Policy Research Working Paper                  9261




   The Transformative Effects of Privatization
                  in China
 A Natural Experiment Based on Politician Career Concern

                             Zhangkai Huang
                                 Jinyu Liu
                              Guangrong Ma
                              Lixin Colin Xu




Development Economics
Development Research Group
May 2020
Policy Research Working Paper 9261


  Abstract
 The serious implications of privatizing state-owned enter-                         discontinuity design, the analysis finds that privatizations
 prises for politicians, managers, and investors make such                          led to productivity gains of more than 170 percent, an
 decisions highly contingent on firm characteristics and past                       order of magnitude larger than the traditional estimates
 performance, complicating the identification of the privat-                        based on the firm fixed effect specification (including its
 ization effects. A unique opportunity for this identification                      random-growth variant). The paper further finds that the
 arises from a rule of promotion of local politicians based on                      privatization effects are significantly larger when the gov-
 age requirements in China. This paper finds that Chinese                           ernment is less involved in the affairs of local firms. The
 cities whose top officials were older than age 58 were 20                          findings underscore the need to deal with the time-varying
 percent less likely to privatize local state-owned enterprises                     selectivity of privatizations and highlight the crucial role
 during the wave of state-owned enterprise restructur-                              that state-owned enterprise privatizations played in China’s
 ing starting in the late 1990s. Relying on the regression                          economic takeoff.




 This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the
 World Bank to provide open access to its research and make a contribution to development policy discussions around the
 world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may
 be contacted at lxu1@worldbank.org.




         The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
         issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
         names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
         of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
         its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                       Produced by the Research Support Team
                      The Transformative Effects of Privatization in China:
                  A Natural Experiment Based on Politician Career Concern 1



                    Zhangkai Huang         Tsinghua University
                           Jinyu Liu       Univ. of International Business and Economics
                     Guangrong Ma          Renmin University of China
                     Lixin Colin Xu        World Bank




Keywords: privatization, career concern, politicians, productivity, China.
JEL Code: D22, D23, L29, H19, P31, P39




1
  Zhangkai Huang: School of Economics and Management, Tsinghua University, Beijing 100084, China (e-mail:
huangzhk@sem.tsinghua.edu.cn); Jinyu Liu: School of Banking and Finance, University of International Business
and Economics, Beijing 100029, China (e-mail: jinyu.liu@uibe.edu.cn); Guangrong Ma: China Financial Policy
Research Center, School of Finance, Renmin University of China, Beijing 100872, China (e-mail: grma@ruc.edu.cn);
Lixin Colin Xu (the corresponding author): World Bank, the Research Group, 1818 H Street, N.W., Washington, DC
20433 (e-mail: lxu1@worldbank.org). Word count: 15294.
      We thank Daniel Xu for excellent editorial service. The results and opinions present in this paper are our own,
and do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent.
The authors declare that they have no relevant or material financial interests that relate to the research described in
this paper.
1. Introduction
After the Thatcher government popularized large-scale privatizations four decades ago, the
policy has been widely used throughout the world, especially by developing and transition
countries with large shares of state-owned enterprises (SOEs). The literature surrounding the
effects of privatization is huge, as shown in several influential surveys across different
generations of research (Megginson and Netter 2001; Djankov and Murrell 2002; Estrin et al.
2009; Megginson 2017). These studies tend to find modest gains in firm performance, while
acknowledging the selectivity of privatizations. In this paper, we revisit this literature by
addressing three questions: What determines a politician’s decision to privatize an SOE? How
large are the causal effects of privatization in China? How do the effects of privatization depend
on local government activism (i.e., its direct involvement in local firms)?
    To make sense of the large literature on privatization effects, we need to consider several
facts. First, privatizations tend to be selective based on past firm performance and
characteristics. The public or private ownership of firms is a key dichotomy in shaping the
basic structure of an economy, and has occupied the attention of generations of economists
who have debated the merits and flaws of state ownership (Bardhan and Roemer 1992, 1993;
Shleifer and Vishny 1994; Stiglitz 1994). Privatizations of SOEs, thus, not surprisingly, reflect
deliberate considerations of many involved parties. When SOEs are profitable, the politicians
in charge benefit from controlling access to these firms’ cash flows, and through arranging jobs
in these lucrative firms. When SOEs are unprofitable, the government shoulders the burden of
keeping them afloat. Thus, privatization decisions are not made in pure pursuit of efficiency.
Instead, they are made to benefit politicians and advance their careers. The types of firms that
can be privatized also depends on investor demand. Highly non-profitable firms do not attract
buyers and are difficult to privatize. Most existing studies on the effects of privatization rely
on the firm fixed effect model, and do not otherwise deal with selectivity of privatization.
Second, the causal effects of privatizations should differ across countries and periods. A
convincing conclusion from the survey of privatization literature is that the effects of
privatization differ by the institutional context of privatizations (Djankov and Murrell 2002;
Estrin et al. 2009): more positive effects or associations, for instance, have been consistently

                                                2
found in CEE (Central and Eastern Europe) rather than CIS countries (i.e., Commonwealth of
Independent States, countries formerly affiliated with the Soviet Union). A thorough
understanding of privatization effects thus requires us to account for the selectivity of
privatization, and the political and economic contexts in which privatizations take place.
       Surveys agree that a glaring shortcoming in the literature surrounding the aftermath of
privatization is a lack of convincing studies on the causal effect of privatization on firm
performance in China. Most studies on privatization focus on CEE and CIS countries, with
occasional exceptions on other large countries such as the United States, Mexico, and India
(Megginson and Netter 2001; Djankov and Murrell 2002; Estrin et al. 2009). Much less work
is done on privatization in China. Estrin et al. (2009) concludes, “There are as yet no TFP
studies using data from China that employ robust methodologies and, perhaps because of this,
the available papers find diverse results, with the effect of nonstate ownership being mostly
positive but sometimes statistically insignificant and sometimes negative” (Estrin et al. 2009,
p. 702).
       This deficiency has not been successfully addressed in the past decade, even though this
period has witnessed strong interest in, and more studies on, privatizations in China. In the
most recent survey on privatization around the world, Megginson (2017) finds that in the new
literature on privatization, papers on China accounted for the largest share among all regions.
However, most of the papers are on share issue privatizations (SIPs), that is, partial
privatization for publicly listed SOEs with the government retaining strong control, 2 and these
account for a tiny share of privatizations in China. Under the slogan of “grabbing the big and
letting go the small,” the vast majority of the privatized SOEs under the SOE restructuring
program near the turn of this century were much smaller and non-listed, and they were
afterwards truly private firms (Xu et al. 2005). Since the literature has suggested that the effects
of full privatization are much more pronounced than partial privatization (Li and Xu 2004,
Megginson 2005), the effects of SIPs are not likely to be representative of full privatization
effects in China. Convincing identification of the privatization effect in China remains rare.



2
     See Sun and Tong (2003), Chen et al. (2008), Berkman, Cole and Fu (2010), Li et al. (2011), Tan et al. (2015),
    Liao, Liu and Wang (2014). See also additional references on SIPs in China in Megginson (2017).
                                                          3
       Yet understanding the effects of China’s privatization is particularly relevant in light of
the diverse interpretation of the Chinese experience. Unbeknown to most, China features the
world’s largest privatization program. Indeed, China’s privatization of SOEs at the turn of the
century is the largest privatization program in human history (Megginson, Nash and Van
Randenborgh 1994). The total amount of worldwide revenues raised by privatization, as
estimated by the end of the last century, was $860 billion (D’Souza and Megginson 1999),
while a conservative estimate of the revenues raised from China’s privatization program alone
is US$700 billion (Gan, Guo and Xu 2017). Despite its gigantic privatization program, China’s
government involvement in the economy has also been exceedingly strong, and China’s
experience has contributed to the popularity of state capitalism. As of 2018, SOEs still
accounted for 27.1 percent of total industrial output, and 13.2 percent of total urban
employment in China. 3 Indeed, China’s post-reform economic system can be characterized as
a dual-track system with both a strong government and strong marketization (Lau, Qian and
Roland 2000). Even though this system was used to jump-start the reform program to reduce
resistance, the dual track of strong marketization and strong government control has remained
intact to this day (Long, Xu and Yang 2020). Is China’s growth a result of privatization,
competition and opening up, or a result of its strong state control, industrial policies and a large
SOE sector?
       Not surprisingly, both of these interpretations have strong proponents. Some pundits
emphasize strong government involvement as fundamental to Chinese growth in the past
decades. Ramo (2004), for instance, coined the term Beijing Consensus, to characterize China’s
unique approach to managing its economy. Others quickly embraced his interpretation and
endorsed strong and direct government involvement in the economy, claiming that the Beijing-
Consensus approach would dominate the 21st century (Halper 2010). 4 Some influential
economists have also embraced the strong government interpretation, including industrial
policies (Lin 2012; Rodrik, 2006; Stiglitz 1994, 2008), or the sufficiency of competition
without changes in state ownership (Lin, Cai, and Zhou 1998). In contrast, many economists
argue that China’s strong growth in the past decades reflects its adoption of standard

3
    The figures are from China Statistical Yearbooks.
4
    See Yao (2015) for a discussion of how the overall Chinese experience corresponds to the Beijing Consensus.
                                                        4
recommendations such as marketization, privatization, and opening up (Yao 2015; Brandt et al.
2016; Megginson, 2017; Wu, 2018; Zhang 2019). Zhang (2019), for instance, offers evidence
that improvement in the level of marketization is strongly and positively associated with
China’s provincial GDP growth rates. Relatedly, Brandt, Kambourov and Storesletten (2016)
offered evidence that the reduction of the state sector contributed to narrowing regional
disparity in the past two decades. Understanding the extent and effects of privatization is thus
critically important for guiding other developing countries to achieve economic growth, and
understanding what China needs to do next. Indeed, Rodrik (2006) implicitly interpreted
China’s experience in the 1990s as evidence against the importance of privatization for
economic success as follows:
        “Rapid economic growth in China, India, and a few other Asian countries has resulted in

        an absolute reduction in the number of people living in extreme poverty. The paradox is

        that that was unexpected too! China and India increased their reliance on market forces, of

        course, but their policies remained highly unconventional. With high levels of trade

        protection, lack of privatization (emphasis added), extensive industrial policies, and lax

        fiscal and financial policies through the 1990s, these two economies hardly looked like

        exemplars of the Washington Consensus. Indeed, had they been dismal failures instead of

        the successes they turned out to be, they would have arguably presented stronger evidence

        in support of Washington Consensus policies” (Rodrik 2006, p. 975).

