Will Government Subsidies Increase Investor Confidence in Listed
Agricultural Companies?
Linlin Guo
a
and Panpan Yang
b
Zhengzhou University of Aeronautics, Zhengzhou, Henan Province, China
Keywords: Government Subsidies, Big Data, Investor Confidence, Web Text.
Abstract: Financial fraud cases of agricultural listed companies have occurred frequently, adversely affecting investor
confidence. By using the big data method, this paper develops the agricultural IC index and analyzes the
impact of different types of government subsidies and tax refunds on the IC of agricultural listed companies
from the perspective of behavioral finance.
1 INTRODUCTION
Agricultural listed companies are at high risk of
financial infidelity and fraud. According to the
Choice of Orient Wealth, from January 2010 to
December 2020, a total of 312 listed companies in
China’s A-share market were subject to public
sanctions or regulatory investigations for disclosing
false information, including 54 agricultural
enterprises (agriculture, forestry, animal husbandry,
fishery, and agricultural product processing),
accounting for about 17% of the A-share market in
China. Poor quality disclosure of information of
agricultural listed companies will cause investors to
worry about possible losses, which seriously
undermines investor confidence (IC) and creates a
risk of a decline in the valuation of agricultural listed
companies or an increase in financing capital. On this
basis, this paper develops an agricultural IC index
with the big data method, and studies whether the
subsidies and supports of the government have a
supportive effect on the investors of agricultural
listed companies from the perspective of behavioral
finance, and further examines the possibility of
increasing IC in the information disclosure of
agricultural enterprises and government supports
easily affecting IC.
a
https://orcid.org/0000-0002-9189-9554
b
https://orcid.org/ 0000-0001-5874-7885
2 LITERATURE REVIEW
In order to study whether government subsidies have
a supportive effect on the agricultural IC, it is
important to first obtain the agricultural IC. The
traditional method is to obtain IC through
investigation. Today, Internet data records the micro-
psychological information and concerns in searching
of investors and provides massive data for research.
Web text data mining and its application in the
economic and financial fields were developed in
foreign countries early on. Ettridge et al. (Ettridge, et
al, 2005) were the first to propose that web search
data have an important value in economic statistical
research. In studying assets pricing. Iresberger
(Iresberger, 2015) used the search engine Google to
collect network data representing the crisis
psychology of investors. Zongyue et al. (Zongyue, et
al., 2017) conducted an emotion analysis based on the
received domestic financial news comments,
expanded the emotion dictionary by using the
clustering method for the news comments, combined
with the time characteristics of the text, and decided
to assess the emotional tendency of the text with the
help of machine learning. Zhang. Zongxin et al.
(Zhang, 2021) conducted an emotion analysis of the
media reporting corpus of individual stocks in Baidu
News, used machine learning to classify text
emotions and develop the emotion value, and studied
the influence of the emotion value of media on the
346
Guo, L. and Yang, P.
Will Government Subsidies Increase Investor Confidence in Listed Agricultural Companies?.
DOI: 10.5220/0011178000003440
In Proceedings of the International Conference on Big Data Economy and Digital Management (BDEDM 2022), pages 346-350
ISBN: 978-989-758-593-7
Copyright
c
2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
securities market. In this paper, it is believed that IC
is highly flexible and diverse. Compared with the
traditional questionnaire interview method, the IC
collection method based on big data technology is
more timely and complete.
3 RESEARCH DESIGN
3.1 Variable Design
1) Investor confidence. The level of investor
confidence (IC) can reflect the optimism of the
market towards agricultural enterprises. This paper
uses web crawler technology to obtain web text data
of investor exchanges on the Stock Bar Forum of
Orient Wealth (China’s first financial website with an
average daily visitors of over 30 million users),
extract the subjects of the text data using the LDA
model, and use machine learning algorithms to
classify investors’ bearish and bullish expectations.
