W 30161

NBER WORKING PAPER SERIES
INVESTING WITH THE GOVERNMENT:

A FIELD EXPERIMENT IN CHINA
Emanuele Colonnelli
Bo Li
Ernest Liu
Working Paper 30161

http://www.nber.org/papers/w30161
NATIONAL BUREAU OF ECONOMIC RESEARCH

1050 Massachusetts Avenue
Cambridge, MA 02138
June 2022
The views expressed in this paper are the authors’ views and should not be attributed to
Zero2IPO or its team. We thank Matt Denes, Will Gornall, John Graham, Zhiguo He, Sabrina
Howell, Niklas Huther, Jessica Jeffers, Steve Kaplan, Song Ma, Elisa Maffioli, Scott Nelson,
Ludovic Phalippou, Tommaso Porzio, Wenlan Qian, Raghu Rajan, David Robinson, Andrei
Shleifer, Michael Song, Amir Sufi, Xuan Tian, Rob Vishny, Wei Xiong, David Yang, Bernard
Yeung, Anthony Lee Zhang, and Lugi Zingales, and seminar participants at Princeton, EIEF,
UChicago, ITAM, Tsinghua University, SFS Cavalcade, ABFER - BFI China Capital Market
Development Series, ABFER Annual Conference, WEFIDEV, the Five Star Junior Conference at
CUHK, and the Kentucly Finance Conference for helpful comments and suggestions. Yiren Ding,
Pranav Garg, Liming Ning, Sixun Tang, Shiqi Yang, and Chun Zhao provided superb research
assistance. We are grateful to The University of Chicago Booth School of Business, the Fama
Research Fund, and the Liew Family Junior Faculty Fellowship. This research was funded in part
by the Tsinghua University - University of Chicago Joint Research Center for Economics and
Finance. The views expressed herein are those of the authors and do not necessarily reflect the
views of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.
© 2022 by Emanuele Colonnelli, Bo Li, and Ernest Liu. All rights reserved. Short sections of
text, not to exceed two paragraphs, may be quoted without explicit permission provided that full
credit, including © notice, is given to the source.
Investing with the Government: A Field Experiment in China
Emanuele Colonnelli, Bo Li, and Ernest Liu
NBER Working Paper No. 30161
June 2022
JEL No. C93,D2,D20,D22,G0,G02,G18,G28,G3,G38,G4,O0,O1,O14,O16,O17,O25,O3,O38,O4,O40,O47
ABSTRACT
We study the demand for government participation in China’s venture capital and private equity
market. We conduct a large-scale, non-deceptive field experiment in collaboration with the
leading industry service provider, through which we survey both sides of the market: the capital
investors and the private firms managing the invested capital by deploying it to high-growth
entrepreneurs. Our respondents together account for nearly $1 trillion in assets under
management. Each respondent evaluates synthetic profiles of potential investment partners,
whose characteristics we randomize, under the real-stakes incentive that they will be introduced
to real partners matching their preferences. Our main result is that the average firm dislikes
investors with government ties, indicating that the benefits of political connections are small
compared to the cons of having the government as an investor. We show that such dislike is not
present with government-owned firms, and this dislike is highest with best-performing firms.
Additional results and follow-up surveys suggest political interference in decision-making is the
leading mechanism why government capital is unattractive to private firms. We feed our
experimental estimates and administrative data into a simple model of two-sided search to discuss
the distributional effects of government participation. Overall, our findings point to a “grabbing
hand” interpretation of state-firm relationships reflecting a desire by the government to keep
control over the private sector.
Emanuele Colonnelli Ernest Liu

Booth School of Business Department of Economics
University of Chicago Princeton University
5807 S Woodlawn Ave 20 Washington Road
Chicago, IL 60637 Princeton, NJ 08540
and NBER and NBER
[email protected] [email protected]
Bo Li
Tsinghua University
43 Chengfu Rd
Beijing, China
100083
[email protected]
An appendix is available at http://www.nber.org/data-appendix/w30161

INVESTING WITH THE GOVERNMENT 1
1. Introduction
Government participation in the economy via direct or indirect ownership of private

sector firms remains pervasive around the world (La Porta, Lopez-de Silanes, and Shleifer,
1999; Bortolotti and Faccio, 2009; Aminadav and Papaioannou, 2020).1 The political and
academic debate around this issue is typically framed through the lens of the model of gov-
ernment which best represents the relationship between state and firms (Shleifer and Vishny,
1998). Under the “grabbing hand” model—commonly used to describe Russia and Eastern
Europe in the 1990s—government interference by bureaucrats and politicians represents a
key friction to the growth of private businesses. In contrast, under the “helping hand” model
the government helps private sector firms overcome market failures (Musgrave, 1959; Stiglitz,
1989).
Understanding what model of state-firm relationships is at play has especially important
implications in China, a massive high-growth economy where the state represents a leading
investor in and minority owner of private firms (Allen, Cai, Gu, Qian, Zhao, and Zhu, 2021).
Given China’s economic success, one narrative is that firms wish to get connected to the
state to grow faster: indeed, many of the most successful Chinese corporations, including
publicly traded ones, have the government as an early investor, as documented by Bai, Hsieh,
Song, and Wang (2020). An alternative view is that what we observe reflects a desire by the
state to keep control over the private sector, suggesting that firms would grow more absent
government interference.
However, understanding the nature of state-firm relationships remains challenging due
to the fundamental difficulty of measuring the private sector demand for government par-
ticipation. That is, do market participants view the government as the grabbing or helping
hand? In this paper, we tackle this question directly by combining a field experiment with
new administrative and survey data to ask whether—all else equal—firms prefer to receive
capital from the government vis-à-vis private investors. Our context is that of venture capital
and private equity (VCPE) in China, representing the second largest market for innovative
and high-growth firms in the world (after the U.S.) and a multi-trillion dollar market—
comparable to the entire GDP of Italy or Brazil, depending on the estimates—where the
government plays a central role in the allocation of capital. Specifically, we focus on the
matching between capital investors, i.e., the Limited Partners (LPs), and profit-seeking
firms, i.e., the fund managers or General Partners (GPs), that manage the invested capital
by deploying it to high-growth entrepreneurs.2
1
See, for example, The Visible Hand (The Economist, 2012).
2
In the paper, for brevity, we will primarily use the standard VCPE terminology of LPs and GPs, even
though we will at times also refer to them as “investors” and “firms,” respectively.
In the first part of the paper, we characterize the role of government in the Chinese
VCPE market by matching data on VCPE investments over the 2015–2019 period with
administrative business registration records, through which we can observe the ownership
structure of all firms (GPs) and investors (LPs) in the data. We establish four main descrip-
tive facts. First, the government—represented by central, provincial, and local government
agencies as well as state-owned enterprises (SOEs)—is the leading investor, with the gov-
ernment as a majority owner of about half of LPs, and government LPs significantly larger
investors than private LPs. Second, the government is also a minority owner of a significant
share (about a third) of GPs. Third, government-owned GPs perform worse than private
GPs. Fourth, there is a pattern of assortative matching, with government LPs investing
disproportionally more in government-owned GPs. These findings are interesting on their
own, but the inability to disentangle demand and supply of government capital means they
could support many different interpretations, thus motivating our experimental approach.3
In the second and central part of this paper, we therefore aim to estimate the firm
demand for government capital. As mentioned, the main challenge with administrative
data is that we typically observe only equilibrium outcomes. Additionally, conditional on
observing certain GPs obtaining capital from government LPs, what we observe might not
be reflective of a preference for government LPs, but rather for certain characteristics of
those LPs that are correlated with being connected to the government (e.g., government
investors are deep-pocketed). We circumvent these empirical challenges by conducting a
field experiment in 2019 in collaboration with the leading VCPE industry service provider
in China, Zero2IPO. Our collaboration led to a new experimental survey of 688 leading
GPs in the market, which together manage nearly $1 trillion. Thanks to the deep industry
connections of our partner, we were able to obtain a response rate of more than 43%, which is
extremely high for the setting. The survey launched as part of a new service by Zero2IPO that
aims to experimentally measure GP preferences so as to help GPs connect to investors. The
experiment has real-world implications as Zero2IPO regularly facilitates the match-making
process between GPs and LPs, and because informal and industry networks and referrals are
the primary way through which investment relationships are formed in VCPE (Hochberg,
Ljungqvist, and Lu, 2007). The surveys were conducted within Zero2IPO’s internal system,
with Zero2IPO also calling each respondent to explain the details of the project and the
incentive structure.
The experiment is inspired by the literature in labor economics and discrimination on
correspondence audit studies (Bertrand and Mullainathan, 2004), and more specifically by
3
For example, these results might be driven by common narratives of the government favoring underperform-
ing politically connected firms, or they might be driven by the government interfering in firms’ operations
in a way that lowers performance. Further considering the equilibrium implications of these various mecha-
nisms, as well as the broader efficiency goals of the government, makes a clean identification of the pros and
cons of receiving government capital from the firms’ perspective quite challenging.
its recent refinement without deception by Kessler, Low, and Sullivan (2019). Akin to the
discrimination literature, the motivation for the experiment is straightforward, as simply
asking respondents about their perceptions of government investors would hardly lead to
accurate measurement, both because the issue is sensitive and because respondents would
not account for other characteristics correlated with being a government investor. As part
of the experiment, GPs are asked to rate 20 profiles of LPs (on a 10-point Likert scale)
along two main dimensions: (i) how interested they would be in establishing an investment
relationship with the LP (under the assumption the LP is interested); and (ii) the likelihood
that the LP would be interested in entering an investment relationship with them if they had
the chance. There is no deception because GPs know the LP profiles are hypothetical. The
profiles are pieces of text that are created to look exactly like those typically available on
the Zero2IPO platform that both GPs and LPs use to research the industry. Additionally,
the incentives to report truthfully are strong because within this high-stakes real context,
Zero2IPO promises to use the ratings of each GP to introduce them to real LPs that match
their preferred characteristics. Importantly, Zero2IPO ensures that our respondents are
exclusively high-level investment managers and partners of leading VCPE firms. The fact
that these respondents are willing to spend 45 minutes on the survey suggests they find the
incentive valuable.
An attractive feature of this setting is that we have full control over the creation of the
LP profiles, which allows us to estimate GP preferences for several randomized characteris-
tics of LPs, while holding other characteristics fixed. We create the profiles together with
the Zero2IPO research team by decomposing real profiles into “components” that profiles
typically consist of, closely following the distribution of profiles on the Zero2IPO platform.
For example, almost all profiles list the headquarters of a given LP, or the amount of capital
they are looking to invest. Importantly, many profiles also list the relationship of the LP to
the government, perhaps because they are SOEs or because they received endorsement by,
say, a provincial government. For each component, e.g., whether an LP has government ties,
we create multiple pieces of text that allow us to signal the LP type in a realistic manner.4
We randomize components to generate the synthetic profiles we use to elicit preferences,
make a few basic changes to the text to ensure language accuracy and realism of the profiles,
and pick a random set to generate the surveys we invite GPs to respond to.
Our main finding is that, on average, GPs dislike LPs with government ties. We also find
that GPs prefer deep-pocketed investors, those headquartered in Beijing, and those that are
4A natural limitation of the approach is that we can only have a finite number of different signals, and thus
we cannot measure, say, GPs’ preferences for all possible government entities. Together with Zero2IPO,
we chose the most representative options that signal relevant characteristics of an LP in our context. For
instance, the sentences indicating a tie to the government include references to SOEs as well as central,
provincial, or local government ownership or supervision of the LP.
not focused on specific industries and stages of investment. Several other components, such
as how long the LP has been investing in the VCPE market, whether they are foreign, and
whether they provide more details about their investment philosophy or corporate governance
practices do not seem to matter. All results are robust to the inclusion of respondent fixed
effects.5
The average effects we uncover indicate that the negatives of receiving capital that is
tied to the government outweigh the positive value GPs may obtain from establishing a link
to a government-related politically connected investor. Our findings are consistent with a
“grabbing hand” interpretation of the government’s involvement in the market. Anecdotally,
consistent with several anonymous discussions with a number of active VCPE firms, a leading
explanation is that government connections of the investors lead to interference in decision-
making that is due to political, rather than profit-maximizing, incentives.
To investigate mechanisms, we first explore the heterogeneity of our findings across
different types of GPs. In particular, we are interested in whether the average dislike for
government capital is driven by the ownership structure of the GP itself. If the presence
of interference in decision-making is seen as unattractive, this should be especially so for
nongovernment-owned GPs that operate according to market principles. On the other hand,
we expect the incentives of government-owned GPs to be more aligned with those of govern-
ment investors, which should result in a more favorable view of government LPs as investment
partners. In our regression of GP interest on LP characteristics, we find that the negative
coefficient on the indicator for the LP having government ties can be fully accounted for by
nongovernment-owned, private GPs. Instead, we find that government ties of the LP do not
matter for the preferences of government-owned GPs. Moreover, the dislike for government
capital is especially pronounced for the best performing private GPs. We find that no other
component of the LP profiles displays a meaningful difference depending on whether the GP
is owned by the government or not. These findings are especially informative considering
that private GPs should be those who would benefit the most from connecting with a govern-
ment investor under the “helping hand” view of government participation. Additional results
point to a largest dislike for central government agencies, followed by provincial government
agencies, and a lower or null dislike by GPs focused on state-dominated industries, where
the helping hand of the government is arguably more valuable. Moreover, we find that our
findings hold independently of whether the respondent GP already has government LPs as
investors at the time of the survey.
We provide additional, largely qualitative evidence on the mechanisms behind the dislike
of private GPs for LPs with government ties using results from a new round of surveys
5Notice that a main worry of a survey like ours in China is that respondents might be wary of rating poorly
entities that appear linked to the government. However, our main finding suggests that such a worry would
only imply we are underestimating the dislike for government capital.
we conducted jointly with Zero2IPO. Designed to obfuscate their specific purpose, these
additional surveys ask respondents to evaluate a list of pros and cons of establishing a
relationship with an investor linked to the government. By and large, GPs lament the
presence of political interference in decision-making by LPs with government ties. To a
lesser extent, GPs also consider the presence of increased policy uncertainty and the lack
of professionalization of teams working for LPs tied to the government to be unattractive
features of government LPs.
We expand on our experimental analysis of the role of government participation in the
matching between GPs and LPs by conducting a contemporaneous analogous survey of the
other side of the market, namely of high-level managers of LPs that are responsible for
selecting the GPs to provide capital to. We are able to survey 312 LPs, with a response rate
of 39%. The survey and the creation of synthetic profiles follow a structure similar to the
GP survey, and the incentive is identical. The profile components are slightly different to
reflect the different type of market participants. Some of the key findings are that LPs prefer
high-performing, foreign, recently established GPs that have a specialized focus in specific
industries. What stands out, however, is that the strongest determinant of LP interest in a
GP is whether that GP already has entities with government ties among its investors. We
also find that LPs value positively GPs whose team members have direct experience in the
government, while industry experience does not matter. Unlike the GP-level analysis, we do
not find much heterogeneity depending on whether the LP is government-owned or not.
In summary, our experimental surveys reveal substantial heterogeneity in preferences
for government participation from both the firm and investor sides of China’s VCPE market.
Given the two-sided nature of the market, changing the extent of government participation
has potentially rich, equilibrium implications on all market participants. To study these
distributional consequences of government participation, based on the elicited preferences,
we build a simple two-sided search and matching model of government and nongovernment
GPs and LPs, discussed in the third and final part of the paper. We calibrate the model
using both our experimental surveys and the administrative data, and we conduct several
policy counterfactuals. First, our model finds and quantifies that increasing government
LP participation will reduce the surplus of nongovernment GPs, while strongly improving
the surplus of government GPs, and in particular that of low-performing government GPs.
Second, we show that counterfactual policies which channel government capital exclusively
to private or well-performing GPs can substantially reduce the average surplus on both
sides of the market, consistent with the dominant role played by government entities as
both investors and firms. One implication of this result is that the empirical regularity
that government investors tend to invest in low-performing firms does not necessarily reflect
capital misallocation by government investors, but may be driven, at least in part, by the fact
that the best-performing firms prefer private capital in the first place. All together, these
findings point to important distributional effects of government participation and highlight
the crucial importance of government entities in driving the allocation of capital in China’s
VCPE market.
Our study is related to a well-established body of work on the role of government par-
ticipation in the economy. Shleifer (1998) reviews the arguments supporting and against
state ownership in a number of economic sectors, and Megginson and Netter (2001) give an
overview of a large literature on privatization. Several studies emphasize the many inefficien-
cies that arise when the government participates in economic activity and financial markets
(Shleifer and Vishny, 1994; La Porta and Lopez-de Silanes, 1999; La Porta, Lopez-de Silanes,
and Shleifer, 2002; Sapienza, 2004; Dinç, 2005; Bai, Lu, and Tao, 2006), with a related and
large literature on the benefits of political connections (Fisman, 2001; Khwaja and Mian,
2005; Faccio, Masulis, and McConnell, 2006) and the costs of corruption (Shleifer and Vishny,
1993; Fisman and Golden, 2017; Colonnelli and Prem, 2022).6 Our approach differs from the
existing literature which, by predominantly studying the effects of government intervention,
leads to findings that typically reflect the combination of the state’s active involvement in
the economy with the selection of firms willing to do business with the state in the first place.
Our key insight and contribution is the estimation of demand for government participation,
by means of a novel field experiment, which puts the spotlight on the pros (e.g., political
connections) and cons (e.g., political interference in decision-making) as seen directly from
the perspective of the private sector. Our results show that the cons outweigh the pros,
thus supporting a “grabbing hand” model of government participation, where government
investors are unattractive to private firms, and especially so to high-performing firms. While
our study does not speak to the broader goals of the state and their overall efficiency impli-
cations, we provide evidence suggesting a desire by the government to keep control over a
specific but rather important part of the private sector.
A related contribution of our work is to provide a comprehensive account of the VCPE
market in China. In particular, despite its size and importance for both innovation and
growth, extremely little is known about preferences of private and government-owned fund
managers and investors, and the role of government participation more broadly (Huang,
Tian, Amstad, Sun, and Xiong, 2020; Cong, Lee, Qu, Shen et al., 2020). This is in stark
contrast with the growing body of evidence regarding the Chinese government’s impact on
6Relatively little is known in the context of high-growth firms. Lerner (2009) provides a critical account
of government policies aimed at spurring innovation, and Bai, Bernstein, Dev, and Lerner (2021) collect
new data to study the characteristics of public entrepreneurial finance programs around the world. Related
papers include Lerner (2000), Howell (2017), Fang, Lerner, Wu, and Zhang (2018), and Babina, He, Howell,
Perlman, and Staudt (2020). Recent work has also looked at the direct provision of venture capital funding
through specific government vehicles in China and around the world (Brander, Du, and Hellmann, 2015;
Cumming, Grilli, and Murtinu, 2017; Fei, 2018).
other sectors of the economy (Young, 2000; Song, Storesletten, and Zilibotti, 2012; Hsieh
and Song, 2015; Xiong, 2018; Liu, 2019; Beraja, Yang, and Yuchtman, 2020; Jia, Lan, and
i Miquel, 2021) and the financial system (Brunnermeier, Sockin, and Xiong, 2020).7 Bai et al.
(2020) and Allen et al. (2021) describe the ownership structure of private firms in China,
uncovering an increasingly blurry distinction between state-owned and privately owned firms
and emphasizing the important implications of disentangling the reasons behind this new
form of state-firm relationships. Our paper provides a novel finding to inform this debate—
that government capital is unattractive to private firms—which has implications for under-
standing the nature of China’s economic growth. Given the tight link between government
participation and development, our paper also naturally relates to earlier work on finan-
cial development and economic growth more broadly (King and Levine, 1993; Rajan and
Zingales, 1998; Levine, 1999; Wurgler, 2000; Levine, 2002).
Finally, we directly contribute to the literature on venture capital and private equity (see
Da Rin, Hellmann, and Puri, 2013 for a review). Bernstein, Lerner, and Schoar (2013) and
Andonov, Hochberg, and Rauh (2018) discuss the role of political investors in the contexts of
sovereign wealth funds and U.S. public pension funds, respectively. Survey evidence on high-
level decision makers in VCPE include Gompers, Kaplan, and Mukharlyamov (2016), Da Rin
and Phalippou (2017), and Gompers, Gornall, Kaplan, and Strebulaev (2020). Few experi-
ments have been conducted in this area, and they largely focus on early stage investments in
the U.S. (Bernstein, Korteweg, and Laws, 2017; Gornall and Strebulaev, 2020; Zhang, 2020).
To our knowledge, ours is the first field experiment that identifies preferences of both GPs
and LPs. We do so in a novel match-making setting, with strong real incentives and a high
response rate, and by targeting a large sample of high-profile managers of leading entities
in the market. In particular, we contribute to the understanding of both the search and
matching process in the VCPE market—with a specific focus on GP-LP matches (Lerner,
Mao, Schoar, and Zhang, 2022) rather than those between GPs and the target investments
(Sørensen, 2007; Ewens, Gorbenko, and Korteweg, 2022)—and of VCPE in emerging mar-
kets, more broadly (Lerner and Schoar, 2005; Kaplan, Martel, and Strömberg, 2007; Lerner,
Schoar, Sokolinski, and Wilson, 2018).
The paper is organized as follows. Section 2 briefly describes the context of VCPE in
China. Section 3 describes the main sources of administrative data and establishes a few key
facts. Section 4 illustrates the details of the experimental design. Section 5 reports the main
results. Section 6 discusses the model and equilibrium impact of government participation.
Section 7 concludes.
7See Amstad, Sun, and Xiong (2020) for an extensive review of the literature.
2. Institutional Context
We study the venture capital and private equity (VCPE) market, which refers to capital
investments in firms that are not publicly listed or traded. While venture capital—which
specifically refers to the funding of high-growth, high-risk companies, typically innovative
entrepreneurial startups—is seen as largely distinct from private equity more broadly in the
U.S. and most other developed economies, such distinctions are quite blurry in China (Huang
et al., 2020). We therefore refer to the general “VCPE” market and investors therein, noting
that the market is characterized primarily by early stage and growth equity investors, which
will be our focus throughout the whole paper. The VCPE market in China is second in size
only to the U.S..
The main players in the VCPE market are the capital providers, which are typically
referred to as Limited Partners (LPs), and the firms that manage the capital invested, namely
the General Partners (GPs), that subsequently deploy the capital by acquiring ownership,
or equity, in other typically high-growth firms. Such investments generate returns to the
investors once the firms’ shares are sold, either publicly through an IPO or privately to
other investors or firms. GPs also capture a share of the profits, in addition to their asset
management fee. Specifically, one or more LPs generally invest capital into a “fund,” which
is the pool of capital raised by a given GP. LPs can invest into more than one fund, and
a GP can raise multiple funds over time. This structure, typical of the U.S. market, is
known as “limited partnership,” and it also became the dominant structure in China with
the Partnership Enterprise Law of 2007. In this context, LPs are considered “passive”
investors, to the extent that their limited liability comes at the cost of not interfering with
the investment allocation decisions of the GP. In practice, however, examples abound about
how LPs can exert a certain degree of influence over how the capital is ultimately allocated.8
As we will show in the next section, a distinctive feature of VCPE in China is the
predominant role played by the government in the allocation of capital. Central government
agencies, local governments, and State-Owned Enterprises (SOEs) supervise or own (partially
or wholly) a large share of the LPs actively operating in the market, thus playing a primary
role in driving high-growth entrepreneurship and private sector development. For instance,
the LP may be a SOE funded by the Provincial People’s Government. Similarly, a local
government may formally approve the establishment of an LP and guide its capital allocation
8While the two-sided nature of the market is the most common in the U.S., China, and around the world,
there are a myriad other nuanced variations of the VCPE model, such as GPs and LPs playing both the
role of investor and fund manager at the same time. For brevity, we abstract away from these details in the
paper, except when clarifications are important for our empirical analysis and arguments. For comprehen-
sive descriptions of all the details and nuances of the VCPE model, from compensation and management
structures to distribution of profits and legal restrictions, see Lerner, Leamon, and Hardymon (2012), among
others.
process. The role of government as an LP is at times made operational by the existence of so-
called “government guided funds,” namely mixed private-public funds created and partially
contributed to by government entities (usually local governments), to which nongovernment
LPs are also expected to contribute. In our paper, for brevity, we will consider LPs as having
government ties if the government is involved in any role in providing capital to any fund
managed by a given GP.
We focus on the matching between GPs and LPs. Within this setting, learning to
deal with government-related entities is often considered a “required course” for VCPE fund
managers.9 Many argue that having the government as an investment partner introduces
inefficiencies in the investment process and can distort the allocation of capital away from
their most productive uses. There are several reasons for why this is the case, as illustrated
through large qualitative evidence gathered in the recent reviews by Malkin (2021) and Lu-
ong, Arnold, and Murphy (2021). First, the government is seen as a more “active” investor
compared to other (commonly passive) LPs as, after the capital is disbursed, it often intro-
duces restrictions on the specific types of investment the GPs can undertake, for example by
trying to favor specific firms, locations, or sectors. Due to political incentives, government
LPs might also want to prioritize projects that are less risky or that can generate returns
within a short time frame. These are all potentially severe forms of interference for GPs,
who tend to look for risky projects with high upsides that often require a long investment
horizon and a high degree of flexibility in decision-making. Moreover, such distortions are
emphasized by the fact that relying on the government as an investor can lead to extra expo-
sure to policy uncertainty, for example because changing government objectives may lead to
unexpected interference in investment decisions. Another source of inefficiency argued by op-
ponents of government participation in the market is the presence of bureaucrats or political
actors, rather than investment professionals, in investment and managerial committees.
There are, on the other hand, several reasons why—from the perspective of fund man-
agers and entrepreneurs alike—having the government as an investor may confer a number
of advantages. Typically, such benefits range from faster regulatory approvals and tax reduc-
tions to better access to information and other favors occurring thanks to political connec-
tions, especially in state-dominated sectors such as construction, mining, or manufacturing.
In particular, government’s support is often seen as necessary to “open doors” for target
firms to grow. For these same reasons, having the government as an investor might be seen
as a positive signal by other investors who are looking for GPs to manage their capital, and
having government-connected individuals in the investment team may prove valuable.10
9See The Chinese state is pumping funds into private equity (The Economist, June 2021).
10From a social perspective, which remains beyond the scope of our paper, the main argument is about
externalities, as the government may allow for capital to flow to projects that would otherwise remain
underfunded (see Lerner (2000) for a discussion). In China, this is reflected in a push by the government for
3. Venture Capital and Private Equity in China: Data and Descriptive