Given the importance of sound policy recommendations based on the Chinese experience, the
continued misunderstanding of China’s privatization, it is important to identify the causal effect
of privatization in China. This requires using comprehensive Chinese SOE data and a credible
identification strategy.
    Importantly and fortunately, the manner in which Chinese SOEs were selected for
privatization grants us a unique opportunity to identify the causal effects of privatization. The
literature has focused on the selectivity of privatization from the perspective of investors, where
private and foreign investors tend to purchase firms with good performance (Estrin et al. 2009).
The Chinese privatization case differs in this aspect. In China, the SOE restructuring program
was decentralized to the local governments, which implemented the privatization programs of

                                                   5
SOEs under their oversight (Huang et al. 2017). The main incentives of the local governments
were to get rid of loss-making SOEs to unburden the governments. Under the central
government’s mandate to completely restructure, most SOEs, invariably possessing valuable
urban land, could be, and had been, sold. We can thus use, with details to be shown later, rule-
based political incentives of local politicians as the instrumental variable for identifying the
privatization effects.
    Using a large data set that covers all Chinese industrial SOEs from 1998 to 2009, we first
document that the rule-based promotion incentives of local politicians (i.e., at the prefectural-
level city) affected their privatization decisions. As a result of the Chinese Communist Party’s
bureaucratic rule that limited promoting local politicians within a specific age range, their
promotion probability drops sharply and robustly when reaching age 58. Indeed, using a
regression discontinuity framework, we find that local politicians become reluctant to privatize
SOEs under their oversight when they get sufficiently “old” and the underlying political
benefits disappear. To exclude the possibility that other unknown local economic factors might
drive the result, we examine the privatization likelihood of SOEs under the oversight of the
provincial or the central government in the same region, and we do not find it related to the
local politician’s age.
    We next study the impact of privatization on firm productivity. First, without dealing with
time-varying selectivity, we use ordinary least squares, fixed-effects and the firm random
growth specifications. We find that privatization in China is associated with productivity
improvements within the range of effects found in the earlier literature on Eastern European
and CIS countries that were formally affiliated with the Soviet Union, with the magnitude on
the high end of the range. Since privatization is not a random event, we then address the time-
varying selectivity of privatizations with the dummy variable of the local politician’s age
exceeding 58. Results from the instrumental variable estimation show that privatizations lead
to dramatic increases in firms’ productivity by more than 170 percent, an order of magnitude
higher than that based on the workhorse model of firm fixed effects. Privatizations also have
strong and positive effects on profitability. Moreover, we find the privatization effects are



                                                6
higher in regions with lower government intervention, which suggests that marketization
facilitates effective privatization.
     Our paper adds to the literature on privatization effects around the world. Most of the
previous studies find that privatization tends to be associated with better firm performance,
while some other studies show that government ownership is not necessarily less efficient than
private ownership. 5 Part of the reason for this diversity in the findings is no doubt due to the
selective nature of privatization (Djankov and Murrell 2002; Estrin et al. 2009; Dinc and Gupta
2011). Perhaps because of the difficulties in finding excludable instruments—and the non-
existence of randomization in privatization of SOEs—most existing studies of privatization
rely on before-after changes or the firm fixed effect specification, or the firm-specific random-
                      6
growth model.              Such specifications cannot deal successfully with the selectivity of
privatization with respect to time-varying firm characteristics that might be important for
                  7
privatizations.           We contribute to the discussion by using the natural experiment of
privatizations caused by the rule-based career incentives of local politicians to identify the
causal effects of privatization. Furthermore, we add to the literature on privatization effects by
adding credible evidence from China, a key region that has seen more privatizations than
anywhere else in the world (Megginson 2017), but is lacking credible evidence on the effect of
privatization on productivity (Estrin et al. 2009). By demonstrating large causal effects of
privatization in China, we offer evidence that the privatization of SOEs was a key factor behind
China’s economic takeoff.
     We also contribute to the literature on the politics of privatization. The role of political
factors in shaping privatization decisions is studied both theoretically (Biais and Perotti 2002)



5
  On positive effects of privatizations, see Megginson, Nash and Van Randenborgh (1994), Boubakri and Cosset
(1998), La Porta and Lopez-de-Silanes (1999), D’Souza and Megginson (1999), Megginson and Netter (2001),
Claessens and Djankov (2002), Gupta (2005), Djankov and Murrell (2002), Li and Xu (2004), and Estrin et al.
(2009). On doubts on positive effects of privatization, see Caves and Christensen (1980), Kole and Mulherin
(1997), and Anderson, Lee and Murrell (2000).
6
  A partial list of the papers relying on the firm fixed effects specification includes Megginson, Nash and Van
Randenborgh (1994), Boubakri and Cosset (1998), La Porta and Lopez-de-Silanes (1999), Frydman et al. (1999),
Claessens and Djankov (2002), D’Souza, Megginson and Nash (2005), Boubakri et al. (2005), Bai, Lu and Tao
(2009), Jefferson and Su (2006), and Brown, Earle and Telegdy (2006, 2016).
7
  Dinc and Gupta (2011), in a more credible identification strategy, use political incentives of local politicians in
India as the instrumental variable for privatization to identify the effects of privatization.
                                                         7
and empirically (Clarke and Cull 2002; Li and Xu 2002; Megginson 2005; Dinc and Gupta
2011). The privatization decisions are found to depend on firm performance (Du and Liu 2015),
local economic conditions (Clarke and Cull 2002), politic costs in terms of local employment
losses (Guo and Yao, 2005; Dinc and Gupta 2011; Gan, Guo and Xu 2017), and political
benefits in terms of strength of pro-privatization interest groups (Li and Xu 2002). We
contribute by demonstrating that in a non-democratic country, the rule-based career concerns
of local politicians materially affect privatization decisions and thus economic efficiency.


2. Institutional Background

The political promotion system in China had a pronounced effect on the implementation of the
privatization program.
Politician Promotion System in China
Different from the election-based selection of officials in democracies, China has a one-party
top-down appointment system where local officials are appointed by the Organization
Department of the Communist Party in the ladder directly above in the hierarchy (McGregor
2010). The promotion of a politician depends on their ability to deliver on key priorities of the
central government such as economic growth, fiscal revenue, the maintenance of political order,
among others (Maskin, Qian, and Xu 2000; Li and Zhou 2005; Xu 2011; Shih, Adolph and Liu
2012; Jia, Kudamatsu and Seim 2015; Yao and Zhang, 2015).
      The incentives of Chinese local politicians change over their political careers, declining
dramatically as they cross certain age thresholds (Kou and Tsai 2014; Gao, Long and Xu 2016).
Age restrictions were first introduced as promotion regulations in the early 1980s, as part of
the effort by Deng Xiaoping to retire elderly politicians and promote the vigor of the cadre
body. 8 Under this regulation, all politicians except members of the Politburo faced strict non-
promotion ages. For senior politicians at the level of provincial governors or ministers, the
mandatory retirement age is 65. For all other politicians at lower levels, the mandatory non-
promotion age is 60. The result is a rigid and step-by-step promotion timeline with few



8
    See The Decision by the Central Committee of the Communist Party of China on the Establishment of a
Retirement System for Old Cadres (No. 13, 1982) (i.e.
                                                   《中共中央关于建立老干部退休制度的决定》通知, 中发 13 号, 1982).
                                                   8
exceptions. Strict non-promotion ages, in combination with designated and lengthy periods of
tenure at each level, cause career stagnation a few years prior to the mandated retirement age.
The retirement age of mayors and party secretaries at the prefectural city level is set at 60, and
this would not change even if they are promoted to the next level of positions, e.g. deputy
provincial governors. If the politician has not finished his five-year tenure when they reach the
retirement age, they are usually given a maximum of two-year extension to finish their tenure.
If the local politician is promoted to the next higher position after age 58, he will not be able
to finish his tenure. To avoid this disorder in the tenure system, when a local politician reaches
age 58, he is unlikely to be promoted—instead, he is likely to retire or be transferred to a
ceremonial position (Xi, Yao and Zhang 2018). As a result, an upper bound of age 57 is
implicitly imposed for officials at the prefecture level or below, and this rule is commonly
referred to as “Seven-Up, Eight-Down (Qishangbaxia).”
Privatization in China
After China started market-oriented reforms in 1978, the profits and taxes per unit of net capital
stock and working capital in industrial SOEs had fallen from 24 percent in 1978 to 7 percent
in 1996 (Qian 2000), and more than one-third of SOEs had losses in 1996. Starting from the
mid-1990s, China started to rejuvenate its ill-performing SOEs by incorporating private and
foreign shareholders. Meanwhile, the affiliation and regulatory power of many SOEs was
commensurately decentralized to the local governments (Xu, Zhu and Lin 2005; Xu 2011). The
rising losses of SOEs and the heavy fiscal burdens on local governments fueled the pace of
large-scale privatization in the late 1990s. In 1997, the central government officially announced
its policy to restructure the state sector by allowing local governments to experiment with
different ways to restructure SOEs. To facilitate economic turnaround, local governments were
encouraged to privatize SOEs. In the Annual Survey of Industrial Firms (ASIF) used in this
study, which accounts for the vast majority of Chinese industrial firms in terms of value added
and employment, roughly 80% of local SOEs had been privatized by 2009.
     Given the extraordinary scale of Chinese privatization and its potential implications, the
number of studies on the impact of Chinese privatization is surprisingly small (Bai, Lu and Tao



                                                9
2009; Du and Liu 2015; Gan, Guo and Xu 2017). 9 One of the reasons for this scarcity might
be the difficulty in dealing with the selection bias in China’s privatization process. The slogan
of the restructuring reform was to “grab the big and to let go of the small.” To “grab the big,”
large and important SOEs were corporatized and consolidated; to “let go the small,” small and
loss-making SOEs were mostly privatized. The government’s discretion and selection in the
privatization process makes the evaluation of the causal impact of privatization complicated.
To identify the causal effects of privatization in China, we explore exogenous factors in the
political promotion mechanism that could be used to address the selection bias associated with
privatizations.
        In China there might be a strong link between local politicians’ promotion incentives and
the privatization decision of local SOEs, which was under the control of the local government
(Huang et al. 2017). Privatization entails significant costs for politicians, depriving their overt
political connections with SOEs under their oversight, and therefore the associated control
benefits. Moreover, privatization is usually followed by massive layoffs as part of the
restructuring efforts, especially for the loss-making SOEs in this period. Indeed, between 1997
and 2002, over 27 million SOE workers, or about 27% of total SOE employment in 1997, had
been laid off (Dong and Xu 2008). These layoffs would cost politicians some local support,
though such support is not essential for their careers. On the other hand, privatization likely
enhances corporate performance and thus delivers higher fiscal revenues and GDP growth
(Chen et al. 2008; Bai, Lu and Tao 2009; Calomiris, Fisman and Yang 2010). Higher fiscal
revenue and GDP growth likely boost the politician’s chance of promotion (Maskin, Qian, and
Xu 2000; Li and Zhou 2005; Xu 2011; Shih et al. 2012; Persson and Zhuravskaya 2016). Thus,
in deciding whether to privatize a local SOE, local politicians face the trade-off between the
benefit of economic gains and the cost of local unemployment and of the loss of control benefits.
At the time of financial losses of SOEs in the late 1990s, due to sharp drops in rent associated
with SOEs, political resistance to privatizations under the Chinese political system should be
much lower than in the democratic societies, and the importance of local politicians looms
especially large for privatization decisions. In particular, local politicians younger than 58,


9
    See in the introduction other references on privatizations, especially share issue privatizations, in China.
                                                           10
facing stronger likelihood of promotion that would be enhanced by better economic
performance coming with privatization, would be more eager to seek privatizing SOEs under
their oversight.


3. Data
Our data set is the Annual Survey of Industrial Firms (ASIF) from 1998 to 2009 collected by
China’s National Bureau of Statistics. It includes all SOEs and all non-state firms with sales
above five million yuan in the industrial sectors, accounting for about 90% of total industrial
output value in China. 10 Since our goal is to study the incentives of local politicians (i.e.,
politicians at the prefectural city level or below), 11 we only keep SOEs that are affiliated with
local governments (defined as those SOEs with the share of local government ownership
exceeding 50 percent) in our baseline regressions. In constructing our panel, we follow the best
practices in fixing some identification inconsistencies. 12
     We construct a dummy variable, Privatized, that equals one for a firm for the privatization
year and the subsequent years if it is privatized during the sample years and 0 otherwise. An
SOE is classified as being privatized in two ways: staying in the database with the state share
dropping below 50 percent; exiting the data set, which implies privatization, and its sales being
below five million yuan or being merged with small private firms. Out of 50,030 SOEs in our
sample, 40,557 had been privatized by the end of our sample period.
     We match the ASIF data of SOEs with a data set on local politicians in China, i.e., party
secretaries in the prefectural-level cities. The politician data set has details of their personal
information (age, gender, and education), as well as their career path (the appointment date, the
next deputation, and promotions/demotions). We use provincial yearbooks for the names of
city-level party secretaries, and search their resumes to identify their personal characteristics
and career trajectory. 13 Macroeconomic variables including the city-level GDP and the

10
   The industrial sector in this data set includes mining and manufacturing.
11
   Whenever referring to “local” we mean the prefectural-level city or below, which includes the county-level.
12
   A common issue that must be addressed when constructing a panel drawn on the ASIF data is that a small
proportion of firms occasionally receives a new identification code due to restructuring, merger, or acquisition.
See Appendix A for details on how we follow best practice in dealing with this issue.
13
   The primary internet sources are resumes of the politicians posted on the government website as well as Baidu
Baike (http://baike.baidu.com/), a Chinese version of Wikipedia.
                                                       11
population are from China City Statistical Yearbook. 14 All time-varying variables are adjusted
to the 1990 constant price with the province-specific GDP deflators from the China Statistical
Yearbook.
     Local party secretaries, with the greatest de jure and de facto power at all government
levels in China (McGregor, 2010), have the ultimate decision power on the privatization of
local SOEs. These officials are thus our focus. For the complete list of the 294 prefectural cities
in China, we could collect information on 883 party secretaries in 282 cities from 1998 to 2009.
     Figures 1a and 1b illustrate the age pattern and the age-of-promotion pattern of the city-
level party secretaries in our sample of 2,631 city-year observations. The age of city-level party
secretaries exhibits a hump-shaped distribution from 39 to 61, peaking around the average and
the median age of 52. Most local party secretaries, with only a few exceptions, leave their
positions in their late 50’s. The incidence of promotion falls dramatically after the age of 55
and drop to zero once a politician reaches the age of 58. Overall, these observations corroborate
our earlier discussion of the role of local politicians’ age in their chance of promotion.
     Panels B and C of Table 1 describe the summary statistics of all main and control
variables. 15 About 88% of the city-level party secretaries have higher education. A small
number of local party secretaries have experience working in the central government (4.1%),
while most of them have experience working in provincial governments (55.1%). The SOEs in
our sample generally displayed poor performance: the average leverage ratio is as high as 72%,
whereas the return to asset (ROA) is low, with the averages being -0.3%, making slight losses.


