Professional Documents
Culture Documents
2
d lnOP
a
2
2
d lnNON
1a
2
2
d lnNET
2a 1a Cov d lnNON; d lnNET 1
where a = NON/(NET + NON) is the noninterest share of net operating revenue, (1 a) is the net interest share, dlnNON is the
log growth rate of noninterest income, dlnNET is the log growth rate of net interest income, a
2
d ln NON
2
is the contribution of
noninterest income to overall revenue volatility, and (1 a)
2
d ln NET
2
is the contribution of net interest income.
The standard portfolio theory implies that if the covariance between noninterest income growth and net interest income
growth is negative, the volatility of the operating revenue growth can be directly reduced. Even if the covariance is positive, the
standard deviation of the net operating revenue will be less than the sum of the weighted average of noninterest income and net
interest income as long as the covariance is not exactly one. Table 2 shows estimates for the components of Eq. (1) for two time
periods: 198799 and 200008. The first column presents the average shares of noninterest income () and net interest income
(1 ), the second column presents the sample variance or covariance of the variables (
d ln OP
2
,
d ln NON
2
,
d ln NET
2
, and
Cov(d ln NON, d ln NET)), and the third column presents the contributions (share-weighted variances), as expressed by the
right-hand side of Eq. (1).
Table 2 shows that the variance of net operating revenue fell from 3.83 (19871999) to 0.91 (20002008), indicating a
decreased volatility of bank revenue. In these two time periods, the variance of net interest income declined substantially (from
4.82 to 0.71), that of noninterest income decreased slightly (from4.66 to 3.84), and the covariance between themfell significantly
(from 1.36 to 0.77), which implies that the entire decline in overall bank revenue volatility is due to these three components,
especially net interest income and the covariance.
4
F-tests reject the null hypothesis of equal standard deviations for all pairwise combinations of the four components of noninterest income except for
investment revenue and exchange gains (p-value = 0.2) and investment revenue and other income (p-value = 0.72).
5
F-tests reject the null hypothesis of equal standard deviations for all pairwise combinations of the four components of noninterest income except for
investment income and exchange gains (p-value = 0.16), exchange gains and other noninterest income (p-value = 0.9), fee and commission income and
investment income (p-value = 0.55), and investment income and other noninterest income (p-value = 0.2).
-40%
-20%
0%
20%
40%
60%
80%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
2
0
0
6
2
0
0
7
2
0
0
8
Growth Rate of Net Interest Income Growth Rate of Noninterest Income
Fig. 3. Growth rate of net interest income and noninterest income (19872008).
156 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
We believe that the decline in net interest volatility after 2000 is mainly due to the stable benchmark interest spread and
macroeconomics. The decreased covariance of the net interest income growth and noninterest income growth indicates
diversification benefits from noninterest activities. One plausible explanation is that the share of noninterest income of the
Chinese banking industry is still low compared to that of the international banking industry, yielding diversification benefits.
In addition, although noninterest income volatility decreased from 198799 to 200008, the variance of the noninterest
income is far greater than the variance of net interest income in 200008 (3.84 vs. 0.71). Moreover, the contribution of
noninterest income to the overall revenue volatility decreases only slightly (from 0.2 to 0.19), which is a much smaller decrease
than that of net interest income (3.05 vs. 0.43), indicating that, as net interest income volatility decreases, noninterest income
may gradually replace net interest income as the main source of overall revenue volatility.
4.2. Aggregate cyclicality
This section further describes the cyclical properties of different types of bank revenue in China using the following regression
model:
d lnX
t
X
5
1
t
d lnX
t
X
5
0
t
d lnGDP
t
t
2
where X
t
is some measure of bank income in period t. The bank income includes net income, net income plus provisions, net
interest income, net interest income less provision, noninterest income, and noninterest income less exchange gains. The GDP is
the real GDP based on the year 2000. The coefficient reflects the impact of lagged bank income on contemporaneous bank
income, reflects the impact of contemporaneous and lagged GDP on contemporaneous bank income, and
t
is the residual.
Net income and net income plus provisions are measures of bank profits. The net income plus provisions is included because
there is some evidence that banks use loan loss provisions to smooth earnings over the cycle, and adding back provisions may
help us to better understand the cyclicality of bank earnings. Net interest income and net interest income less provisions are
-400%
0%
400%
800%
1200%
1600%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
2
0
0
6
2
0
0
7
2
0
0
8
Fees and Commission Income Exchange Gains
Investment Revenue Other Income
-100%
-50%
0%
50%
100%
150%
200%
250%
300%
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
2
0
0
6
2
0
0
7
2
0
0
8
Fees and Commission Income Exchange Gains
Investment Revenue Other Income
Fig. 4. Growth rates of four components of noninterest income (19872008).
