March 2008 working paper no. 330 Issn 1975-5163 Joon-Ho Hahm

3. Non-Interest Income Ratio and Bank Profitability and Risks 
Now we turn to our main empirical question of whether diversification and 
expansion toward non-interest income contributes to a higher and more stable profit for 
commercial banks. In order to answer this question, we adopt four distinct dependent 
variables: return on asset (ROA) as a measure of profitability, standard deviation of ROA 
as a measure of profit variability, equity-asset ratio as a measure of capital adequacy 
soundness, and z-score as a measure of bank insolvency risk. 
The z-score is constructed by dividing the sum of average ROA and average 
equity-asset ratio by standard deviation of ROA. It can be understood as a measure of 
insolvency risk since it represents the number of standard deviations that negate mean 
earning plus capital. It evaluates a bank’s capital and profit-generating capacity against its 
risk-taking level. A lower z-score value indicates that the bank is exposed to higher 
insolvency risk. 
Note that, in actual estimations, we need to define intervals during which, the 
average and standard deviation variables are constructed. Our first approach is to exploit 
three distinct sub-periods: 1992-96, 1997-2001, and 2002-06. For each bank, we 
computed sub-period mean ROAs, standard deviations of ROA, mean equity-asset ratios, 
and z-scores. We also computed the sub-period mean values for bank specific factors as 
well as macroeconomic variables, and estimated the following regressions: 

---

18 
t
i
i
t
t
i
t
i
t
i
t
i
CT
PD
MCF
BSF
NIIR
BPF
,
,
,
,
,
'
'













(2-1) 
t
i
i
t
t
i
t
i
t
i
t
i
CT
PD
MCF
BSF
NIIR
BPF
,
1
,
1
,
1
,
,
'
'
















(2-2) 
i
i
i
i
i
i
CT
MCF
BSF
NIIR
BPF











'
'
(2-3) 
Where, BPF denotes bank performance variables including ROA, standard deviation of 
ROA, equity-asset ratio, and z-score. NIIR is non-interest income ratio of individual banks, 
and BSF and MCF are vectors of bank specific and macroeconomic variables as before. 
PD and CT are period and country dummy variables respectively. 
The regressions are estimated using the pooled OLS method. Note that, while we 
use contemporaneous explanatory variables in (2-1), we lag explanatory variables by one 
sub-period in (2-2) to avoid potential endogeneity problems. Note also that, in regression 
(2-3), we estimate the simple cross-section regressions across banks using the mean and 
standard deviation variables computed over the whole sample period of 1992-2006. 
The bank-sub-period regression estimation results on (2-1) are summarized in 
Table 3. Note that the table is organized by three different sets of regressions – first, bank 
specific variables only, second, with macroeconomic variables included, and third, by 
adding an interaction variable between the stock market capitalization to GDP ratio and 
non-interest income ratio. This interaction variable should capture any non-linear 
accelerating effect of non-interest income through capital markets. 
First, consider the regression results with bank specific variables only. Bank 

---

19 
profitability as measured by ROA is positively associated with non-interest income ratio 
after controlling for the effect of bank size, equity capital to asset ratio and loan ratio. 
Note also that the non-interest income ratio is significantly positively associated with 
profit variability as measured by standard deviation of ROA, while the bank asset size is 
negatively associated with standard deviation of ROA. The non-interest income ratio is 
also significantly positively associated with equity-asset ratio, while it has no impact on 
the z-score. In sum, the expansion toward non-interest income seems to enhance bank 
profitability and capital adequacy. However, since it increases profit variability as well, it 
has little impact on bank insolvency risk. 
When we include macroeconomic variables, non-interest income becomes 
insignificant in the ROA regression, while it remains significant in the ROA standard 
deviation and equity-asset ratio regressions. It is interesting to note that real GDP growth 
and stock market capitalization are positively associated with bank ROA, while inflation 
and real interest rates are negatively associated with ROA. As for the profit variability, 
Real GDP growth is negatively associated with standard deviation of ROA, while inflation 
and real interest rates are positively related to the standard deviation of ROA. Hence, 
economic growth seems to favorably shift banks’ return-risk tradeoff, while higher 
inflation shifts the tradeoff negatively. When we include the interaction dummy, non-
interest income becomes insignificant while the interaction term becomes significant. 
In sum, the evidence in this table indicates that the expansion toward non-interest 
income tends to increase profit variability rather than stabilizing bank revenue. However, 

---

20 
since the expansion also raises equity-asset ratio and possibly ROA of banks, there seems 
to be no significant impact on the insolvency risk of banks as measured by z-score. 
Moreover, in the channel through which non-interest income tends to be associated with 
higher profit variability, stock market developments seem to play an important role. 
Regression results with explanatory variables lagged one period are summarized in 
Table 4. In general they are consistent with the results reported in Table 3 despite the fact 
that we lost almost a third of observations. Two major differences are that the interaction 
term now becomes insignificant, and that the size of bank assets tends to be negatively 
associated with z-score. Hence, the non-linear effect of non-interest income through stock 
markets discovered above should be interpreted with due caution. 
Table 5 reports estimation results for the cross-section regressions across banks 
based upon whole period average and standard deviation variables. In general, the 
evidence is consistent with our previous findings. The non-interest income ratio is 
significantly positively associated with ROA, standard deviation of ROA and equity-asset 
ratio even in the presence of macroeconomic controlling variables. The regression also 
shows a non-linear, positive correlation between the non-interest income ratio and ROA 
variabilitiy potentially through stock markets. Note again, since the non-interest income 
raises both the numerator (ROA and equity-asset ratio) and the denominator (standard 
deviation of ROA), it has no significant impact on z-score of banks. 

---

21