4. Effects of Politicians’ Age on Privatization
We provide evidence that privatization decisions are closely related to the age of city party
secretaries. We first employ an ordinary least square (OLS) regression of the privatization
dummy on a series of age dummies and control variables as follows: 16
                       ������������������������������������������������������������������������������������������������������������������������������������������������������������������������ = ������������0 + ∑������������ ������������������������ ������������������������������������������������������������������������ + ������������������������������������������������������������ + ������������������������ + ������������������������ + ������������������������ + ������������������������������������   (1)

14
   Four municipalities, Beijing, Shanghai, Tianjin and Chongqing are under the direct administration of the central
government and have the same rank as provinces. Thus, our sample does not include these four municipalities.
15
   We report summary statistics for two samples, one used for the determinants of privatization, which cut the
post-privatization observations for a firm, and another used for the effects of privatization, which keep all post-
privatization observations.
16
   The results based on probit analysis are of course very similar and not reported.
                                                                                             12
The dependent variable, ������������������������������������������������������������������������������������������������������������������������������������������������������������������������ , is the dummy variable indicating whether an SOE i

in city c and two-digit industry j is privatized in year t. Since privatization is rarely reversed,
each SOE’s observations after the year of privatization are dropped from the sample for the
current empirical exercise. ������������������������������������������������������������������������ refers to a series of age dummies of the party secretary,
being one if the party secretary in city c and year t is s years old, with s taking the values of 51
and above. We treat party secretaries younger than age 51 as the reference group. We include
city, industry and year fixed effects, and cluster the standard errors at the city level to
accommodate for correlation in the error term within a city.
       We control for firm characteristics, the party secretary’s background, and the city’s macro
conditions (i.e., ������������������������������������������������ ). The firm-level characteristics, all once-lagged, include labor
productivity (Labor prod, the logarithm of the ratio of sales to the number of employees), the
leverage ratio, the logarithm of total employees, and the return on asset (ROA). The
characteristics of party secretaries include the indicator of having college education (College),
the indicator of having work experience in the central government (Central Experience), and
in the provincial government (Provincial Experience). The macroeconomic conditions in the
city include the GDP growth rate and GDP per capita at the city level.
      According to the OLS results in Table 2, the probability of the privatization of local SOEs
drops abruptly when the age of the local party secretary passes the threshold of 58. The
coefficients of the age dummies for age between 51 and 57 are all statistically insignificant,
indicating no differences in the tendency to pursue privatization relative to that of a younger
politician. However, when he becomes 58 years or older, the probability of an SOE proceeding
with privatization drops by 4 percentage points. This is a large magnitude: relative to the mean
of the privatization dummy of 0.209, it represents an increase of the probability of the
privatization of 20.6 percent. 17 The results are robust to the inclusion of various sets of control
variables, such as lagged firm performance, city-level macroeconomic conditions, and the
party secretary’s education and work experience. Local politicians are thus much less likely to
pursue privatization of local SOEs after age 58, when they are passed over for promotion.



17
   The coefficient of the dummy variable of age 60 is insignificant, perhaps reflecting that this age cell has too
few observations (see Figure 1a).
                                                                13
     There is apparent selectivity of privatizations based on firm characteristics. Smaller SOEs
are more likely to be privatized, consistent with the government’s mandate of restructuring
SOEs by “grabbing the big and letting go the small.” SOEs with higher leverage, representing
higher fiscal burden, are more likely to be privatized. More importantly, worse-performing
SOEs are more likely to be privatized, as shown by the negative signs of lagged labor
productivity and profitability. Since all these firm characteristics are time-varying, controlling
for firm fixed effects is unlikely to be sufficient in dealing with the endogeneity of
privatizations.
Regression Discontinuity Design
The OLS estimation may yield biased estimates when there are unobservable differences
between different cities governed by young or old politicians. Because, as discussed earlier,
city party secretaries experience a significant drop in their probability of promotion from age
58 forward, we could rely on the regression discontinuity design (RDD) to address the
selectivity associated with cities governed by young and old politicians. Assuming that
unobserved variables vary smoothly around the age threshold (i.e., age 58), any discontinuity
in the probability of privatization around this threshold should reflect the causal effects of the
age-promotion rule. We use the following specification for the RDD estimation:
         ������������������������������������������������������������������������������������������������������������������������������������������������������������������������ = ������������0 + ������������1 ������������������������������������(������������������������������������ ≥ 58)������������������������ + ������������(������������������������������������������������������������ ) + ������������������������������������������������������������ + ������������������������ + ������������������������ + ������������������������ + ������������������������������������ (2)

Where ������������������������������������(������������������������������������ ≥ 58)������������������������ is a dummy variable that equals one if the local official’s age is
equal to or older than 58, and 0 otherwise. ������������������������������������������������������������ is the running variable, and ������������(������������������������������������������������������������ )
represents a flexible function of the official’s age to account for different slopes on the two
sides of the cutoff point. ������������1 captures the decline in the probability of privatizations after the
cutoff age of 58 for the city party secretary.
      Estimating equation (2) could be done in two complementary ways, which provide a
mutually reinforcing specification check (Lee and Lemieux 2010). The first is the parametric
global polynomial approach in which we can control for a parametric function, i.e., a high-
order polynomial (second or third order), in the running variable, and we use all the available
data to estimate these equations. The second is the nonparametric local linear approach, in
which we use a narrow bandwidth near the cutoff and control for a linear polynomial. We limit

                                                                                                           14
the sample to consist of observations with the age of a city secretary in the range of 55-61 years
old, using a bandwidth of ±3 years of age. 18
     The validity of the RDD approach relies on the “smoothness assumption,” i.e., all the
predetermined firm-level characteristics vary smoothly across the threshold (Imbens and
Lemieux 2008). Therefore, to ensure the validity of the RDD, we conduct balance tests on the
following variables: labor productivity, the leverage ratio, the logarithm of total employees,
and the return on sales, all of which are once-lagged. Specifically, we estimate regressions of
the form described in equation (2) using these characteristics as dependent variables. We use a
local linear approach, and cluster the standard errors at the city level. The results in Table 3
show that the dummy variable of the party secretary being age 58+ is never significant in the
balance test regressions, indicating that these characteristics do not significantly differ across
this age threshold. Predetermined firm-level characteristics are thus balanced between those
firms associated with the group of the older and of the younger party sectaries.
     Columns (1) to (4) of Table 4 report the RDD results. The third-order results are shown in
columns (1) to (2). 19 We focus on the coefficient of the indicator of the party secretary’s age
being 58 and older, which should be significantly negative if there is a sharp decline in the
tendency of local politicians to pursue privatizations. This pattern is confirmed: the coefficient
of the indicator of the party secretary’s age being 58 and older is -0.050 and statistically
significant, representing a drop in the privatization probability at the mean by about a quarter.
The RDD results using the non-parametric local linear approach, in columns (3) and (4) of
Table 4, remain similar. In column (4), the coefficient for the indicator of the party secretary’s
age being 58 and older is significant at -0.048. The results suggest that under China’s age-
promotion profile rule, the career concerns of city-level politicians have profound influence on
their decisions to privatize local SOEs.
Placebo Test: Central and Provincial SOEs
We have shown that city-level politicians, once crossing the age threshold of non-promotion,
are reluctant to privatize local SOEs. It is possible that some city-level factors correlated with


18
   For simplicity, we manually set the regression bandwidth. Note that we have very limited choices of bandwidth
here with the upper bound of age being 62 years old.
19
   The results using the quadratic specification of the running variable are similar.
                                                      15
the age of politicians might play a role in the privatization process. If the age-cutoff effect only
reflects local politicians’ incentives, it should appear only in the privatizations of SOEs under
the control of the local politicians. In other words, the privatization of SOEs under the oversight
of the central or provincial governments, although located in the same cities, was not
determined by local politicians, and these SOEs’ privatization should not be affected by the
age and promotion incentives of city-level politicians.
    We thus conduct a placebo test by exploiting the central and provincial SOEs sample to
rule out the effect of any unobservable city-level factors. Among the sample of 9,813 central
and provincial SOEs, 8,695 of them had been privatized in the sample period. We re-conduct
the RDD regressions with the new sample and expect that the relationship between the age of
the local politician and the privatization of these non-locally-governed SOEs does not change
abruptly. The results are shown in Table 5. The coefficients of the local party secretary’s age
being 58 and older are insignificant in both the parametric and the non-parametric RDD. The
magnitude of all point estimates is much smaller than those reported in Table 4. The placebo
test thus suggests that unobservable local factors do not explain our previous results on the age
threshold effect of local politicians on privatization of local SOEs.
Robustness Test: Other Definitions of Privatization
In our baseline results, we use a broad definition of SOE privatization: an SOE is considered
to be privatized when the firm’s state shares dropping below 50 percent, or when they exit the
database. Since the ASIF data include all SOEs and all private firms with sales above 5 million
yuan, when an SOE exits the database, it could only be three cases: (i) merging with private
firms, which is clearly privatization; (ii) the SOE is privatized and the post-privatization sales
drop below 5 million yuan, which again is privatization; (iii) merging with an SOE, in which
case classifying the exited firm as being privatized causes misclassification. Case (iii) is a likely
a small-probability event, and it can still be regarded as one way of SOE restructuring.
Nevertheless, it is a misclassification of privatization. To check the robustness of our
conclusions, we re-conduct the main regressions using a narrow definition of privatization, i.e.
the criterion of the state share dropping below 50 percent for firms remaining in the database.
We reach similar results using this narrow definition of privatization in both the OLS regression

                                                 16
and the RDD regressions (see appendix B for details). The drops in the probability of
privatization for ages 58 and 59 remain large and statistically significant, with the magnitude
being similar for age 59, and slightly smaller for age 58. When using the various RDD
approaches, the qualitative and quantitative conclusions on the age threshold on the
privatization likelihood also remain robust.