Table 2
Decomposition of variance of net operating revenue.
Average share Variance/covariance Contribution to variance
19871999
Net operating revenue 3.83
Net interest income 79.50 4.82 3.05
Noninterest income 20.50 4.66 0.20
Net interest/noninterest 1.36 0.22
20002008
Net operating revenue 0.91
Net interest income 77.70 0.71 0.43
Noninterest income 22.30 3.84 0.19
Net interest/noninterest 0.77 0.13
Notes: (1) All data in this table are multiplied by a factor of 100; (2) contributions do not exactly add up because the shares change through the sub-periods.
157 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
indicators of traditional banking activities, and subtracting provisions may provide a more accurate measure of the return to
traditional lending activities, as it accounts for expected defaults. Noninterest income and noninterest income less exchange gains
measure nontraditional activities in the form of noninterest income, subtracting exchange gains because the volatility of
exchange gains is largest in the four components of noninterest income, allowing the cyclicality of the other three components of
noninterest income to be further examined.
Table 3 shows estimates of Eq. (2). From column 1, the contemporaneous and lagged GDP coefficients are jointly significant at
the 5% level (p-value = 0.039), indicating that a significant linear relationship exists between net interest income growth and
GDP growth. Adding back provisions in column 2 weakens the significance (p-value = 0.247), suggesting that income smoothing
does not exist in the Chinese banking industry. We further examine the other two components of net operating revenue, namely,
the growth rates of net interest income and noninterest income. Column 3 shows that the contemporaneous and lagged GDP
coefficients are not jointly significant (p-value = 0.816), indicating that there is no significant relationship between net interest
income growth and GDP growth. Column 4 shows a higher p-value (p-value = 0.535) between the contemporaneous and lagged
GDP coefficients, but this value is still not significant. Thus, despite our consideration of the provisions, net interest income is
barely affected by the macroeconomic fluctuations. Column 5 shows that the contemporaneous and lagged GDP coefficients are
jointly significant at the 10% level (p-value = 0.057). Removing exchange gains from noninterest income in column 6 weakens
the link between contemporaneous and lagged GDP, which implies that macroeconomic fluctuations are more likely to affect the
exchange gains than the other three components of noninterest income. According to these regressions, noninterest income
shows a more obvious cyclicality than net interest income. That is, an increase in noninterest income may extend, rather than
smooth, the impact of macroeconomic fluctuations on bank revenue.
A second way to address this issue is using a simple vector autoregression (VAR) framework. Fig. 5 shows the impulse
response function from a VAR with GDP, net interest income, and noninterest income from 1987 to 2008. The VAR is estimated in
log-levels for real GDP, real net interest income, and real noninterest income in a trend. The impulse response function shows
how a shock in GDP propagates through the system and affects net interest income and noninterest income. Neither net interest
income nor noninterest income exhibits a response in the first quarter. Net interest income continues to show no response to an
innovation in GDP after the first quarter. However, noninterest income shows a significant positive response in the second
quarter, but then declines sharply and becomes negative.
Table 3
Link between aggregate bank income growth and GDP growth.
Dependent variable (X
t
)
Bank profits Traditional revenue Non-traditional revenue
Net
income
Net income
plus provisions
Net interest
income
Net interest income
less provisions
Noninterest
income
Noninterest income
less exchange gains
X
t 1
0.268
(1.167)
0.400
(1.005)
0.484
(1.005)
0.215
(0.479)
0.574
(2.460)
0.522
(2.247)
X
t 2
0.137
(0.703)
0.081
(0.160)
0.246
(0.303)
0.761
(1.083)
0.166
(0.746)
0.181
(0.766)
X
t 3
0.193
(1.098)
0.190
(0.440)
0.076
(0.096)
0.354
(0.571)
0.657
(2.848)
0.662
(2.496)
X
t 4
0.816
(3.979)
0.206
(0.570)
0.326
(0.355)
0.443
(0.605)
0.716
(2.476)
0.540
(1.834)
X
t 5
0.432
(1.931)
0.256
(0.689)
0.510
(0.589)
0.328
(0.503)
0.666
(2.584)
0.742
(2.803)
GDP
t
10.921
(3.106)
12.235
(1.780)
0.843
(0.069)
12.731
(1.355)
8.254
(2.264)
7.189
(1.610)
GDP
t 1
1.130
(0.217)
9.628
(1.310)
1.069
(0.085)
6.657
(0.536)
9.500
(2.411)
10.905
(2.327)
GDP
t 2
4.029
(0.935)
12.115
(1.268)
0.910
(0.060)
12.592
(0.849)
5.932
(1.476)
4.380
(0.895)
GDP
t3
5.128
(1.138)
2.752
(0.237)
1.865
(0.247)
5.155
(0.642)
2.652
(0.762)
3.461
(0.804)
GDP
t 4
8.297
(2.260)
5.812
(0.577)
0.508
(0.092)
2.820
(0.533)
9.335
(3.273)
9.948
(2.845)
GDP
t 5
9.458
(3.771)
5.162
(0.754)
3.364
(0.651)
1.094
(0.245)
1.089
(0.434)
2.379
(0.809)
Constant 2.172
(3.528)
1.137
(1.171)
0.441
(0.268)
1.363
(0.892)
2.240
(2.723)
2.036
(2.142)
Jt. sig. of lagged X 0.009 0.396 0.812 0.860 0.099 0.153
Jt. sig. of GDP and lagged GDP 0.039 0.247 0.816 0.535 0.057 0.087
Adjusted R
2
0.817 0.276 0.591 0.315 0.666 0.588
No. obs. 22 22 22 22 22 22
Notes: (1) t-value in parentheses; (2) Jt. sig. reports the p-values associated with the null hypothesis that the set of independent variables are jointly insignificant;
(3) ***, **, and * indicate statistical significance at the 99%, 95%, and 90% levels, respectively.