Analysis
In this section, we describe the main sources of administrative data we use throughout
the paper. First, we describe the administrative data from Zero2IPO on General Partners
(GPs), Limited Partners (LPs), and Venture Capital (VC) and Private Equity (PE) invest-
ments (Section 3.1). We then illustrate the data on the ownership structure of GPs and LPs
and related measures of government connections (Section 3.2). Finally, in Section 3.3, we
discuss basic summary statistics of our sample and establish a few descriptive facts.
3.1. Administrative Data on Venture Capital and Private Equity. Our primary
source of administrative data is the full database created and maintained by our research
partner Zero2IPO, which collects data on VCPE firms and their investments in a number
of ways. First, they continuously aggregate multiple sources of data, from administrative
registries such as those of the Asset Management Association of China (AMAC) and the
National Enterprise Credit Information Publicity System (NECIPS), and those of stock ex-
changes and regional equity markets, as well as from several industry associations and com-
peting data platforms, and including information announcements from government agencies
and news press releases in VCPE-focused publications.
These data cover GPs and LPs actively operating in the market, but the lack of formal
reporting requirements makes them imperfect with respect to coverage of deals and their
performance, a typical issue in markets for private capital around the world. To alleviate
this issue, Zero2IPO collects its own data through a range of quarterly and annual online
surveys, which are regularly validated through in-person meetings and follow-ups with re-
spondents via phone and at leading conferences, workshops, and similar events throughout
the year. Finally, Zero2IPO has a dedicated research team to cross-check and standardize
the information, not only across data sources but also by verifying the information reported
by multiple parties (e.g., GP and LPs in a given deal). Overall, despite some limitations
that are standard given the context, the data collection and validation process of Zero2IPO
is largely similar to that of leading and widely trusted data providers in the VCPE space in
the U.S., such as PitchBook and Preqin.
Because of the nature of the data collection, the database provides accurate information
about the identity of GPs, LPs, and the funds they are associated with, together with registry
information such as company name, founding date, headquarters location, and registered
capital. We match GPs and LPs using the fund-level data, which indicates the GP managing
the fund and the LPs that committed capital to the fund. For each of the entities in the
data, the Zero2IPO data platform also provides a text-based profile description of the entity.
capital flows to strategic sectors and locations that private LPs, such as wealthy families and individuals or
large corporations, are not targeting.
We design the synthetic profiles used for the experimental surveys to mimic these real-world
textual profile descriptions, a point we return to in detail in Section 4. Finally, for a subset
of the sample we have access to data at the deal level, which includes information on the
target company, deal’s size and date, and round of fundraising, among others.
3.1.1. Measuring Performance. A common issue with VCPE data is that observing perfor-
mance measures is difficult, because the data often remain confidential and because there
are several weaknesses associated with various measurement approaches, not least due to
the dependence on data from unrealized private investments (see Phalippou (2008), Cole,
Melecky, Mölders, and Reed (2020), and Jeffers, Lyu, and Posenau (2021) for discussions of
these issues).
Similarly to most standard U.S.-focused datasets, our data also lack the universe (and
respective timing) of cash-flows between GPs, LPs, and funds that is ideally needed to
compute returns. However, our close collaboration with Zero2IPO allows us to construct
a measure of returns, which they label “comprehensive return” (henceforth, CR), and that
is typically unavailable to researchers. The CR is a weighted average of various measures
Zero2IPO collects, such as funds raised, investments, and exits, among others that are ob-
tained directly from the entities and that remain confidential. Because the magnitude of
this measure is not directly interpretable, in our analysis we use each GP’s corresponding
quantile of CR as a performance measure between 0 and 1. While also subject to many
of the common concerns, the CR is relevant to the extent that it is used by Zero2IPO to
compile its yearly rankings of GPs in China, which are the most authoritative in the market
and relied upon by many investment professionals. Whenever we split GPs in terms of high
versus low quality in the paper, we do so by cutting the sample at the median of CR, and
considering a GP as high quality if it has above-median CR or if it was ever ranked as a top
GP by Zero2IPO.
Finally, despite the fact that they are sensitive to the timing of cash flows, whenever
using performance data, we further report robustness results that use the simpler measure
of internal rates of return (IRRs), which are reported by the GPs directly to Zero2IPO for
a subset of the data.
3.2. Measuring Government Ownership. We measure whether GPs and LPs are par-
tially or wholly owned by the government using business registration data from NECIPS,
as in Bai et al. (2020). We access the database through a dedicated API provided by the
commercial company Tianyancha. The database contains the ownership structure of each
legal business entity in China. That is, for each entity, we can observe its shareholders, and
the shareholders of each shareholder, until we reach the ultimate owners and their respective
shares in the given entity.
To define government ownership, we search for ultimate owners that are either state-
owned enterprises (SOEs) or (central, provincial, or local) government agencies. We obtain
the most comprehensive list of SOEs from the State-owned Assets Supervision and Adminis-
tration Commission (SASAC), which we match to the business registration data. To identify
government agencies, we proceed in two steps. First, we create a list of agencies from the
State Council and from each provincial government’s website, respectively. Second, starting
from these lists, we extract the primary keywords in their names that are indicative of a
government agency, such as “department,” “administration,” “bureau,” and “government,”
and search for these keywords in the business registry data. We do a similar search for the
list of city names in the data, as many local governments are city administrations. We then
manually go over the results from the searches to screen out false positives, and to categorize
government agencies into central, provincial, and city (hereafter, local) level agencies, for
a total of 124 central, 220 provincial, and 1,110 local government agencies in the business
registration data. We complement these data with data collected by Zero2IPO itself through
their regular surveys regarding the ownership and government relation of LPs and funds.
Our main analyses consider GPs and LPs as government-owned if they have a positive
share of government ownership: that is, if any of their ultimate owners are a government
entity, we consider a GP or LP as government-owned.11
3.3. Descriptive Analysis. The main starting administrative data sample we rely on
throughout the paper consists of all GPs that are labeled as “active” by our partner and data
provider, Zero2IPO, as of December 2019. This includes all GPs that have made at least an
investment or managed a VCPE fund in the 5-year period 2015–2019, and that Zero2IPO
flagged as GPs for which confidence regarding data quality is high.12 The data do not include
individual investors, and so the focus is only formal business entities, which account for the
bulk of VCPE capital in the market. We have a total 6,308 active GPs, which include all
respondents to our survey, described in Section 4. We then define as “active” all LPs that
have ever invested in a fund managed by an active GP. We have a total of 7,974 active LPs,
which also include all respondents to our survey. We were able to collect ownership infor-
mation for the near-universe of these GPs and LPs.13 Overall, our sample can be considered
as having high-quality coverage of the main players in the VCPE ecosystem in China. On
11We report robustness to another commonly used definition to capture corporate control, according to which
we define as government-owned only those entities where the government owns at least 20% of the shares
(Aminadav and Papaioannou, 2020). For brevity, we add to the the Appendix only the tables corresponding
to the main analysis tables. These robustness tables are Appendix Tables A1, A4, A5, and A14.
12For example, GPs that appear to have made an investment in the same 5-year period but that Zero2IPO
was unable to reach to validate the information are typically not included.
13The only exception are the GPs that are registered as foreign entities. We classify these GPs as privately
(i.e., nongovernment) owned. Because our respondents are not foreign, we remove foreign-owned GPs and
LPs from the descriptive statistics reported below.
the other hand, smaller local players are less likely to be labeled as active by Zero2IPO or
to have any data reported in the Zero2IPO database.
As discussed later, in Section 4.1, 688 GPs and 312 LPs responded to our surveys. Of
these, we drop from the main analysis 11 GPs and 2 LPs that did not fully complete the
surveys. Unfortunately, for confidentiality reasons, we are unable to observe the sample
of 1,600 GPs and 790 LPs that received an invitation to participate to our study. In the
tables discussed in this section, as well as in Appendix Figures A1 and A2, we report a
comparison of the basic characteristics of our respondents to the other GPs and LPs in our
main dataset. Similarly to the VCPE studies of Gompers et al. (2016) and Gompers et al.
(2020), our sample selection leads to a final sample of respondents that is more representative
of large and active players in the market.
We present a few main facts to characterize the VCPE market in China, focusing the
discussion on all active GPs and LPs over the period 2015–2019.14
Fact 1: The government is the leading VCPE investor. Table 1 reports summary statistics
on our main data sample, showing the characteristics of LPs (Panel A) split by government-
owned and nongovernment owned entities. The first fact we point to in the data is the
dominant role of government investors in the market. First, about half of the entire set
of investors consists of government-owned LPs, as shown in the first row of the LP panel.
Second, there is a large difference in size between government-owned investors and other
investors, with the former investing significantly larger amounts of capital (about six times
more than a nongovernment owned LP) and investing in more VCPE funds on average.
We characterize the role of government investors in several additional ways. Table 2
reports a more detailed breakdown of government ownership shares across different layers of
the government. The government is typically a majority owner of the LPs: in Panel A of
Table 2, we find that conditional on having at least one government shareholder, the median
LP ownership share by the government is 82.62%. The additional statistics by government
layer indicate the distribution of ownership conditional on the LP having at least a positive
ownership share by that government type (central, provincial, or local), pointing to the
pervasive presence of local governments in the market.15
We further report the distribution of LP types in Appendix Table A2, using the internal
classification of Zero2IPO and weighting by the total investment amount of each LP type
over 2015–2019. Not only are the majority of entities dedicated VCPE institutions, but there
is also a range of players typical of other leading international VCPE markets. Importantly,
14The facts established in this section apply similarly to the sample of respondents only. In addition to the
output discussed below for facts 1 and 2, we report also Appendix Tables A7 and A9 to show that facts 3
and 4 hold in the sample of respondents only as well.
15In Appendix Table A3 we also show what share of LPs is owned by central, provincial, or local government
agencies, respectively.
while the government does have wholly owned entities such as government bureaus and
guided funds, which do not have a counterpart among private investors, we find a large
overlap across other entity types. This evidence speaks to the government functioning in
many regards—by means of its widespread ownership stakes—as a typical LP in the market
looking to provide capital with GPs to obtain investment returns.
Finally, Figure 1 display the distribution of headquarters location, investment region,
and investment industries among active LPs, and Figure 2 instead illustrates the differences
across government-owned versus other entities. Relative to private investors, government
investors are more focused on traditional industries (e.g., manufacturing) and less developed
regions (e.g., inland China). However, we still observe a large degree of overlap across regions
and industries.
Fact 2: The government is a minority owner of a significant share of VCPE fund managers.
Moving the focus to the GP-side of the market, we establish that a striking 38% of these
fund managers also have a positive share of government ownership, as shown in Panel B of
Table 1.
Akin to the LP analysis, we find that government-owned GPs are also larger, as they
have higher assets under management (AUM). As reported in Table 2, however, the gov-
ernment is typically a minority owner of the GPs, with the median government-owned GP
having a 41.97% government ownership share. Appendix Tables A2 and A3 and Figures
Figures 1 and 2 report additional summary statistics analogous to the previous analysis of
LPs.
Fact 3: Government-owned fund managers perform worse than their private counterparts.
We find that government-owned GPs have a lower performance compared to privately owned
GPs. While this is already apparent in the raw summary statistics of Table 1, which show
a much lower internal rate of returns (IRR), we can also analyze it more precisely when
controlling for other characteristics. In Table 3 we observe that government-owned GPs
have lower comprehensive returns (CR, introduced in Section 3.1.1) as well as lower internal
rates of return (IRR), even after controlling for size (AUM) and location (headquarters fixed
effects). While these performance measures are imperfect—the performance of GPs tends to
be multidimensional and not easily quantifiable—these patterns are nevertheless suggestive
that government-owned entities tend to underperform in terms of generating financial returns
on investments. These findings are consistent with other work on government funding in
China, as reviewed by Cong et al. (2020).
Fact 4: There is assortative matching, as the government invests disproportionally more in