Finally, this paper uses the classified emotion value
and the quantified feature word weights to develop
the Stock Bar investors’ bearish and bullish
expectation index, of which, combined with Baidu
search confidence index and stock turnover rate, will
be used to construct the final IC index using principal
component analysis. The specific methods are as
follows.
First of all, the Python crawler software was used,
and the hot post page of Orient Wealth Finance and
Economics Review Bar (Stock Bar) was considered
as the entry page. The basic information crawled
includes page information, post title, publication
time, number of comments, page views, and
comment contents, and the crawled information was
stored by date to form the web text database of the
Stock Bar with a certain scale.
Secondly, the text data was processed by cleaning
and standardizing the initial Stock Bar comment text
data, which was then vectorized using the vector
space model. Assume the web text as Q
k
D
, k
D
…,k
D
, D
as the feature word of
Text Q, and k
as the weight of the feature word, the
feature value of each text was determined using the
information gain algorithm, and the cosine value of
the angle between vectors was used to indicate the
similarity between Text Q. Then, the LDA subject
extraction model was used to extract the subject
information representing the emotional intensity of
investors from the text data set, and it was represented
by simplified text. Finally, machine learning
classification technology was used to calculate the
emotional tendency value of the text data.
Thirdly, the development of the bearish and
bullish expectation index of investors of the Stock
Bar. The text emotion value was represented by the
product of the emotional tendency value of the
bearish and bullish of each subject and the sum of the
feature words weights, and then the expectation
values of all the texts were summed to obtain the
expectation values of investors of the Stock Bar for
that day. The expectation index was developed
according to the following formula.
Estock =
∑
sent
∑
Temperature
T
(1)
In which, M=
1，2， … ，M
represents the
number of daily text comment subjects of the Stock
Bar, sent
represents the bearish and bullish
expectation value of the M-th subject, and
∑
Temperature
T
represents the sum of
feature words weights in M subjects.
The TF-IDF model was used to calculate the
weight of each feature word. Assume TF
as the
frequency of Feature Word D
in Document Q
,
IDF
as the inverse text frequency of Feature Word
D
in Document Q
, and n
as the number of
training samples of D
, the calculation formula of the
weight of the feature value is as follows.
Temperature
T
= TF
× IDF
=
log
freq
+
1
× log
n n
⁄
(2)
In conclusion, the Stock Bar agricultural stock
investors’ bearish and bullish expectations index can
be expressed as:
Estock =
∑
sent
∑
log
freq
+1
×
log
n n
⁄
(3)
Fourthly, the development of the IC index. The
bearish and bullish expectation index of investors of
the Stock Bar (Estock), Baidu search index (BIX),
and stock turnover (STM) were fitted by principal
component analysis method to develop an IC index.
For the Baidu search index, the monthly Baidu search
volume of the stock code of agricultural listed
companies was used instead of the search volume of
the company name. Because the company name can
be divided into abbreviations and full name, it is
impossible to determine which name investors used.
In addition, the company name may be searched on
Baidu for different purposes. The investors may not
search the company name because they care about the
stock of the company, but the company’s stock code
is searched, and can directly indicate that investors
actively search and pay attention to the stock of the
company
In the principal component analysis, the following
standard was strictly complied with to select the
information which could ensure the explanation of
principal components and all the indexes. The
Will Government Subsidies Increase Investor Confidence in Listed Agricultural Companies?
347
cumulative variance contribution rate reaches over
85%, and therefore, the first two principal
components were selected, and the weighted average
was obtained according to the variance contribution
rate. The weight was the ratio of the variance
contribution rate of the principal component to the
sum of the variance contribution rates of the three
principal components, and the final result was as
follows.