158 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
These results indicate that net interest income does not significantly respond to macroeconomic fluctuations, while
noninterest income responds significantly to changes in real output. Based on this analysis, we conclude that the main reason that
net interest income is not subject to macroeconomic fluctuations is government investment, which stimulates demand and spurs
economic growth during economic depressions in China. In addition, the government encourages banks to provide large-scale
credit to entity economy, especially the government financing platform. When the benchmark interest spread changes little,
traditional interest activities can ensure a steady growth, even during a recession. In contrast, noninterest income exhibits
cyclicality, being affected by some market factors. For instance, during economic depressions, the scarcity of funds and investment
channels will result in a decrease in noninterest income, while the availability of sufficient funds and investment channels during
economic booms will lead to an increase in noninterest income.
5. Bank-level correlation and diversication effect
The analysis above is based on the aggregate data and concludes that there are diversification benefits from increasing
noninterest income. However, based on a deeper assessment, the most direct way to examine the diversification effect of an asset
portfolio is to examine the correlation of each asset in the portfolio. To examine the diversification benefits from the increase
in noninterest income, this section examines the correlation between net interest income and noninterest income using
cross-sectional correlation and bank-specific correlation based on the bank-level data. The cross-sectional correlation measures
the correlation between the net interest income growth and noninterest income growth across banks at a point in time, whereas
bank-specific correlation measures the correlation between net interest income growth and noninterest income growth across
time for each bank.
5.1. Cross-sectional correlation
The cross-sectional correlation is defined as
6
t
Corr d lnNET
i;t
; d lnNON
i;t
t
X
n
i1
d lnNET
i;t
E d lnNET
t
d lnNON
i;t
E d lnNON
t
h i
X
n
i1
d lnNET
i;t
E d lnNET
t
2
!
1=2
X
n
i1
d lnNON
i;t
E d lnNON
t
2
!
1=2
3
where
t
is estimated for each year from 1987 to 2008. d ln NET
i,t
and d ln NON
i,t
are the growth rate of net interest income and
noninterest income, respectively, for bank i in year t. E(d ln NET
t
) and E(d ln NON
t
) are the average growth rate of the net interest
income and noninterest income, respectively, for all banks in year t.
The cross-sectional correlation is estimated separately for each year and describes the correlation between net interest income
and noninterest income across banks in a particular year. That is, it shows whether a bank with above average noninterest income
growth typically has above average net interest income growth. If noninterest income has a strong diversification effect on bank
revenue, then
t
is expected to be negative. On the other hand, if
t
is strongly positive, noninterest income provides little
diversification benefit for bank revenue.
We estimate
t
for three sets of banks: all banks, large banks (assets greater than 3 trillion in 2000 RMB), and small
banks(assets less than 3 trillion in 2000 RMB). Fig. 6 plots the time series of cross-sectional correlations,
t
, for the three sets of
banks. From Fig. 6,
t
of all three sets of banks fluctuates between positive and negative. For all banks, the number of negative
years is 12, the number of positive years is 10, the average
t
across all years is 0.064, and the mean
t
is 0.008 between 1987
and 1999 and 0.144 between 2000 and 2008. The results indicate that noninterest income provides diversification benefits in
the Chinese banking industry. For large banks, the number of negative years is 12, the number of positive years is 10, the average
t
across all years is 0.1105, and the mean
t
is 0.024 between 1987 and 1999 and 0.236 between 2000 and 2008. For small
6
This formula is cited from Stiroh (2004).