government-owned fund managers. Among the actual GP-LP matches, there is sorting along
the dimension of government ownership: government-owned GPs are significantly more likely
to receive capital from government-owned LPs, and conversely, government-owned LPs are
significantly more likely to invest in government-owned GPs.
These patterns are illustrated in Table 4, where we report the likelihood ratio index for
each pair of LP and GP types. The likelihood ratio index for each GP of type i and LP of
type j, with i, j ∈ {government, nongovernment} is defined as
P r(GP of type i matches with LP of type j)
s(i, j) = .
P r(a random GP has type i) × P r(a random LP has type j)
The measure s(i, j) benchmarks the empirically observed frequency of matches relative to the
frequency that would have occurred by chance. If GPs and LPs form matches at random—
without sorting by type—then the likelihood ratio should be equal to one in a large sample.
A likelihood ratio s(i, j) above one indicates that matches between type-i GPs and type-j
LPs occur more likely than could be attributable to chance, suggesting a preference to match
on both sides relative to potential partners of other types. Conversely, s(i, j) < 1 indicates
that type-i GPs and type-j LPs may have a dislike to be matched with each other.
4. Experimental Design
The previous section establishes a few basic facts regarding the matching between GPs
and LPs. Yet, the equilibrium nature of the observational data makes it difficult to tease
out the demand and supply of government capital. In this section, we describe our main
experimental survey design, which aims to estimate fund managers’ demand for different
sources of capital, and specifically for capital coming from investors with government ties.
Estimating preferences for government capital versus capital from private sources is
empirically challenging for several reasons. First, it is difficult to separate capital coming
from government investors from other confounding factors, such as the fact that they tend to
have deep pockets, as we established earlier. That is, that the investor has government ties
is correlated with a host of other traits of the investor. Second, government investors may
be more or less inclined to provide capital to a given GP, relative to other investors. As a
result, GPs may have differential expectations about whether the government investor would
provide capital to them in the first place. Third, any match between GPs and investors in
observational data would reflect both preferences as well as the endogenous matching process
during which the GP observes several other characteristics of the investor that are unobserved
by the econometrician.
Therefore, the objective of our experiment is to create an environment where we can
randomize whether an investor is connected to the government while holding fixed other
characteristics, and where we can isolate GPs’ preferences for investors independent of the
likelihood of a match. To do so, we ask GPs to rate synthetic LP profiles by providing a strong
incentive that aligns our research interests with the interests of the GPs. The incentive for
GPs is to be matched with real LPs by Zero2IPO—a partner that respondents trust and that
can make credible promises—based on their ratings of the synthetic profiles. Such a design
is inspired by the work of Kessler et al. (2019) and Low (2021) to measure preferences for
individual characteristics without deception in the hiring and dating settings, respectively.16
This design provides a deception-free alternative to correspondence audit studies, common in
the literature on discrimination in labor markets, which are especially difficult to realistically
conduct in high-stakes contexts like ours where trust is of major consideration. The setting
also allows us to go beyond typical binary outcome variables based on “call-back” rates, as
we are able to ask respondents to rate investors on multiple dimensions while providing them
with specific instructions about factors that should not enter into their rating.
Our research design is further explained in what follows. We introduce the survey we
conducted with Zero2IPO in Section 4.1, focusing on the recruitment process and inventive
structure. In Section 4.2, we illustrate how we create the pool of realistic, synthetic profiles
of GPs and LPs, including details on the specific features we include in the profiles. In
Section 4.3, we discuss the questions we ask respondents to rate potential partners, which
will be used as dependent variables in our analysis.
4.1. The China Equity Investment Survey: Recruitment and Incentives. The core
of our paper are new experimental surveys of a large number of GPs and LPs we conducted
in collaboration with Zero2IPO, widely considered the leading integrated service and data
provider in the China VCPE market since its founding in 2001. We conducted these surveys
in the last quarter of 2019.17 Specifically, we designed a new survey instrument, which
we labeled the “Chinese Equity Investment Survey,” designed to be filled in by high-level
managers or partners of the targeted organizations.
The process of recruiting respondents is managed directly by Zero2IPO, which regularly
conducts surveys of GPs (and LPs) in the VCPE market in China. Zero2IPO has also recently
started to play the important role of facilitating the matching between GPs and LPs, by
means of face-to-face events and introductions made among various industry players. To
this end, our survey is marketed as a joint collaboration between Zero2IPO and Tsinghua
University PBC School of Finance, with the objective of using machine learning techniques
to improve the matching between GPs and LPs.18 Specifically, the respondents are truthfully
told that survey responses, namely their rating of synthetic investment partner profiles, would
be used to introduce them to real LPs matching their preferred characteristics. Importantly,
Zero2IPO further conducted follow-up phone calls with the GPs after the survey links were
16See Harrison and List (2004) for a broader discussion of “framed field experiments.”
17The surveys were conducted before the first case of COVID-19 was reported on December 31st, 2019.
18As a result, in addition to the sample selection discussion in Section 3.3 regarding our respondents being
typically large and active players in the market, our incentive structure also likely leads to the selection of
players that are looking for investment partners at the time of the survey.
sent, further explaining the project’s goal and reiterating the main participation incentive
of introductions to potential capital providers. Zero2IPO also explained the details of the
synthetic rating part of the survey, ensuring respondents understood both the incentive and
the rating questions.19 We report the full recruitment script sent to respondents, translated
to English, in Figure 3.
Zero2IPO sent the surveys to a total of 1,600 GPs and 790 LPs, respectively. All GPs
surveyed are profit-driven. We obtained a total of 1,000 responses, 688 from GPs and 312
from LPs, for an average response rate of 42%. This response rate is high for this setting,
especially considering our large sample size.20 The response rate combined with the fact
that the main incentive to participate in the survey consists of being introduced to potential
capital providers gives us confidence that GPs value this incentive, as participating in a
45-minutes survey is costly for VCPE fund managers. That such introductions are valuable
is not surprising in a context like that of GP-LP matching, where the lack of a central
marketplace and survey evidence suggest that introductions by trusted third-parties are a
common tool to establish investment partnerships (Hochberg et al., 2007; Gompers et al.,
2020).
The survey is organized as follows. First, we show an introductory page describing the
goals of the survey and the incentives to participate, while also providing survey instructions
to the participants. Second, respondents are asked to rate 20 profiles of potential investment
partners along several dimensions,with this being the core part of our experiment. Third,
we ask a series of questions regarding the respondent’s company and individual role in the
organization, which are used by Zero2IPO to ensure that responses are accurate and filled
in by a high-level manager of the organization.
4.2. Creating Partner Profiles. We estimate GPs’ preferences for LPs by asking each of
them to evaluate 20 unique, synthetic profiles. These profiles are brief textual descriptions
of LPs summarizing their key features. We create the synthetic LP profiles in direct collab-
oration with the Zero2IPO research team, using a combination of automated programming
and manual checks and changes.
The first step of the process consists of a structured analysis of all text-based descrip-
tions of LPs on the Zero2IPO platform. In particular, we aim to first identify general text
19Following extensive discussions with Zero2IPO, we opted not to specify the number of introductions that
would be made. While the instructions also mention the research focus of the survey, this is pitched as
secondary. Respondents are also promised a summary of the results.
20For example, the response rates for other survey-based studies of investors are 13.8% for Da Rin and
Phalippou (2017), 10.3% for Bernstein, Lerner, and Mezzanotti (2019), 6.5% for Gornall and Strebulaev
(2020), 11.6% for Denes, Howell, Mezzanotti, Wang, and Xu (2020), 0.5% for Zhang (2020), and 2.5-4%
for Giglio, Maggiori, Stroebel, and Utkus (2021). The highest response rates in the literature are those by
Gompers et al. (2016) (47%) and Gompers et al. (2020) (21%). Relatedly, in the seminal survey work on
the practices of Chief Financial Officers, Graham and Harvey (2001) obtain a response rate of 8.9%.
organization patterns that we can use to create realistic profiles, for example by studying
how long the profile description typically is, how it is organized in terms of paragraphs, and
the order in which certain pieces of information appear. Second, we identify the pieces of
information, i.e., “components,” that a profile typically consists of, and their approximate
probability distribution. For example, we observe that LP profiles nearly always display in-
formation about their size, location, and their relation to SOEs or other government agencies.
Third, we create a few pieces of text that are often used to characterize each component,
which we generate by manually reading several hundred profiles for each component identi-
fied in the previous step. In this way we are able to ensure that survey respondents observe
realistic variation in the profiles they are evaluating, which would not be possible if all the
information was mechanically presented using the same exact sentence or words in each
profile.
Table 5 reports the variables we create from the text of the synthetic LP profiles (column
1), together with a brief explanation of what each variable captures. We expand on the
description of all profile components from which the analysis variables are generated in
Table 6, where we report all possible ways through which a given component may appear
in the text of the synthetic profile. Column 1 of Table 6 also reports in parenthesis the
unconditional probability that a given component is randomly drawn to be included in a
profile. For a given component, each piece of text has equal probability of being drawn,
conditional on the component appearing in the synthetic profile. For a given component,
certain pieces of text (displayed in bold) indicate when the dummy variable in our regression
takes value 1, while the others indicate when the variable takes value 0, as reported in the
second column of Table 6 that refers to the specific numbered text boxes.21
To illustrate, consider our main LP characteristic of interest, namely “Government Ties,”
which is drawn to appear in a synthetic profile with 80% probability. Conditional on appear-
ing, the LP displays the related text-based information in 11 possible different ways (as per
column “Options” in Table 6). Of these 11 pieces of text, 7 of them (i.e., those in bold) would
capture an LP that has government ties (i.e., GovernmentT ies = 1), while 4 of them would
indicate the LP is not linked to the government (i.e., GovernmentT ies = 0) using analogous
pieces of text. For example, a synthetic profile would suggest the LP has government ties
when it reads: “It is an investment organization established by a state-owned firm funded
by the provincial government, [...].” Meanwhile, a LP synthetic profile that does not have
government ties reads: “This company aims to give full play to the role of the market in allo-
cating resources and expand private capital investments in innovation and entrepreneurship,
[...].”22
21Respondents only see text in Chinese, but we report a translated version in English as well.
22While we discuss the other components of the profiles in more detail when reporting the results in Section
5, it is worth noting that the information provided is mostly qualitative in nature, rather than quantitative.
The second step of the process consists of randomly generating synthetic profiles of LPs
by mixing and matching the profile components according to the respective probabilities
of appearance. Staying somewhat close to the real probability distribution is important so
that respondents evaluate profiles they deem realistic. Relatedly, notice that the creation of
the final synthetic profiles involves a certain degree of manual adjustments and minor text
additions, which are carried out by the Zero2IPO team. In particular, the probabilities of
appearance of each component and the specific pieces of text used to characterize a given
component are ultimately decided by Zero2IPO. There are two reasons for this. First, text-
based profiles are not available for all LPs. Second, only Zero2IPO (and not the researchers)
was aware of the specific pool of GPs that would receive the survey invitation. As a result,
the Zero2IPO team was able to ensure that the synthetic profiles would look realistic and
be a good fit with respect to the specific sample in our study, an issue of crucial importance
as also highlighted by Kessler et al. (2019) in the context of employers screening CVs they
deem relevant to them.23
The process of actually generating the synthetic profiles is then straightforward. Fol-
lowing the probability distribution in place, a Python program would randomly generate all
possible profiles by putting together the randomly selected pieces of text for each compo-
nent that is drawn to appear in a given profile. Second, we randomly draw from this pool
the total number of profiles needed to generate the surveys that would be sent out to the
potential respondents. Because our survey was sent to 1,600 GPs, a total of 32,000 profiles
were created. Finally, the research team at Zero2IPO and a large team of research assistants
from the University of Chicago and Tsinghua University manually went over each and every
profile to make small manual changes needed to ensure perfect readability of each profile.24
An example of a synthetic LP profile (with government ties) shown to GPs is the following:
The investment institution has a total registered capital of RMB 1 billion, was
established at the beginning of 2007, and is located in Guangdong to promote
stronger domestic enterprises in the Greater Bay area. It is an investment
organization established by a state-owned firm funded by the provincial gov-
ernment. It mainly focuses on investment, financing, and asset management.
This is necessary to ensure that the profiles look realistic. As explained to us by Zero2IPO, the structure
of these profiles resembles that of warm introductions between GPs and LPs that would be made via email,
for example. This necessary choice involves a trade-off, as we are unable to provide a perfect monetary
quantification of preferences over each component, a point we return to in Section 6.
23For similar reasons, in their seminal study on labor market discrimination, Bertrand and Mullainathan
(2004) avoid constructing CVs that would make the candidates overqualified or that would include unusual
combinations of components that might make respondents suspicious.
24Notice that the order in which components are shown is typically fixed to best reflect the profiles in
Zero2IPO. With reference to the components described in Table 6, the order of appearance is: Registered
Capital, Founding Year, Location of HQ, Government Ties, Investment Philosophy, Industry, Stage Focus,
Fund Size and Management, Corporate Governance.
The investments target late stage projects which can facilitate the IPO of inno-
vative companies. The total size of the funds it provided capital to reached 700
Million yuan, with 15 RMB funds in total. The capital went to 20 startups,
8 of which are now listed companies.
4.3. Rating Profiles of Investment Partners. We measure GPs’ interest in LPs by

asking the GPs to rate 20 synthetic LP profiles. We use a 10-point Likert scale to measure the
rating, which allows us to observe GPs’ preferences towards characteristics of inframarginal
LP profiles. The respondents are instructed that the responses to both questions would be
used to generate their LP matches. Our main dependent variable is captured by the following
question:
1 “Are you interested in establishing an investment relationship with this investment
partner?”
We measure the response on a scale of 1 to 10, where 1=“Not interested” and 10=“Ex-
tremely interested.” We indicate the answers to these questions as Partner Rating, and they
represent our main dependent variable to capture how interested a GP is in a given LP
profile. We also specify: “Assume that the investment partner is already interested in es-
tablishing an investment relationship with your organization—therefore please only consider
your views on the quality of the investment partner.” Importantly, the additional emphasis
on assuming that the LP is interested allows us to separate the GPs’ interest from their
beliefs about the likelihood that the LP would want to provide capital to them. This was a
key aspect of the study that Zero2IPO emphasized to the respondents.
We then ask an additional question whose primary purpose is to further encourage GPs
to focus only on their interest in establishing an investment partnership with the given LP
when answering the main question. On its own, this additional question allows us to also
explore GPs’ beliefs about the likelihood that an LP would want to provide investment
capital to the GP if given the chance. The question asks the following:
2 “How likely do you think it is that this investment partner would want to enter an
investment relationship with your organization?”
We measure the response on a scale of 1 to 10, where 1=“Not likely” and 10=“Extremely
likely”. We also specify: “Assume that you have already expressed interest in the investment
partner—therefore please only consider whether you think the partner is interested in estab-
lishing an investment relationship with your organization.” We indicate the answers to these
questions as Expected Interest, and we report results for this measure in the Appendix.
We also measure whether the GP is interested in meeting an LP with the given synthetic
profile with a simple additional question: “Would you like to be introduced to this investment
partner?” The binary answer to this question has intuitive appeal and is akin to what the
resume audit literature typically captures in hiring settings (Bertrand and Mullainathan,
2004), but a concern is that it conflates GP interest in an LP with the GP’s expectation
that the LP would be interested in establishing an investment relationship if they had the
chance (Kessler et al., 2019). We report results for this measure in the Appendix as well.
5. Results
This section describes our main empirical results. We begin with Section 5.1 by outlining
the econometric specifications used to analyze our survey experiment. In Section 5.2, we
report the main results on the GPs’ preferences for LP characteristics, and specifically for LPs
with government ties. We then discuss mechanisms in Section 5.3, by showing heterogeneous
effects for government-owned GPs, additional heterogeneities, and new qualitative surveys
asking GPs about the pros and cons of receiving capital from LPs with government ties.
Finally, in Section 5.4, we analyze the results of our experimental surveys of LPs’ preferences
for GPs.
5.1. Estimating Equations. We estimate specifications of the following form:

