Commodity prices and bank lending in low-income countries-Isha.… · Commodity prices and bank...
Transcript of Commodity prices and bank lending in low-income countries-Isha.… · Commodity prices and bank...
Commodity prices and bank lending
in low-income countries∗
Isha Agarwal† Rupa Duttagupta‡ Andrea F. Presbitero§
August 15, 2016
Preliminary and incomplete draft – Comments welcome
Abstract: The recent sharp decline in global commodity prices since mid-2014 has hit manydeveloping countries hard. While the direct effect from commodity price changes to exportingfirms’ profitability and real GDP growth is well understood, prospects of additional strains tothe real economy through a deterioration in bank health and decline in intermediation are lessexamined, particularly in low-income countries (LICs). This paper uses data on more than 900banks for a sample of 46 LICs and quantifies the transmission of commodity price fluctuationsthrough the bank lending channel. Our results show that a reduction of commodity pricesreduces bank lending by domestic bank in commodity-dependent countries, and this effectis stronger for weakly capitalized banks and banks with low retail funding. For a typicalcommodity-dependent LIC, with commodity exports to total exports around 15 percent, a dropin commodity prices in the order of what registered in 2015 reduces banking lending by weaklycapitalized banks by an additional 2 percentage points.
JEL Codes: G01; G21; J23; J63.
Keywords: Bank lending; Commodity prices; Bank capital; Low-income countries.
∗This research is part of a project on Macroeconomic Research in Low-Income Countries (project id: 60925) sup-ported by the U.K.’s Department for International Development. This paper should not be reported as representingthe views of the IMF or of DFID. The views expressed in this paper are those of the author and do not necessarilyrepresent those of the IMF, IMF policy, or of the DFID. We wish to thank Christopher Adam, Maria Soledad Mar-tinez Peria, Camelia Minoiu and participants at a seminar at the IMF for helpful comments and suggestions. Wealso thank Bertrand Gruss for kindly sharing his data on country-specific terms-of-trade. The usual disclaimersapply.†International Monetary Fund and Cornell University. E-mail: [email protected].‡International Monetary Fund. E-mail: [email protected].§International Monetary Fund and MoFiR. E-mail: [email protected].
1 Introduction
Developing countries have traditionally been extremely vulnerable to adverse exogenous shocks,
with severe consequences in terms of output growth and macroeconomic and political stabil-
ity (Deaton, 1999; Loayza et al., 2007; Raddatz, 2007; Brückner and Ciccone, 2010; Dabla-Norris
and Bal Gunduz, 2014; Bazzi and Blattman, 2014). While developing countries’ resilience to
economic crises has increased since the 2000s—mostly because of good policies and a lower
incidence of external and domestic shocks (Abiad et al., 2015)—the recent sharp fall in com-
modity prices and the global slowdown are again putting several developing economies un-
der stress (International Monetary Fund, 2015): growth in commodity exporting developing
countries has slowed significantly from 6 percent to 0.8 percent between 2013 and 2016, while
growth in diversified exporters remained stable around 6.3 percent.
In particular, commodity prices and terms of trade volatility have been associated with
lower GDP growth and output fluctuations (Mendoza, 1997; Bleaney and Greenaway, 2001).
The impact of external shocks is likely to be higher on lower-income developing countries
given that challenges of low diversification and poor institutional capacity tend to get exac-
erbated at lower income levels. In fact, looking at the historical experience of 40 low-income
countries (LICs), Raddatz (2007) show that, even if most of the variance in real per capita GDP
is explained by domestic factors, the effect of external shocks is economically meaningful and,
among external shocks, changes in commodity prices are the most important source of output
fluctuations. Indeed, Figure 1 (left-hand side panel) shows a strong co-movement between real
GDP growth and changes in commodity prices in a sample of commodity exporting LICs.
