0.0198*** (0.0746) (0.0481) (0.00588)# of CC politicians connected 0.0461 -0.0170 0.00181 (0.0368) (0.0213) (0.00355)# of SC politicians connected -0.0588 0.0897** 0.00615 (0.0706) (0.0373) (0.00615)Observations 52,048 52,052 52,046 Number of firms 9898 9898 9,897 R-squared 0.104 0.048 0.015 Year dummy YES YES YES Firm FE YES YES YES Industry*Year dummy YES YES YES Province*year dummy YES YES YES Connection pattern * year dummy YES YES YES Ownership type*year dummy YES YES YES F-test for connection variables jointly zero 3.088 2.469 4.267 Prob>F 0.0260 0.0600 0.00510

Robust standard errors clustered by firm in parentheses. *** p<0.01, ** p<0.05, * p<0.1

24

Table 5 Political connection and other firm characteristics (1) (2) (3) (4) VARIABLES Employee Yearly Sales Total Capital Profit # of CCSC politicians connected -0.00685 0.0263 -0.0228 30,309 (0.0146) (0.0199) (0.0157) (23,919) # of CC politicians connected 0.0191** 0.0187* 0.00923 -478.8 (0.00800) (0.0105) (0.00724) (3,185) # of SC politicians connected 0.00454 -0.00610 0.0204 -7,405 (0.0134) (0.0175) (0.0127) (6,626) Observations 52,052 52,049 52,044 52,052 R-squared 0.151 0.144 0.069 0.262 Number of idpanel 9,898 9,898 9,898 9,898 Year dummy YES YES YES YES Firm FE YES YES YES YES Industry*Year dummy YES YES YES YES Province*year dummy YES YES YES YES Connection pattern*year dummy YES YES YES YES Ownership type*year dummy YES YES YES YES Robust standard errors clustered by firm in parentheses *** p<0.01, ** p<0.05, * p<0.1

25

Table 6 Heterogeneous effects: firm Characteristics and more resources (1) (2) VARIABLES State Capital VARIABLES Subsidy # of CCSC politicians connected 0.252*** # of CCSC politicians connected 0.0262 (0.0852) (0.0485) # of CC politicians connected 0.0477 # of CC politicians connected -0.0164 (0.0371) (0.0216) # of SC politicians connected -0.0626 # of SC politicians connected 0.0944** (0.0721) (0.0422) # of CCSC politicians connected_t* -0.124 # of SC politicians connected_t* -0.671*** (0.310) (0.225) CCSC_t 0

-0.287 CCSC_t 0 0.376 (0.460) (0.236)

0.123** -0.0309 # of CCSC politicians connected_t* X employee_t*-1 (0.0608)

# of SC politicians connected_t* X employee_t*-1 (0.0379)

-8.84e-08** -6.26e-08 # of CCSC politicians connected_t* X profit_t*-1 (3.52e-08)

# of SC politicians connected_t* X profit_t*-1 (8.76e-08)

-0.0799* 0.0784*** # of CCSC politicians connected_t* X yearly sales_t*-1 (0.0464)

# of SC politicians connected_t* X yearly sales_t*-1 (0.0288)

0.0318 0.00241 # of CCSC politicians connected_t* X total other capital_t*-1 (0.0196)

# of SC politicians connected_t* X total other capital_t*-1 (0.0126)

CCSC_t 0 X employee_ t 0-1 0.0539 SC_t 0 X employee_ t 0-1 -0.0550 (0.0864) (0.0462) CCSC_t 0 X profit _ t 0-1 2.40e-06* SC_t 0 X profit _ t 0-1 -2.55e-08 (1.39e-06) (1.56e-07) CCSC_t 0 X yearly sales _ t 0-1 0.0340 SC_t 0 X yearly sales _ t 0-1 -0.0169 (0.0582) (0.0328) CCSC_t 0 X total other capital _ t 0-1 -0.0237 SC_t 0 X total other capital _ t 0-1 0.0151 (0.0347) (0.0124) Observations 51,531 Observations 51,535 R-squared 0.104 R-squared 0.049 Number of idpanel 9,829 Number of idpanel 9,829 Year dummy YES Year dummy YES Firm FE YES Firm FE YES Industry*Year dummy YES Industry*Year dummy YES Province*year dummy YES Province*year dummy YES Connection pattern* year dummy YES Connection pattern* year dummy YES Ownership type*year dummy YES Ownership type*year dummy YES Robust standard errors clustered by firm in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: t* denotes the year into connection, t*-1 denotes one year before get into connection t 0 denotes the year out of connection, t 0-1 denotes one year before lose connection