N
(5.1) yij = αi + β × GovernmentTiesj + γm × Characteristicjm + ϵij ,
X
m=1
where i indicates the GP who is responding to the survey, and j indicates the synthetic
LP profile that is evaluated. y is one of our main dependent variables described in Section
4.3, such as Partner Rating. The main parameter of interest is β, which measures the average
effect of rating an LP that is connected to the government. The parameters γm capture all
other characteristics that we randomized in the synthetic LP profiles, as discussed in Section
4.2. We report results both with and without αi , which are the GP fixed effects that account
for different average ratings across respondents.
The set of other characteristics included in the regression is discussed next together with
the analysis of the results, while Table 5 summarizes the main variables that we create from
the synthetic profiles. All regressors are indicator variables equal to 1 or 0, depending on
the piece of text included in the synthetic profile, as indicated in Table 5 and Table 6.25
5.2. GPs’ Preferences for LPs. We report our main experimental results in Table 7.
In particular, we show regression results where the dependent variable is Partner Rating,
which measures the GP interest in LP profiles on a scale of 1–10. The coefficients in the
top row show that, on average, GPs dislike LPs with Government Ties. The coefficient is
-0.114 on the Likert scale, which indicates that the average respondent GP is willing to give
25If the profile component we use to construct our variables of interest does not appear in the profile, the
variable takes value 0.
up nearly $70 million in potential investment from the given LP.26 The negative coefficient
on Government Ties is significant both in our specification without (column 1) and with
(column 2) GP fixed effects. This is a key result we return to in extensive detail in the next
subsection to further discuss economic mechanisms.
Other LP characteristics are also valued positively. GPs are attracted to deep-pocketed
LPs, as indicated by the positive coefficients on Large Investor—which captures LPs that
have allocated at least 1 billion yuan to VCPE—and High Registered Capital—which captures
LPs with at least 1 billion yuan in registered capital. These results are intuitive as, all
else equal, GPs are unsurprisingly attracted to LPs that could generate larger influxes of
capital to their funds. We also find that GPs have a preference for LPs with Headquarter
In Beijing. On the other hand, we observe a dislike for LPs depicted to have a focus on
specific industries (Industry Information) or stages of investments (Stage Focus). These
latter findings are consistent with the average GP in the VCPE market in China having a
wide spectrum with regards to its investment focus. More broadly, the findings on preferences
with respect to these standard characteristics of the LPs seem to be largely uncontroversial,
which is reassuring to the extent that we can interpret them as a signal that GPs are indeed
evaluating the synthetic profiles according to their true preferences.
We also find that several other components of the LP profiles do not seem to affect GP
preferences. We do not observe a statistically significant differential preference for Young LPs
established after 2010, for LPs with Headquarter in Foreign Country, or for profiles displaying
information about the Investment Philosophy or the Corporate Governance practices of the
LP.27
As described earlier, our surveys also include a separate question that captures the
likelihood that the (synthetic) LP would want to provide investment capital to the GP if
given the chance. While this is included primarily to ensure that our measure of partner
26To compute the dollar values of the Likert coefficient we rely on the variable Large Investor, whose coeffi-
cient is 0.147, which has a more quantitative interpretation. This latter coefficient implies that the value of
Government Ties can be approximated by 0.147 0.114
times the value obtained from a Large Investor. To construct
the dollar value of the Large Investor coefficient we first divide the total assets invested by the large LPs
(using the average value that respondents see in the synthetic profiles of Table 6 when Large Investor=1)
by the average number of funds large LPs contributed to according to the 2015–2019 administrative data.
We then substract from this amount the analogous number we compute for the complementary group of
small LPs. Notice that the cutoff we use to split large and small LPs both in the synthetic profiles and the
administrative data is RMB 1 billion. We finally multiply the difference by 0.1140.147 to obtain a value of $70
million for the Government Ties coefficient.
27
The latter regressors are the outcomes of several discussions with Zero2IPO and primarily aim to make the
profiles look realistic, based on typical descriptions of potential investment partners that GPs see, e.g., on
Zero2IPO’s platform. Despite having been randomized independently of each other, they are at times similar
in nature. For instance, a piece of text for Investment Philosophy would be “As a long-term investor, the
investment philosophy is to achieve market return while controlling for risk.” Similarly, Corporate Governance
is equal to one if the profile includes, for instance, “The goal is to achieve the highest possible returns at
acceptable levels of risk, so as to generate strong returns in the long-term.”
rating is not confounded with concerns that the LP would be interested in the GP in the first
place, it is also of interest on its own. We explore what influences GPs’ expected likelihood
that a given LP would provide capital to them in Appendix Table A10. We find that
GPs report LPs with government ties to be less likely to provide them investment capital,
albeit the coefficient becomes statistically marginally insignificant when GP fixed effects are
included.
Robustness. As our main specifications are ordinary least squares (OLS) regressions, we are
implicitly making a linearity assumption regarding the 10-point Likert scale ratings. In
Appendix Table A11, we show that our results are robust to relaxing this assumption by
running ordered probit regressions, which only require that GPs, on average, value a higher
rating more highly than a lower rating. Appendix Table A12 reports the analysis using
as dependent variable the 0–1 indicator for Cooperation Interest, namely the answer to the
question “Would you like to be introduced to this investment partner?” as discussed in
Section 4.3. Appendix Table A13 reports the main analysis clustering the standard errors
at the respondent level.
5.3. Why Do GPs Dislike Government LPs? The results in Table 7 show that, on
average, GPs dislike LPs with government ties, suggesting that in our context the negatives
of receiving capital that is tied to the government outweigh the positives. In particular,
our results indicate that typical political connections considerations, which would make
government investors attractive, are not strong enough to outweigh the cons of dealing with
government LPs. As discussed earlier, a leading explanation for our findings is one in which
investors linked to the government might interfere in the investment decisions of GPs due
to political motives, which is seen as unattractive by GPs, considering that they are profit-
seeking entities interested in maximizing financial returns.
Importantly—by design—our findings are obtained after controlling for a number of
factors that might confound the above interpretation. For instance, real government-related
LPs are different along many dimensions compared to private LPs, such as size and preference
for certain regions and industries. Without controlling for these differences, our estimates
might be suggestive of both a dislike for government interference in investment decisions, for
example, or a general dislike for other characteristics of the investor that are correlated with
the investor having government ties. For instance, a dislike for government investors might
simply be driven by a general dislike for certain industries or regions that are not considered
attractive investment opportunities. Since both industry and regions of focus are randomized
across LP profiles, these concerns are largely muted in our setting. Moreover, notice that
our findings are unlikely to be explained by a differential expectation that government LPs
would actually invest in the GP. Indeed as discussed in Section 4.1, the instructions of the
experiment, which are explained in detail by Zero2IPO also via phone calls, make clear that
the respondent should assume that the LP would provide funding to them if they expressed
interest.
Below, we dig deeper into the potential economic mechanisms at play in three additional
ways. First, we report an analysis that shows how the effects vary depending on whether the
GP is also government-owned. Second, we show additional descriptive heterogeneities across
government types and sectors. Third, we discuss the findings from additional qualitative
surveys we conducted on a sample of our respondents that allow us to shed light on aspects
of this setting that are impossible to measure with the experiment or administrative data
alone.
5.3.1. Government-Owned versus Private GPs. To further investigate mechanisms, we start

by studying the heterogeneity of our main results along a key margin, namely whether
the respondent GP is also government-owned or not. If the dislike for government-related
investors is due to the distortions the government introduces after providing investment
capital, we should see stronger (i.e., more negative) effects for GPs that have no existing link
to the government and that operate according to market principles. On the other hand, we
expect the incentives of government-owned GPs to be more aligned with those of government
investors, which should result in a more favorable view of government LPs as investment
partners. These views are vastly confirmed by anecdotal evidence from both government and
private sources, as summarized by Luong et al. (2021) among others. Government ownership
of GPs, as for many other private sector entities, is pervasive in China, as we illustrated in
3.3. Importantly, however, all GPs in our sample, independent of their ownership structure,
are profit-driven, as discussed earlier.
We report the analysis for the sample of government-owned GPs versus private GPs in
Table 8, where we focus on our main dependent variable, Partner Rating. We find that the
negative coefficient on the indicator for the LP having government ties can be fully accounted
for by private GPs. Instead, we find that government ties of the LP do not matter for the
preferences of government-owned GPs. Interestingly, we find that no other component of
the LP profiles displays a meaningful difference depending on whether the GP is owned by
the government or not.
A caveat of this analysis is that while all components of the LP profiles are randomized
and all GPs are incentivized in an identical way, it is plausible that government-owned GPs
are more likely to focus on regions or industries that are a better match with the focus of
government-related LPs. To account for this, we report in Appendix Table A15 a version of
Table 8 where we also control for whether the GP has a region and/or industry of focus that
matches that of the given synthetic LP profile under evaluation.28 We find that our main
results remain strong, thus indicating that independently of whether the LP’s investment
28
Appendix Table A16 reports instead a version of our main table which includes these additional controls.
focus aligns with that of the GP, the GP prefers to receive funding from LPs that do not
have government ties.
A further possible explanation is that government-owned versus privately owned GPs
have prior differential exposure to government LPs. If this were the case, the differential
effects we observe might be driven by a differential expectation regarding the costs and
benefits of having the government as an investor. We therefore report our analysis also
controlling for whether the respondent GP ever had a government LP as an investor in the
last three years. As shown in Appendix Table A17, we find that our results are mostly
unchanged. Similarly, as reported in Appendix Table A18, we find that GPs with prior
experience working with a government LP do not have significantly different preferences
compared to other GPs.
We then conduct a heterogeneity analysis where, in addition to studying how the effects
vary depending on the ownership structure of the GPs, we further augment the analysis
using data on whether GPs are high- or low-performing firms. To do so, we rely on data
on GP performance introduced in Section 3.1.1, which allow us to observe a measure of
returns Zero2IPO uses to create their annual rankings, that is the comprehensive returns
(CR) for a subset of the respondent GPs. Using these data, we categorize respondents into
High Quality or Low Quality, depending on whether they have above or below median CR
in the sample. We then report, in Table A19, the results for a specification analogous to
equation 5.1, where we interact all possible splits by government ownership and performance
of the GP with our main regressor of interest, Government Ties. All estimates of these
heterogeneities are therefore relative to the preference of private low-performing GPs for
government LPs. Interestingly, we find that the strongest dislike for government LPs is
driven by high-performing private GPs.
Overall, the evidence in Table 8 seems consistent with a “grabbing hand” view of the
government according to which—all else equal—government investors introduce distortions
in the investment process which are particularly unattractive to high-performing private
firms.
5.3.2. Heterogeneity across Layers of Governments and Sectors. As discussed earlier, a lim-
itation of our approach is that we cannot observe preferences for all possible government
entities. In China, the pros and cons of being connected to the government can vary strongly
depending on whether the connection is to the central, provincial, or local government. In
particular, local government connections are often especially important for the growth of
early stage firms typically targeted by VCPE investors. In the Appendix, Table A20, we
explore whether the dislike for government investors is especially pronounced for certain
types of government entities.29 When estimating a specification analogous to equation 5.1,

but where the main regressor is split into different indicators for each level of government,
we find that the dislike is strongest for investors related to the central government, and it
is also present when focusing only on provincial government ties. On the other hand, we do
not find evidence of a dislike for government investors linked to local governments.30
Another important margin of heterogeneity for which having a government investor
might be particularly important is the focus of a GP’s investments. Indeed, if government
connections were important to open doors, they should be particularly so in state-dominated
sectors, as also discussed in Bai et al. (2020). We therefore explore whether GPs focused on
specific sectors might have a stronger preference for investors with government ties relative to
other GPs. To do so, we first categorize GPs into their specific sector of focus, by picking the
sector in which at least 50% of their 2015–2019 investments were made.31 We subsequently
estimate a nonparametric causal forest model to measure heterogeneous treatment effects
following the methodology of Wager and Athey (2018) and Athey and Wager (2019).32 We
report the Conditional Average Treatment Effects (CATE) for the various sectors in Figure
4 together with 95% confidence intervals. Despite the noise in the estimation, we observe a
pattern suggestive of a lower dislike for sectors where the government plays a more domi-
nant role, such as Construction and Real Estate, Manufacturing, Mining, and Finance and
Insurance, relative to sectors with a smaller government role such as Cleantech and Health,
among others.33
29
To do so, instead of using just a single dummy variable, we assign specific pieces of text related to Govern-
ment Ties in Table 6 to create a dummy for Central (option 5), Provincial (options 6-9), or Local (options
10-11) ties to to the government.
30
In Appendix Table A21, we further report the differential dislike of government-owned versus private GPs
for investors with ties to central, provincial, or local levels of the government. We find the strongest difference
occurring in preference for investors with ties to provincial governments.
31
As a result, for this specific test, we drop sector-agnostic GPs to which we cannot assign a specific sector
of focus, and are left with a sample of 236 respondent GPs. We use the coarsest categorization of sectors in
the Zero2IPO administrative data, which was also used in the creation of Figures 1 and 2. Some sectors do
not enter our analysis if the sample of respondent GPs listing that sector as their primary investment area
is too small.
32
In practice, in our setting we take the Government Ties dummy as the “treatment” and Partner Rating
as the “outcome.” In addition to the sector dummies of interest, we control for all regressors listed in Table
5, as well as whether the GP is government-owned or not, whether it is focused on the same region and/or
same industry as the synthetic LP profile, and whether the respondent GP had ever received capital from
the government in the past.
33
As discussed by Bai et al. (2020), there are several formal documents pointing explicitly to the importance
of specific sectors such as Construction and Real Estate, Manufacturing, Mining, and Finance and Insurance
for the state. See, for example, State-owned economy should maintain absolute control over seven industries
(China’s SASAC, 2006), Notice of the General Office of the State Council on Printing and Distributing
the Special Management Measures for the Market Entry of Foreign Investment in Pilot Free Trade Zones
(China’s State Council, 2015), and China names key industries for strong state control (China Daily, Dec
2006).
5.3.3. Surveying GPs on Pros and Cons of Investors with Government Ties. Our analysis
so far points to an explanation according to which the government introduces frictions in
the investment process of GPs, therefore making government capital unattractive. Yet,
while difficult to pin down experimentally, we conducted an additional round of surveys of
our respondents to provide additional more granular evidence. These surveys, which are
not experimental but rather qualitative in nature, were conducted in the last quarter of
2021. These new surveys were pitched as a research study to understand the advantages
and disadvantages introduced by government participation as an LP. The surveys were not
incentivized, except for the promise of a general summary of the results. We were able to
reach a total of 361 GPs, which are a subset of the respondents to our main 2019 survey.34
We take several steps to ensure that responses reflect the accurate, unbiased beliefs of
the respondents regarding the role of government in the capital allocation process. First,
all responses were promised to be used only for research purposes and anonymized, and all
questions were framed by detaching the respondent from the questions. That is, following
the literature on measuring sensitive issues such as corruption (Sequeira, 2012), we ask
respondents to state not what they think, but rather what they think are the main advantages
and disadvantages of having government-related entities as LPs from the perspective of typical
GPs in the market. Second, even though our interest is to primarily identify the reasons why
the government might not be an attractive LP to GPs, we attempt to alleviate the issue that
respondents might be wary of speaking negatively about the government. To do so, we do
not use explicitly negative language in the introductory messages, and we ask respondents
to first state the “advantages” that government LPs can bring. Only afterwards we ask
for what “improvements” might make the government a better investment partner. The
survey defines government-related LPs government entities or SOEs, and those sponsoring
a government-guided fund. We report the full recruitment script sent to respondents and
translated to English in Figure 5.
Our survey frames the pros and cons of government investors based on the anecdotal
evidence discussed in Section 2 alongside several discussions with Zero2IPO’s expert team.
A few key findings emerge from our new survey, as illustrated in Figure 6. First, as shown
in Panel A, we find that GPs rank post-investment interference in the investment process
as the main negative of receiving capital from government LPs. To a lesser extent, GPs
also list the presence of increased policy uncertainty and the lack of professionalization of
teams working for LPs tied to the government as unattractive features of government LPs.
On the other hand, the GPs are less concerned about differential requirements in terms
of project risk or investment horizon with government LPs. Second, as shown in Panel B,
34
We analyze the attrition between the original survey and the new qualitative survey in Appendix Table
A22. We observe a limited extent of selection bias, with those who responded to both surveys having made
more investments on average.
when analyzing what are considered the main advantages of receiving government capital, we
observe that GPs find the ability to obtain more favorable local government support to be the
most attractive feature of having government-related entities as investors. All together, our
qualitative surveys add color to our experimental analysis, by illustrating specific frictions
that may account for the peculiar pros and cons associated to the role government LPs in
the market.
5.4. LPs’ Preferences for GPs. We conduct a contemporaneous experimental survey of

LPs to study LP preferences for GP characteristics. This additional survey allows us to
study both sides of the market, a unique feature of our experimental setting which we return
to when discussing the equilibrium impact of government participation in the next section.
The survey, recruitment, and incentive structure are analogous to the survey of GPs. We
were able to reach a total of 312 LPs. We report the details of the variables used in the
analysis and the randomized components of synthetic GP profiles in Appendix Tables A23
and A24, respectively.
The analysis follows the same structure as the previous analysis of GP preferences. The
results are presented in Table 9. Some of the key findings are that LPs prefer high-performing,
foreign, recently established GPs that have a specialized focus in specific industries. What
stands out, however, is that the strongest determinant of LP interest in a GP is whether
that GP already has entities with government ties among its investors. We also find that
LPs value positively GPs whose team members have direct experience in the government,
while industry experience does not matter.35
Unlike the GP-level analysis, we find little heterogeneity depending on the ownership
structure of the LP itself, as illustrated in Appendix Table A28, even though private LPs
have a slightly stronger preference for high-performing GPs.
6. Equilibrium Impact of Government Participation
Our experimental surveys reveal substantial heterogeneity in preferences for government

participation from both the firm and investor sides of China’s VCPE market. Given the two-
sided nature of the market, the allocation of government capital must be co-determined in
equilibrium by both the ability of government LPs to find GPs and the preferences and
demand for capital on the GP side. To better understand and quantify the equilibrium and
distributional consequences of government participation, in Section 6.1 we build a simple
two-sided search and matching model of government and nongovernment GPs and LPs.
We calibrate the model using both our experimental surveys and the administrative data
35
Appendix Table A25 shows the analysis with Expected Interest as dependent variable. Appendix Table
A26 shows robustness to an ordered probit specification, while Appendix Table A27 reports the analysis
clustering the standard errors at the respondent level.
in Section 6.2. We conduct policy counterfactuals on changing the nature and extent of
government participation in Section 6.3.
6.1. Model Setup. We model the formation of GP-LP partnerships as a two-sided search
and matching process in continuous time. There are discrete I types of GPs and J types
of LPs looking for potential partners. While the matching between GPs and LPs can be
many-to-many—each GP can be funded by multiple LPs, and each LP can invest in many
GPs—we abstract away from the intensive margin of investment size and instead assume
each GP has a discrete number of investment slots to be fulfilled with funding. Each slot can
be fulfilled by a partner LP, and each LP has the capacity to invest in potentially multiple
slots. The matching process between GPs and LPs at the investment slot-to-capacity level
is one-to-one. If GP of type i ∈ I and LP of type j ∈ J jointly decide to form a partnership,
then the GP obtains value xij + ϵ and the LP obtains value yij + δ, where xij and yij
are type-specific values from the partnership and ϵ, δ are idiosyncratic values that capture
heterogeneity within GP and LP types i and j.
Meeting a potential partner takes time and is thus costly; we let r denote the discount
rate. Let {ni }Ii=1 and {mj }Jj=1 respectively denote the distribution of GP and LP types
waiting to be matched in the market, with Ii=1 ni = Jj=1 mj = 1. Meetings arrive following
P P
a Poisson process. The types of GP and LP at each meeting are independently drawn, with
probability ni mj a meeting takes place between GP of type i and LP of type j. Both parties
then decide whether to form a partnership—the LP decides whether to invest in the GP and
the GP decides whether to accept the investment. A partnership is formed if and only if
both parties prefer the match over rejecting the counterparty. If either prefers to wait for
another match, both parties go back to the market.
Let ui and vj denote the value functions of umacthed GPs and LPs, respectively. The
value functions are characterized by the following Hamilton-Jacobi-Bellman (HJB) equations:
 