IC=0.6928×Estock+0.6003×BIX+0.5221×STM (4)
2) Government subsidies and tax refunds. The
government will support the development of
agricultural enterprises through various supporting
policies (such as government subsidies, tax
incentives, and direct investment). Combined with
relevant data on agricultural listed companies, this
paper mainly considers two kinds of government
support, that is, government subsidies and tax
refunds. For government subsidies, according to the
provisions of Accounting Standards for Business
Enterprises No.16-Government Subsidies
promulgated in 2017, this paper sets GOVNOI, the
ratio of government subsidies included in non-
operating income to operating income, to measure the
government’s non-operational subsidies to
agricultural enterprises, and GOVOI, the ratio of
government subsidies included in other income to
operating income, to measure the government
operational subsidies to agricultural enterprises. As
for tax refunds, in order to support the development
of agricultural enterprises, the government will give
subsidies in the form of tax refunds. This paper sets
GOVTR, the ratio of tax refunds to operating income,
to measure the tax refunds received by agricultural
enterprises.
3) Moderating variables. From the perspective of
corporate governance, government support is
affected by internal governance and third party
supervision. this paper sets GOVEI as the moderating
variable to measure the shareholding ratio of state-
owned capital investment in agricultural enterprises.
A high shareholding ratio indicates that state-owned
capital has a strong motivation to participate, which
may increase the supervision of agricultural
enterprises and the supporting effect of government
support.
4) Control variables. Control variables mainly
include enterprise size (Size), enterprise profitability
(ROE), enterprise financial leverage (LEV), listing
age (Age), stock return (SR), dummy variable of the
year (Year), and dummy variable of the sector
(Sector). If it belongs to the Beijing Stock Exchange,
the value is 1, otherwise it is 0.
3.2 Model Design
In order to examine whether government subsidies
and tax refunds can increase IC in agricultural listed
companies, this paper sets up a double fixed-effect
model of individuals and time points based on the
data obtained.
IC
,
= β
GOVNOI
,
+ β
GOVOI
,
+ β
GOVTR
,
+
β
Size
,
+ β
ROE
,
+ β
Age
,
+ β
Age
,
+
β
SR
,
+ β
Sector
,
+ β
+ θ
+ε (5)
Where β
is the fixed effect of an individual, θ
is the fixed effect of a time point, and the other
variables have been introduced in the preceding part
of this paper. In addition, from the perspective of
corporate governance, we explored whether state-
owned capital shareholding and audit supervision
improved or worsened the relationship between
government subsidies and IC. In Model (2), the
moderating variables state-owned capital
shareholding ratio were added to test.
IC
,
= β
GOVNOI
,
+ β
GOVOI
,
+ β
GOVTR
,
+α
,
GOV
,
× GOVEI
,
+ β
Size
,
+ β
ROE
,
+ β
Age
,
+ β
Age
,
+ β
SR
,
+β
Sector
,
+ β
+ θ
+ε (6)
Where GOV_SUB
,
is the vector consisting of
government subsidies GOVNOI and GOVOI, and tax
refund GOVTR; β
is the fixed effect of an
individual; θ
is the fixed effect of a time point.
3.3 Data Sources
The data studied in this paper is from agricultural
enterprises listed on the Shanghai Stock Exchange,
Shenzhen Stock Exchange, and Beijing Stock
Exchange (the former National Equities Exchange
and Quotations selected layer and innovative layer)
from 2015 to 2020. A total of 1,432 agricultural listed
companies in the agriculture, forestry, animal
husbandry, and fishery industries were collected from
the Shanghai and Shenzhen Stock Exchanges,
excluding alcohol, ST, and samples with less than 3
years of listing. At the same time, a total of 226
agricultural companies of the Beijing Stock
Exchange were collected and classified based on the
management classification results of listed companies
in the National SME Share Transfer System.
Table 1: Descriptive statistics of variables.