5. Effects of Privatization on Firm Productivity
We first present estimates on the association of privatizations and productivity based on the
conventional specification, with and without, firm fixed effects. We then rely on the promotion
discontinuity in the age of local politicians to identify the causal effects of privatization. We
finally investigate how the privatization effects differ by local government activism.
Effects of Privatization Based on the Conventional Estimators
When estimating the effects of privatization on firm performance, each SOE’s post-
privatization observations are included in the sample. As noted above, our baseline definition
of privatization is broad, that is, an SOE is considered to be privatized when the firm’s state
share dropping below 50 percent, or when they disappear from the database. However, when
examining the effects of privatization, since we must have observations of post-privatization
performance, the “SOE-exiters” are deleted in the sample here, and our broad and narrow
definitions of privatization coincide. Our estimation of the privatization effects is thus subject
to a caveat. The estimated effects do not represent the effect of privatization on privatized firms
that were merged or those privatized firms whose sales fell under the threshold of five million
yuan.
        We first estimate two conventional specifications from the literature: the ordinary least
square (OLS) model, and the firm fixed effect (FE) model:

                    ������������������������������������������������������������ = ������������0 + ������������1 ������������������������������������������������������������������������������������������������������������������������������������������������������������������������ + ������������������������������������������������ + (������������������������ ∗ ������������������������ )′ ������������ + ������������������������ + ������������������������������������ + ������������������������������������   (3)

Here ������������������������������������������������������������ is the performance measures of firm i of city c and industry j in year t. We measure

the performance of the firm using several measures of the total factor productivity, in particular,
those based on the OLS approach (TFP_OLS), those based on the Olley-Pakes approach
(TFP_OP), and those based on the index function approach (TFP_IN); all three measures are

                                                                                                      17
closely correlated, with pair-wise correlation coefficients around 0.95. See appendix C for
details of the constructions of these and other TFP measures. 20 The various measures of TFP
have the virtue of capturing productive efficiency, and using various popular ways to measure
TFP allows us to ensure the robustness of the productivity results. We in addition consider two
alternative outcomes: labor productivity (Labor Prod, measured as the logarithm of sales per
employment) and ROA, respectively. Labor productivity is more transparent, but it does not
capture differences in inputs. We also look at the ROA because improving profitability was an
objective of the SOE restructuring program. The dummy variable, ������������������������������������������������������������������������������������������������������������������������������������������������������������������������ , equals one

for all the years after firm i has undertaken privatization and 0 otherwise. We control for ������������������������������������ ,
the city-level macro variables (i.e., GDP Per capita and GDP Growth). We further control for
industry-year fixed effects, ρjt, and the interaction terms between the firm-level pre-treatment
characteristics, ������������������������ , and year dummies, ������������������������ , in the regressions, thereby flexibly controlling for
the time-varying effects of such characteristics on the outcome variables. In the OLS
specification, we do not, but in the firm fixed-effects specification, we do control for the firm-
level fixed effects γi. The FE specification has the advantage of controlling for all firm-specific
time-invariant factors, but it cannot account for the endogeneity of privatization based on
selection on time-varying firm factors.
      According to the OLS results in Panel A of Table 6, privatizations are invariably associated
with a large effect on firm productivity or profitability. Privatizations are associated with a
higher TFP_OP or TFP_IN by 20-22 log points, 21 and a higher labor productivity by 26 log
points. Privatizations are also associated with an increase in profitability by 1.1 percentage
points.
      According to the firm-FE results in Panel B of Table 6, the magnitudes of the privatization
effects are in the same ballpark as the OLS estimates, though slightly smaller. Privatization is
associated with an increase in TFP_IN and in TFP_OP by 14 and 16 log points, respectively.
The TFP effects are all slightly smaller than the labor productivity effects of 17.7 log points.

20
   In constructing the index function TFP (TFP_IN), we rely on factor shares based on OECD countries (Bentolila
and Saint-Paul 2003). We have also tried using the factor shares based on our own data and obtained qualitatively
similar conclusions. Using factor shares based on our own data is inappropriate since the factor shares reflect
choices associated with privatization itself and are thus contaminated. See appendices C and D.
21
   We focus on TFP_OP and TFP_IN because they better deal with various biases associated with estimating
TFP using the OLS method. But results using all three TFP measures are qualitatively very similar.
                                                                 18
The FE estimate of the privatization effects on productivity is located at the high end of the
spectrum of the privatization effects in the literature. For instance, Brown, Earle and Telegdy
(2006) find the productivity effects to be around 15-50 percent in Romania, 8-28 percent in
Hungary, 2-16 percent in Ukraine, and -5 to 14 percent in Russia. Estrin et al. (2009) similarly
would place our FE estimate of around 15 percent at the high-end of the estimates in Eastern
European and the CIS countries. In addition, privatization is positively associated with ROA
by 1.2 percentage points, or an increase of 0.1 standard deviation (11.2 percent).
    The preferred specification in two key papers of the literature on privatization effects
(Brown, Earle and Telegdy 2006; Brown, Earle, and Gehlbach 2009), likely an important
improvement over the firm fixed effects model, is to allow for firm-specific random growth
rates, as in the impact evaluation literature (Heckman and Hotz 1989). In particular, besides
firm fixed effects, this fixed-effects and fixed-trend (FE-FT) specification allows for firm-
specific fixed effects and growth rates, as follows:
         ������������������������������������������������������������ = ������������1 ������������������������������������������������������������������������������������������������������������������������������������������������ + ������������������������������������������������ + (������������������������ ∗ ������������������������ )′ ������������ + ������������������������ + ������������������������������������ + ������������������������ + ������������������������������������   (5)

In practice, the FE-FT model is estimated in two steps: first detrending all variables for each
firm separately, and then re-estimating the model with these detrended variables. The results
are shown in Panel C of Table 6.
    The effects based on the FE-FT specification are qualitatively similar to the FE estimation,
but significantly smaller. For instance, the effect of privatization based on TFP_IN is 2.6 log
points, roughly 20 percent of the FE estimate. Similarly, the FE-FT estimate of the effect of
privatization on labor productivity is 5.7 log points, or about 30 percent of the FE estimate.
These FE-FT estimates would place the productivity effects of privatization in the middle of
the range found in the literature (see Estrin et al. 2009).
    Based on the OLS, the FE, and the FE-FT estimations, privatizations in China had positive
associations with productivity, and the magnitude ranges from moderate (such as 3 log points
based on the FE-FT specification) to large (such as 16 log points based on the FE specification).
Instrumental Variable Estimation
As shown before, the privatization of SOEs is strongly related to time-varying firm
characteristics. To deal with the endogeneity issue, we use the following two-stage, least

                                                                                                         19
squares specification under a fuzzy regression discontinuity framework. We use the dummy
variable of the local politician’s age being 58 and older (i.e., Dum(Age≥58)) as the instrument
variable for the privatization decision. We control for the second-order polynomial of the
running variable (i.e., the politician’s age) and other control variables as in equation (3).

                                                                                                                                                                                                  ′
������������������������������������������������������������������������������������������������������������������������������������,������������������������������������ = ������������0 + ������������1 ������������������������������������(������������������������������������ ≥ 58)������������������������ + ������������1 (������������������������������������������������������������ ) + ������������1 ������������������������������������ + (������������������������ ∗ ������������������������ ) ������������������������ + ������������1������������ + ������������1������������������������ + ������������������������������������

������������������������,������������������������������������ = ������������0 + ������������1 ������������������������������������������������������������������������������������������������������������������������������������,������������������������������������ + ������������2 (������������������������������������������������������������ ) + ������������2 ������������������������������������ + (������������������������ ∗ ������������������������ )′ ������������������������ + ������������2������������ + ������������2������������������������ + ������������������������������������       (6)

            The IV regressions are reported in Panel D of Table 6. The first-stage results show that
Dum(Age≥58) is negatively related to privatization, as documented before. The F-statistics of
Dum(Age≥58) in the first stage are close to 20, suggesting that the instrument variable is not
weak (Staiger and Stock 1997). The second-stage estimation suggests that after dealing with
endogeneity, privatization improves TFP_IN by around 100 log points (or 173 percent),
TFP_OP by 143 log points (or 318 percent), an order of magnitude larger than the FE- or FE-
FT-based estimates. Similar patterns also exist for labor productivity and ROA. The effects of
privatization on profitability, for instance, is many times higher than those based on the FE or
the FE-FT specifications.
            Since the literature places the effects of privatization somewhere between no effects (or
even negative effects) and large (i.e., more than 15 percent) (Brown, Earle and Telegdy 2006;
see also Estrin et al. 2009), our estimates of the causal effects of privatization on productivity
is an order of magnitude larger than what the literature finds in the rest of the world. Thus, the
OLS, the firm-FE and the FE-FT specifications likely substantially underestimates the causal
effects of privatization. This is not surprising, since we know from the institutional background
that privatization is likely negatively selected on time-varying firm characteristics: under the
Chinese SOE restructuring/privatization program, local governments explicitly tried to let go
of loss-making SOEs, and had strong incentives to keep well-performing SOEs for control
benefits. The findings here suggest that the causal effects of privatization in China are, in
magnitudes, not marginal or modest, as suggested by the conventional estimators based on the
OLS of the firm-FE estimation, but transformational based on the plausible instrumental
variable.
Government Influence and the Effects of Privatization on TFP
                                                                                                                       20
Chinese local governments have been heavily involved in local firms, especially in SOEs and
privatized firms (Cull et al. 2015; Harrison et al. 2019). As discussed earlier, pundits have
argued over whether Chinese growth has been due to heavy-handed government guidance or
allowing the market to work. To shed light on this key concern, here we examine how the
effects of privatization depend on government activism in dealing with firms in general. The
more the government remains active in dealing with firms, the less privatized firms act like
true private firms, and thus the effects of privatization are likely smaller. Indeed, Harrison et
al. (2019) offer evidence that Chinese privatized firms behave like a mixture of SOEs and
private firms, still enjoying subsidies from the government, and performing more poorly than
true private firms but nevertheless having significantly better performance than SOEs. Thus,
where the government is more active, privatized firms are more likely to behave like SOEs,
and we expect the effects of privatization to be more muted.
     Our proxy of the extent of government activism is based on the World Bank Enterprise
Survey data in 120 Chinese cities in 2005. In the survey, the firm is asked to rate “the percent
of officials in various government departments that facilitate the development of the firm,”
which captures the strength of interactions between the government and the firm. This variable
is averaged to the city level to capture regional variations. In Panel A of Table 7, we provide
the firm-FE and the IV estimates of the effect of privatization that hinges on our proxy of
government activism. 22
     Consistent with our expectation, the effects of privatization are significantly lower where
local government activism is stronger. Relying on the IV estimate and TFP_IN, the average
effects of privatization on productivity is 102 log points at the mean level of local government
activism (i.e., 0.407), and it is 180 log points at one standard deviation (i.e., 0.213) below the
mean. Government activism thus reduces the effects of privatization.




22
   Using the same logic as before, the privatization dummy’s instrumental variable is the dummy of the age of the
local party secretary is 58 or older, and the instrumental variable for the interaction term of privatization and local
government activism is the latter times the dummy of the age of the local party secretary being 58 or older.
                                                          21
6. Conclusion
Our paper deals with two critical issues in the literature of privatization: the lack of convincing
identification of privatization effects due to the selectivity on time-varying firm characteristics,
and the lack of understanding of the causal effects of privatization on productivity in the
country where the scale of privatization has been the greatest. To address these issues, we take
advantage of the bureaucratic rules surrounding the promotions of local politicians, wherein
their chance of promotion drastically drops once crossing the age threshold of 58. After
empirically confirming this pattern, we use the dummy variable of city party secretaries
crossing the age threshold as the instrument to identify the effects of privatization in firm
productivity regressions, while allowing the age of the city party secretaries to have direct
effect. We find that, without controlling for time-varying selectivity and using the OLS, the
firm fixed-effects and/or the firm random growth specifications, privatization in China is
associated with productivity improvements within the range of effects found in earlier literature
on Eastern European and CIS countries, on the relatively high end of the spectrum of those
estimates. However, once addressing the time-varying selectivity of privatization with our
RDD instrument, the effects of privatization increase to more than 170 percent, an order of
magnitude higher than the workhorse fixed-effects estimate. Privatizations also have strong
and positive effects on profitability. Privatizations thus drastically transform the loss-making
SOE sectors. Moreover, we find the privatization effects to be much higher in regions with
lower government activism, which implies that marketization and privatization are
complements.
    Our findings have several implications. First, in light of the large disparity between our
RDD-based instrumental variable estimates and those of conventional fixed-effects-based
specifications, the causal effects of privatization were likely substantially underestimated in
the past. In other words, negative selectivity based on time-varying characteristics of SOEs
could be of first-order importance. Future studies of the effects of privatization should seriously
address the importance of selectivity on time-varying factors. This selectivity is indeed intuitive:
SOE performances vary greatly over time, depending on the entry barriers to non-state firms,
market competition, among other factors, all of which vary over time. Only when SOEs

                                                 22
perform badly and pose serious fiscal burdens does the government then have strong incentives
to privatize, as in the case of China in the late 1990s in the Zhu Rongji era. Second, because
China features the largest privatization program in the world, and we have found the
privatization effects to be transformative, the privatization program must be viewed as a critical
component for China’s growth since the mid-1990s. Recently, in many countries and perhaps
especially in China, state capitalism where the government is strongly involved in running
corporations has become more popular (Megginson 2017). Reevaluating the role of
privatization in China’s growth has thus become even more important. In light of SOEs’ non-
trivial share of the Chinese economy (Huang et al. 2017) and China’s slowing down in growth,
further privatizations should clearly be considered for the future. Relatedly, our finding that
stronger local government activism in firm affairs reduces the effects of privatization also
underscores the importance of reducing the interference of governments in firms’ business
affairs.
     Our findings here should not be interpreted as implying huge causal effects of privatization
in every institutional context. The privatization literature has emphasized repeatedly that the
effects of privatization depend critically on the underlying institutional background. When
China’s then premier Zhu Rongji undertook the privatization program in the late 1990s, SOEs
were on average losing money, and their performance was at the lowest level possible. Political
opposition to privatization was low, and local governments could experiment with various
ways to privatize SOEs. Moreover, the liberalization reforms in the 1980s set up a solid
platform for privatized firms to prosper (Li, Li and Zhang, 2000; Huang 2012). The huge causal
effects of privatization for the Chinese SOE privatization program thus have their favorable
pre-conditions and are likely to hold only in countries with a large loss-making SOE sector and
favorable pre-privatization conditions such as strong competition. Nevertheless, our findings
imply that in many contexts, the effects of privatization are likely under-estimated in important
ways. Thus, future research in different institutional contexts that relies on credible
identification strategies is important to uncover the distribution of causal effects of
privatizations around the world.