J  
(6.1) rui = ρG mj qij E [max (ui + ϵ0 , xij + ϵ) − ui ] .
X

j=1
 
I  
(6.2) rvj = ρL ni pij E [max (vj + δ0 , yij + δ) − vj ] .,
X
 
i=1
where ρG and ρL are the Poisson rates at which a GP and an LP meet a potential partner,
respectively. We allow these meeting rates to differ (ρG ̸= ρL ), reflecting the fact that the
total supply of investment capacity may differ from the available number of investment slots.
To interpret the HJB equations, consider equation (6.1). r is the opportunity cost of
waiting, and rui is thus the flow value of a GP waiting to be matched. With Poisson rate ρG ,
the GP gets to meet an LP of type j randomly drawn from distribution {mj }. Upon meeting,
both parties learn about each other and then decide whether to form a partnership—LP
decides whether to invest in the GP and the GP decides whether to accept the investment.
From the GP’s perspective, its continuation value is xij + ϵ if forming the partnership, and
is ui + ϵ0 if it continues to search, where ϵ0 denotes the change in continuation value despite
rejecting the potential partner; ϵ0 could reflect the information the GP gathers from the
meeting, potentially about its own investment prospects or about the market more broadly.
A partnership is formed when both parties prefer the match over rejecting the counter-
party. Conversely, both parties have to continue the search if either party decides against
forming a partnership. In equation (6.1), the term qij captures the probability that the LP of
type j prefers the match. In that case, the GP’s continuation value is max (ui + ϵ0 , xij + ϵ)
and the expected change in value is thus E [max (ui + ϵ0 , xij + ϵ) − ui ]. Otherwise, if the LP
rejects the GP, the GP’s continuation value is ui + ϵ0 as it has no choice but to continue the
search, with the expected change in value being zero.
Whether to form a partnership or continue to search is the only decision that each party
gets to make. The probabilities of preferring to match (pij and qij ) follow
(6.3) pij = E [ui + ϵ0 ≥ xij + ϵ] , qij = E [vj + δ0 ≥ yij + δ] .
We normalize the ex-ante expected value of ϵ0 to be zero. The expected change in the
GP’s continuation value, conditioning on matching a random LP of type j, is therefore
qij E [max (ui + ϵ0 , xij + ϵ) − ui ]. The right-hand side of equation (6.1) calculates the un-
conditional expected change in value by integrating the conditional change in value over the
distribution of LPs and then multiplying by the Poisson rate of matching. The HJB equation
(6.2) for LP has a similar interpretation.
We take as model primitives the type-specific values from partnerships (xij and yij ),
the rate at which meetings occur ({ρG , ρL }), the discount rate r, and the distribution of
unmatched types (ni , mj ). That is, we study a stationary equilibrium where a constant
stream of new GPs and LPs enter the search market to replace those that leave after having
found a partner, such that the total mass and distribution of participant types are time-
invariant. Given the model primitives, the probabilities of preferring to match (pij and qij )
follow (6.3), and the value of unmatched entities (ui and vj ) are the fixed point solutions to
the HJB equations (6.1) and (6.2) and are therefore endogenous outcomes of the matching
equilibrium. We later consider counterfactual changes to the model primitives as we conduct
policy experiments.
We impose the standard assumption in the discrete choice context that the idiosyncratic
values (ϵ’s and δ’s) are drawn from type-I extreme value distributions, implying
eui
(6.4) pij = , E [max (ui + ϵ0 , xij + ϵ)] = ln (eui + exij ) ,
eui + exij
evj
qij = , E [max (vj + δ0 , yij + δ)] = ln (evi + eyij ) .
e +e
vj yij
6.2. Calibration. We now describe how we leverage both our experimental surveys and the
administrative data to recover the model primitives and conduct policy counterfactuals. Mo-
tivated by our reduced form evidence, we categorize GPs into I = 4 four types, according to
their government ownership ∈ {gov, non-gov} and quality ∈ {high, low}. In our calibration,
we cut the sample along the quality dimension by the median as measured by comprehensive
returns. We categorize LPs into J = 2 types according to government ownership only.
We exploit the two main questions we ask respondents as part of the experimental
survey, namely [1] “Are you interested in establishing an investment relationship with this
investment partner?” and [2] “How likely do you think it is that this investment partner
would want to enter an investment relationship with your organization?”. We interpret the
answers to question [1], from both GPs and LPs, as informative of xij and yij , respectively.
For question [2], we assume each response provides a noisy signal to the probability of
being preferred to match. Specifically, for each GP respondent ν of type i rating a synthetic
LP profile of type j, we assume ν’s answer to question [2] is a noisy monotone transformation
to the expected probability of counterparty’s cooperation interest qij . We parametrize the
monotone transformation using log-likelihood ratio to ensure that the underlying probability
lies between zero and one:
qij
(6.5) ansGP2 (v|i, j) = α + β ln +ξ
1 − qij
where ξ are i.i.d. mean-zero errors, and α and β are parameters to be calibrated. By
collapsing the survey responses to each type-pairs, we purge the i.i.d. errors and obtain the
average GP type i’s assessment of LP type j’s interest to cooperate:
qij
anw
¯ GP
2 (i, j) = α + β ln .
1 − qij
We assume LP respondents’ answers are symmetrically informative of pij .36
From the administrative data, we observe the distribution of existing matches between
GPs and LPs across each type-pair, {µij }, as reported in Table A6. We now argue {µij }
is informative of the distribution of unmatched GPs and LPs, {ni } and {mj }. Specifically,
consider a meeting between a GP and an LP. The likelihood that the meeting involves GP
type i and LP type j is ni mj ; the likelihood of the meeting turning into a partnership is
ni mj pij qij . In a stationary environment, the distribution of existing matches must satisfy:
µij ni mj pij qij
(6.6) = .
µi′ ,j ′ ni′ mj ′ pi′ j ′ qi′ j ′
36LP preferences for GPs by ownership and quality types (i, j) are reported in Table A29.
That is, upon a meeting taking place, if the pair ij has a greater likelihood of being drawn
(higher ni mj ) and forming a partnership (higher pij qij ) than the alternative pair i′ j ′ , the
former pair must have a proportionally bigger presence in the existing matches. Equation
(6.6) implies that, given the preference probabilities {pij , qij } and the observed distribution
of existing matches {µij }, one can recover the distribution of GP and LP types that are still
waiting to find a partner, exploiting the fact that the distributions integrate to one:
µij (pij qij )−1 µij (pij qij )−1
(6.7) ni = PI −1 , mj = PJ −1 .
i′ =1 µi′ j (pi′ j qi′ j ) j ′ =1 µij ′ (pij ′ qij ′ )
We now describe the calibration strategy. We will calibrate parameters {α, β} as well as
ρ /r and ρG /r.37 α and β translate survey response to counterparty’s cooperation interests
G
{pij , qij }, which enables us to recover the value of unmatched entities {ui , vj } using equations
(6.4) and, along with the observed distribution of matches µij in the administrative data, also
recover the distribution {ni , mj } of unmatched types in the search market using equation
(6.7). Since the value of unmatched entities {ui , vj } must satisfy the HJB equations (6.1)
and (6.2), which provide I + J = 6 (4 GP types and 2 LP types) moment conditions. We
thus have an over-identified system with six moments and effectively four parameters, which
calibrate to minimize the sum of square errors in the moment conditions.
Based on our calibration, a typical entity is willing to cooperate with about one in
five potential partners it encounters. GPs on average have 30% shorter waiting time than
LPs (ρG /ρL ≈ 1.29), implying that there are more LPs funding investments than GPs with
investment opportunities. Given an annual cost of funds at r = 20%, our estimates imply
that an average GP (LP) meets 34 (26) potential partners a year.
As a validation exercise, Table A8 Panel A shows the model-implied distribution of
meetings (ni ×mj ) between unmatched GPs and LPs by their types, recovered from equations
(6.7). The assortative matching along government ownership types, shown as motivating
evidence in Table 4, is also reflected in the probability to form partnerships conditioning on
meetings, as shown in Panel B of Table A8. Controlling for quality, a government LP is about
50% more likely to form a partnership with a government GP than with a nongovernment
GP. Conversely, a nongovernment LP is bewteen 65% (if the GP’s IRR is above-median)
and 110% (if the GP’s IRR is below-median) more likely to form a partnership with a
nongovernment GP than a government one.
6.3. Results and Counterfactuals. We now exploit the calibrated model to perform coun-
terfactuals in order to assess the equilibrium impact of government participation in China’s
VCPE market.
37Note the Poisson meeting rates are not separately identified from the discount rate r, as the HJB equations
continue to hold if ρL , ρG , and r are multiplied by the same factor. Nevertheless, by identifying the meeting
rates relative to discount rate, we can already perform the policy counterfactuals in Section 6.3.
In the first set of counterfactuals, reported in Panel A of Table 10, we consider changing
the composition of government versus nongovernment investors in the economy. In column
(1), we assume a 10-percentage-point increase in the share of LPs that are government-owned.
Because of assortative matching, government GPs—especially low-quality ones—experience
surplus gains (increase in ui ). In contrast, nongovernment GPs—especially high-quality
ones—experience surplus losses. We report the magnitudes of the surplus changes in terms
of Likert points, consistent with our survey design, in Table 10. Extrapolating the coefficients
in Table 7, we find that the impact of this policy experiment is quantitatively significant. For
example, the effect on government GPs’ surplus is equivalent to raising the capital allocated
by their LP investors by $26 and $43 million, respectively, for high- and low-quality GPs.
On the LP side, the surplus of unmatched government LPs declines by 0.012 Likert points
(equivalent to a one-percentage-point decline in internal rates of return of all potential GP
partners). On the other hand, because of their now-higher relative scarcity, the surplus of
unmatched nongovernment LPs increases, albeit only marginally. In column (2), we consider
a more extreme version of changing government participation, by computing the equilibrium
impact of replacing all government investors with nongovernment ones. This large shock
reduces the surplus of government GPs by over 0.25 points on the Likert scale (equivalent
to more than $144 million decline in the capital allocated by their investors) and raises the
surplus of nongovernment GPs by an analogous amount. Importantly, the distributional
effects of replacing all government investors with private ones are such that they are the
most negative for low-quality GPs, while they are the most positive for high-quality GPs,
consistent with our reduced-form evidence.
One common narrative is that the government misallocates funds by favoring underper-
forming politically connected firms. However, a nuanced view of our experimental results is
that to the extent that high-performing, privately owned GPs have a dislike for government
capital, any empirical observation about government investors possibly misallocating funds
(e.g., as shown in the correlations of Table 4) might suggest, at least in part, their inability
to attract the best firms rather than poor decision-making due to corruption, favoritism, or
incompetence (Murphy, Shleifer, and Vishny, 1993; Shleifer, 1998; Lerner, 2009; Colonnelli,
Prem, and Teso, 2020). We therefore next consider a set of policy experiments that channel
government capital exclusively to nongovernment or well-performing GPs, reported in Panel
B of Table 10. First, we consider the policy restriction that government LPs must invest
in nongovernment GPs. How does this mandate affect the quality of investments that are
ultimately funded, and what are the distributional effects on market participants? To inves-
tigate this, we hold constant all other model primitives, and we specify that when a meeting
between a GP and an LP takes place, and if both parties are drawn to be government-
owned, then the meeting dissolves immediately and a new pair of GP and LP are drawn
to meet. Column (1) in Panel B of Table 10 shows that government GPs and LPs experi-
ence very large declines in their equilibrium surplus, whereas nongovernment-owned entities
experience more moderate increases. This is intuitive, as the policy experiment effectively
lowers the rate at which government-owned entities meet any potential partner while raising
the rate at which nongovernment entities meet potential partners. Altogether, the average
surplus of entities on both sides experiences substantial declines. In terms of magnitudes,
unmatched low-quality, government-owned GPs experience the largest decline in surplus of
-1.6 Likert points. Extrapolating the coefficients in Table 7, this is equivalent to reducing
the capital allocated by their investors by nearly $1 billion. Under the policy mandate,
government LPs invest in better-performing GPs with significantly higher (6.8 percentage
points) returns. However, the overall effect on returns of funded GPs is small, namely a 0.4-
percentage-point increase. This is because even though government LPs direct investments
away from government GPs, these GPs can and do substitute towards nongovernment LPs
to fulfill their funding needs. Our model shows that the mandate reduces the average returns
of GPs funded by nongovernment LPs by 6.3 percentage points. We then consider a similar
experiment in column (2) of Table 10 Panel B, where we study the policy that enforces
government LPs to invest only in those GPs with above-median returns. Unsurprisingly,
the policy drastically raises the performance of GPs receiving investment from government
LPs (by 17.2 percentage points). The net effect on the average returns of all funded GPs
is however lower (an increase of 3.7 percentage points), due to substitution by low quality
GPs towards investments from nongovernment LPs. Despite the increase in average returns,
market participants’ equilibrium value again decreases on average for entities on both the
LP and GP side of the market, consistent with the large surplus losses of government-owned
entities. Overall, these results indicate that the empirical regularity that government LPs
tend to invest in low-performing GPs does not necessarily reflect only capital misallocation.
Instead, it might be at least in part driven by the preferences of the top-performing GPs for
private capital, which makes it challenging for government investors to match with the best
firms in the first place.
7. Conclusion
In China, as well as in many other, typically developing economies around the world,
the government plays a key role as an investor in and owner of private sector firms. In
light of this fact—which we establish using rich administrative data within the context of
the second-largest market for investment in high-growth firms and entrepreneurs, namely
that of venture capital and private equity (VCPE) in China—two opposing models of state-
firm relationships have naturally different implications for our understanding of the growth
path of these economies. We show that a “grabbing hand” model where the state aims to
keep control over the private sector fits our context better relative to a model where the
government acts as a “helping hand” that allows firms to overcome other first-order frictions
and grow faster.
Our main contribution to the literature consists of the design of a non-deceptive field
experiment to estimate the demand for government participation. In collaboration with
the leading industry organization, we conduct 1,000 experimental surveys of both sides of
the market: the capital investors and the private firms that manage the invested capital
by deploying it to high-growth entrepreneurs. The experimental design, which is inspired
by studies of discrimination in the labor market, allows us to overcome typical empirical
difficulties, which in our context are that we observe only equilibrium matching outcomes
and that government investors differ from other investors along a multitude of dimensions.
We document that the average firm dislikes investors with government ties, that such dislike
is not present for government-owned firms, and that it is highest for the best-performing
firms and lowest for firms operating in state-dominated industries. Consistent with the
experimental evidence, we also conduct new qualitative surveys which directly point to
political interference in decision-making as a leading mechanism why government capital
is unattractive to private firms. We conclude the paper by quantifying the distributional
implications of government participation using an equilibrium model of matching between
government and nongovernment firms and investors.
Our study has several implications. On the one hand, by providing direct evidence
of the private sector perspective of the advantages and, in particular, the disadvantages of
government investors, we help advance the recent debate aimed at understanding the nature
of China’s model of economic growth grounded on the dominance of state economic actors
(Bai et al., 2020). On the other hand, our paper makes the simple point that the demand
for government capital differs across different types of firms. As a result, to the extent that
how capital is allocated depends on the agents receiving it, understanding the demand side is
important to fully capture the efficiency implications of government participation. We believe
this is an aspect of the debate that has been largely neglected but that is crucial for both
theory and policy. That is, analyzing the efficiency outcomes and potential misallocation
consequences of government participation requires understanding the demand for what the
government offers. Such an implication is natural in the context of government as an investor,
like the one we study, and in contexts where we aim to estimate the impact of government
programs, such as financial assistance to businesses, among many others. More broadly,
there are several contexts, such as that of public procurement or foreign direct investments,
where the differential selection of firms willing to engage with the state in the first place has
direct economic consequences.
Our paper also naturally has limitations that future research should build on. First, our
experiment only focuses on a specific market largely characterized by sophisticated investors,
and on a context, that of China, that is certainly unique. There are reasons to believe several
of the pros and cons that typically accompany government investments are prevalent in the
broader debate about how governments around the world should foster entrepreneurship and
innovation, and whether governments are well-equipped to do so in the first place (Bai et al.,
2021), but establishing external validity to other contexts should be an important next step.
Similarly, our survey was conducted in 2019, and to the extent that state-firm relationships
evolve dynamically, it would be valuable to measure changes in such relationships over time.
Second, in the interest of realism, our design favors simplicity to the detriment of a perfect
quantification of magnitudes. Third, our study is silent regarding the broader efficiency goals
of the government. For example, the state might want to control private sector firms as a
way to channel resources to regions and industries where the social value of investments, such
as poverty reduction, might be higher. These are first order issues that should be studied in
future work, and for which we hope our study can have important lessons for.
References
Allen, F., J. Cai, X. Gu, J. Qian, L. Zhao, and W. Zhu (2021): “Centralization or Decentralization?
The Evolution of State-Ownership in China,” working paper.
Aminadav, G. and E. Papaioannou (2020): “Corporate control around the world,” The Journal of
Finance, 75, 1191–1246.
Amstad, M., G. Sun, and W. Xiong (2020): The Handbook of China’s Financial System, Princeton
University Press.
Andonov, A., Y. V. Hochberg, and J. D. Rauh (2018): “Political representation and governance:
Evidence from the investment decisions of public pension funds,” The Journal of Finance, 73, 2041–2086.
Athey, S. and S. Wager (2019): “Estimating treatment effects with causal forests: An application,”
Observational Studies, 5, 37–51.
Babina, T., A. X. He, S. T. Howell, E. R. Perlman, and J. Staudt (2020): “The color of money:
Federal vs. industry funding of university research,” Tech. rep., National Bureau of Economic Research.
Bai, C.-E., C.-T. Hsieh, Z. M. Song, and X. Wang (2020): “Special Deals from Special Investors: The
Rise of State-Connected Private Owners in China,” Tech. rep., National Bureau of Economic Research.
Bai, C.-E., J. Lu, and Z. Tao (2006): “The multitask theory of state enterprise reform: Empirical
evidence from China,” American Economic Review, 96, 353–357.
Bai, J., S. Bernstein, A. Dev, and J. Lerner (2021): “Public Entrepreneurial Finance around the
Globe,” Tech. rep., National Bureau of Economic Research.
Beraja, M., D. Y. Yang, and N. Yuchtman (2020): “Data-intensive innovation and the State: evidence
from AI firms in China,” Tech. rep., National Bureau of Economic Research.
Bernstein, S., A. Korteweg, and K. Laws (2017): “Attracting early-stage investors: Evidence from a
randomized field experiment,” The Journal of Finance, 72, 509–538.
Bernstein, S., J. Lerner, and F. Mezzanotti (2019): “Private equity and financial fragility during the
crisis,” The Review of Financial Studies, 32, 1309–1373.
Bernstein, S., J. Lerner, and A. Schoar (2013): “The investment strategies of sovereign wealth funds,”
Journal of Economic Perspectives, 27, 219–38.
Bertrand, M. and S. Mullainathan (2004): “Are Emily and Greg more employable than Lakisha and
Jamal? A field experiment on labor market discrimination,” American economic review, 94, 991–1013.
Bortolotti, B. and M. Faccio (2009): “Government control of privatized firms,” The Review of Financial
Studies, 22, 2907–2939.
Brander, J. A., Q. Du, and T. Hellmann (2015): “The effects of government-sponsored venture capital:
international evidence,” Review of Finance, 19, 571–618.
Brunnermeier, M. K., M. Sockin, and W. Xiong (2020): “China’s model of managing the financial
system,” Tech. rep., National Bureau of Economic Research.
Cole, S., M. Melecky, F. Mölders, and T. Reed (2020): “Long-run Returns to Impact Investing in
Emerging Markets and Developing Economies,” Tech. rep., National Bureau of Economic Research.
Colonnelli, E. and M. Prem (2022): “Corruption and firms,” The Review of Economic Studies, 89,
695–732.
Colonnelli, E., M. Prem, and E. Teso (2020): “Patronage and selection in public sector organizations,”
American Economic Review, 110, 3071–99.
Cong, L. W., C. M. Lee, Y. Qu, T. Shen, et al. (2020): “Financing entrepreneurship and innovation
in China,” Foundations and Trends® in Entrepreneurship, 16, 1–64.
Cumming, D. J., L. Grilli, and S. Murtinu (2017): “Governmental and independent venture capital
investments in Europe: A firm-level performance analysis,” Journal of corporate finance, 42, 439–459.
Da Rin, M., T. Hellmann, and M. Puri (2013): “A survey of venture capital research,” in Handbook of
the Economics of Finance, Elsevier, vol. 2, 573–648.
Da Rin, M. and L. Phalippou (2017): “The importance of size in private equity: Evidence from a survey
of limited partners,” Journal of Financial Intermediation, 31, 64–76.
Denes, M. R., S. T. Howell, F. Mezzanotti, X. Wang, and T. Xu (2020): “Investor tax credits and
entrepreneurship: Evidence from us states,” Tech. rep., National Bureau of Economic Research.
Dinç, I. S. (2005): “Politicians and banks: Political influences on government-owned banks in emerging
markets,” Journal of financial economics, 77, 453–479.
Ewens, M., A. Gorbenko, and A. Korteweg (2022): “Venture capital contracts,” Journal of Financial
Economics, 143, 131–158.
Faccio, M., R. W. Masulis, and J. J. McConnell (2006): “Political connections and corporate
bailouts,” The journal of Finance, 61, 2597–2635.
Fang, L., J. Lerner, C. Wu, and Q. Zhang (2018): “Corruption, government subsidies, and innovation:
Evidence from China,” Tech. rep., National Bureau of Economic Research.
Fei, C. Y. (2018): “Can Governments Foster the Development of Venture Capital?” Available at SSRN
3221997.
Fisman, R. (2001): “Estimating the Value of Political Connections,” The American Economic Review, 91,
1095–1102.
Fisman, R. and M. A. Golden (2017): Corruption: What everyone needs to know, Oxford University
Press.
Giglio, S., M. Maggiori, J. Stroebel, and S. Utkus (2021): “Five facts about beliefs and portfolios,”
American Economic Review, 111, 1481–1522.
Gompers, P., S. N. Kaplan, and V. Mukharlyamov (2016): “What do private equity firms say they
do?” Journal of Financial Economics, 121, 449–476.
Gompers, P. A., W. Gornall, S. N. Kaplan, and I. A. Strebulaev (2020): “How do venture
capitalists make decisions?” Journal of Financial Economics, 135, 169–190.
Gornall, W. and I. A. Strebulaev (2020): “Gender, race, and entrepreneurship: A randomized field
experiment on venture capitalists and angels,” Available at SSRN 3301982.
Graham, J. R. and C. R. Harvey (2001): “The theory and practice of corporate finance: Evidence from
the field,” Journal of financial economics, 60, 187–243.
Harrison, G. W. and J. A. List (2004): “Field experiments,” Journal of Economic literature, 42, 1009–
1055.
Hochberg, Y. V., A. Ljungqvist, and Y. Lu (2007): “Whom you know matters: Venture capital
networks and investment performance,” The Journal of Finance, 62, 251–301.
Howell, S. T. (2017): “Financing innovation: Evidence from R&D grants,” American Economic Review,
107, 1136–64.
Hsieh, C.-T. and Z. M. Song (2015): “Grasp the Large, Let Go of the Small: The Transformation of the
State Sector in China,” Brookings Papers on Economic Activity.
Huang, Z., X. Tian, M. Amstad, G. Sun, and W. Xiong (2020): “15. China’s Venture Capital Market,”
in The Handbook of China’s Financial System, Princeton University Press, 383–418.
Jeffers, J., T. Lyu, and K. Posenau (2021): “The Risk and Return of Impact Investing Funds,” .
Jia, R., X. Lan, and G. P. i Miquel (2021): “Doing business in China: Parental background and
government intervention determine who owns busines,” Journal of Development Economics, 151, 102670.
Kaplan, S. N., F. Martel, and P. Strömberg (2007): “How do legal differences and experience affect
financial contracts?” Journal of financial intermediation, 16, 273–311.
Kessler, J. B., C. Low, and C. D. Sullivan (2019): “Incentivized resume rating: Eliciting employer
preferences without deception,” American Economic Review, 109, 3713–44.
Khwaja, A. I. and A. Mian (2005): “Do lenders favor politically connected firms? Rent provision in an
emerging financial market,” The Quarterly Journal of Economics, 120, 1371–1411.
King, R. G. and R. Levine (1993): “Finance and growth: Schumpeter might be right,” The quarterly
journal of economics, 108, 717–737.
La Porta, R. and F. Lopez-de Silanes (1999): “The benefits of privatization: Evidence from Mexico,”
The quarterly journal of economics, 114, 1193–1242.
La Porta, R., F. Lopez-de Silanes, and A. Shleifer (1999): “Corporate ownership around the world,”
The journal of finance, 54, 471–517.
——— (2002): “Government ownership of banks,” The Journal of Finance, 57, 265–301.
Lerner, J. (2000): “The government as venture capitalist: the long-run impact of the SBIR program,” The
Journal of Private Equity, 3, 55–78.
——— (2009): Boulevard of broken dreams, Princeton University Press.
Lerner, J., A. Leamon, and F. Hardymon (2012): Venture capital, private equity, and the financing of
entrepreneurship: The power of active investing, Wiley Hoboken, NJ.
Lerner, J., J. Mao, A. Schoar, and N. R. Zhang (2022): “Investing outside the box: Evidence from
alternative vehicles in private equity,” Journal of Financial Economics, 143, 359–380.
Lerner, J. and A. Schoar (2005): “Does legal enforcement affect financial transactions? The contractual
channel in private equity,” The Quarterly Journal of Economics, 120, 223–246.
Lerner, J., A. Schoar, S. Sokolinski, and K. Wilson (2018): “The globalization of angel investments:
Evidence across countries,” Journal of Financial Economics, 127, 1–20.
Levine, R. (1999): “Law, finance, and economic growth,” Journal of financial Intermediation, 8, 8–35.
——— (2002): “Bank-based or market-based financial systems: which is better?” Journal of financial
intermediation, 11, 398–428.
Liu, E. (2019): “Industrial Policies and Economic Development,” Quarterly Journal of Economics.
Low, C. (2021): “Pricing the Biological Clock: The Marriage Market Costs of Aging to Women,” .
Luong, N., Z. Arnold, and B. Murphy (2021): “Understanding Chinese Government Guidance Funds,”
Center for Security and Emerging Technology, March.
Malkin, A. (2021): “China’s Experience in Building a Venture Capital Sector: Four Lessons for Policy
Makers,” working paper.
Megginson, W. L. and J. M. Netter (2001): “From state to market: A survey of empirical studies on
privatization,” Journal of economic literature, 39, 321–389.
Murphy, K. M., A. Shleifer, and R. W. Vishny (1993): “Why is rent-seeking so costly to growth?”
The American Economic Review, 83, 409–414.
Musgrave, R. A. (1959): The theory of public finance; a study in public economy, Kogakusha Co.
Phalippou, L. (2008): “The hazards of using IRR to measure performance: The case of private equity,”
Available at SSRN 1111796.
Rajan, R. and L. Zingales (1998): “Financial Dependence and Growth, American Economic Review,” .
Sapienza, P. (2004): “The effects of government ownership on bank lending,” Journal of financial econom-
ics, 72, 357–384.
Sequeira, S. (2012): “Advances in measuring corruption in the field,” New advances in experimental
research on corruption.
Shleifer, A. (1998): “State versus private ownership,” Journal of economic perspectives, 12, 133–150.
Shleifer, A. and R. W. Vishny (1993): “Corruption,” The Quarterly Journal of Economics, 108, 599–
617.
——— (1994): “Politicians and firms,” The Quarterly Journal of Economics, 995–1025.
——— (1998): The grabbing hand: Government pathologies and their cures, Harvard University Press.
Song, Z. M., K. Storesletten, and F. Zilibotti (2012): “Growing Like China,” American Economic
Review.
Sørensen, M. (2007): “How smart is smart money? A two-sided matching model of venture capital,” The
Journal of Finance, 62, 2725–2762.
Stiglitz, J. E. (1989): “Markets, market failures, and development,” The American economic review, 79,
197–203.
Wager, S. and S. Athey (2018): “Estimation and inference of heterogeneous treatment effects using
random forests,” Journal of the American Statistical Association, 113, 1228–1242.
Wurgler, J. (2000): “Financial markets and the allocation of capital,” Journal of financial economics, 58,
187–214.
Xiong, W. (2018): “The mandarin model of growth,” Tech. rep., National Bureau of Economic Research.
Young, A. (2000): “The razor’s edge: Distortions and incremental reform in the People’s Republic of
China,” The Quarterly Journal of Economics, 115, 1091–1135.
Zhang, Y. (2020): “Discrimination in the Venture Capital Industry: Evidence from Two Randomized
Controlled Trials,” arXiv preprint arXiv:2010.16084.
A LP: Headquarters B LP: Investment Region C LP: Investment Industry