Variable mean SD
minimu
m
maximum
IC -0.181 1.783 -3.110 3.459
GOVNO
I
0.015 0.037 0 0.790
GOVOI 0.009 0.255 0 0.457
BDEDM 2022 - The International Conference on Big Data Economy and Digital Management
348
GOVTR 0.007 0.378 0 0.101
GOVEI 0.017 0.047 0 0.845
Size 22.30 0.891 0 45.680
ROE 0.011 0.045 -1.00 0.671
LEV 0.49 0.300 0.072 0.920
A
g
e 11.45 5.261 4 17
SR 0.039 0.155 -1.231 1.458
4 EMPIRICAL TEST
4.1 Test of the Effect of Government
Subsidies and Tax Refund on IC
According to the test results in Table 2, the full
samples of government subsidies included in “non-
operational income” unrelated to daily operation does
not have a significant effect on IC, while government
subsidies and tax refunds included in “other income”
related to daily operation have a positive effect on IC.
Compared with Shanghai and Shenzhen stocks, the
effect on the IC of agricultural enterprises listed on
the Beijing Stock Exchange is relatively big, perhaps
because most companies listed on the Beijing Stock
Exchange are small and medium-sized companies
whose corporate governance and corporate operation
should be improved. In addition, the information
disclosure requirements of enterprises listed on the
Beijing Stock Exchange are different from those of
enterprises listed on the main board. Therefore,
investors do not trust agricultural enterprises listed on
the Beijing Stock Exchange, which affects the IC.
However, the results show that government subsidies
can positively increase the IC in agricultural
enterprises of the Beijing Stock Exchange, with a
greater supporting effect. Secondly, the test results
also show that government subsidies unrelated to the
operation may not have a supporting effect on
investors. Compared with government subsidies
unrelated to daily operation, tax refunds and
government subsidies related to daily operation
included in “other income” have a greater positive
effect on IC, that is, to increase IC. Domestic scholars
have also found that different classifications of
government subsidies have different influences on
R&D expenditures, innovation ability, and other
aspects of enterprises. Seen from control variables,
the profitability, financial leverage, and other aspects
of agricultural enterprises have no statistically
significant impact, whereas the enterprise size and
listing age have a positive impact on IC. Finally, we
can see from the test results that the effect of tax
refund (with an impact coefficient of 0.0318) is
greater than that of the government subsidies
included in “other income” related to daily operation
(with an impact coefficient of 0.0245), perhaps
because agricultural enterprises have certain
conditions to meet the requirements for tax refunds
with a stronger supporting effect.
Table 2: Regression results of the effect of government
subsidies on IC.
Var ia bl e
IC IC IC
full HS BJ
GOVNOI
0.0031
(0.129)
0.0024
(0.170)
0.0036
(0.137)
GOVOI
0.0276
*
(0.062)
0.0117
**
(0.041)
0.0210
*
(0.077)
GOVTR
0.0325
*
(
0.078
)
0.0206
*
(
0.063
)
0.0389
*
(
0.066
)
Size
0.0012
*
(0.039)
0.0010
*
(0.044)
0.0018
(0.109)
ROE
0.0036
(0.901)
0.0054
(0.862)
0.0051
(0.151)
LEV
-0.9290
(
0.890
)
-1.4456
(
1.237
)
-1.3786
(
0.972
)
Age
0.0010
*
(0.048)
0.0007
*
(0.044)
0.0020
*
(0.021)
SR
0.0335
*
(0.072)
0.0298
**
(0.083)
0.0451
*
(0.027)
Individual
and Yea
r
Controlled Controlled Controlled
Adj.
R
0.3002 0.3461 0.2217
F 107.15
***
110.28
***
100.22
***
N 1647 1421 226
***, ** and * denote statistical significance at the 1%, 5%, and 10% level.
4.2 Test of the Moderating Effect of
State-owned Capital Shareholding
Based on the above test results, we further explored
whether state-owned capital shareholding could
enhance the supporting effect of government
subsidies and tax refunds. The results in Table 3 show
that the regression coefficient of GOVNOI×GOVEI
is not significant, that of GOVOI×GOVEI is 0.0572
(𝜌 <0.1), and that of GOVTR×GOVEI is 0.0656
(𝜌 <0.1) when the interaction terms of state-owned
capital shareholding ratio, government subsidies, and
tax refund are added in Models 5, 6, and 7. Consistent
with the test results in Table 2, the state-owned capital
shareholding cannot enhance the influence of non-
operational government subsidies . The regression
coefficient of the interaction terms of state-owned
capital shareholding, government subsidies, and tax
refunds related to daily operation is significantly
positive, which indicates that a higher proportion of
state-owned capital shareholding can enhance the
influence of government subsidies and tax refunds
related to daily operation on IC. The above-
Will Government Subsidies Increase Investor Confidence in Listed Agricultural Companies?
