fixed effects models to test the correlation between the two from more dimensions.
Risks
2021
,
9
, 99
14 of 27
Table 3.
Benchmark model regression results.
(1)
Random Effect
(2)
Individual Fixed Effect
(3)
Time-Fixed Effect
(4)
Two-Way Fixed Effect
Fintech 1
−
0.00419 **
−
0.00977 ***
−
0.0141 ***
−
0.0102 *
(
−
2.39)
(
−
3.40)
(
−
3.15)
(
−
1.76)
Size
−
0.00233 ***
0.00524 *
−
0.00222 ***
0.00504
(
−
3.29)
(1.68)
(
−
3.21)
(1.26)
DAR
0.00113 ***
0.00101 ***
0.00111 ***
0.000994 **
(4.31)
(2.70)
(4.26)
(2.48)
CAR
0.000105 ***
0.000136 ***
0.000108 ***
0.000134 ***
(3.34)
(4.21)
(3.44)
(4.04)
Netprf
−
0.000402 ***
−
0.000452 ***
−
0.000389 ***
−
0.000455 ***
(
−
7.27)
(
−
4.85)
(
−
6.85)
(
−
4.61)
CIR
−
0.000129 **
−
0.0000276
−
0.000109 *
−
0.0000292
(
−
1.98)
(
−
0.22)
(
−
1.64)
(
−
0.22)
NIRR
0.0000638 **
0.0000711 *
0.0000516 **
0.0000709 *
(2.45)
(1.85)
(1.97)
(1.84)
SAR
−
0.000182 ***
−
0.000108
−
0.000158 ***
−
0.0000972
(
−
3.58)
(
−
1.15)
(
−
3.12)
(
−
1.03)
PGDP
−
0.00180 *
−
0.00257
−
0.000612
−
0.00281
(
−
1.93)
(
−
1.23)
(
−
0.59)
(
−
1.33)
FinDev
0.000192 ***
0.000144
0.000195 ***
0.000117
(3.30)
(1.54)
(3.25)
(1.21)
M2
−
0.000268
0.000322
−
0.00487 ***
−
0.000635
(
−
0.56)
(0.87)
(
−
2.97)
(
−
0.22)
_cons
0.00226
−
0.0915
0.102 **
−
0.0724
(0.08)
(
−
1.51)
(2.28)
(
−
0.68)
Observations
846
846
846
846
R2
0.148
0.161
0.164
0.176
Note: z statistics are shown in brackets; ***, ** and * mean significant at the level of 1%, 5% and 10%, respectively; the same below. Source:
own calculations. Software: Stata.
The four regression results show that the development of regional Fintech has a
significant inhibitory effect on bank risk taking. Under the two-way fixed effect model, the
correlation coefficient between Fintech 1 and bank risk Z-Score is
−
0.0102, and significantly
negative at the 10% level. This shows that the development of regional Fintech can reduce
the risks faced by commercial banks, and the overall operating stability of banks will be
improved. From this point of view, Fintech has promoted a series of changes such as the
expansion of banking business boundaries and the transformation of banking business
models, improved the bank’s operating efficiency and profit income, and also improved
the bank’s operating stability. This can be proved by the correlation between net operating
rate of Netprf and bank risk Z-score. Due to its regression coefficient is
−
0.000455, and
significant at the significance level of 1%, indicating that an increase in the operating net
profit margin can inhibit the risk taking of commercial banks. For bank size (size), two
models show a significant positive relationship, one model shows a significant negative
relationship, and one model shows no significance. Therefore, it is necessary to carry
out group regression according to the size of banks and investigate the true relationship
Risks
2021
,
9
, 99
15 of 27
between bank size and bank risk. The bank’s asset-liability ratio (DAR) has a significant
positive impact on the bank’s risk exposure. Excessive bank debt levels will increase the
bank’s risk-taking. All four models show that the bank’s capital adequacy ratio, CAR,
has a significant positive correlation with banks’ risk-taking, and banks’ increased capital
adequacy will actually increase the bank’s risk-taking. We believe that a bank’s higher
capital adequacy ratio means that the bank receives relatively less capital support from the
inside when carrying out asset business, and the external capital cost is higher than the
internal, which reduces the bank’s operating profit margin. Their pursuit of high profits
will increase their risk appetite, thus making the overall risk of banks present an upward
trend. There is a significant positive relationship between bank non-interest income ratio
(NIRR) and bank risk exposure. For banks, interest income is their most traditional and
major source of income. The development of Fintech has narrowed the spread between
deposits and loans, forcing banks to expand operating income beyond spread income.
However, these businesses have higher instability and risk, which has increased the risk
exposure of banks. Bank cost-to-income ratio CIR, liquidity SAR, and bank risk generally
show a negative relationship. In addition, the regression results show that there is no
significant correlation between the control variables at the city level and the macro level
and the bank’s risk taking. This shows that for regional small and medium banks such
as city commercial banks and rural commercial banks, the quality of the meso-economic
and macro-economic environment has little impact on their operations, and factors at the
individual level of the bank have a decisive role in their risk behavior. We performed a
Wald test of the benchmark regression model, Please see Appendix
E
.
6.2. Endogenous Inspection
The level of Fintech development is a macro variable at the city level, which is little
affected by the risk behaviors at the individual level of banks, but there may still be some
uncontrollable factors that cause endogenous problems (
Li et al. 2020
). Therefore, we
use the instrumental variable method to control the endogenous problem.
2
Specifically,
this article selects the Urban Innovation Index (Innovation) compiled by the Industrial
Development Research Center of Fudan University as a tool variable. In order to avoid
the influence of outliers, tail shrinking is carried out. The index is compiled based on
the updated information of the legal status of authorized patents for micro-inventions of
the state intellectual property office, as well as the annual fee structure data of invention
patents of different ages, which can well reflect urban patent renewal behavior and quality
differences. There are two main reasons for choosing this index as an instrumental variable:
(i) first, the index is compiled based on the number of urban patents. These patents
include not only the financial industry, but also high-tech industries, such as electronic
equipment, computers, and communications, and these are the hardware facilities and
technical foundations on which Fintech development depends. Therefore, the index is
highly correlated with the degree of Fintech development; (ii) second, this indicator is not
an economic and financial indicator, it is not directly related to bank behavior, therefore,
there is a certain extraneous relationship between the index and bank risk-taking.
Based on the double fixed effect model, this paper chooses two-stage least squares
(2SLS) method for regression estimation. The variable settings are shown in Table
4
.
2
This paper also uses one-stage lag independent variable to weaken the endogenous problem. Due to space limitations, we will not go into details.