de
ai il Tra
ngh
a i ngh Reta
Sha Sha ale
and
ous
ing
oles areh
Wh d W
on a
n ices
n rtati Serv
egio spo
ion nd R Tran Oth
er
Reg Inla
Industry Group
Region Group
Region Group
nd g
Inla ufac
turin
ong a n
ngd M
olog
y
n g Gua n Te
chn
n gdo rma
ti o
Hea
lt h
Gua Info
tries e
oun ra n c
nC Insu
ig and
Fore nce eral
Re gion Fina and
Min
stal egio
n
Ene
rgy
state
Coa s tal R ea l E
dR
Coa ctio
n a n
ante
ch
stru Cle
ing ing Con
Beij Beij ecre
atio
n
d R
an ing
ent, Fish
rt ainm and
0 10 20 30 0 10 20 30 , En
te str y, 0 10 20 30
Arts Fore
Share (%) Share (%) ure, Share (%)
cult
Agri
D GP: Headquarters E GP: Investment Region F GP: Investment Industry

e
i Trad
i ha Reta
il
gha Sh ang and
S han ale usin
g
h oles W a reho
W d
n an ices
n atio Serv
egio sport
nd R er
n Tra n
Oth
egio Inla
Industry Group
nd R
Region Group
Region Group
Inla ring
factu
ong Ma n u
ngd logy
g Gua Tec
hno
on tion
ngd rma Hea
lth
Gua Info
tries e
oun Insu
ranc
ignC and l
ion Fore Fina
n c e
in e ra
eg n dM
tal R n a
rgy
Coa
s egio Ene Esta
te
s tal R R eal
Coa
d
n an ante
ch
ctio Cle
stru
ing Con
Beij Be ijing reati
on
Rec
n t, and h in g
e is
ainm nd F
0 10 20 0 10 20 30 tert y, a 0 10 20 30
, En restr
Arts , Fo
Share (%) Share (%) cu lture Share (%)
Agri
Figure 1. Distribution of Headquarters Location, Investment Region, and

Investment Industry
Notes: This figure reports the distribution of headquarters location, investment region, and investment industry for
the sample of active LPs and GPs. We have 7,974 active LPs and 6,308 active GPs. We exclude foreign entities from
this analysis. Panels A and D show the distribution of headquarters for LPs and GPs, respectively. Panels B and
E show the proportion of investment in each region group for LPs and GPs, respectively. In the Region Group of
Panels A, D, B and E, we map all regions into 6 categories for visualization, Beijing, Shanghai, Guangdong, Inland
Region, Coastal Region and Foreign Countries, in which Coastal Region indicates that the area belongs to a province
adjacent to the sea, while Inland Region is the opposite. Panels C and F show the proportion of investment in each
industry group for LPs and GPs, respectively.
A LP: Headquarters B LP: Investment Region C LP: Investment Industry

de
ai il Tra
ngh
a i ngh Reta
Sha Sha ale
and
ous
ing
oles areh
Wh d W
on a
n ices
n rtati Serv
egio spo
ion nd R Tran Oth
er
Reg Inla
Industry Group
Region Group
Region Group
nd g
Inla Government Government ufac
turin Government
ong a n
ngd M
olog
y
n g No Gua No n Te
chn No
n gdo rma
ti o
Hea
lt h
Gua Info
Yes n tries Yes n c e Yes
Cou Insu
ra
ign and
eral
gion Fore Fina
nce Min
Re and
stal gion Ene
rgy
state
Coa s tal R
e
ea l E
dR
Coa ctio
n a n
ante
ch
g stru Cle
in ing Con
Beij Beij ecre
atio
n
d R
an ing
ent, Fish
rt ainm and
0 10 20 30 0 10 20 30 , En
te str y, 0 10 20 30
Arts Fore
Share (%) Share (%) ure, Share (%)
cult
Agri
D GP: Headquarters E GP: Investment Region F GP: Investment Industry

e
i Trad
ha Reta
il
Shanghai
Sh ang and g
ale usin
h oles W a reho
W d
n an ices
n atio Serv
egio sport
nd R er
Tra n
Inland Region Oth
Inla
Industry Group
Region Group
Region Group
ring
Government Government factu Government
ong Ma n u
ngd logy
Guangdong No Gua No Tec
hno
No
tion lth
rma Hea
Info
Yes tries Yes e Yes
oun Insu
ranc
ignC and l
Coastal Region Fore Fina
n c e
dM
in e ra
a n
n rgy
egio Ene Esta
te
s tal R R eal
Coa
d
n an ante
ch
ctio Cle
Beijing stru
Con
Be ijing reati
on
Rec
n t, and h in g
e is
ainm nd F
0 10 20 30 0 10 20 30 tert y, a 0 10 20 30
, En restr
Arts , Fo
Share (%) Share (%) cu lture Share (%)
Agri
Figure 2. Distribution of Headquarters Location, Investment Region, and

Investment Industry (by Government Ownership)
Notes: This figure reports the distribution of headquarters location, investment region, and investment industry
for the sample of active LPs and GPs, split by government-owned versus nongovernment-owned entities. We have
3,969 government-owned active LPs and 4,005 nongovernment-owned active LPs. We have 1,812 government-owned
active GPs and 4,496 nongovernment-owned active GPs. We exclude foreign entities from this analysis. Panels A
and D show the distribution of headquarters for LPs and GPs, respectively. Panels B and E show the proportion of
investment in each region group for LPs and GPs, respectively. In the Region Group of Panels A, D, B and E, we
map all regions into 6 categories for visualization, Beijing, Shanghai, Guangdong, Inland Region, Coastal Region and
Foreign Countries, in which Coastal Region indicates that the area belongs to a province adjacent to the sea, while
Inland Region is the opposite. Panels C and F show the proportion of investment in each industry group for LPs and
GPs, respectively.
2019 Chinese Equity Investment Survey
Zero2IPO and Tsinghua University PBC School of Finance are studying how to improve the resource
allocation in China's private equity investment market more effectively, establish an efficient and reliable
market-based investment system, and better promote technological innovation. The purpose of the survey
is to use machine learning technology to introduce general partners (GP) and limited partners (LP), and to
help GP and LP form a more effective match by identifying important characteristics of different
institutions. We sincerely hope that we could receive strong support and assistance from your
organization. Please take the time to fill out the survey questionnaire accurately.
We hope you could evaluate the profiles of hypothetical investment partners. Your choices will be used to
provide you with recommendations of and make introductions with actual partners you may be interested
in that closely match your preferences. In the survey questionnaire, you will see descriptions of 20
hypothetical partners. Please evaluate each profile based on the following questions:
0）Would you like to meet this investment partner?
1）Are you interested in establishing an investment relationship with this investment partner?
(On a scale of 1-10, 1=“Not interested”; 10=“Extremely interested”)
2) How likely do you think it is that this investment partner would want to enter an
investment relationship with your organization? (On a scale of 1-10, 1=“Not likely”; 10=“Extremely
likely”)
Question 1) seeks to measure your interest in this partner. Assume that the investment partner is
already interested in establishing an investment relationship with your organization—therefore please
only consider your views on the quality of the investment partner.
Question 2) seeks to measure the likelihood that this partner wants to establish a business
relationship with your organization. Assume that you have already expressed interest in the investment
partner—therefore please only consider whether you think the partner is interested in establishing an
investment relationship with your organization.
* All the data you fill in will be kept strictly confidential, and we will also send you anonymous
summary research and related policy reports.
In order to thank your institution for participating, we will provide you with:
1） An introduction between the (real) general partner (GP) and the (real) limited partner (LP) to
form more effective matches;
2） An early research report from this survey.
Figure 3. 2019 Experimental Survey: Recruitment Email

Notes: This figure shows the recruitment email sent to respondents by Zero2IPO for the 2019 survey. Respondents
would read this page before they start the surveys and Zero2IPO would guide them with phone calls and in case they
have any questions during the whole process.
08 Construction and Real Estate | |
07 Manufacturing | |
06 Mining | |
05 Finance and Insurance | |
04 Information Technology | |
03 Arts, Entertainment, and Recreation | |
02 Health | |
01 Cleantech | |
−1.0 −0.5 0.0 0.5

CATE
Figure 4. GP Dislike for LPs with Government Ties: Heterogeneity by In-

vestment Sector
Notes: This figure shows the heterogeneity of GP preferences for LP government ties by GP’s industry focus, using
the causal forest machine learning model suggested by Wager and Athey (2018) and Athey and Wager (2019). 95%
confidence intervals are reported. We define the “industry focus” of a GP as the industry that accounts for more
than half of the GP’s total investment deals. Five industries (Agriculture, Forestry, and Fishing, Other, Services,
Transportation and Warehousing, and Wholesale and Retail Trade) were dropped due to small sample size. The
conditional average treatment effects are estimated on a sample of 236 GP respondents.
2021 China Equity Investment Market Research Survey
About this survey

Zero2IPO Research Center and PBC School of Finance of Tsinghua University are jointly studying how to more effectively improve the allocation of resources in China's venture capital (VC) and private equity (PE) market, so as to establish an efficient and reliable market-based
investment system that can promote technological innovation. Your institution has previously strongly supported and participated in the "2019 China Equity Investment Survey". After rigorous machine-lerning analysis, we have helped GPs and LPs form effective matches with each other.
A sizable share of investment in the Chinese VC and PE market comes from the government or from enterprises with state-owned equity, which have the purpose of supporting entrepreneurship and technological innovation, especially among young and small to medium sized firms. We
would like you to respond to the questions below, based on the general perceptions from the perspective of typical GPs in the market, about government-related LPs (such as government agencies or state-owned firms, or government entities investing in guided funds) and evaluate (1) the
advantages of receiving funding from government-related LPs, and (2) how to improve the efficiency in the investment of government-related funding.
* After completion, we will summarize the research, and write policy reports and proposals that can inform relevant regulatory authorities to improve the system. All the information you fill in will be kept strictly confidential, and we will also send you anonymous summaries of the
research and related policy reports. We sincerely hope that we can continue to receive strong support and assistance from your organization. Please take the time to fill out the survey questionnaire and send it back within the next two weeks.
1: The advantages of government-related LPs (10=extremely important, 1=not important at all) Please mark the most important advantage among the 5 options below.
1 To speed up regulatory approvals and obtain tax reductions Please choose: a value between 1-10
INVESTING WITH THE GOVERNMENT

To obtain larger shares of returns from the government, receive timely funding when facing shortages of private funds in the market, reduce the pressure of fundraising, and obtain
2 Please choose: a value between 1-10
follow-up funds more easily
3 To obtain faster access to reliable information/relevant future policies/industry resources Please choose: a value between 1-10
4 Government LPs can obtain support from the local government and bring local investment opportunities Please choose: a value between 1-10
5 Please choose: a value between 1-10

To help attract potential investors and follow-up investment from private capital
Other, please specify: Please provide comments or suggestions:
2: What can be improved by government-related LPs (10=extremely important, 1=not important at all) Please mark the most important one among the 5 options below.
1 Need less post-investment restrictions on usage of funds in specific regions and industry and on the ratio of investment from private LPs Please choose: a value between 1-10
2 Need more tolerance of investment risks, and more focus on profit maximization with high-return/high-quality/competitive projects Please choose: a value between 1-10
3 Need to extend the investment horizon and the requirements on when to exit Please choose: a value between 1-10
4 Need a more professional team and a more professional approach to make investment decisions so that value can be added post-investment Please choose: a value between 1-10
5 Need to reduce exposure to policy uncertainty and have more clear investment objectives Please choose: a value between 1-10
Other, please specify: Please provide comments or suggestions:
Figure 5. 2021 Qualitative Survey

Notes: This figure shows the recruitment email sent to respondents by Zero2IPO for the 2021 survey. Respondents would read this page before they start the
surveys and Zero2IPO would guide them with phone calls and in case they have any questions during the whole process.
45
A: Main Disadvantages
ty
r tain
nce
c yU
Poli
su re to
xpo
5. E
eam
al T
s sion
fP rofe
a ck o
4. L
izon
Hor
ent
stm
Inve
h or t
3. S
e
ranc
Tole
o Risk
2. N
e
enc
rfer
Inte
ent
estm
1 . Inv
0 10 20
Share (%)
B: Main Advantages
tor s
ves
al In
tenti
t Po
ttrac
5. A
pp or t
t Su
men
vern
a l Go
Loc
btain
4. O
n
atio
form
ss to In
cce
3. A
ss ure
Pre
ing
drais
ce Fun
edu
2. R
ns
ctio
R edu
Tax
and
als
p prov
ry A 0 5 10 15 20 25
lato
egu
1. R Share (%)
Figure 6. Survey on Pros and Cons of Government Investors