-.15
-.10
-.05
.00
.05
.10
.15
1 2 3 4 5 6 7 8 9 10
Response of Net Interest Income to Cholesky
One S.D. GDP Innovation
-.3
-.2
-.1
.0
.1
.2
.3
1 2 3 4 5 6 7 8 9 10
Response of Noninterest Income to Cholesky
One S.D. GDP Innovation
Fig. 5. Impulse response function of net interest income and noninterest income to an innovation in GDP.
159 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
banks, the number of negative years is 9, the number of positive years is 7, the average
t
across all years is 0.008, and the mean
t
is 0.027 between 1993 and 1999 and 0.006 between 2000 and 2008. The results showthat for both large banks and small banks,
noninterest income provides diversification benefits in the Chinese banking industry. It is worth noting that the estimate
indicates a greater diversification effect for large banks than small banks, which is due to the smaller contribution of noninterest
income to net operating income for large banks relative to small banks. The noninterest income to net operating income ratio is
27.74% for large banks and 28.57% for small banks. Furthermore, half of the ratios of noninterest income to net operating income
for small banks exceeded 30% during 19932008, compared to none of those of large banks. Therefore, for large banks,
noninterest income has a greater diversification effect. The results also indicate that as the noninterest income increases, the
marginal revenue brought about by diversification declines.
We further analyze the correlation between various components of noninterest income and net interest income. Fig. 7 plots
t
between various components of noninterest income and net interest income. From Fig. 7, we observe that the correlation
between various components of noninterest income and net interest income fluctuates between negative and positive, showing
no regularity. The results, ranked from strongest to weakest, are exchange gains (mean of 0.071), fee and commission income
(0.023), other income (0.0465), and investment revenue (0.0629).
5.2. Bank-specic correlation
The bank-specific correlation defined as
7
i
Corr d lnNET
i;t
; d lnNON
i;t
i
X
T
t1
d lnNET
i;t
E d lnNET
i
d lnNON
i;t
E d lnNON
i
h i
X
T
t1
d lnNET
i;t
E d lnNET
i
2
!
2
X
T
t1
d lnNON
i;t
E d lnNON
i
2
!
2
4
where
i
is estimated for each bank during the examination period. d ln NET
i,t
and d ln NON
i,t
are the growth rates of net interest
income and noninterest income, respectively, for bank i in year t. E(d ln NET
i
) and E(d ln NON
i
) are the average growth rates of
net interest income and noninterest income, respectively, for bank i in all years.
The bank-specific correlation is estimated separately for each bank and describes the correlation between net interest income
and noninterest income over time for a particular bank. This is a relatively traditional method, as it directly measures whether the
noninterest income has diversification benefits. A negative correlation would suggest strong potential diversification benefits.
The calculation results based on Eq. (4) show that, for all banks, the average
i
is 0.017, the median is 0.020, and the standard
deviation is 0.270. The mean is 0.005 for large banks and 0.023 for small banks. The results indicate that noninterest income
has evident diversification benefits in the Chinese banking industry, which is in agreement with the conclusion above.
Fig. 8 plots the distribution of
i
. From Fig. 8,
i
is distributed between 0.519 and 0.390 with a median of 0.020. For all banks,
the number of negative banks is 7, and the number of positive banks is 8. Although the numbers of both negative and positive
banks are almost the same, for the 7 negative banks, the absolute average value is far greater than that for the 8 positive banks
(0.252 vs. 0.188). As shown in Fig. 8, the distribution of negative
i
is wider than that of positive
i
.
We further analyze bank-specific correlation using the following regression model
8
:
i
1
ln A
i
2
D
A
!
3
d ln A
i
4
NONSH
i
2
5
FEESH
i
6
EXCHSH
i
7
OTHERSH
i
8
MULTI
i
i
5
7
This formula is cited from Stiroh (2004).
8
When the noninterest income's share of the net operating revenue is 0 or 1, there is no relationship between noninterest income and net interest income
(
i
= 0). Therefore, we expect that a non-linear relationship exists between the noninterest income's share of the net operating revenue and
i
. Thus, we
consider both linear and non-linear relationships between the noninterest income's share of the net operating revenue and
i
(using NONSH
2
as an independent
variable). However, the regression results do not reveal a signicant inverted-U-shaped relationship between
i
and NONSH
2
. Thus, we rule out NONSH
2
. The
cause of this result may be insufcient data.
-1.0
-0.5
0.0
0.5
1.0
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
2
0
0
6
2
0
0
7
2
0
0
8
All Banks Large Banks Small Banks
Fig. 6. Cross-sectional correlation between noninterest income growth and net interest income growth. Notes: The data set for small banks begins in 1993.