Given the high dependence on commodity exports of many low-income countries—the
share of commodities in total exports in the average low-income country in our sample is 55
percent—changes in commodity prices can directly affect output through a worsening of the
current account and of the fiscal balance. Domestic exporting firms could be under stress be-
cause of lower export revenues and they could reduce their demand for credit. This is the
demand channel depicted in Figure 2. A second, indirect, channel through which low com-
modity prices could dampen economic activity is through the contraction in the supply of
credit to the economy. This channel has been so far overlooked by the literature.1 However,
stylized facts show a strong association between the evolution of commodity prices and the
ratio of credit to the private sector over GDP (Figure 1, right-hand side panel), especially for
1To the best of our knowledge, the only exception is a recent work by Kinda et al. (2016), who look at theeffect of commodity prices on financial fragility in a sample of emerging and developing countries and show thatnegative commodity price shocks are associated with higher non-performing loans, lower profitability, and a higherlikelihood of banking crises.
2
commodity exporters, where credit growth slowed down after the 2008 reduction in commod-
ity prices. Intuitively, a fall in commodity prices can lead to lower credit growth through
a deterioration of bank health: lower deposits due to a contraction in export revenues and
higher unemployment could generate a funding shock to domestic banks, coupled with gov-
ernment arrears and weak revenue growth of commodity dependent firms, which may render
them unable to service their loans, thereby worsening asset quality and reducing bank capital.
As a result, bank lending to the economy will slow down—this is the supply channel shown in
Figure 2. Consistent with the findings of Kinda et al. (2016), the evidence based on our sample
of banks shows that non-performing loans indeed increase in response to a worsening of com-
modity prices in low-income countries particularly dependent on commodity exports, while
bank profitability and bank capital contract.
This paper focuses on the supply channel of the transmission of commodity prices to bank
lending in LICs, as a mechanism through which the commodity cycle can amplify the business
cycle over and above the direct effect of commodity prices on GDP growth. In particular, we
are interested in the heterogeneous response of banks to commodity price changes, depending
on their exposure to the price shock and on their characteristics (ownership and health).
Since there is wide variation in the commodity basket exported by LICs, a change in com-
modity prices of the same magnitude can have significantly different effects on the terms of
trade of different countries depending on the importance of that commodity in the country’s
total exports. It is important for our analysis to capture this variation since the extent to which
the banking system is affected by movements in commodity prices would depend on how
strongly the shock permeates to the revenues of the commodity exporters, which is contingent
on the share of different commodities in total exports. We measure commodity price changes
with a country-specific commodity price index and the bank-specific exposure to commodity
prices by the share of commodity exports over GDP of the country where the bank is located.
The assumption is that fluctuations in commodity price will equally affect all banks in a given
country: while this assumption may look quite strong, we show that our measure is highly
correlated with a bank-specific measure based on bank lending in the syndicate market.
The transmission of commodity price changes to lending could be dampened by the high
level of profitability and capitalization of the banking system in low-income countries, reflect-
ing impediments to market competition (Beck et al., 2010; Ghosh, 2016)—the median regulatory
capital to risk weighted assets ratio for low-income countries since 2008 has been above 20.2
Also, foreign banks can insulate themselves against funding shocks and capital shortfalls more
2For instance, Abuka et al. (2015) use loan level data on Uganda to show that the transmission of monetarypolicy to the supply of credit is weaker for better capitalized banks.
3
than domestic banks, since they can rely on the internal capital market and on the support from
parent banks. Therefore, we exploit the heterogeneity across banks to inspect whether the im-
pact of swings in commodity prices on lending is only a function of their exposure to the shock
or whether bank ownership and health also matter for the transmission.
Our results show that commodity prices significantly affect bank lending only for weakly
capitalized domestic banks. There is no evidence that commodity prices are associated with
bank lending across the board, after controlling for bank characteristics and a standard set of
macroeconomic indicators which should proxy for changes in demand for credit. However,
the effect of commodity prices on lending is increasing with the exposure of banks to com-
modity prices and it becomes positive and significant for domestic banks which have low retail
funding and are weakly capitalized. In economic terms, in a commodity-dependent country
like Nigeria, one standard deviation reduction in commodity prices—which corresponds to
roughly half of the reduction in the commodity price index between 2008 and 2009—is asso-
ciated with a 3 percentage point contraction in loan growth for a weakly capitalized domestic
bank, a relatively large effect given that the average loan growth for weakly capitalized banks
in the sample is 7.4 percent.
