280 play a more active role in the corporate governance management of their portfolio firms in the
post-IPO period, and this continues to positively foster the post-IPO performance of VC-backed firms.
Thus,  these  two  models  predict  opposite  portfolio  firm  performance  in  the  IPO  and  post-IPO process. In the IPO time, the certification model proposes lower IPO cost and less underpricing
for  VC-backed  IPOs  while  the  adverse  selection  model  expects  higher  IPO  costs  and  greater underpricing  because  VC-backed  IPOs  have  higher  risk.  Regarding  the  IPO  period,  the
monitoring  model  assumes  worse  operating  performance  of  VC-backed  companies  while  the adverse selection model expects same or superior performance to ensure the IPO success. As to
post-IPO market and operating performance, the monitoring model predicts that VC-backed ones are superior to non-VC-backed counterparties though the difference will decline as firm ages and
venture  capitalists  exit.  In  contrast,  the  grandstanding  model  predicts  that  VC-backed  firms perform worse and as firm age increase the potential risk would come into truth.
3. Data and methodology
The objection of this paper is to explore the degree that the IPO and post-performance of Jiangsu listed firms is affected by the participation of VC investors. We intend to compare a match-pair
sample of 89 non-VC-backed and 102 VC-backed listed firms in Jiangsu Province.
3.1 Sample selection
We extract raw information from the CV Sources Database to set up a dataset of 190 VC-backed firms. We exclude firms, from which VCs exit by merger and acquisition. In other words, we
focus  on  the  Chinese  VC-backed  companies  that  succeeded  in  IPOs.  Then  according  to  each industry, we download a group of 2230 firms from the CSMAR database. After excluding the
VC-backed listed firms, the rest 2040 is the total sample of non-VC-funded listed firms. Also, we exclude the VC-funded firms which are listed in the NYSE, DASDAQ, Korea Stock Exchange
KSE, Singapore Exchange Limited SGX and Taiwan Stock Exchange Corporation TSEC. The rest are all listed on the Shanghai Securities Exchange SSE.
Then,  companies  with  incomplete  have  deleted  from  the  sample,  as  well  as  some  abnormal variables. For example, firm stock code starting with ‘900’ and ‘200’ were deleted since related
closing  price  could  not  be  found  from  the  CSMAR  database.  After  deducting  those  data mentioned  above,  the  whole  sample  consists  of  1681  observations  from  1993  the  year  time
when  the  first  VC-backed  firm  was  listed  to  2014  the  most  recent  accessible  data  for  this research.
3.1.1 Methodological solution to selection bias adjusted
Based on previous studies, it is expected that venture-backed enterprises may differ significantly from non-funded even in the same industry so we cannot simply set up treatment and control
group by industry. Two major reasons lead to a statistical bias. First of all, venture capitalists are interested  in  those  superior  firms  only  they  have  survived  from  pre-investment  screening.
Secondly, firms that are not worthy to being invested may not finance by VC. As a result, they normally do not involve in the screening process Engel and Keilbach, 2007. As suggested by
Rosenbusch et al. 2013, it is found that VC investment is positively related to performance of backed  companies  but  after  control  the  industry  selection  bias  the  positive  influence  is
eliminated. This selection bias is not sufficient to be accounted for positive factor of venture fund to firm
281 earning  performance.  We  need  to  control  for  the  selection  biases  so  that  we  construct  a  new
control group of non-VC-backed firms using the propensity score matching Croce et al., 2013; Rosenbaum and Rubin, 1983. The aim is to find the non-VC-backed firms that have the most
similar probability of receiving VC for each VC-backed firm. Following the notation in Lee and Wahal 2004, we define 1 as VC-backed and 0 as non-VC-
backed.  Hence Y and Y  separately  denote  the  daily  return  of  non-VC-backed  IPO  and  VC- backed IPO. Our intention is the impact of VCs on IPOs, which is Y − Y . Since one firm is
either  to  be  funded  or  not  be  funded, Y − Y  is  unobtainable.  Thus  we  estimate  the  average treatment effect ATT of VCs, which is as follows.
E[Y − Y |VC = , X] = E[Y |VC = , X] − E[Y |VC = , X] X  is  denoted  as  industry  features  that  are  related  to  the  selection  procedure. E[Y |VC = , X]
defines the average daily return for VC-backed IPOs but the E[Y |VC = , X] is unobtainable. From the conditional independence assumption, the traditional way is to apply
[Y |VC = , X
] = [Y |VC = ,X
] The  conditional  independence  assumption  Rubin,  1977  is  firms  vary  in  target  factor
�
regardless whether realizations of
�
is the same or not. If the above assumption stands, the ATT is as follows,
E[Y |VC = , X] − E[Y |VC = , X]. Unfortunately, venture fund is not invested randomly but is an endogenous choice. It means the
existence of bias, which is �
= E[Y |VC = , X] − E[Y |VC = , X] To eliminate the bias, propose the propensity score method PSM is proposed by  Rosenbaum
and Rubin 1983 to match the treated and control groups. A probit model is estimated for the endogenous  selection  variable  Y=1  for  venture  fund  and  Y=0  otherwise  with  a  set  of  X
variables including firm age, education of founders, numbers of employees, total assets, gross sales and industry dummies. The firms’ conditional possibility to obtain VC fund is employed as
the  propensity  score  and  VC-funded  IPO  is  matched  with  the  non-VC-funded  IPO  with  the nearest propensity score. As PSM we used in STATA is a one-to-one matching technique, the
limitation is it drops potentially valuable sample.