160 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
where i is bank i, A is the assets of the bank,
D
A
is the leverage ratio, d ln(A) is the annual asset growth rate, NONSH is the
noninterest income's share of the net operating revenue, FEESH is the fee and commission income's share of the noninterest
income, EXCHSH is the exchange gains' share of the noninterest income, and OTHERSH is the other income's share of the
noninterest income. All of the variables marked by an upper bar are averages.
9
MULTI is a dummy variable that equals either 1
(large banks) or 0 (non-large banks). The estimation results are presented in Table 4.
From Table 4, columns 1and 2, the coefficient of NONSH is positive and significant at a 95% confidence level, indicating that as
noninterest income increases, the correlation between net interest income and noninterest income increases, and the
diversification benefits decrease. That is, although the noninterest income has diversification benefits in the Chinese banking
industry at the present stage, with the increase in shares of noninterest income, the marginal revenue brought about by
diversification gradually declines. At the same time, the correlation between MULTI and
i
is insignificant, indicating that large
banks do not have higher or lower
i
than small banks. Furthermore, from column 2, the correlation between the components of
noninterest income and
i
is not significant, showing that the individual components of noninterest income do not have a
significant impact on the correlation between net interest income and noninterest income. Moreover, the correlation between the
three control variables (ln(A),
D
A
, and d ln(A)) and
i
is significant. Banks with higher A and d ln(A) show a higher correlation
between net interest income and noninterest income, while banks with higher
D
A
show a lower correlation.
We also estimate the correlation between net interest income and noninterest income for large banks and small banks using
Eq. (5). Because there are only five large banks, which is insufficient for regression, we only calculate the estimate for small banks.
From Table 4, columns 3 and 4, the only significant correlation is between d ln(A) and
i
. The results differ from our expectations,
which is likely due to the insufficient data.
6. Noninterest income's impact on bank risk and return
After examining the diversification benefits using correlation analysis, we examine the link between noninterest income
growth and bank revenue and risk using models. We first graph the relationship between noninterest income shares and two
measures (bank performance and risk). We use the risk-adjusted returns (Sharp ratio) to measure the bank performance. The
Sharp ratio is the average return on equity (ROE) divided by the standard deviation of the ROE. The lower the Sharp ratio, the
worse the bank performance and the higher the risk. The Z-score (or Z) is the number of standard deviations below the mean by
which profits must fall to bankrupt the firm. Z is defined as
10
Z
X
T
t1
2
t
= A
t
A
t1
!
=T
X
T
t1
E
t
E
t1
= A
t
A
t1
!
=T
( )
=
ROA
6
where A
t
is the total assets in period t,
t
is the net income after taxes in period t, E
t
is the total equity, and 2
t
= A
t
A
t1
is the
return on assets in year t.
ROA
is the estimated standard deviation of the return on assets. As the formula indicates, the higher the
mean rate of return on assets and the higher the ratio of equity to assets, the higher Z is. Hence, higher values of Z are associated
with lower probabilities of failure. The more volatile the asset returns, the lower the Z-score.
According to the noninterest income's share of operating revenue for all banks, we plot the curve of the Sharp ratio and Z-score
from small to large in Fig. 9. Based on Fig. 9, with an increasing proportion of noninterest income in the net operating income, the
Sharp ratio and Z-score do not exhibit an obvious increase or decrease, making it difficult to determine the impact of the share of
noninterest income on the bank performance and risk.
9
One must arbitrarily drop one of the four components of noninterest income to avoid perfect collinearity. Investment revenue was chosen because it has the
largest mean share of noninterest income.
10
This formula is cited from Lown et al. (2000).
-1.2
-0.8
-0.4
0.0
0.4
0.8
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
2
0
0
5
2
0
0
6
2
0
0
7
2
0
0
8
Fees and Commission Income Exchange Gains Investment Revenue Other Income
Fig. 7. Cross-sectional correlation between net interest income growth and growth in components of noninterest income.
161 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
Next, we establish an econometric model to analyze the impact of noninterest income on bank risk and return. The
model is
Y
i
1
ln A
i
2
D
A
!
3
d ln A
i
4
NONSH
i
2
5
FEESH
i
6
EXCHSH
i
7
OTHERSH
i
8
MULTI
i
i
7
where i is bank i and Y
i
is the dependent variable, including the mean and standard deviation of net income growth, the mean and
standard deviation of ROE, the Sharp ratio, and the Z-score. The other variables are the same as those in Eq. (5). The estimation
results are shown in Table 5.