2 Data and stylized facts
2.1 Data
For the empirical analysis, we use bank-level data from Bankscope, a global database of banks’
financial statements which covers about 90 percent of the total assets of the banking system in
each country. The sample is constrained by the availability of bank-level data in Bankscope; in
particular, we limit our analysis to countries with data for at least 5 active banks in any year.
As a result, our sample consists of 46 low-income countries with 928 active banks for the time
period 2004-2015.3 We use the following bank-level variables from Bankscope: loan growth,
the ratio of equity over assets, retail funding, liquidity, size, returns-on-assets, and the ratio of
non-performing loans to gross loans.4
Table 1 presents the summary statistics for loan growth by bank characteristics. The av-
erage loan growth for the sample is 27.4% with a very high variation across banks—the loan
growth of the upper quartile is approximately five times as high as the loan growth of the low-
est quartile. There are also interesting differences in loan growth across bank characteristics.
Banks with high capital ratios (equity to assets) have higher average loan growth than banks
3See Table A1 in the Appendix for the list of countries and the number of banks in the sample.4The definition of each variable is in Table A2; all variables are winsorized at the 5th and 95th percentile to
remove outliers.
4
with low capital ratios. The average loan growth of small banks is almost twice as much as
loan growth of large banks. This could be a artifact of the low competition in the banking
system in less developed countries.
In the spirit of Deaton and Miller (1995) and Bazzi and Blattman (2014), we consider a
country-specific measure of commodity prices based on 33 commodities. More specifically, we
use the commodity price index constructed by Gruss (2014) as follows:
PRICE INDEX =J
∑j=1
∆Pj,tωi,j,τ (1)
where, ∆Pj,t is the logarithm of the relative price of commodity j in period t within year τ and
the weights are predetermined and calculated as a three-year average:
ωi,j,τ =13
3
∑s=1
xi,j,τ−s
∑Jj=1 xi,j,τ−s
(2)
The key advantage of such a measure, compared to a more standard commodity price index,
is that, being country-specific, it can take into account that prices of different commodities
have been moving quite differently in past years, so that not all countries have been equally
hit by the slowdown in commodity prices. Also, having predetermined weights, rather than
fixed ones, takes into consideration the rapid change in the composition of export products in
several LICs.
Data on macroeconomic variables—GDP, domestic interest rates, and inflation—are from
the IMF’s World Economic Outlook and International Financial Statistics Database. See Table
A2 for variable definitions and data sources.
2.2 Stylized Facts
The banking sector in low-income countries is highly concentrated (Beck et al., 2010; Ghosh,
2016). In our sample, the median asset share of the three largest banks was more than 80
percent in 2004 and has not declined significantly over the years. The banking landscape is
a mix of domestic and foreign banks, with the former accounting for 71.5 percent of the total
number of banks in the sample, and an asset share of 78 percent.5
Loan growth in LICs tends to be sensitive to movements in commodity prices, especially
for commodity exporters, which witnessed a fall in their loan growth of approximately 10
percentage points following the commodity price bust of 2008 and are still experiencing per-
sistent declines in loan growth since the second round of commodity price bust began in 2011
and gathered pace in mid-2014 (Figure 3, panel a).
5These numbers are in line with what reported by Claessens and Van Horen (2014).
5
There also seems to be some evidence that commodity prices affect bank profitability in
commodity exporting countries. Figure 3 (panel b) shows that the median bank profitability
in commodity exporting countries has been falling since the recent commodity price bust and
has declined even more sharply since 2014 while the bank profitability for diversified exporters
has largely remained unchanged.
The effect of commodity price movements on bank lending can differ across banks de-
pending on how well capitalized the bank is and how vulnerable the bank is to funding and
liquidity dry-up in the event of shocks. In our sample of low-income countries, the median
retail funding ratio for commodity exporters has been close to 90 percent. It declined some-
what following the 2008 episode of fall in commodity prices and has declined significantly
since 2011 (Figure 3, panel c) suggesting that banks may become vulnerable to funding shocks
following commodity price busts. However, banks in diversified exporters did not experience
a similar decline in retail funding. Similarly, Figure 3 (panel d) shows that the equity-to-assets
ratio in commodity exporting countries witnessed a sharp decline in 2008 and 2014 while it
remained stable for diversified exporters.