Notes: This figure shows the distribution of responses from the 2021 surveys, and specifically the shares of each
option marked as the most important reason by the respondent. Panel A shows the main advantages of government
LPs. Panel B shows the main disadvantages of government LPs.
Table 1. Summary Statistics
Active Respondents
All Gov NonGov All Gov NonGov
Panel A: LPs
Share Government-Owned (%) 50.11 100.00 0.00 77.52 100.00 0.00
Capital Invested ($ millions) 50.36 98.95 16.18 399.59 471.71 207.33
Funds Invested 1.98 2.53 1.43 9.24 10.18 4.45
Panel B: GPs
Share Government-Owned (%) 38.63 100.00 0.00 32.05 100.00 0.00
AUM ($ millions) 741.30 993.02 607.21 1001.76 1491.76 691.78
IRR (% median) 27.64 23.48 31.16 32.34 25.78 36.57
Funds 2.54 2.77 2.38 3.32 4.22 2.81
Investments 13.42 11.72 14.47 48.40 44.36 50.35
Exits 5.91 6.82 5.37 9.36 11.86 8.06
Notes: This table reports summary statistics for both LPs and GPs, using Zero2IPO administrative data for the
period 2015–19. We have 7,974 active LPs of which 312 LPs are respondents, and 6,308 active GPs of which 688 GPs
are respondents. We exclude foreign entities from this analysis. The Panel A includes variables for LPs. The Panel
B includes variables for GPs. Share Government-Owned (%) is the share of entities that have at least one ultimate
owner that is affiliated either with a government agency or a state-owned enterprise, Capital Invested ($ millions)
is the amount of capital the LP invested in funds (in Million USD), Funds Invested is the number of funds the LP
invested in; AUM ($ millions) is the assets under management (in Million USD), IRR (% median) is the median
internal rate of return, Funds is the number of funds managed by the GP, Investments is the number of investments
made by the GP, Exits is the number of exit events for the GP investments. Capital Invested ($ millions), AUM ($
millions) and IRR (% median) are winsorized at the top 95%.
Table 2. Government Ownership of Investors and Fund Managers

Active Respondents
Min p10 p25 Median Mean p75 p90 Max Min p10 p25 Median Mean p75 p90 Max
Panel A: LPs
Total Gov Share 0.00 0.00 0.00 0.00 29.76 76.01 100.00 100.00 0.00 0.00 1.66 63.96 55.98 100.00 100.00 100.00
Gov Share (within Gov Entities) 0.00 1.09 12.80 82.62 61.70 100.00 100.00 100.00 0.14 18.79 43.39 99.93 73.16 100.00 100.00 100.00
Central Gov Share 0.00 0.03 0.46 2.84 17.35 17.60 68.37 100.00 0.00 0.14 1.74 9.60 23.55 39.31 70.89 100.00
Provincial Gov Share 0.00 0.03 0.27 4.97 27.99 45.73 100.00 100.00 0.00 0.09 3.29 22.02 42.56 100.00 100.00 100.00
Local Gov Share 0.00 0.19 1.52 46.38 51.04 100.00 100.00 100.00 0.00 0.29 3.03 50.83 53.01 100.00 100.00 100.00
Panel B: GPs
Total Gov Share 0.00 0.00 0.00 0.00 13.87 1.86 69.27 100.00 0.00 0.00 0.00 0.00 11.56 2.17 48.04 100.00
Gov Share (within Gov Entities) 0.00 1.25 8.07 41.97 48.24 100.00 100.00 100.00 0.00 0.89 4.50 31.24 38.56 67.87 100.00 100.00
Central Gov Share 0.00 0.06 0.60 3.22 19.45 29.73 69.19 100.00 0.00 0.36 0.35 2.00 14.92 17.88 45.76 100.00
Provincial Gov Share 0.00 0.05 0.27 5.99 25.02 35.13 100.00 100.00 0.00 0.02 0.14 2.17 20.65 31.02 90.00 100.00
Local Gov Share 0.00 0.10 0.98 8.20 31.69 53.12 100.00 100.00 0.00 0.09 0.95 4.67 20.68 33.70 69.91 100.00
Notes: This table reports the summary statistics of government ownership for both LPs and GPs. We have 3,969
active government-owned LPs (out of 7,974 active LPs) of which 238 government-owned LPs are respondents (out of
312 LP respondents), and 1,812 active government-owned GPs (out of 6,308 active GPs) of which 216 government-
owned GPs are respondents (out of 688 GP respondents). We exclude foreign entities from this analysis. For this
analysis, we omit government-owned entities whose government ownership was identified but for which the precise
government ownership share value was missing. Total Gov Share is computed using all entities, with nongovernment-
owned entities having 0 government ownership share. Gov Share (within Gov Entities) is computed using only the
government-owned entities; Central Gov Share is computed using only the sample of entities with at least some
central government ownership; Provincial Gov Share is computed using only the sample of entities with at least some
provincial government ownership; Local Gov Share is computed using only the sample of entities with at least some
local government ownership. Government-owned entities are those with at least one ultimate government owner, as
described in the paper.
Table 3. Government-Owned GPs Perform Worse
(1) (2) (3) (4) (5) (6) (7) (8)

CR CR CR CR IRR IRR IRR IRR
Gov GPs -0.012∗∗∗ -0.006∗∗ -0.014∗∗∗ -0.008∗∗ -12.871∗∗∗ -10.529∗∗ -17.211∗∗∗ -15.112∗∗∗
(-3.74) (-2.21) (-3.25) (-2.22) (-3.13) (-2.51) (-3.52) (-3.00)
AUM 0.000∗ -0.000 -0.001 -0.002
(1.84) (-0.13) (-1.09) (-1.60)
Observations 1104 1104 683 683 984 984 631 631
HQ FEs No Yes No Yes No Yes No Yes
Notes: This table illustrates the association between GPs’ government ownership status and GP performance. The
specification is yj = αi + β × GovGPsj + γ × AUMj + ϵij .The sample includes all active GPs with non-missing data
for CR (columns 1-4) and IRR (columns 5-8). GovGPs is a dummy indicating whether a GP is government-owned.
CR is comprehensive return, which is standardized to 0-1. IRR is winsorized at the 95% percentile. AUM is the total
asset under management in USD millions, and is winsorized at the 95% percentile. Columns 1 and 5 show the basic
models. Columns 2 and 6 show the results with headquarters FEs. Columns 3 and 7 show the results with AUM as
controls. Columns 4 and 8 show the results with both headquarters FEs and AUM controls. t statistics are presented
in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Table 4. Assortative Matching Between Government-Owned GPs and LPs
Gov LP Non-Gov LP ColRatio

Gov GP 1.608 0.633 2.540
( 33.54 %) ( 13.46 %) ( 0.000)
Non-Gov GP 0.828 1.001 0.827
( 23.75 %) ( 29.25 %) ( 0.000)
RowRatio 1.941 0.632
( 0.000) ( 0.000)
Assortative Index 1.254
Homogeneity Test(p-value) 0.000
Notes: This table presents the distribution of links between different GPs and LPs grouped by government
ownership, illustrating assortative matching patterns. The likelihood ratio index is calculated as s(pGP , pLP ) =
P r(GGP =pGP ,GLP =pLP )
P r(GGP =pGP )P r(GLP =pLP )
We define P r(GGP = p) as the ratio of type p GP among all GPs with at least one link,
.
e.g., if p is government owned, then the probability is the ratio of government owned GPs among GPs with at least
one link. P r(GGP = GLP = p) is defined as the ratio of links where GP and LP both belong to group p among all
links in the sample. The number in the parentheses is the fraction of links among all links formed between GP and
LP with ownership information. Assortative index is calculated as the weighted average of the diagonal elements.
ColRatio is calculated as column 1 divided by column 2 in the same row. RowRatio is calculated as row 1 divided
by row 2 in the same column. The numbers in the parentheses under the ColRatios and RowRatios are the p-values
of the binomial test within the corresponding rows and columns respectively, under the null hypothesis of random
matching. The p-value of the homogeneity test is a Chi-square test. Government GP and government LP are defined
as entities that have at least one ultimate government owner, as described in the paper.
Table 5. Variables in Synthetic LP Profiles
Variables Description
Government Ties A dummy indicating whether the LP has ties
to the government.
Large Investor A dummy indicating whether the LP has size
above 1 billion yuan.
High Registered Capital A dummy indicating whether the registered
capital of the LP is > 1 billion yuan.
Industry Information A dummy indicating whether the LP profile
displays industry information.
Young LP A dummy indicating whether the LP is a
young LP (founded after 2010).
Headquarter in Foreign Country A dummy indicating whether the LP is
headquartered in a foreign country.
Headquarter in Beijing A dummy indicating whether the LP is lo-
cated in Beijing.
Corporate Governance A dummy indicating whether the LP pro-
file displays description of corporate gover-
nance.
Investment Philosophy A dummy indicating whether the LP profile
displays description of investment philoso-
phy.
Stage Focus A dummy indicating whether the LP profile
displays the targeted stage of investments.
Notes: This table illustrates the coding of regressors based on original profile components. The first column shows
the main regressors. The second column gives a brief description of the variables. See Table 6 for details on all profile
components.
Table 6. Description of LP Profiles Randomized Components

Variable Categorical Value Options
Government 1 2 3 4
Government Ties:
Ties This company aims to give full With the help of the private cap- This organization is one of the This organization channels capi-
1 if with government ties [5-11].
(0.8) play to the role of the market ital market and modern man- earliest market-oriented financ- tal to independent innovative en-
in allocating resources and ex- agement practices, this organiza- ing platforms in China. The terprises nationwide. It aims to
pand private capital investments tion channels capital to sectors of management team is commit- better promote technological in-
in innovation and entrepreneur- strategic importance and benefi- ted to increasing investors’ as- novation through its professional
ship, so as to promote the devel- cial to social development. set value, using modern man- business model and its efficient
opment of emerging industries. 公司依托资本市场，运用现代管理 agement methods to protect in- and reliable market-oriented in-
公司目标是充分发挥市场资源配置战略，吸引各类社会资金投资具有 vestors’ rights. vestment system.
和扩大社会资本投资创新创业的作战略意义和促进社会发展的领域。公司是最早的以市场为导向的投资机构面向全国自主创新企业，并为
用，以促进新兴产业发展。融资平台之一，管理者运用现代管其提供资金支持，通过专业的经
理方式维护投资者权力，致力于为营模式，高效可靠的市场化投资体
投资者资本保值增值。系，更好地推动科技创新。
5 6 7 8
It is a platform for the This organization is estab- It is an investment organiza- It is funded by the provin-
central government to hold lished by the CCP provincial tion established by a state- cial government. It aims to
shares of large enterprises committee and the provin- owned firm funded by the attract social capital invest-
and supervise SOEs, thus cial government. It accel- provincial government. It ment into enterprises.
supporting the restructuring erates economic transforma- mainly focuses on invest- 得到省政府财政资金的支持，目的
of SOEs and the adjustment tion mainly through sup- ment, financing, and asset 是吸引社会资本投资到企业。
of industrial structure. porting the development of management.
中央政府通过持股大型企业来直接 venture capitals and attract- 经省人民政府资助的国有独资公司
监督国有企业，为国有企业的改制 ing social capital into ven- 批准而设立的投资机构，主要从事
和产业结构调整搭建平台。 ture capitals. 投资，融资和资产管理。
经省委、省政府批准而设立的投资
机构，主要支持风险投资企业的发
展，引导社会资金进入风险投资领
域，加快转型。
9 10 11
The provincial government Its establishment and opera- Funded and managed by the
established this organization tion is approved by the local local government, it oper-
and guides its capital opera- government. Its main busi- ates in a market-oriented
tion, equity investment and nesses include the invest- manner.
asset management. ment, operation, and man- 由当地政府出资和运作，并以市场
由省政府批准成立设立，指导资本 agement of state-owned as- 化的方式进行运作。
运作、股权投资和资产管理。 sets.
经当地政府批准，主营业务包括国
有资产的投资，经营和管理。
Fund Size 1 2 3 4
Large Investor:
and The amount under management It established 20 RMB funds The total size of the funds it pro- A total of 21 funds were set up,
1 if fund size >1 billion [7-12].
Management is 200 Million yuan allocated to with a total size of 600 Million vided capital to reached 700 Mil- which led to about 650 Million
(0.8) a total of 12 funds, with in- yuan. lion yuan, with 15 RMB funds yuan of social funds.
vestments in more than 12 star- 管理20多支人民币基金，总资金管 in total. The capital went to 设立风险投资基金21支，带动社会
tups, including 5 of them that 理规模达到6亿元。 20 startups, 8 of which are now 资金约6.5亿元。
are listed in domestic and foreign listed companies.
capital markets. 管理总规模达到7亿元，设立15支
管理基金投资项目2亿多元，拥有基投资基金，已投资接近20家创业公
金12支，已投资12家创业公司，其司，通过股权投资平台，成功培育
中5家在国内外资本市场上市。了8家上市公司。
5 6 7 8
It established more than 25 As of end of 2018, it managed The assets under manage- As of December 2018, it
funds, with a total committed a capital of 800 Million yuan, ment are over 1.5 Billion established 8 direct invest-
capital of over 750 Million yuan with 23 completed investment yuan, with investments in 25 ment platforms, and had
and more than 20 accumulated projects, and 9 listed companies projects, and a total amount committed capital of 2.5 Bil-
investment projects. that have been fostered by the in- invested of 900 million yuan. lion yuan, with investments
已参与设立25多支基金，总管理资 vestment platform. 管理资产超过 15亿
亿元，完成投资项 in over 25 enterprises.
金超过7.5亿元，累计投资项目超截至2018年，已管理8亿元资金，完目 25个
个，总投资为9亿亿元。截至 2018年
年12月月，共有 8个个直接投
过20个。成投资项目23个，通过投资平台，资平台、管理资金达到 25亿
亿元，累
培育了9家上市公司。计投资企业超过 25家家。
9 10 11 12
By the end of March 2018, As of June 2018, it con- By the end of 2018, the com- The target scale of the fund
it invested in 15 funds, for a tributed to 22 funds for a to- pany invested in 30 funds, to invest in is 15 Billion
total of 2.5 Billion yuan, and tal size of 6 Billion yuan. including industrial invest- RMB, and in the past it in-
overall available assets of 4.5 截至 2018年
年6月
月，公司共设立 22支
支 ment funds and platform invested in 30 funds, and 45 in-
billion yuan. 基金，总资金规模达到60亿亿元。 vestment funds. novative small and medium-
截至 2018年
年3月月底，已投资 15支
支基截至 2018年
年底，公司管理了 sized enterprises, effectively
金，总规模达 25亿
亿，总资产为45亿亿支基金，包括产业投资基金和
近 30支 playing the exemplary role
元。平台投资基金。 of guiding the funds to pro-
mote innovation and en-
trepreneurship.
基金目标规模为 150亿
亿人民币，共
设立了 30家
家创业投资基金，已投
资 45家
家创新型中小企业，有效发挥
了推动创新创业的示范作用。
Registered 1 2 3 4
High Registered Capital:
Capital The institution has a registered The financing platform has ini- The investment institution has a The registered capital of the
1 if >1 Billion [5-9].
(1) capital of 100 million yuan, tial total assets of 500 million total registered capital of RMB 1 government-guided fund reaches
公司的注册资本为1亿元人民币， RMB, billion, RMB 1 billion,
融资平台的起始资本为5亿人民币，投资公司的注册资本是10亿元人民引导基金的注册资本是10亿元人民
币，币，
5 6 7 8
The guided fund has a reg- The investment institution The investment institution The guided fund has a regis-
istered capital of 3 billion has a registered capital of 3 has a registered capital of tered capital of RMB 5 bil-
yuan, billion yuan, RMB 5 billion, lion,
引导基金资本的注册为 30亿
亿元人民投资机构的注册资本为 30亿
亿元人民风险投资的注册资本为 50亿
亿元人民政府引导基金的注册资本为 50亿
亿元
币，币，币，人民币，
9
The government guided
fund, which provides strong
support to advanced in-
dustries, has a registered
capital of RMB 8 billion,
一家为促进先进产业发展而设立的
政府引导基金，注册资本为 80亿
亿元
人民币，
Table 6 (cont.): Description of LP Profiles Randomized Components

Industry: 1 2 3 4
Industry
1 if show industry information It focuses on the Internet in- In the past, the institution The institution targets in- The firm seeks to invest in
(0.5)
[1-21]. dustry and provides financ- has successfully funded sev- vestments in information Bio and Healthcare indus-
ing service for enterprises in eral investments in Social technology and related sec- tries and actively seeks eq-
the industry. Network and Media. tors such as Blockchain, uity investments or strategic
重点投向互联网行业，并为企业提已投资项目聚集在社交网络和媒体 Big Data, Artificial Intel- buyouts.
供金融服务。行业。 ligence, Robot, or Human 致力于投资生物和医疗保健领域，
Face Recognition. 并积极寻求股权投资和战略性收
重点投向区块链、大数据、人工智购。
能、机器人或人脸识别等信息技术
相关的高科技领域。
5 6 7 8
The primary industries of The investment scope in- The main direction of the Over the past years, the
past investments include cludes advanced manufac- company’s investments is in- investment focus has been
high-tech, high growth com- turing, modern agriculture, frastructure investment and been on new opportunities
panies in clean technology, and the maritime economy. the development of elec- in the wealth management
healthcare, and advanced 重点投资领域包括先进制造业，现 tric power, gas, water pro- industry.
manufacturing sectors. 代农业和海洋经济。 duction and supply, railway 一直以来，机构专注财富管理行业
优先投资领域为清洁技术、医疗保 transportation and other in- 的新发展环境和机遇。
健、先进制造业。 dustries.
主要投资方向是基础设施投资和电
力、天然气、水生产和供应、铁路
运输和其他行业。
9 10 11 12
Core businesses include ven- The institution prefers in- The investments currently The focus is on strategic
ture capital broadly, and vestments in fast-moving focus on education and emerging industries such as
sectors related to fund man- consumer products (Food training. biotech, internet, new en-
agement, assets manage- and Beverages) and the 公司目前的投资重点是教育和培 ergy, new materials, new
ment, project assessment, broader services industry. 训。 generation of information
and financial advisory in fi- 机构主要投资于快速消费品（食品 technology, cultural creativ-
nance. 和饮料）和服务业。 ity, energy conservation, and
公司核心业务包括风险投资、基金 environmental protection.
管理、资产管理、项目评估和金融专注于生物、互联网、新能源、
行业财务咨询等服务。新材料、新一代信息技术、文化创
意、节能环保等战略性新兴产业。
13 14 15 16
The institution focuses on The institution seeks op- The incubation and invest- The institution is equipped
investments in Aerospace re- portunities in information ment in the transforma- with specialized investment
lated industries, as well as technology, energy conser- tion of scientific and tech- teams that produced suc-
industries such as life and vation and environmental nological achievements in- cessful exits in various in-
health, ocean, military in- protection, new energy, cludes information technol- dustries, such as agriculture,
dustry, robots, wearable, new materials, biotechnol- ogy, life sciences and Biolog- chemical engineering, en-
and intelligent equipment. ogy, high-end equipment ical Medicine. ergy, pharmaceutics, health-
致力于对航空航天相关行业的投 manufacturing and other 投资科技成果转化的孵化，包括信 care, and information tech-
资，同时投向大健康、海洋，军事 national strategic emerging 息技术、生命科学和生物医学。 nology.
工业、机器人、可穿戴设备和智能 industries. 机构拥有专业的团队，在农业、化
设备等行业。在信息技术、节能环保、新能源、学工程、能源、制药、医疗保健、
新材料、生物技术、高端装备制造信息技术等行业创造大量成功案
等国家战略性新兴产业投资。例。
17 18 19 20
The fund pays important The investment areas are It regularly invests in To promote local high-tech
attention to intelligence- very extensive, and include satellite applications, in- industry, the institution fo-
sensitive services, advanced software and hardware com- formation technology, new cuses on new materials,
manufacturing, environment panies, production compa- materials and new energy, new equipment, new energy,
protection, and energy sav- nies and technology service aerospace special technolo- new communication tech-
ing industries. companies, including home gies, automation and special nologies, marine tech, en-
重点投向先进制造业、环保和节能 and business mobile commu- vehicles and other fields. ergy conservation and en-
产业。 nications. 一直以来投资于航空航天、卫星应 vironmental protection, and
投资领域非常广泛，包括软件和用、信息技术、新材料和新能源、 life and health.
硬件公司、生产公司和技术服务公航空航天特殊技术应用、汽车零部重点投向新材料、新设备、新能
司，涵盖家庭和企业移动通信。件和特种车辆等领域。源、新一代通信技术、海洋高科
技、节能环保、生命健康等领域，
促进当地高新技术产业发展
21
The portfolio covers a broad
spectrum of industries: fi-
nancial services, telecommu-
nications, media technology,
energy resources, and life
sciences.
投资项目涉及广泛的行业，金融服
务、电信、媒体技术、能源和生命
科学。
Founding 1 2 3 4
Young LP:
Year founded in June 2000, formally established in 2002, had more than 15 years of expe- established at the beginning of
1 if founded after 2010 [5-9].
(0.8) 2000年成立， 2002年成立， rience in assets management, 2007,
拥有超过15年的资金管理经验， 2007年初成立，
5 6 7 8
founded in December 2010, established in 2011, founded in 2012, established in 2015,
年12月
2010年月成立，年成立，
2011年年成立，
2012年年成立，
2015年
9
was recently established in
2016,
年新成立，
2016年