                                                23
Reference
Anderson, J. H., Y. Lee, and P. Murrell. 2000. Competition and privatization amidst weak
  institutions: evidence from Mongolia. Economic Inquiry 38(4), 527-549.
Bai, C. E., J. Lu, and Z. Tao. 2009. How does privatization work in China? Journal of Comparative
  Economics 37(3), 453-470.
Bardhan, P., and J. E. Roemer. 1992. Market socialism: a case for rejuvenation. Journal of Economic
  Perspectives 6(3), 101-116.
Bardhan, P., and J. E. Roemer. 1993. Market socialism: the current debate. Oxford University Press,
  USA.
Bentolila, S., and G. Saint-Paul. 2003. Explaining movements in the labor share. Contributions in
  Macroeconomics 3(1).
Berkman, H., R. A. Cole, and L. J. Fu. 2010. Political connections and minority shareholder
  protection: Evidence from securities-market regulation in China. Journal of Financial and
  Quantitative Analysis 45(6), 1391-1417
Biais, B., and E. Perotti. 2002. Machiavellian privatization. American Econ. Rev. 92(1), 240-258.
Boubakri, N., and J. C. Cosset. 1998. The financial and operating performance of newly privatized
  firms: evidence from developing countries. Journal of Finance 53(3), 1081-1110.
Boubakri, N., J. C. Cosset, K. Fischer, and O. Guedhami. 2005. Privatization and bank performance
  in developing countries. Journal of Banking and Finance 29(8-9), 2015-2041.
Brown, J. David., J. S. Earle, S. Gehlbach. 2009. Helping hand or grabbing hand? state bureaucracy
  and privatization effectiveness. American Political Science Rev. 103(2), 264-283.
Brown, J. D., J. S. Earle, and A. Telegdy. 2006. The productivity effects of privatization:
  longitudinal estimates from Hungary, Romania, Russia, and Ukraine. Journal of Political
  Economy 114(1), 61-99.
Brown, J. D., J. S. Earle, A. Telegdy. 2016. Where does privatization work? Understanding the
  heterogeneity in estimated firm performance effects. J. of Corporate Finance 41(1), 329-362.
Brandt, L., J. Van Biesebroeck, and Y. Zhang. 2012. Creative accounting or creative destruction?
   firm-level productivity growth in Chinese manufacturing. Journal of Development
   Economics 97(2), 339-351.
Brandt, L., G. Kambourov, and K. Storesletten. 2016. Firm entry and regional growth disparities:
   the effect of SOEs in China. Working Paper, University of Toronto.
Calomiris, W. C., R. Fisman, and Y. Yang. 2010. Profiting from government stakes in a command
  economy: evidence from Chinese asset sales. Journal of Financial Economics 96(3), 399-412.
Caves, D. W., C. Laurits. 1980. The relative efficiency of public and private firms in a competitive
  environment: the case of Canadian railroads. J. of Political Economy 88(5), 958-976.
Chen., G., M. Firth, Y. Xin, and X. Liping. 2008. Control transfers, privatization, and corporate
  performance: Efficiency gains in China's listed companies. Journal of Financial and
  Quantitative Analysis 43(1), 161-190.
Clarke, G. R., and R. Cull. 2002. Political and economic determinants of the likelihood of privatizing
   Argentine public banks. The Journal of Law and Economics 45(1), 165-197.

                                                  24
Claessens, S., and S. Djankov. 2002. Privatization benefits in Eastern Europe. Journal of Public
   Economics 83(3), 307-324.
Cull, R., W. Li, B. Sun, and L. C. Xu. 2015. Government connections and financial constraints:
  evidence from a large representative sample of Chinese firms. Journal of Corporate Finance
  32(1), 271-294.
Dinc, I. S., and N. Gupta. 2011. The decision to privatize: finance and politics. The Journal of
  Finance 66(1), 241-269.
Djankov, S., and P. Murrell. 2002. Enterprise restructuring in transition: a quantitative
   survey. Journal of Economic Literature 40(3), 739-792.
Dong, X. Y., and L. C. Xu. 2008. The impact of China’s millennium labor restructuring program
  on firm performance and employee earnings. Economics of Transition 16(2), 223-245.
D'souza, J., and W. Megginson. 1999. The financial and operating performance of privatized firms
   during the 1990s. The Journal of Finance 54(4), 1397-1438.
D'Souza, J., W. Megginson, and R. Nash. 2005. Effect of institutional and firm-specific
   characteristics on post-privatization performance: evidence from developed countries. Journal
   of Corporate Finance 11(5), 747-766.
Du, J., and X. Liu. 2015. Selection, staging, and sequencing in the recent Chinese privatization. The
  Journal of Law and Economics 58(3), 657-682.
Estrin, S., J. Hanousek, E. Kŏcanda, and J. Svenjnar. 2009. Effects of privatization and ownership
   in transition economies. Journal of Economic Literature 47(3), 699-728.
Frydman, R., C. Gray, M. Hessel, and A. Rapaczynski. 1999. When does privatization work? The
   impact of private ownership on corporate performance in the transition economies. The
   Quarterly Journal of Economics 114(4), 1153-1191.
Gan, J., Y. Guo, and C. G. Xu. 2017. Decentralized privatization and change of control rights in
  China. The Review of Financial Studies 31(10), 3854–3894.
Gao, N., C. X. Long, and L. C. Xu. 2016. Collective leadership, career concern, and the Housing
  market in China: the role of standing committees. Review of Development Economics 20(1), 1-
  13.
Guo, K., Y. Yao. 2005. Causes of privatization in China: Testing several hypotheses. Economics of
  Transition 13(2), 211-238.
Gupta, N. 2005. Partial privatization and firm performance. Journal of Finance 60(2), 987-1015.
Halper, S., 2010, The Beijing consensus: how China's authoritarian model will dominate the twenty-
  first century, Basic Books.
Harrison, A., M. Meyer, P. Wang, L. Zhao, and M. Zhao. 2019. Can a tiger change its stripes?
  Reform of Chinese state-owned enterprises in the penumbra of the state. NBER Working Paper
  No. 25475.
Heckman, J. J., and V. J. Hotz. 1989. Choosing among alternative nonexperimental methods for
  estimating the impact of social programs: the case of manpower training. Journal of the
  American Statistical Association 84(408), 862-874.
Huang, Y. 2012. How did China take off? Journal of Economic Perspectives 20(4), 147-170.


                                                 25
Huang, Z., L. Li, G. Ma, and L. C. Xu. 2017. Hayek, local information, and commanding heights:
  decentralizing state-owned enterprises in China. American Economic Review 107(8), 2455-78.
Imbens, G. W., and L. Thomas. 2008. Regression discontinuity designs: a guide to practice. Journal
  of Econometrics 142(2), 615-635.
Jefferson, G. H., and J. Su. 2006. Privatization and restructuring in China: evidence from
   shareholding ownership, 1995–2001. Journal of Comparative Economics 34(1), 146-166.
Jia, R., M. Kudamatsu, and D. Seim. 2015. Political selection in China: the complementary roles of
    connections and performance. Journal of the European Economic Association 13(4), 631-668.
Kole, S. R., and J. H. Mulherin. 1997. The government as shareholder: a case from the US. Journal
  of Law and Economics 40(1), 1-22.
Kou, C. W., and W. H. Tsai. 2014. “Sprinting with small steps” towards promotion: solutions for
  the age dilemma in the CCP cadre appointment system. The China Journal 71(1), 153-171.
La Porta, R. and F. Lopez-de-Silanes. 1999. The benefits of privatization: evidence from
  Mexico. The Quarterly Journal of Economics 114(4), 1193-1242.
Lau, L. J., Y. Qian, and G. Roland. 2000. Reform without losers: An interpretation of China's dual-
  track approach to transition. Journal of Political Economy 108(1), 120-143.
Lee, D. S., and T. Lemieux. 2010. Regression discontinuity designs in economics. Journal of
  Economic Literature 48(2), 281-355.
Li, H., and L. Zhou. 2005. Political turnover and economic performance: the incentive role of
   personnel control in China. Journal of Public Economics 89(9–10), 1743–1762.
Li, K., T. Wang, Y. Cheung, and P. Jiang. 2011. Privatization and risk sharing: evidence from the
   split share structure reform in China. Review of Financial Studies 24(7), 2499-2525.
Li, S., L. Shuhe, and W. Zhang. 2000. “The Road to Capitalism: Competition and Institutional
   Change in China.” Journal of Comparative Economics 28: (2):269–292
Li, W., and L. C. Xu. 2002. The political economy of privatization and competition: cross-country
   evidence from the telecommunications sector. Journal of Comparative Economics 30(3), 439-
   462.
Li, W., and L. C. Xu. 2004. The impact of privatization and competition in the telecommunications
   sector around the world. Journal of Law and Economics 47(2), 395-430.
Liao, L., B. Liu, and H. Wang. 2014. China’s secondary privatization: perspectives from the split-
   share structure reform. Journal of Financial Economics 113(3), 500-518.
Lin, JY. 2012. Demystifying the Chinese Economy. Cambridge: Cambridge University Press.
Maskin, E., Y. Qian, and C. Xu. 2000. Incentives, information, and organizational form. The Review
  of Economic Studies 67(2), 359-378.
McGregor, R. 2010. The party: the secret world of China's communist rulers. New York: Harper
  Collins Press.
Megginson, W., and J. Netter. 2001. From state to market: a survey of empirical studies on
  privatization. Journal of Economic Literature 39(2), 321-389.
Megginson, W. 2005. The financial economics of privatization. Oxford University Press.


                                                26
Megginson, W. 2017. Privatization, state capitalism, and state ownership of business in the 21st
  century. Foundations and Trends in Finance 11(1-2), 1-153.
Megginson, W., R. C. Nash, and M. Van Randenborgh. 1994. The financial and operating
  performance of newly privatized firms: an international empirical analysis. The Journal of
  Finance 49(2), 403-452.
Persson, P., and E. Zhuravskaya. 2016. The limits of career concerns in federalism: evidence from
   China. Journal of the European Economic Association 14(2), 338-374.
Pomeranz, K. 2000. The great divergence: China, Europe, and the making of the modern world
  economy. Princeton: Princeton Press.
Qian, Y. 2000. The institutional foundations of China's market transition. Annual World Bank
   Conference on Development Economics 1999. The World Bank, 289-310.
Ramo, J.C., 2004, The Beijing Consensus. Foreign Policy Centre.
Rodrik, D. 2006. “Goodbye Washington Consensus, Hello Washington Confusion? A Review of
  the World Bank’s Economic Growth in the 1990s: Learning from a Decade of Reform. Journal
  of Economic Literature XLIV, 973-987.
Shih, V., C. Adolph, and M. Liu. 2012. Getting ahead in the communist party: explaining the
   advancement of central committee members in China. American Political Science Review 106(1),
   166-187.
Shleifer, A., and R. W. Vishny. 1994. The politics of market socialism. Journal of Economic
   Perspectives 8(2), 165-176.
Staiger, D., and J. H. Stock. 1997. Instrumental variables regression with weak instruments.
   Econometrica 65(3), 557-586.
Stiglitz, J. E. 1994. Whither socialism. Cambridge: MIT Press.
Stiglitz, J. 2008. Is there a Post-Washington Consensus Consensus? In The Washington Consensus
   Reconsidering: Toward a New Global Governance. Edited by N. Serra and J. Stiglitz, 41-56.
   New York: Oxford University Press.
Sun, Q., and W. H. Tong. 2003. China share issue privatization: the extent of its success. Journal of
  Financial Economics 70(2), 183-222.
Tan, Y., X. Tian, X. Zhang, and H. Zhao. 2015. The real effects of privatization: evidence from
  China’s split share structure reform. Kelley School of Business Research Paper: 2014-33.
Williamson, J. ed. 1990. Latin American Adjustment: How Much has it Happened? Washington,
  D.C.: Institute for International Economics.
Wu, J. 2018. China’s Economic Reform Process. (Zhongguo Jinji Gaige Jincheng.) Beijing: China
  Macropaedia Press (Zhongguo dabaike quanshu chubanshe).
Xi, T., Y. Yao, and M. Zhang. 2018. Competence versus incentive: evidence from city officials in
   China. Journal of Comparative Economics 46(4), 1046-61.
Xu, C. 2011. The fundamental institutions of China’s reforms and development. Journal of
  Economic Literature 49(4), 1076–1151.
Xu, L. C., T. Zhu, and Y. Lin. 2005. Politician control, agency problems, and ownership reform:
  evidence from China. Economics of Transition 13(1), 1-24.