2.3 Commodity prices and bank health
While the stylized facts discussed in the previous section point towards an underlying theme
which suggests that changes in commodity prices may have an effect on the financial sector
stability in commodity exporting LICs, which may eventually impact credit growth, we con-
duct an empirical analysis to corroborate the first step in the transmission process. To formally
assess the impact of commodity prices on the health of the banking system, we estimate a set
of simple panel regressions in which a set of bank health indicators—equity over assets, the
return on assets, and non-performing loans to gross loans—are function of the lagged price
shock and its interaction with a measure of bank exposure to commodity price changes.6 The
inclusion of GDP growth and interest rates accounts for macroeconomic factors which may
affect bank performance, while bank and year fixed effects absorb bank-specific unobserved
heterogeneity and the effect of global shocks on bank performance.
In line with what is discussed by Kinda et al. (2016), we find that commodity prices are
not correlated with bank health indicators in the overall sample, but they do have a signifi-
cant association with bank health depending on the how exposed the bank is to commodity
price movements (Table 2). In particular, we find that commodity prices are positively asso-
ciated with equity and return on assets and negatively with non-performing loans. Having
6We measure the exposure of the bank to commodity price shocks by the share of commodity exports in totalexports of the country in which it is located.
6
established that commodity prices have an impact on financial health of banks which are more
exposed to price changes, we now explore the second link in the transmission channel—the
effect of commodity prices on bank lending.
3 Empirics
The key problem we face when estimating the effect of commodity prices on bank lending is
disentangling demand from supply effects. The estimated coefficient on the commodity price
index, in fact, will capture both a change in the supply of bank lending due to changes in
prices, but also a shift in demand for credit resulting from a change in prices. For instance,
a sharp improvement in the terms of trade would induce an expansion of economic activity
of exporters, which will demand more bank credit, but it could also improve bank asset qual-
ity (e.g. through a reduction in NPLs due to better economic conditions), so that banks will
expand their balance sheets.
Our strategy to identify the supply effect of commodity prices on bank lending relies on
the differential exposure of banks in LICs to variations in commodity prices. In particular,
we assume that banks located in commodity-dependent countries are the ones affected by the
shock, while the ones located in diversified exporters are the non-affected ones. More precisely,
we adopt a continuous measure of exposure, which is the time-varying country-specific ratio
of commodity exports over GDP (see Figure 4). For the median bank, the measure of exposure
is around 6 percent, but the exposure varies substantially across banks—the third quartile is
above 15 percent and, on average, commodity exports account for more than 20 percent of
GDP in a number of LICs (e.g. Bolivia, Congo, Mauritania, Mongolia, Nigeria, Tajikistan,
Vietnam, Yemen and Zambia). The drawback of such a measure is that it is country-specific,
rather than bank-specific, so that it relies on the assumption that all banks in the same country
will be equally affected by variations in commodity prices. Controlling for a set of bank fixed-
effects and time-varying bank characteristics should take into account different exposure of
different banks to the exporting sector, mitigating concerns that our measure pools together
banks differently exposed to price changes (for instance, large and foreign banks are generally
more likely to deal with large exporting companies). For a subset of banks (less than 20 percent
of the sample) we are able to match banks’ balance sheets from Bankscope with their lending
activity in the syndicated loan markets through Dealogic and compute a time-invariant bank-
specific measure of exposure to commodity sectors. While the limited sample size prevents us
form using this measure in the empirical analysis, we can observe that its correlation with the
country-specific measure of exposure is as high as 70 percent.