349
mentioned results show that state-owned capital
shareholding can enhance the supervision of the
governance environment of agricultural enterprises,
thereby further increasing IC.
Table 3: Test results of the moderating effect of state-
owned capital shareholdings.
Va ri ab le
Model
(
5
)
Model
(
6
)
Model
(
7
)
IC IC IC
GOVNOI
0.0008
(0.134)
GOVEI
0.0302
*
(0.077)
GOVNOI×GOVEI
0.0405
(
0.184
)
GOVOI
0.0334
***
(0.004)
GOVEI
0.0355
**
(0.085)
GOVOI×GOVEI
0.0572
*
(
0.090
)
GOVTR
0.0201
*
(
0.089
)
GOVEI
0.0142
*
(0.092)
GOVTR×GOVEI
0.0656
*
(0.097)
control variable
Yes Yes Yes
Adj. R
0.3711 0.3223 0.3039
F
27.021
***
28.119
***
27.458
***
N
1647 1647 1647
***, ** and * denote statistical significance at the 1%, 5%, and 10% level.
Some control variables that will affect IC in
agricultural enterprises may be omitted in this paper.
In order to solve the possible endogeneity problem of
the model and perform the robustness test, in this
paper, the following methods were used. Firstly, by
adjusting the characteristic variables of agricultural
enterprises, the debt-to-assets ratio (Debt) and
operating profit margin (Margins) were added. Then,
the IC index (ICQ) was transformed, and Tobin Q
value, turnover rate, and Baidu search index were
used for re-fitting. Following that, the dynamic GMM
model was used for estimation test. The test results
are almost the same to the main research conclusions
of this paper.
5 CONCLUSIONS
From the perspective of behavioral finance, this paper
develops the IC index by combining big data and
traditional indexes. The bearish and bullish
expectation index of investors of the Stock Bar, Baidu
search index, and stock turnover (STM) were fitted
by principal component analysis to extract the
common components of IC. Compared with the
existing method, this method can measure the IC
more timely and comprehensively.
According to the research results, government
non-operational subsidies cannot increase IC.
Therefore, in order to increase the IC in listed
agricultural companies of the market, when giving
subsidies to agriculture, the government should make
more customized subsidy policies related to daily
operation according to the characteristics of
agricultural operation. In addition, compared with
government subsidies, agricultural listed companies
may need to truly operate to receive the tax refunds.
Therefore, we can also consider decreasing direct
fund subsidies and supporting agricultural enterprises
from taxes and dues.
ACKNOWLEDGEMENTS
We are grateful to the editor for their comments that
helped improve the paper. This paper is supported by
the Henan Province Soft Science Research
(192400410384).
REFERENCES
Ettredge, M., J. Gerdes & G. Karuga. (2005). Using web-
based search data to predict macroeconomic statistics.
J. Communications of the ACM. 48, 87–92.
Irresberger, F. (2015). Explaining bank stock performance
with crisis sentiment. J. 59, 311-329.
Zhang. Z. X. (2021). Media's Emotional Contagion and
Analyst Optimistic Bias:Evidence Based on the
Technique of Machine Learning. J. 37, 170-185.
Zongyue, W. & Q. Sujuan. (2017). A Sentiment Analysis
Method of Chinese Specialized Field Short
Commentary. IEEE International Conference on
Computer and Communications.
BDEDM 2022 - The International Conference on Big Data Economy and Digital Management
350