Headquarter in Foreign Country: 1 2 3 4
Location of
1 if headquarter in Foreign Country located in Jiangsu Province. set up in the Guizhou Province. headquartered in Shanghai. mainly invests in Shanghai and
HQ
[13,14]. 位于江苏省。设立于贵州省。总部位于上海市。 Yangtze River Delta.
(1)
Headquarter in Beijing: 投资主要覆盖上海和长三角地区。
1 if headquarter in Beijing [15,16]. 5 6 7 8
located in Guangdong to pro- investment headquartered in located in the Shenzhen-Hong set up 10 business centers in 8
mote the development of the Guangzhou. Kong Business Cooperation cities including Beijing, Shang-
Greater Bay area. 投资总部设在广州。 Zone. hai, Guangzhou, Shenzhen and
位于广东，致力于推动大湾区发地点位于深港业务合作区。 Chongqing.
展。在北京，上海，广州，深圳，重庆
等8个城市分别设立了10个业务中
心。
9 10 11 12
has 15 branches in 10 regions which invests all provinces and established in Fujian Province an influential investment insti-
across the whole China. cities across the country. as one of the most important in- tution in Shandong Province.
在中国内地10个地区设有15家分支投资项目覆盖全国各省市。 vestment platforms. 是目前山东省具有影响力的投资机
机构。是福建省重要的投资平台之一。构。
13 14 15 16
headquartered in the Silicon based in Singapore and located in Beijing. headquartered in Beijing, it
Valley. concentrated on Asia and 位于北京。 has offices in Europe and
总部位于硅谷。 growth markets. North America.
位于新加坡，主要关注亚洲和新兴总部位于北京，在欧洲和北美设有
市场。分支机构。
Investment Investment Philosophy: 1 2 3 4
Philosophy 1 if investment philosophy is Its investment philosophy To attract to the local area It aims to enhance in- It aims to enrich the struc-
(0.63) included [1-10]. is to promote new indus- high-quality venture capi- dependent innovation abil- ture of financial products
trialization through science tal firms, projects, technolo- ity through attracting ven- through technological and
and technological develop- gies, and talents, it focuses ture capital investment into management innovation,
ment. It also takes advan- on cultivating strategic and SMEs, especially science thus enlarging the space for
tage of the amplifying effect emerging industries. and technology SMEs, and economic development and
of financial leverage and en- 投资专注于培育战略和新兴产业， taking advantage of the am- social reform.
forces professional manage- 以吸引高质量的风险投资、项目、 plifying effect of financial 经营宗旨是：不断推进技术和管
ment. 技术和人才到当地。 leverage. 理创新，丰富金融产品结构，为经
投资理念是依靠科学技术的发展为了发挥资金的杠杆放大效应，济发展和社会改革提供更广阔的空
和高新技术的开发，推动新型工业吸引风险投资公司对中小企业的投间。
化，发挥杠杆放大效应，实施专业资，特别是科技型企业，以提高自
化管理。主创新能力。
5 6 7 8
It aims to promote the de- Accelerating the improve- Its long-term goal is to It aims to attract social
velopment of the venture ment of industrial structure promote the development capital to follow its invest-
capital market, thus accel- through the integration of of high-tech industries in ment, including prestigious
erating the improvements of high-quality social resources China through providing venture capital institutions
financing environment and is its investment objective. value-added services related from both within and out-
economic structure. 投资坚持 “集
集合优质社会资源，促 to venture capital invest- side the local province.
发展目标是促进风险投资环境的进产业优化 ”的
的宗旨。 ment, thus nurturing strate- 引导基金吸引社会资本的范围包括
发展，推进融资环境和经济结构调 gic industries and promot- 省内外的一流风险投资公司。
整。 ing the economic transfor-
mation.
长期目标是通过提供风险投资相关
的增值服务，促进中国高科技产业
的发展，培育战略性产业，实现经
济转型的目标。
9 10
It implements a manage- As a long-term investor,
ment system that sepa- it has the investment phi-
rates management decision- losophy of achieving the
making from the govern- targeted return rate while
ment; Its operation princi- keeping the risks low.
ples are “government guid- 作为长期投资者，投资理念是在实
ance, market operation, am- 现回报目标同时控制风险。
plification through leverage,
and risk prevention”.
机构实行管理决策与政府分离的管
理体制，按照 “政
政府引导，市场运
作，杠杆放大和风险防范 ”的
的原则
运作。

Corporate Corporate Governance: 1 2 3 4
Governance 1 if corporate governance [1-7]. The organization adopts a This firm implements strict With a professional team With the goal of accelerat-
(0.5) rigorous auditing and com- risk management with mod- and an open cultural at- ing industrial advancement
pliance system on par with ern corporate governance mosphere, this firm offers and social development, this
international standards to practices; it closely fol- comprehensive and profes- firm has the following codes
better serve the interests of lows the core values of “in- sional financial services to of conduct: professionaliza-
investors. tegrity, professionalization, the clients, and is commit- tion, innovation, rigor, and
公司引入严格的审计和合规体系， standardization, and innova- ted to becoming the most re- efficiency.
以更好地服务于投资人的利益。 tion”. liable, the most sustainable, 公司坚持专业、创新、严谨、高效
公司坚持 “诚
诚信，专业，规范，创 and the most advanced firm 的准则，以促进产业进步和社会发
新 ”的
的核心价值观，依靠现代化的 in the industry. 展发挥积极作用。
公司治理，严格的实施风险控制和依靠专业的员工团队，开放的文化
管理。氛围，为投资者提供全面和专业
的金融服务，致力于成为行业 “最
最
可靠，最可持续的和最领先的公
司 ”。
。
5 6 7
Bringing long-term returns To help start-ups estab- This firm operates, invests,
at an acceptable level of risk lish a leading position in manages, and withdraws in
is its long-standing invest- their industries, this firm a market-oriented way.
ment philosophy. established a standard and 以市场导向的方式运作、投资、管
长期以来的投资理念是在可接受的 rigorous investment and 理和撤回。
风险水平下带来长期回报。 risk management system,
introduced advanced man-
agement philosophy and
professional methods, and
built an experienced and
high-quality investment
team.
建立规范、严谨的投资管理体系和
风险控制体系，引入先进管理理念
和专业管理手段，培育了一支经验
丰富、高素质的投资团队，确立创
业型企业的领先地位。
Stage Focus Stage Focus: 1 2 3
(0.5) 1 if show stage focus [1-3] The purpose is to channel It frequently provides fi- The investments target late
capital to angel projects to nancing for investments in stage projects which can fa-
help finance early stage en- the growth and expansion cilitate the IPO of innovative
terprises. stage, but it also invests se- companies.
集中投资企业的早期天使项目，并 lectively in early and late 投资着重于后期阶段，以帮助创新
发挥对资本的撬动作用。 stage projects. 企业实现上市目标。
一直以来为企业的扩张阶段提供融
资，同时投资早期和晚期的项目。
Table 7. GP Preferences for LPs
Partner Rating
(1) (2)
Government Ties -0.114∗∗∗ -0.079∗∗
(-2.92) (-2.14)
Large Investor 0.147∗∗∗ 0.167∗∗∗
(4.21) (5.03)
High Registered Capital 0.196∗∗∗ 0.185∗∗∗
(5.52) (5.53)
Industry Information -0.231∗∗∗ -0.178∗∗∗
(-6.68) (-5.39)
Young LP -0.004 -0.010
(-0.11) (-0.29)
Headquarter In Foreign Country 0.034 -0.022
(0.55) (-0.35)
Headquarter In Beijing 0.208∗∗∗ 0.175∗∗∗
(4.04) (3.51)
Corporate Governance 0.013 0.055∗
(0.37) (1.67)
Investment Philosophy 0.014 0.039
(0.40) (1.14)
Stage Focus -0.085∗∗ -0.086∗∗
(-2.44) (-2.57)
Observations 13375 13375
Unique GPs 679 679
GP FEs No Yes
Model OLS OLS
DV Mean 6.448 6.448
DV SD 2.016 2.016
Notes: This table shows GP preferences for LP synthetic characteristics. The specification is yij = αi + β ×
PN
GovernmentTiesj + m=1 γm × Characteristicjm + ϵij . The sample includes all GP respondents participating in the
experiments who gave at least one valid answer to each question. GovernmentTies is a dummy indicating whether
the LP profile displays a link to the government. Details of the remaining characteristics are illustrated in Table 5.
Partner Rating is on a scale of 1-10. Column 1 shows the baseline OLS. Column 2 shows the regression adding GP
respondents fixed effects. t statistics are presented in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Table 8. GP Preferences for LPs: Heterogeneity by Government-Owned GPs
(1) (2) (1)=(2) (3) (4) (3)=(4)

Gov Non-Gov P-Value Gov Non-Gov P-Value
Government Ties 0.016 -0.173∗∗∗ 0.026 0.008 -0.119∗∗∗ 0.104
(0.22) (-3.68) (0.13) (-2.70)
Large Investor 0.186∗∗∗ 0.131∗∗∗ 0.470 0.186∗∗∗ 0.157∗∗∗ 0.682
(2.95) (3.11) (3.08) (3.94)
High Registered Capital 0.210∗∗∗ 0.189∗∗∗ 0.782 0.163∗∗∗ 0.194∗∗∗ 0.664
(3.28) (4.44) (2.66) (4.85)
Industry Information -0.255∗∗∗ -0.222∗∗∗ 0.658 -0.172∗∗∗ -0.181∗∗∗ 0.893
(-4.09) (-5.33) (-2.84) (-4.59)
Young LP 0.010 -0.012 0.774 -0.007 -0.013 0.931
(0.16) (-0.28) (-0.11) (-0.33)
Headquarter In Foreign Country 0.027 0.039 0.926 -0.091 0.011 0.431
(0.24) (0.52) (-0.81) (0.15)
Headquarter In Beijing 0.281∗∗∗ 0.175∗∗∗ 0.349 0.226∗∗ 0.151∗∗ 0.486
(2.98) (2.84) (2.46) (2.54)
Corporate Governance 0.047 -0.003 0.503 0.123∗∗ 0.024 0.160
(0.75) (-0.08) (2.05) (0.62)
Investment Philosophy 0.008 0.020 0.882 0.050 0.036 0.852
(0.13) (0.45) (0.80) (0.88)
Stage Focus -0.083 -0.084∗∗ 0.985 -0.115∗ -0.071∗ 0.531
(-1.31) (-1.99) (-1.90) (-1.78)
Observations 4221 9154 4221 9154
Unique GPs 214 465 214 465
GP FEs No No SUR Yes Yes SUR
Model OLS OLS OLS OLS
DV Mean 6.452 6.445 6.452 6.445
DV SD 2.038 2.006 2.038 2.006
Notes: This table compares government GP and nongovernment
PN GP preferences for LP synthetic characteristics.
The specification is yij = αi + β × GovernmentTiesj + m=1 γm × Characteristicjm + ϵij . We run separate regressions
for government GPs and nongovernment GPs. Gov-GPs are defined as GPs with government owners. The sample
includes all GP respondents participating in the experiments who gave at least one valid answer to each question.
GovernmentTies is a dummy indicating whether the LP profile displays a link to the government. Details of the
remaining characteristics are illustrated in Table 5. Partner Rating is on a scale of 1-10. Columns 1 and 2 show the
basic models for government GPs and nongovernment GPs respectively. Column 3 shows the difference in coefficients
in columns 1 and 2 using SUR model. Columns 4 and 5 show regressions with GP respondents fixed effects. Column
6 shows the difference in coefficients in columns 4 and 5 using SUR model. t statistics are presented in parentheses.
*** p<0.01, ** p<0.05, * p<0.1.
Table 9. LP Preferences for GPs
Partner Rating
(1) (2)
Government Investors 0.652∗∗∗ 0.692∗∗∗
(7.27) (7.60)
Team Government Experience 0.196∗∗ 0.191∗∗
(2.40) (2.31)
Team Industry Experience 0.050 0.041
(0.61) (0.49)
High AUM 0.025 0.056
(0.35) (0.76)
High IRR 0.153∗∗ 0.159∗∗
(2.46) (2.50)
Exits 0.151∗∗ 0.160∗∗
(2.27) (2.35)
Ranked GP -0.271 -0.252
(-1.22) (-1.12)
Industry Information 0.631∗∗∗ 0.637∗∗∗
(10.85) (10.69)
Young GP 0.172∗∗∗ 0.137∗∗
(2.60) (2.02)
Headquarter In Foreign Country 0.490∗∗∗ 0.466∗∗∗
(3.87) (3.62)
Headquarter In Beijing 0.069 0.065
(0.87) (0.81)
VC 0.019 -0.010
(0.23) (-0.12)
Market Approach 0.111 0.106
(1.55) (1.45)
Investment Philosophy -0.029 -0.042
(-0.50) (-0.71)
Investment Stage 0.076 0.072
(1.06) (1.00)
Investment Horizon -0.101∗ -0.094
(-1.65) (-1.50)
Serial Fund Manager 0.042 0.007
(0.47) (0.08)
Observations 6220 6220
Unique LPs 311 311
LP FEs No Yes
Model OLS OLS
DV Mean 4.284 4.284
DV SD 2.326 2.326
Notes: This table shows LP preferences for GP synthetic characteristics. The specification is yij = αi + β ×
PN
GovernmentInvestorsj + m=1 γm × Characteristicjm + ϵij . The sample includes all LP respondents participating in
the experiments who gave at least one valid answer to each question. GovernmentInvestors is a dummy indicating
whether the GP profile indicates the GP already had government investors. Details of the remaining characteristics
are illustrated in Appendix Table A23. Partner Rating is on a scale of 1-10. Column 1 shows the basic models.
Column 2 shows regressions adding LP respondents fixed effects. t statistics are presented in parentheses. ***
p<0.01, ** p<0.05, * p<0.1.

W 30161

Uploaded by

Copyright:

Available Formats

W 30161

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

W 30161

Uploaded by

Copyright:

Available Formats

NBER WORKING PAPER SERIES

INVESTING WITH THE GOVERNMENT:

Working Paper 30161

NATIONAL BUREAU OF ECONOMIC RESEARCH

Emanuele Colonnelli Ernest Liu

An appendix is available at http://www.nber.org/data-appendix/w30161

Government participation in the economy via direct or indirect ownership of private

3. Venture Capital and Private Equity in China: Data and Descriptive

Fact 4: There is assortative matching, as the government invests disproportionally more in

4.3. Rating Profiles of Investment Partners. We measure GPs’ interest in LPs by

5.1. Estimating Equations. We estimate specifications of the following form:

5.3.1. Government-Owned versus Private GPs. To further investigate mechanisms, we start

types of government entities.29 When estimating a specification analogous to equation 5.1,

5.4. LPs’ Preferences for GPs. We conduct a contemporaneous experimental survey of

6. Equilibrium Impact of Government Participation

Our experimental surveys reveal substantial heterogeneity in preferences for government

(6.3) pij = E [ui + ϵ0 ≥ xij + ϵ] , qij = E [vj + δ0 ≥ yij + δ] .

A LP: Headquarters B LP: Investment Region C LP: Investment Industry

D GP: Headquarters E GP: Investment Region F GP: Investment Industry

Figure 1. Distribution of Headquarters Location, Investment Region, and

A LP: Headquarters B LP: Investment Region C LP: Investment Industry

D GP: Headquarters E GP: Investment Region F GP: Investment Industry

Figure 2. Distribution of Headquarters Location, Investment Region, and

2019 Chinese Equity Investment Survey

Figure 3. 2019 Experimental Survey: Recruitment Email

08 Construction and Real Estate | |

05 Finance and Insurance | |

03 Arts, Entertainment, and Recreation | |

−1.0 −0.5 0.0 0.5

Figure 4. GP Dislike for LPs with Government Ties: Heterogeneity by In-

About this survey

INVESTING WITH THE GOVERNMENT

5 Please choose: a value between 1-10

Other, please specify: Please provide comments or suggestions:

Other, please specify: Please provide comments or suggestions:

Figure 5. 2021 Qualitative Survey

Figure 6. Survey on Pros and Cons of Government Investors

Table 1. Summary Statistics

Table 2. Government Ownership of Investors and Fund Managers

Table 3. Government-Owned GPs Perform Worse

(1) (2) (3) (4) (5) (6) (7) (8)

Table 4. Assortative Matching Between Government-Owned GPs and LPs

Gov LP Non-Gov LP ColRatio

Table 5. Variables in Synthetic LP Profiles

Table 6. Description of LP Profiles Randomized Components

Table 6 (cont.): Description of LP Profiles Randomized Components

Table 6 (cont.): Description of LP Profiles Randomized Components

Table 6 (cont.): Description of LP Profiles Randomized Components

Table 7. GP Preferences for LPs

Table 8. GP Preferences for LPs: Heterogeneity by Government-Owned GPs

(1) (2) (1)=(2) (3) (4) (3)=(4)

Table 9. LP Preferences for GPs

You might also like