26

Table 7 Capital structure change around years into and out of connection with CCSC (1) (2) (3) (4) (5) (6) (7) VARIABLES State K Collective K Private K Legal person K HK & Makao K Foreign K total K CCSC__t*-1 -0.851** -0.393* -0.00816 1.027** -0.139 0.167 -0.0576 (0.421) (0.234) (0.156) (0.445) (0.127) (0.140) (0.0475) CCSC__t* 2.243*** -0.583** -0.184 -1.389** -0.205 0.126 -0.121** (0.528) (0.294) (0.229) (0.545) (0.173) (0.188) (0.0603) CCSC__t*+1 2.114*** -0.264 -0.0510 -1.395** -0.0102 0.0786 -0.00113 (0.549) (0.295) (0.223) (0.546) (0.189) (0.207) (0.0644) CCSC__t*+2 0.289 -0.102 0.208 -0.290 0.00623 -0.137 -0.0916 (0.507) (0.234) (0.214) (0.482) (0.124) (0.109) (0.0657) CCSC_t 0 0.576 0.200 -0.299 -0.668 -0.0214 0.0891 -0.0421 (0.529) (0.309) (0.195) (0.502) (0.148) (0.144) (0.0636) CCSC_t 0+1 0.431 0.360 -0.0177 -0.697 0.0709 -0.328* -0.121 (0.565) (0.407) (0.287) (0.588) (0.230) (0.194) (0.0912) CCSC_t 0+2 -0.721 0.334 0.286 0.0856 -0.0619 0.0422 -0.0112 (0.485) (0.360) (0.177) (0.506) (0.206) (0.205) (0.0626) Observations 3,748 3,748 3,748 3,747 3,748 3,748 3,747 R-squared 0.325 0.331 0.397 0.299 0.589 0.568 0.206 Number of firms 524 524 524 524 524 524 524 Year dummy YES YES YES YES YES YES YES Firm FE YES YES YES YES YES YES YES Industry*Year dummy YES YES YES YES YES YES YES Province*year dummy YES YES YES YES YES YES YES Connection pattern* year dummy YES YES YES YES YES YES YES Ownership type*year dummy YES YES YES YES YES YES YES

Robust standard errors clustered by firm in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: t* denotes the year into connection, t*-1 denotes one year before get into connection, t*+1 one year after, etc. t 0 denotes the year out of connection, t 0+1 denotes one year after the year losing connection, etc. Sample limit to firms who do get(lose) more state capital when step into(out of) connection with CCSC

27

Table 8 Firm Characteristics change around year into/out of connection with CCSC (1) (2) (3) VARIABLES Employee Yearly sales Profit CCSC__t*-1 0.0340 -0.0256 -9,227 (0.0525) (0.0677) (6,945) CCSC__t* -0.0375 -0.00967 -25,152 (0.0647) (0.0878) (21,471) CCSC__t*+1 -0.0164 -0.000982 -22,156 (0.0788) (0.0909) (23,041) CCSC__t*+2 -0.0181 0.0665 2,071 (0.0582) (0.0743) (8,475) CCSC_t 0 0.0186 0.0669 -17,421 (0.0507) (0.0883) (15,418) CCSC_t 0+1 -0.000117 0.0447 -43,304 (0.0684) (0.0986) (28,067) CCSC_t 0+2 0.0189 0.0740 -28,790 (0.0548) (0.0867) (31,900) Observations 3,748 3,748 3,748 R-squared 0.282 0.264 0.440 Number of idpanel 524 524 524 Year dummy YES YES YES Firm FE YES YES YES Industry*Year dummy YES YES YES Province*year dummy YES YES YES Connection pattern*year dummy YES YES YES Ownership type*year dummy YES YES YES

Robust standard errors clustered by firm in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: t* denotes the year into connection, t*-1 denotes one year before get into connection, t*+1 one year after, etc. t 0 denotes the year out of connection, t 0+1 denotes one year after the year losing connection, etc. Sample limit to firms who do get(lose) more state capital when step into(out of) connection with CCSC

28

Table 9 Firm Characteristics change around year into/out of connection with SC (1) (2) (3) VARIABLES Employee Yearly sales Profit SC__t*-1 -0.00427 0.119 5,048 (0.0770) (0.0741) (8,096) SC__t* 0.0528 0.0917 23,572** (0.0756) (0.0825) (11,187) SC__t*+1 -0.0291 0.0681 11,546 (0.0760) (0.0757) (15,501) SC__t*+2 -0.0292 0.0231 -1,128 (0.0678) (0.0731) (17,542) SC_t 0 -0.00404 -0.165* -21,591 (0.0708) (0.0863) (13,139) SC_t 0+1 -0.0561 -0.203* -18,282 (0.0764) (0.112) (27,864) SC_t 0+2 -0.0110 -0.177* -2,374 (0.0924) (0.0960) (13,344) Observations 2,878 2,878 2,878 R-squared 0.353 0.350 0.467 Number of idpanel 389 389 389 Year dummy YES YES YES Firm FE YES YES YES Industry*Year dummy YES YES YES Province*year dummy YES YES YES Connection pattern*year dummy YES YES YES Ownership type*year dummy YES YES YES

Robust standard errors clustered by firm in parentheses *** p<0.01, ** p<0.05, * p<0.1 Note: t* denotes the year into connection, t*-1 denotes one year before get into connection, t*+1 one year after, etc. t 0 denotes the year out of connection, t 0+1 denotes one year after the year losing connection, etc. Sample limit to firms who do get(lose) more state capital when step into(out of) connection with SC

29