From Table 5, in all regression equations, we observe that the relationship between NONHS and the dependent variables is not
significant. That is, from a statistical perspective, the relationship between noninterest income growth and bank risk and return is
not significant. This finding may be a result of insufficient data. Nevertheless, it is enlightening to analyze the relationship
between NONHS and the dependent variables to some extent. Columns 1 and 2 suggest a positive relationship between NONHS
and the mean and standard deviation of net income growth, indicating that noninterest income growth induces net income
growth and an increase in the volatility of net income growth. Columns 3, 4, and 5 suggest a negative relationship between
NONHS and both the mean ROE and Sharp ratio and a positive relationship between NONHS and the standard deviation of ROE,
indicating that the noninterest income growth increases the volatility of the net income growth but not the average revenue. That
is, banks with high shares of noninterest income usually have lower risk-adjusted returns (lower profits per unit of risk). Column
6 suggests a positive relationship between NONHS and Z-score, indicating that noninterest income growth does not lead to bank
insolvency risk.
0.00
0.04
0.08
0.12
0.16
0.20
0.24
-1.00 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75 1.00
R
e
l
a
t
i
v
e
F
r
e
q
u
e
n
c
y
Rho
Fig. 8. Bank-specific correlation between noninterest income growth and net interest income growth.
Table 4
Determinants of correlation between net interest income growth and noninterest income growth.
All banks Small bank
Log assets (ln(A)) 0.287
(3.344)
0.310
(2.534)
0.145
(0.793)
0.720
(1.581)
Debt/assets
D
A
15.465
(3.625)
20.236
(2.446)
15.873
(1.675)
24.325
(1.799)
Growth in assets (dln(A)) 3.051
(3.374)
3.663
(3.698)
3.342
(2.886)
3.718
(2.217)
NONSH 1.743
(2.912)
1.855
(2.897) 0.517
(0.420)
4.103
(1.318)
FEESH 0.254
(0.441)
1.697
(1.285)
EXCHSH 2.371
(1.743)
5.202
(1.247)
OTHERSH 0.168
(0.476)
0.221
(0.369)
MULTI 0.194
(0.852)
0.416
(1.551)
Constant 2.859
(3.470)
3.104
(2.785)
1.456
(0.990)
6.585
(1.599)
Adjusted R
2
0.604 0.618 0.616 0.533
Jt. sig.: component shares
of noninterest income
0.417 0.632
No. obs. 15 15 10 10
Notes: (1) t-value in parentheses; (2) ***, **, and * indicate statistical significance at the 99%, 95%, and 90% levels, respectively; (3) Jt. sig. reports the p-values from
the F-test of the joint significance, where the hypothesis of the joint test is that the coefficient of the dependent variable is not significant.
162 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
Regarding the components of noninterest income, only other revenue is negatively correlated with the mean ROE and Sharp
ratio and positively correlated with the standard deviation of ROE, indicating that higher shares of other revenue lead to greater
volatility and lower returns. The higher the growth in assets is, the higher the growth in net income is. Additionally, the higher the
leverage (debt/assets) is, the lower the profits per unit of risk are.
Based on an analysis of Fig. 8 and Table 5, we can see that the impact of noninterest income on bank risk and return is not
significant. However, we should note that an increase in noninterest income may lead to an increase in volatility (higher risk) but
not higher returns. In other words, although noninterest income does not have a significant impact on the bank risk/return
trade-off from a statistical viewpoint, it may worsen the risk/return trade-off.
7. Conclusion
This paper explores the relationship between noninterest income growth and bank risks based on the data for the Chinese
banking industry from1986 to 2008. First, we examine the volatility and cyclicality of the sample data on an aggregate level. Next,
we examine the correlation between net interest income and noninterest income through cross-sectional correlation and
bank-specific correlation. Furthermore, we establish econometric models to analyze the impact of noninterest income on bank
risk and return.
At the aggregate level, in the Chinese banking industry, the net operating income's volatility tended to decline during 1986 and
2008. In particular, the volatility of net interest income, the volatility of noninterest income, and their covariance decrease.
-1
0
1
2
3
4
C
C
B
I
C
B
C
A
B
C
B
O
C
O
M
H
X
B
S
P
D
B
S
D
B
C
E
B
C
M
B
G
D
B
C
N
C
B
C
I
B
C
M
B
C
B
O
C
E
B
0
5
10
15
20
25
Sharp Ratio (Left Axis) Z-score (Right Axis)
Fig. 9. Relationship between risk measures and noninterest income share.
Table 5
Noninterest income shares as determinants of bank risk and return.