7
Therefore, we look at the response of bank lending to changes in commodity prices by
estimating the following model, based on the traditional specifications used to estimate the
reaction of bank lending to monetary policy shocks (Kashyap and Stein, 1995; Gambacorta,
2005; Gambacorta and Marques-Ibanez, 2011; Aiyar et al., 2016):
∆LOANSbct = αPRICE INDEXct−1 + βPRICE INDEXct−1 × EXPOSUREbt−1 +
+2
∑i=1
COUNTRYict−1 +
4
∑j=1
BANK jbt−1 + δb + τt + εbct (3)
where ∆LOANSbct is the growth rate of outstanding loans of bank b, located in country c, in
year t; PRICE INDEXct−1 is the country-specific commodity price index presented in equation
1 for country c in the previous year (t − 1); EXPOSUREbt−1 is the lagged exposure of bank b
to commodity prices, as discussed above; COUNTRYct−1 is a set of country-specific control
variables including real GDP growth and the logarithm of domestic interest rates; BANKbt is a
set of time-varying bank-specific controls, lagged one period, including measures of liquidity,
size, capitalization, and reliance on retail funding; δb and τt are bank and year fixed effects; and
εbct is the standard error term.7 Bank fixed effects control for the possibility that a systematic
matching between banks and firms confound the identification of the effect of commodity
prices. For instance, there is a consistent literature showing that large banks often lend to
large firms (Berger and Udell, 2002) and that the latter are more likely to be exporters (see, for
instance, Rankin et al., 2006, for evidence on Africa) and affected by commodity prices. Hence,
the change in prices would affect the demand for credit, rather than the supply.
4 Results
4.1 Baseline
Table 3 presents the results from our baseline model, estimated on a common sample of 584
banks in 40 LICs, because of limited data availability for some control variables. Column 1
includes year fixed effects and no controls. In column 2 and 3, we add country and bank fixed
effects respectively. Column 4 includes the bank-level controls in addition to the bank and
time fixed effects, column 5 adds country-level controls. Finally, in column 6, we add our main
variable of interest, which is the interaction between the lagged value of the commodity price
index and the measure of exposure to commodity exports of the country in which the bank is
located.
Our results indicate that there is a positive, unconditional, association between commod-
7Compared to the traditional literature cited above, our data are at annual frequency, rather the quarterly, so thatwe simply take all the explanatory variables lagged one year, rather than including a more complex lag structure.
8
ity prices and bank lending in LICs (column 1), which becomes smaller once we control for
country- and bank-specific unobserved heterogeneity. According to the estimate reported in
column 3, one standard deviation decline of the commodity price index translate into 1.6 per-
centage point fall in loan growth. While for the overall sample, 1.6 percentage point fall in
loan growth may not seem important, it may be quite significant for select countries, given the
wide variation in loan growth across the sample and the diverse characteristics of the banks
in low-income countries. However, the effect of commodity prices on loan growth becomes
not significantly different from zero once we include bank characteristics and macroeconomic
factors, which may affect bank lending through the the demand channel (column 4 and 5).
The results reported in the last column, instead, show that the effect of commodity prices
on loan growth depends on the intensity of banks’ exposure to commodity exports. The coeffi-
cient on the interaction term PRICE INDEX × EXPOSURE is significant and positive, which
implies that banks which are located in countries more exposed to commodity price shocks
curtail lending in response to a fall in commodity prices over and above what is explained by
a fall in demand for credit. While the coefficient on the interaction term is statistically signif-
icant, in economic terms the effect of commodity prices on loan growth is still not significant
for a bank with the average exposure, and it becomes positive (but still relatively small) for a
bank at the 90th percentile of the exposure distribution (i.e. a bank located in a country where
commodity exports account for almost 27 percent of GDP).
The dependence of the effect of commodity prices on banks’ exposure could be the result
of the credit supply channel that operates through the link between falling commodity prices
and their impact on bank health. In this respect, the effect is likely to be stronger for weaker
banks and for domestic ones. Thus, in the next section we deepen our analysis to explore the
differential effect of commodity prices on bank lending across bank characteristics.
4.2 Bank heterogeneity
To understand the differences in the transmission of commodity prices to lending across bank
characteristics, we focus on bank ownership (foreign vs domestic)8, bank size—measured by
the logarithm of total assets—and two measures of bank health—retail funding and bank
capital— partly driven by our findings on the the impact of commodity prices on bank health
in section 2.3. When considering bank size and health, we divide the banks into high and low
groups, where low refers to the banks in the lowest quartile of the bank-characteristic in ques-
tion while high refers to the remaining banks. We enrich the model in equation 3 by splitting
8We classify a bank as a foreign bank if the country code of the global ultimate owner of the bank is differentfrom the country code where the bank operates.