                                                 27
Yao, Y.. 2015. Beijing Consensus. In The Wiley Blackwell Encyclopedia of Race, Ethnicity, and
  Nationalism. https://onlinelibrary.wiley.com/doi/abs/10.1002/9781118663202.wberen577.
Yao, Y., M. Zhang. 2015. Subnational Leaders and Economic Growth: Evidence from Chinese
  Cities. Journal of Economic Growth 20(4), 405-436.
Zhang, W. 2019. The China model view is factually false. Journal of Chinese Economic and
  Business Studies, DOL: 10.1080/14765284.2019.1663696.
Zhu, X. 2012. “Understanding China’s Growth: Past, Present and the Future.” Journal of
  Economic Perspectives 26(4), 103-124.




                                               28
                                                                 Figure 1a: Distribution of local politicians’ age
                           300
Frequency of politicians


                           250

                           200

                           150

                           100

                                  50

                                          0
                                                   39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
                                                                                   Age of polititicians



                                                         Figure 1b: The age of local politicians in the year of promotion
                                              45

                                              40

                                              35
                   Frequency of politicians




                                              30

                                              25

                                              20

                                              15

                                              10

                                              5

                                              0
                                                    39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
                                                                         Age of politicians in the year of promotion


Figure 1a depicts the distribution of local officials’ ages on the city-year level. Here local officials refer to
the prefecture-level city party secretaries. There are 2,631 city-year observations in the sample. Figure 1b
shows the number of officials that were promoted to a higher-level position at a certain age. A higher-level
position means a vice-provincial status, such as a vice provincial governor.




                                                                                         29
                        Table 1: Variable definitions and summary statistics

  This table provides a brief description of the variables and the summary statistics.


  Panel A. variable definitions
Variable                   Variable Description
Privatized                 A dummy variable indicating the privatization of the firm: Privatized equals 1
                              if its state share falls below 50% or it exits from the database.
Age                        The age of the city party secretary.
College                    The dummy variable equals 1 if the city party secretary has college education
                              and 0 otherwise.
Central Experience         The dummy variable indicating the city party secretary having experience
                              serving in the central government.
Provincial Experience      The dummy variable indicating the city party secretary having experience
                              serving in the provincial government.
Labor prod                 The labor productivity of the firm calculated as ln (sales/employment).
Leverage                   The leverage ratio of the firm, calculated as total debt/total asset.
GDP Per capita             GDP per capita (of thousand RMB in the 1990 constant price) of the city.
GDP Growth                 The real growth rate of GDP (%) of the city.
Government activism        The city-level average “percent of the officials in the department who facilitate
                              the development of the company” in city. Data is from the World Bank
                              Enterprises Survey for China (2005). We obtain the city average from firm-
                              level data.
TFP_OP                     Total factor productivity (TFP) of the firm calculated using the Olley-Pakes
                              method (see Appendix C for details).
TFP_OLS                    TFP of the firm calculated using the OLS method (see Appendix C for details).

TFP_IN                     TFP of the firm calculated using the index number method (see Appendix C for
                             details).
TFP_CS                     TFP of the firm calculated using the index function method, with the cost shares
                             generated from the ASIF data (see Appendix C for details).




                                                     30
      Panel B. Summary statistics of the sample before and in the year of privatization
                                       Obs.       Mean       Median     Std. Dev.     Min          Max
Privatizedit                         151,269      0.268       0.000       0.443      0.000         1.000
Age t-1                              151,269      51.822     52.000       4.092      39.000        61.000
Labor prod t-1                       151,269      4.056       4.042       1.383      -0.113        8.001
Leverage t-1                         151,269      0.722       0.696       0.402      0.001         2.669
ROA t-1                              151,269      -0.003      0.000       0.111      -0.426        0.878
GDP per capita t (thousand RMB)      151,269      7.483       5.103       8.636      1.500         85.829
GDP growth (%)t                      151,269      9.532      10.710       6.286      -17.600       26.300
College t                            151,269      0.881       1.000       0.324      0.000         1.000
Central Experience t                 151,269      0.041       0.000       0.199      0.000         1.000
Provincial Experience t              151,269      0.551       1.000       0.497      0.000         1.000


      Panel C. Summary statistics of the sample including post-privatization observations
                                        Obs.       Mean       Median     Std. Dev.         Min       Max
TFP_OP                                 99,703      1.767       1.872       1.433          -3.377     5.176
TFP_OLS                               125,407      0.820       0.856       1.365          -4.010     4.382
TFP_IN                                125,407      1.324       1.443       1.543          -3.360     5.033
TFP_CS                                169,143      -0.463      -0.388      1.487          -5.943     3.494
Government activism                   119,492      0.407       0.351       0.213          0.095      0.994




                                                   31
                 Table 2: Official age and privatization of local SOEs
The table reports the effects of local politicians’ age on the tendency of privatization by including a
series of age dummies of local politicians. The dummy variable Privatized equals to 1 when the
SOE experiences privatization in the specific year and 0 otherwise. As privatization is mostly
irreversible in our sample, we delete the observations after Privatized has turned 1, leaving 151,269
observations in the regressions. Control variables include macroeconomic conditions and officials’
individual characteristics (education level, prior experience of working in the central authorities and
provincial level). Industry and city fixed effects are controlled. All standard errors clustered at the
city level are reported in parentheses. *, **, and *** represent statistical significance at the 10%,
5%, and 1% level, respectively.

 Dependent Variable: Privatized                (1)            (2)             (3)             (4)
 Dum(Age=51)                                 -0.008         -0.007          -0.006          -0.006
                                            (0.010)         (0.009)         (0.010)        (0.010)
 Dum(Age=52)                                  0.015          0.015           0.014           0.016
                                            (0.011)         (0.011)         (0.011)        (0.011)
 Dum(Age=53)                                  0.001          0.001           0.001           0.002
                                            (0.012)         (0.011)         (0.011)        (0.011)
 Dum(Age=54)                                 -0.006         -0.006          -0.006          -0.004
                                            (0.014)         (0.013)         (0.013)        (0.013)
 Dum(Age=55)                                 -0.001         -0.002          -0.002           0.001
                                            (0.012)         (0.011)         (0.011)        (0.012)
 Dum(Age=56)                                  0.017          0.017           0.017           0.020
                                            (0.016)         (0.016)         (0.016)        (0.015)
 Dum(Age=57)                                 -0.006         -0.006          -0.007          -0.004
                                            (0.014)         (0.014)         (0.014)        (0.015)
 Dum(Age=58)                               -0.042**        -0.042**        -0.042**       -0.039**
                                            (0.017)         (0.017)         (0.017)        (0.018)
 Dum(Age=59)                                -0.047*        -0.049**        -0.049**        -0.043*
                                            (0.024)         (0.024)         (0.024)        (0.024)
 Dum(Age=60+)                                 0.004         -0.002          -0.002           0.008
                                            (0.039)         (0.037)         (0.037)        (0.037)
 Labor prodt-1                                            -0.021***       -0.021***      -0.021***
                                                            (0.002)         (0.002)        (0.002)
 ROAt-1                                                   -0.147***       -0.147***      -0.146***
                                                            (0.018)         (0.018)        (0.018)
 Leveraget-1                                              0.044***        0.044***       0.044***
                                                            (0.005)         (0.005)        (0.005)
 Ln(employmentt-1)                                        -0.035***       -0.035***      -0.035***
                                                            (0.002)         (0.002)        (0.002)
 GDP Per capitat-1                                                          -0.001          -0.001
                                                                            (0.010)        (0.009)
 GDP Growtht-1                                                               0.001           0.001
                                                                            (0.001)        (0.001)
 College                                                                                    0.025*
                                                                                           (0.013)
 Central Experience                                                                         -0.019
                                                                                           (0.020)
 Provincial Experience                                                                       0.004
                                                                                           (0.009)
 Industry FE and City FE                      Yes            Yes             Yes              Yes
 N                                          151,269        151,269         151,269         151,269
 R-square                                    0.097          0.113           0.113            0.114


                                                  32
                                      Table 3: Balance tests
This table reports the results of the balance tests. We limit the sample to consist of observations
with the age of a city secretary in the range of 55-61 years old. The dependent variables are once-
lagged labor productivity (Labor prod), firm’s leverage ratio (Leverage), the logarithm of total
employees (Ln(employment)), and the return on sales (ROA) respectively. The purpose of the
balance tests is to show that the control variables vary smoothly around the age threshold, thus
discontinuities around the threshold reflect the causal effect of the decrease of promotion probability.
Definitions of other variables are the same as the previous tables. Standard errors clustered at the
city level are reported in parentheses. *, **, and *** represent statistical significance at the 10%,
5%, and 1% level, respectively.

                                    (1)                (2)                 (3)                   (4)
 Dependent Variable:
                               Labor prodt-1       Leveraget-1       Ln(employmentt-1)          ROAt-1
 Dum(Age≥58)                       0.005            0.014*                0.027                -0.004
                                 (0.026)            (0.008)              (0.026)              (0.003)
 Age-58                          -0.021*            -0.006*              0.021*                -0.000
                                 (0.011)            (0.003)              (0.013)              (0.001)
 (Age-58)* Dum(Age≥58)            -0.004            0.014*               -0.093                0.004
                                 (0.054)            (0.007)              (0.056)              (0.005)
 Industry FE and City FE            Yes               Yes                  Yes                   Yes
 N                                43,591            43,591               43,591               43,591
 R-square                          0.313             0.111                0.172                0.094




                                                  33
       Table 4: Privatization and age of local officials: Regression Discontinuity Design
This table reports the parametric and non-parametric results of the Regression Discontinuity Design
(RDD). The dependent variable, Privatizedit, is the dummy variable indicating whether a firm i is
privatized in year t. Dum(Age≥58) is a dummy variable equaling 1 if the official’s age is 58+ and 0
otherwise. In columns (1)-(2) we report the results of RDD with the global parametric polynomial
approach. We control for the third order polynomial in the running variable. Columns (3) and (4) report
the non-parametric local linear approach results of RDD. Standard errors clustered at the city level are
reported in parentheses. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level,
respectively.

        Dependent Variable:                      (1)          (2)              (3)         (4)
        Privatized                           Cubic Parametric RDD             Local Linear RDD
        Dum(Age≥58)                          -0.050**    -0.051***          -0.046**    -0.048**
                                              (0.020)      (0.019)           (0.022)     (0.020)
        Age-58                                  0.004        0.005           -0.001       0.000
                                              (0.008)      (0.008)           (0.007)     (0.007)
        (Age-58)* Dum(Age≥58)                  -0.103       -0.084            0.017       0.014
                                              (0.085)      (0.080)           (0.020)     (0.020)
        (Age-58)^2                              0.001        0.001
                                              (0.001)      (0.001)
        (Age-58)^2* Dum(Age≥58)                 0.123        0.098
                                              (0.089)      (0.082)
        (Age-58)^3                              0.000        0.000
                                              (0.000)      (0.000)
        (Age-58)^3*Dum(Age≥58)                 -0.031       -0.023
                                              (0.021)      (0.019)
        Other controls                           No           Yes              No           Yes
        Industry FE and City FE                  Yes          Yes             Yes           Yes
        N                                     151,269     151,269            43,742       43,742
        R-square                                0.096        0.113           0.105         0.125




                                                    34
           Table 5: Placebo test: Privatization of central and provincial SOEs
This table reports the result on the relationship between age structure and privatization using a
placebo sample of provincial and central SOEs. Columns (1)-(2) show the parametric global
polynomial results, and columns (3) and (4) show the non-parametric local linear results. All the
variables and specifications are the same as the previous tables. Standard errors clustered at the city
level are reported in parentheses. *, **, and *** represent statistical significance at the 10%, 5%,
and 1% level, respectively.