Dependent variable
Net income growth Return on equity (ROE) Z-score
Mean Std. dev. Mean Std. dev. Sharp ratio
Log assets (ln(A)) 0.017
(0.226)
0.045
(0.249)
0.080
(0.559)
0.231
(0.365)
0.119
(0.249)
1.556
(0.384)
Debt/assets
D
A
3.086
(0.573)
11.532
(1.116)
15.614
(1.625)
58.324
(2.653)
89.443
(3.658)
365.782
(1.557)
Growth in assets (dln(A)) 1.902
(3.122)
1.315
(0.894)
1.505
(1.290)
7.342
(1.428)
3.943
(1.017)
40.471
(1.234)
NONSH 0.053
(0.133)
0.627
(0.660)
0.770
(1.022)
2.457
(0.740)
3.280
(1.309)
24.567
(1.159)
FEESH 0.040
(0.113)
0.165
(0.193)
0.983
(1.449)
4.313
(1.443)
2.227
(0.987)
2.633
(0.138)
EXCHSH 0.297
(0.355)
0.304
(0.150)
0.456
(0.285)
0.385
(0.055)
1.280
(0.240)
36.739
(0.815)
OTHERSH 0.221
(1.017)
0.107
(0.205)
1.062
(2.557)
4.075
(2.227)
3.518
(2.550)
8.326
(0.713)
MULTI 0.223
(1.349)
0.760
(1.909)
0.011
(0.036)
0.098
(0.071)
2.015
(1.920)
6.841
(0.770)
Constant 0.190
(0.277)
0.810
(0.489)
1.158
(0.882)
3.409
(0.590)
2.543
(0.583)
7.699
(0.209)
Adjusted R
2
0.451 0.061 0.173 0.104 0.560 0.058
Jt. sig.: component shares of noninterest income 0.757 0.982 0.156 0.207 0.174 0.740
No. obs. 15 15 15 15 15 15
Notes: (1) t-value in parentheses; (2) ***, **, and * indicate statistical significance at the 99%, 95%, and 90% levels, respectively; (3) Jt. sig. reports the p-values from
the F-test of the joint significance, where the hypothesis of the joint test is that the coefficient of dependent variable is not significant.
163 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
However, the decrease in the volatility of the net operating income is mainly subject to the decrease of the net interest income
and the decrease of the covariance between the net interest income and noninterest income. The decline of the covariance
between the net interest income and noninterest income indicates that the development of noninterest income in the Chinese
banking industry will bring diversification benefits. Although the noninterest income volatility has declined compared to 1987
1999, when the variance of noninterest income was less than that of net interest income, the variance of noninterest income was
much larger than the variance of net interest income during 20002008. Compared to the sharp decrease in the contribution of
the net interest income to the variance of the net operation income, the decrease in the contribution of noninterest income to the
variance of net operating income is not significant, indicating that as the volatility of net interest income decreases, noninterest
income may gradually replace net interest income as the main cause of the volatility of operating income. We also find that,
compared to the net interest income, noninterest income show obvious cyclicality. That is, an increase in noninterest income may
extend, rather than smooth, the impact of macroeconomic fluctuations on bank revenue. At the aggregate level, our conclusion is
different fromthe finding of Stiroh (2004). Although we also find that the declining volatility of net operating revenue reflects the
reduced volatility of net interest income, we report that it reflects the reduced covariance between net interest income and
noninterest income, which indicates to some extent that the development of noninterest income in the Chinese banking industry
will bring diversification benefits. Stiroh (2004) finds that the covariance between net interest income and noninterest income
increases during the sample period and concludes that at the aggregate level, the shift from traditional business activities to
nontraditional business activities does not bring diversification benefits in the U.S. banking industry.
At the bank level, the cross-sectional correlation and bank-specific correlation between net interest income and noninterest
income show that, in the Chinese banking industry, noninterest income growth has diversification benefits. However, with the
increase in shares of noninterest income, the marginal revenue brought about by diversification gradually declines. Further
analysis shows that the impact of noninterest income growth on bank risk and return is not significant. However, we should note
that an increase in noninterest income may lead to an increase in bank risk, albeit not with higher returns. At the bank level, our
conclusion is also different from the findings of Stiroh (2004). Stiroh (2004) reports that the growth rates of net interest income
and noninterest income become more correlated during a sample period; however, we find that the correlation between the two
growth rates becomes small, which reflects diversification benefits of the noninterest income growth in the Chinese banking
industry.
This study shows that in the Chinese banking industry, an increase in noninterest income has diversification benefits.
However, because noninterest income has higher volatility and cyclicality than net interest income, with the increase in shares of
noninterest income, the marginal revenue brought about by diversification gradually declines. Therefore, relying more on
noninterest income may worsen the risk/return trade-off.
Compared to the extant literature, our contribution is that there is a reasonable boundary of noninterest income growth.
Unreasonably higher shares of noninterest income may increase risks rather than bring profits. Only by considering the scale of
the bank's business development can we determine the optimal proportion of noninterest income.
References
Acharya, V.V., Hasan, I., Saunders, A., 2002. The Effects of Focus and Diversification on Bank Risk and Return: Evidence from Individual Bank Loan Portfolios.
Mimeo, New Jersey Institute of Technology, March 15.