  Dependent Variable:                         (1)         (2)                (3)            (4)
  Privatized                             Cubic Parametric RDD                Local Linear RDD
  Dum(Age≥58)                               -0.026      -0.026             -0.016         -0.020
                                           (0.021)     (0.021)            (0.022)        (0.021)
  Age-58                                    -0.003      -0.003             -0.003         -0.002
                                           (0.009)     (0.009)            (0.008)        (0.008)
  (Age-58)* Dum(Age≥58)                   -0.197**    -0.185**             0.000          -0.006
                                           (0.090)     (0.081)            (0.018)        (0.017)
  (Age-58)^2                                -0.000      -0.000
                                           (0.001)     (0.001)
  (Age-58)^2* Dum(Age≥58)                  0.235**     0.217**
                                           (0.098)     (0.090)
  (Age-58)^3                                -0.000      -0.000
                                           (0.000)     (0.000)
  (Age-58)^3*Dum(Age≥58)                 -0.061***   -0.057***
                                           (0.023)     (0.021)
  Other controls                              No         Yes                 No             Yes
  Industry FE and City FE                    Yes         Yes                 Yes            Yes
  N                                        34,801      34,801              10,668         10,668
  R-square                                  0.178       0.190               0.241          0.249




                                                  35
             Table 6: Effects of privatization: OLS, FE, FE-FT and IV regression
This table reports the effects of privatization by regressing the performance measures, TFP, labor
productivity (Labor Prod) and ROA on the privatization indicator Privatized. Here we use three
measures of TFP calculated using the Olley-Pakes methods (TFP_OP), OLS (TFP_OLS), and Index
Number methods (TFP_IN), respectively. Note that in these panels, the post-privatization
observations of privatized firms are included in the sample. Since the data source does not contain
the necessary data to compute TFP for 2008 and 2009, our analysis of firm outcomes in columns (1)
to (5) use the data ending in 2007. Panels A and B report the regression results without and with
firm fixed effects, respectively. Panel C reports the effects of privatization using FE-FT model. We
regress the detrended performance measures (y), ΔTFP, ΔROA and labor productivity (ΔLabor
Prod) on the detrended dummy variable, ΔPrivatized. Control variables are the same as those in
Panel B. In Panel D we adopt the fuzzy RD framework by using a dummy variable, Dum(Age≥58),
as the instrumental variable of privatization. We control for quadratic polynomials in the regressions.
In all regressions through Panel A-D, we control for the industry-year fixed effects, and the
interaction of year dummies with initial sales, initial ROA, and initial TFP, as well as the city-level
macro variables (i.e., GDP Per capita and GDP Growth). Standard errors clustered at the firm level
are reported in parentheses. *, **, and *** represent statistical significance at the 10%, 5%, and 1%
level, respectively.

Panel A. OLS regressions
                                       (1)          (2)           (3)          (4)             (5)
                                    TFP _OP      TFP _OLS      TFP _IN     Labor Prod         ROA
Privatized                          0.220***     0.188***      0.202***     0.258***        0.011***
                                     (0.013)      (0.012)       (0.012)      (0.009)         (0.001)
GDP Per capitat-1                   0.017***     0.018***      0.021***     0.047***       -0.001***
                                     (0.005)      (0.005)       (0.004)      (0.004)         (0.000)
GDP Growtht-1                       0.002***     0.002***      0.002***     0.003***        0.000***
                                     (0.001)      (0.001)       (0.001)      (0.000)         (0.000)
Year FE*Initial firm attributes       YES          YES           YES          YES             YES
Year FE*Industry FE                   YES          YES           YES          YES             YES
N                                    95,212       119,954       119,954     187,822          187,822
R-square                              0.629        0.615         0.704        0.680           0.344


Panel B. The FE regressions

                                      (1)          (2)           (3)           (4)            (5)
                                  TFP _OP      TFP _OLS      TFP _IN      Labor Prod         ROA
Privatized                        0.161***      0.136***     0.143***      0.177***        0.012***
                                   (0.012)       (0.011)      (0.011)       (0.008)         (0.001)
GDP Per capitat-1                  0.016**        0.008       0.016**      0.024***       -0.002***
                                   (0.007)       (0.006)      (0.006)       (0.004)         (0.001)
GDP Growtht-1                       0.001         0.000        0.000         -0.001        0.000***
                                   (0.001)       (0.001)      (0.001)       (0.000)         (0.000)
Year FE*Initial firm attributes      YES          YES           YES           YES            YES
Year FE*Industry FE                  YES          YES           YES           YES            YES
Firm FE                              YES          YES           YES           YES            YES
N                                  89,314        113,485      113,485       178,601         178,601
R-square                            0.786         0.783        0.835         0.841           0.614



                                                  36
     Panel C. The FE-FT regressions

                                       (1)          (2)                (3)             (4)              (5)
                                   ΔTFP _OP     ΔTFP _OLS          ΔTFP _IN        ΔLabor Prod         ΔROA
    Δprivatized                     0.032**        0.019            0.026**         0.057***         0.005***
                                    (0.014)       (0.013)           (0.013)          (0.008)          (0.001)
    ΔControls                         YES          YES                YES             YES               YES
    Year FE*Initial firm
                                 YES                YES              YES              YES             YES
             attributes
    Year FE*Industry FE          YES                 YES             YES              YES             YES
    Firm FE                      YES                 YES             YES              YES             YES
    N                           62,090              80,740          80,740           130,257         130,257
    R-square                     0.181               0.174          0.172             0.202           0.160
     Here Δ means first difference.

     Panel D. The IV regressions

                                              (1)            (2)             (3)             (4)            (5)
                                                                                            Labor
                                        TFP _OP         TFP _OLS        TFP _IN                           ROA
                                                                                            Prod
2nd stage
Privatized                               1.431**           1.191**          1.003*       1.755***       0.145***
                                         (0.592)           (0.534)         (0.522)        (0.494)         (0.053)
GDP Per capitat-1                         0.003             -0.003           0.008       0.022***       -0.003***
                                         (0.010)           (0.008)         (0.008)        (0.005)         (0.001)
GDP Growtht-1                             0.000             -0.000          -0.000       -0.001**         0.000*
                                         (0.001)           (0.001)         (0.001)        (0.001)         (0.000)
Quadratic polynomials of age               YES               YES             YES           YES             YES
Year FE*Initial firm attributes            YES               YES             YES           YES             YES
Year FE*Industry FE                        YES               YES             YES           YES             YES
Firm FE                                    YES               YES             YES           YES             YES
N                                        89,314            113,485         113,485        178,601        178,601
R-square                                  0.740             0.752            0.819         0.768           0.534
1st stage
Dum (Age≥58)                            -0.034***       -0.032***       -0.032***        -0.025***      -0.027***
                                          (0.007)         (0.006)         (0.006)          (0.004)        (0.004)
GDP Per capitat-1                        0.010***        0.009***       0.009***            0.002       0.005***
                                          (0.003)         (0.002)         (0.002)          (0.002)        (0.002)
GDP Growtht-1                              0.000           0.000           0.000         0.001***          0.000
                                          (0.000)         (0.000)         (0.000)          (0.000)        (0.000)
Quadratic polynomials of age               YES             YES             YES              YES            YES
Year FE* Initial firm attributes           YES             YES             YES              YES            YES
Year FE*Industry FE                        YES             YES             YES              YES            YES
Firm FE                                    YES             YES             YES              YES            YES
N                                         89,314          113,485        113,485          178,601        178,601
R-square                                   0.702           0.704           0.705            0.700          0.696
First stage F-statistic                   25.241          24.528          24.813           39.210         12.012




                                                      37
              Table 7. Government Involvement and the effects of privatization on TFP
       This table reports the interaction between government activism and the effects of privatization. We
       include the measure of government activism, and its interaction terms with the privatization dummy
       variable, Privatized. The definitions of other variables and control variables are the same as those
       defined in Table 6. We report the results of the FE regressions in the column (1)-(3) and IV
       regressions in (4)-(6) in each of the panels. We also control for firm-level fixed effects, and the
       industry-year fixed effects, and the interaction of year dummies with initial sales, initial ROA, and
       initial TFP, as well as the city-level macro variables (i.e., GDP Per capita and GDP Growth).
       Standard errors clustered at the firm level are reported in parentheses. *, **, and *** represent
       statistical significance at the 10%, 5%, and 1% level, respectively.



                                               Firm FE                           IV Regression (2nd stage)
Dependent variable:                  (1)          (2)           (3)           (4)          (5)            (6)
                                  TFP _OP     TFP _OLS       TFP _IN       TFP _OP     TFP _OLS        TFP _IN
Privatized * Government             0.072        0.058         0.055       -4.569**     -3.846**       -3.671**
              activism             (0.057)      (0.056)       (0.054)       (1.933)      (1.556)        (1.512)
Privatized                        0.127***     0.107***      0.116***      3.247***     2.794***       2.517***
                                   (0.027)      (0.027)       (0.026)       (1.165)      (0.968)        (0.939)
Quadratic polynomials of age         No           No            No           YES          YES            YES
Other controls                      YES          YES           YES           YES          YES            YES
Year FE*Initial firm attributes     YES          YES           YES           YES          YES            YES
Year FE*Industry                    YES          YES           YES           YES          YES            YES
Firm FE                             YES          YES           YES           YES          YES            YES
N                                  60,865       73,915        73,915        60,865       73,915         73,915
R-square                            0.783        0.781         0.828         0.702        0.721          0.791




                                                        38
                Appendix A: Construction of the panel from the ASIF data

Construction of the panel from the ASIF data. In the dataset, every firm is given a unique firm
code. A small number of firms may have changed their firm codes within the sample period but
remained in the sample. To address this issue, we follow Brandt et al. (2012) and Yang (2015) to
obtain unique firm codes based on the firm’s name, zip code, telephone number, and founding year.
We clean the data as follows. First, if the year t observation of a firm cannot be matched to any
firm’s observation in year t+1 based on the firm code, we try to find a firm with the same name in
year t+1, and match them by giving the year t+1 observation the same firm code as the year t
observation. Second, for those firms that cannot be matched by the code or name, we rely on the
combinations of the zip code, telephone number and the founding year to match them. We delete
firms with missing key information, i.e. assets, fixed assets, sales and employment. Table A1
presents the frequency with which we can link the observations in different years for both SOEs and
non-SOEs.

                         Table A1. Evolution of the raw panel over time

  Year    Total firms    Entrants             Incumbent, linked using              Exiting (in the
                                        NBS ID           Other information           next year)
  1998      164,452                                                                    28,709
  1999      161,439      25,696         130,863                 4,880                  27,672
  2000      162,350      28,583         130,538                 3,229                  36,395
  2001      170,780      44,825         117,526                 8,429                  24,356
  2002      181,149      34,725         142,950                 3,474                  28,378
  2003      196,204      43,433         146,605                 6,166                  51,295
  2004      274,750      129,841        137,681                 7,228                  45,085
  2005      271,819      42,154         226,675                 2,990                  25,819
  2006      301,943      55,943         243,728                 2,272                  28,485
  2007      336,742      63,284         271,629                 1,829

Note: Entrants are those that first appear in the sample in the specific year. Exiting means dropping
out of the sample in the next year. The ASIF dataset includes all SOEs, and all non-state firms with
sales exceeding five million yuan. Thus, a firm's entry year may differ from its establishment year.
Similarly, a firm’s exiting year may differ from its death year.




                                                  39
               Appendix B: Robustness test: Sample of narrowly-defined privatization
 This table reports the main results using a more conservative definition of privatization: Privatized
 equals 1 only when the state share drops below 50% but staying in the database. All the variables
 and specifications are exactly the same as the previous tables. t-statistics are reported in
 parentheses. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level,
 respectively.

 Panel A. OLS Regression with age dummies

Dependent Variable: Privatized                                                                   (1)          (2)
Dum(Age=51)                                                                                    -0.007       -0.007
                                                                                               (0.005)     (0.004)
Dum(Age=52)                                                                                     0.003        0.002
                                                                                               (0.005)     (0.005)
Dum(Age=53)                                                                                     0.000       -0.002
                                                                                               (0.006)     (0.005)
Dum(Age=54)                                                                                    -0.002       -0.004
                                                                                               (0.007)     (0.006)
Dum(Age=55)                                                                                   -0.010*     -0.013**
                                                                                               (0.005)     (0.005)
Dum(Age=56)                                                                                     0.004        0.002
                                                                                               (0.007)     (0.006)
Dum(Age=57)                                                                                    -0.001       -0.003
                                                                                               (0.007)     (0.007)
Dum(Age=58)                                                                                  -0.028***   -0.028***
                                                                                               (0.010)     (0.009)
Dum(Age=59)                                                                                   -0.031*       -0.027
                                                                                               (0.019)     (0.017)
Dum(Age=60)                                                                                    -0.021       -0.020
                                                                                               (0.029)     (0.029)
Controls                                                                                         No           Yes
Industry FE and City FE                                                                          Yes          Yes
N                                                                                             151,269      151,269
R-square                                                                                        0.166        0.184

 Panel B. Regression Discontinuity Design

                                                                                               Cubic
Dependent Variable:                                                                                      Local linear
                                                                                            parametric
Privatized                                                                                                  RDD
                                                                                               RDD
Dum(Age≥58)                                                                                  -0.029**      -0.033**
                                                                                              (0.012)       (0.013)
������������������������������������ # , ������������������������������������ # ∗ ������������������������������������(������������������������������������ − 58)                               Yes           Yes
All quadratic terms related to the above                                                        Yes           No
All cubic terms on
                                                                                               Yes           No
                   ������������������������������������ # , ������������������������������������ # ∗ ������������������������������������(������������������������������������ − 58)
Controls, Industry FE and City FE                                                              Yes            Yes
N                                                                                            151,269        43,742
R-square                                                                                      0.166          0.196
     Note. ������������������������������������ # is (Age-58).

                                                                                       40
                                        Appendix C. Estimating TFP
    Here we describe how we estimate the firm-level TFP in three ways.
    We use a standard log-linear Cobb-Douglas production function to estimate the firm-level TFP.
Specifically, the TFP of firm i in year t is the estimated residual from the regression:
                                           ������������������������������������ = ������������0 + ������������������������ ������������������������������������ + ������������������������ ������������������������������������ + ������������������������������������   (A1)
where ������������������������������������ is the logarithm of value-added, and ������������������������������������ and ������������������������������������ are the logarithms of capital and labor,
respectively. To allow for different factor intensities across industries, we estimate equation (A1)
separately for each two-digit industry. TFP can be interpreted as the relative productivity of a firm
within its industry.
      Real value added is obtained by subtracting the real input from the real output. We use the two-
digit ex-factory price index from China Urban Living and Price Statistics to deflate the output. The
input deflator is calculated based on the available output deflators at the two-digit industry level and
information from the National Input-Output (IO) tables in 1997, 2002, and 2007. From the IO tables,
we know how much inputs are needed to produce a unit of output. Then the average input price
index is the weighted average of the price indices of those inputs. Thus, to obtain the input deflator
for each industry, we calculate a weighted average of the input deflators, using as weights the
coefficients in the IO table.23
      In the ASIF dataset, firms report the total annual employment, but they do not report the real
capital stock. Instead, the firms report the value of their fixed capital stock at the original purchase
prices. As these book values are the sum of the nominal values for different years, they are not equal
to the real capital stock and are not comparable across time and across firms. Since we do not have
all past investments of a firm to construct the real capital stock, we follow Brandt, Van Biesebroeck
and Zhang (2012) and make several assumptions to convert the value of their capital stock at the
original purchase prices into the real values using the following procedures.
      First, we estimate the nominal value of the capital stock for each year between a firm’s birth
year and the first year in which the firm appears in our data set. We assume that it is 1998, the first
year of our panel. We assume that the growth rate of the nominal capital stock of each firm equals
to the growth rate of the nominal capital stock in the corresponding two-digit industry as reported
in the China Statistical Yearbooks. 24 We then calculate the nominal capital stock in 1998 as follows:
                                           ������������������������1998 = ������������������������������������ ∏1998
                                                                        ������������=������������ (1 + ������������
                                                                                          ������������ )                                 (A2)

Where ������������������������1998 is the nominal captial stock in 1998 reported in the ASIF data, s indicates the firm’s
first year of operation, ������������������������������������ is the nominal captial stock of the firm in its birth year, and ������������������������ is the
growth rate of the nominal capital stock in the two-digit industry in year t, as reported by the China
Statistical Yearbooks. From equation (A2), we calculate the nominal stock in each year between the
firm’s birth year and 1998.
      Second, the annual nominal investment ������������������������������������ is the change in the nominal capital stock between
two consecutive years, that is,
                                                ������������������������������������ = ������������������������������������ − ������������������������������������−1                    (A3)



23
   The 1997 IO table is used to construct the input deflators of 1998-2000, the 2002 IO table is used to
construct the input deflators of 2001-2005, and the 2007 IO table is used to construct the input deflators
in 2006-2007.
24
   Since China Statistical Yearbooks report the growth rate of nominal capital stock in the two-digit
industry from 1986, we assume that firms established before 1986 are established in 1986.
                                                                                 41
     Third, we derive the real capital stock for each year between the firm’s birth year and 1998.
We deflate the annual nominal investment in each year ������������������������������������ into the real value ������������������������������������ using the
investment deflator, which is in China Statistical Yearbooks from 1990. For years 1986-1989, we
use the investment deflator constructed by Perkins and Rawski (2008).
     Fourth, we obtain the real capital stock in 1998 with the perpetual inventory method.
Specifically,
                                                     ������������������������������������ = (1 − ������������)������������������������������������−1 + ������������������������������������                                                  (A4)
Where ������������������������������������ is the real capital stock in year t, and ������������ is the depreciation rate as estimated by:

                                             Accumulated depreciation reported in 1998
                                                                                                                                             /������������������������1998   (A5)
                                                                                    1998−s


              Finally, we obtain the annual real investment and the real capital stock after 1998. For years
after 1998, we use the observed change in the firm's nominal capital stock at the original purchase
prices as our estimate of the nominal annual investment, that is, the nominal annual investment
������������������������������������ is still obtained from ������������������������������������ − ������������������������������������−1 . The real fixed investment ������������������������������������ is obtained by deflating
������������������������������������ with the investment deflator in China Statistical Yearbooks. The real capital stock is
constructed using the perpetual inventory method, that is,
                                       ������������������������������������ = ������������������������������������−1 − ������������������������������������������������������������������������������������������������������������������������������������������������������������ + ������������������������������������                   (A6)

������������������������������������������������������������������������������������������������������������������������������������������������������������ is annual depreciation that is reported in ASIF, again deflated by the investment
deflators in China Statistical Yearbooks.
              We estimate equation (A1) by ordinary least squares (OLS). We call this TFP-OLS. While this
approach is commonly used in the literature, the existing research has argued that the OLS estimates
suffer from two endogeneity issues: simultaneity of input choices and selection biases. These two
issues will generate biased estimates of ������������������������ and ������������������������ , and therefore biased estimates of the TFP. A
variety of techniques have been suggested to address these issues. We use the widely-used method
proposed by Olley and Pakes (1996). We call this TFP-OP.
              As another key measure of TFP, we use a straightforward index number approach, which does
not require estimating any parameters. To implement, the industry-specific wage share in the output
is used to measure ������������������������ . One minus this share is used to measure ������������������������ . Here the assumption is that a
cost-minimizing firm will make sure that the relative factor price ratio equals the local elasticity of
substitution between the inputs of the production technology. To avoid potential bias from the cost
shares using our own data—the cost shares in our data might reflect the ownership restructuring
directly—we rely on the estimates of the factor shares at the two-digit industry level from Saint-
Paul and Bentolila (2003), as in Bloom, Sadun and Van Reenen (2012). We call this TFP-IN.
              Overall, these three approaches yield quite similar results. The correlations of these
productivity measures are fairly high (see Table C1 in this appendix): that between TFP-OLS and
TFP-IN is 0.955; that between TFP-IN and TFP-OP is 0.937. Thus, it is not surprising that our
results do not hinge on how we measure productivity.
          Our index function approach relies on factor shares from Saint-Paul and Bentolila (2003). This
has the virtue of not relying on the cost shares estimated from our own data. After all, the cost shares
from the dataset itself might reflect the allocation effects of ownership restructuring itself, and thus
bias the TFP measurements based on the cost shares from the data. However, it is also possible that
the cost share based on our own data might better capture the underlying technologies and using
them might result in more accurate measurements of TFP. We thus also experimented with the

                                                                                        42
measure of firm-level TFP based on the cost shares from the data, as in Hau, Huang and Wang
(2019). We call this TFP-CS, which appears to be less closely related to other measures of TFP.

               Table C1: The Correlation matrix of different TFP measures
                               TFP_OP       TFP_OLS       TFP_IN       TFP_CS
                 TFP_OP            1
                 TFP_OLS         0.935          1
                  TFP_IN         0.937        0.955          1
                  TFP_CS         0.432        0.485        0.435           1

Reference
Bentolila, S., and G. Saint-Paul. 2003. Explaining movements in the labor share. Contributions in
     Macroeconomics 3(1).
Bloom, N., R. Sadun, and J. Van Reenen. 2012. The organization of firms across countries.
     Quarterly Journal of Economics 127(4), 1663-1705.
Brandt, L., J. Van Biesebroeck, and Y. Zhang. 2012. Creative accounting or creative destruction?
     Firm-level productivity growth in Chinese manufacturing. Journal of Development Economics
     97(2), 339–351.
Hau, H., Y. Huang, and G. Wang. 2019. Firm response to competitive shocks: evidence from China’s
     minimum wage policy. Working paper.
Olley, G. S., and A. Pakes. 1996. The dynamics of productivity in the telecommunications equipment
     industry. Econometrica 64(6), 1263–1297.
Perkins, D., and T. Rawski. 2008. Forecasting China’s Economic Growth to 2025. China’s Great
     Economic Transformation, edited by Brandt, L., and T. Rawski, Cambridge University Press,
     829-886.
Saint-Paul, G. and S. Bentolila. 2003. Explaining Movements in the Labor Share. B.E. Journal of
     Macroeconomics 3(1), 1-33.




                                                43
              Appendix D. Estimates of the privatization effects using TFP_CS


                Table D: Effects of privatization: Alternative TFP measures
This table re-estimate the economic effects of privatization in Table 6 using the index function TFP
using cost shares from ASIF data, i.e., TFP_CS (see Appendix C for details). Specifications are the
same as those in Table 6. Columns (1) and (2) reports the simple OLS regression results with and
without firm fixed effects, respectively. Column (3) reports the economic effects of privatization
using FE-FT model (Brown, Earle and Telegdy 2006; Brown, Earle, and Gehlbach 2009). We
regress the detrended performance measures the detrended dummy variable of Privatized. In column
(4), we use a dummy variable, Dum(Age≥58), as the instrumental variable of privatization, which
equals 1 when the age of the official is equal to or above 58 and 0 otherwise. We control for quadratic
polynomials and city-level macro variables (GDP Per capita and GDP Growth). The control
variables are the same as those in Table 6. We also control for firm-level fixed effects (except for
column (1)), and the industry-year fixed effects, and the interaction of year dummies with initial
sales, initial ROA, and initial TFP. For the lack of enough data to calculate TFP, the observations
in (1) -(4) columns ends in 2007. Standard errors clustered at the firm level are reported in
parentheses. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level,
respectively.



                                OLS                                                  IV Regression
 Dependent variable:                               Firm FE         FE-FT model
                          (without firm FE)                                            (2nd stage)
 TFP_CS
                                 (1)                  (2)               (3)                (4)
 Privatized                   0.153***             0.132***          0.037***           1.616**
                               (0.014)              (0.011)           (0.012)            (0.666)
 Other controls                  Yes                  Yes               Yes                Yes
 N                             163,758              155,379           112,887           155,379
 R-square                       0.995                0.998             0.998              0.997




                                                  44