Baele, Lievan, de Jonghe, Olivier, Vander Vennet, Rudi, 2007. Does the stock market value bank diversification? J. Bank. Financ. 31, 19992023.
Boyd, John H., Graham, S.L., 1988. The profitability and risk effects of allowing bank holding companies to merge with other financial firms: a simulation study.
Quarterly Review, 12(2). Federal Reserve Bank of Minneapolis, pp. 320 (Spring).
Boyd, John H., Graham, S.L., Hewitt, R.S., 1993. Bank holding company mergers with nonbank financial firms: effects on the risk of failure. J. Bank. Financ. 17, 4363.
Brewer, E., 1989. Relationship between bank holding company risk and non-bank activity. J. Econ. Bus. 1 (4), 337353 (November).
Busch, Ramona, Kick, Thomas, 2009. Income diversification in the German banking industry. Working Paper (www.bundesbank.de/download/bankenaufsicht/
dkp/200909dkpb_.pdf).
Canals, J., 1993. Competitive Strategies in European Banking. Oxford University Press, Oxford.
Chiorazzo, V., Milani, C., Salvini, F., 2008. Income diversification and bank performance: evidence from Italian banks. J. Financ. Serv. Res. 33, 181203.
DeYoung, R., Rice, T., 2004. Noninterest income and financial performance at U.S. Commercial Banks. Financ. Rev. 39 (1), 101127.
DeYoung, Robert, Roland, Karin P., 2001. Product mix and earnings volatility at commercial banks: evidence from a degree of total leverage model. J. Financ.
Intermed. 10, 5484.
Eisemann, P., 1976. Diversification and the congeneric bank holding company. J. Bank Res. 7 (1), 6877 (Spring).
Esho, N., Kofmann, P., Sharpe, I.G., 2005. Diversification, fee income, and credit union risk. J. Financ. Serv. Res. 27 (3), 259281.
Gallo, J., Apilado, V., Kolari, J., 1996. Commercial bank mutual fund activities: implications for bank risk and profitability. J. Bank. Financ. 20 (10), 1,7751,792 (December).
Kwast, Myron, 1989. The impact of underwriting and dealing on bank returns and risk. J. Bank. Financ. 13, 101125.
Lepetit, L., Nys, E.C., Rous, P., Tarazi, A., 2008. The expansion of services in European banking: implications for loan pricing and interest margins. J. Bank. Financ. 32,
23252335.
Lown, Cara S., Osler, Carol L., Strahan, Philip E., Sufi, Amir, 2000. The changing landscape of the financial service industry: what lies ahead? Economic Policy
Review, 6(4). Federal Reserve Bank of New York, pp. 3954.
Mercieca, S., Schaeck, K., Wolfe, S., 2007. Small European banks: benefits from diversification? J. Bank. Financ. 31, 19751998.
Radecki, L.J., 1999. Banks' payments-driven revenue. FRBNY Econ. Policy Rev. 5369.
Rose, Peter S., 1989. Diversification of the banking firm. Financ. Rev. 24 (2), 251280.
Rosen, Richard J., Lloyd-Davies, Peter R., Kwast, Myron L., Humphrey, David B., 1989. A portfolio analysis of bank investment in real estate. J. Bank. Financ. 13 (3), 355366.
Santomero, Anthony W., Chung, Eek-june, 1992. Evidence in support of broader banking powers. Financ. Mark. Inst. Instrum. 1 (1), 169.
Saunders, Anthony, Walters, Ingo, 1994. Universal Banking in the United States: What Could We Gain? What Could We Lose? Oxford University Press, New York, NY.
Smith, R., Staikouras, C., Wood, G., 2003a. Non-interest income and total income stability. Working Paper (www.bankofengland.co.uk/publications/
workingpapers/wp198.pdf).
Smith, R., Staikouras, C., Wood, G., 2003b. Non-interest income and total income stability. Working Paper, No. 198. Bank of England.
Stiroh, Kevin J., 2004. Diversification in banking: is noninterest income the answer? J. Money Credit Bank. 36 (5), 853882.
164 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165
Stiroh, Kevin J., 2006. New evidence on the determinants of bank risk. J. Financ. Serv. Res. 30, 237263.
Stiroh, Kevin J., Rumble, A., 2006. The dark side of diversification: the case of US financial holding companies. J. Bank. Financ. 30, 21312161.
Templeton, William K., Severiens, Jacobus T., 1992. The effect of nonbank diversification on bank holding companies. Q. J. Bus. Econ. 31 (4), 316.
Zhou, Haowen, Wang, Jing, 2008. A review of the fluctuation of non-interest income of Chinese commercial banks from the perspective of portfolio theory. Econ.
Surv. 4, 155158.
165 L. Li, Y. Zhang / Journal of Empirical Finance 24 (2013) 151165