9
the coefficient on the interaction term PRICE INDEX × EXPOSURE between the two groups
of banks (domestic and foreign, and low and high size and health), as follows:
∆LOANSbct = αPRICE INDEXct−1 + βPRICE INDEXct−1 × EXPOSUREbt−1 +
γhPRICE INDEXct−1 × EXPOSUREbt−1 × BANKHIGHbt−1
γl PRICE INDEXct−1 × EXPOSUREbt−1 × BANKLOWbt−1
+2
∑i=1
COUNTRYct−1 +4
∑j=1
BANKbt−1 + δb + τt + εbct
In this model, we are interested in the coefficients γh and γl , which quantify the differential
effect of commodity prices on lending across bank characteristics.
Table 4 summarizes the results. Column 1 duplicates the results from column 6 of Table 3 to
facilitate comparison with the baseline specification. In column 2, we look at the effect across
domestic and foreign banks; column 3, 4 and 5 show the results for bank size, equity and retail
funding respectively. All columns have bank and year fixed effects, in addition to bank and
country-level controls.
The results unveil interesting asymmetries in the transmission process. With respect to
ownership, we find that credit growth is not affected by changes in commodity prices for for-
eign banks while it is positively associated with commodity prices for domestic banks (column
2). This finding is consistent with the hypothesis laid out in the section 1 which suggests that
foreign banks may be better able to insulate themselves from funding shocks and capital short-
falls in the aftermath of commodity price bust as they can rely on parent banks for funding and
hence, may not curtail lending. Domestic banks, on the other hand, are more likely to experi-
ence difficulties in access to cheap funding and deteriorating capital base as commodity prices
fall and overall economic conditions worsen and may end up reducing the supply of credit.
Compared to the results for the overall sample, the effect of commodity prices on lending is
already positive for banks with an exposure at the 75th percentile (i.e. commodity exports
accounting for about 18 percent of GDP).
We also find some evidence of differential effects of commodity prices across bank size, eq-
uity and retail funding, conditional on the exposure to commodity exports. Large banks and
banks with high retail funding ratios do not seem to be affected by commodity price move-
ments. (column 4 and 6). Loan growth for banks in the high and low capital group is posi-
tively associated with commodity prices, however, the correlation is much stronger for weakly
capitalized banks as compared to better capitalized banks (column 5).
Given that we find no effect of commodity prices on lending for foreign banks, we fo-
cus our attention to the sample with only domestic banks (which constitute two-thirds of the
10
total banks in our sample) and explore the heterogeneity across bank characteristics in this
sub-sample. The results are presented in Table 5. We find that commodity price shocks are
positively associated with bank lending for domestic banks and the economic significance is
much larger as compared to the overall sample. For a typical commodity-dependent LIC, with
commodity exports to total exports around 11 percent, a one standard deviation drop in the
commodity price index reduces loan growth by an additional 0.8 percentage points. This effect
increases to 1.3 percentage points for domestic banks located in countries where commodity
exports account for 18 percent of GDP.
In terms of differential effects across bank characteristics, we find that bank size does not
matter for the transmission process while retail funding and capital are important channels of
transmission (column 4 and 5). In particular, weakly capitalized banks and banks with low
retail funding exacerbate the impact of commodity price movement on bank lending. In eco-
nomic terms, in a commodity-dependent country like Nigeria, one standard deviation reduc-
tion in commodity prices—which corresponds to roughly half of the reduction in the commod-
ity price index between 2008 and 2009—is associated with a 3.3 percentage point contraction
in loan growth for a weakly capitalized domestic bank, a relatively large effect given that the
average loan growth for weakly capitalized banks in the sample is 7.4 percent. At the extreme,
in a country like Republic of Congo, a similar reduction in commodity prices will reduce bank
lending by 4.4 percentage points for the average domestic banks and by 8 percentage points for
weakly capitalized banks. On the other hand, the effect of a similar fall in commodity prices
for banks in countries with a moderate exposure to commodity prices is much smaller, even
considering weakly capitalized banks (in a country like Senegal loan growth will slow down
by about one percentage point).
4.3 Robustness exercises
Our key results are robust to: 1) clustering the standard errors at the bank level, to allow for
serial correlation within banks, 2) restricting the sample to banks for which there are at least
4 yearly observations in the panel, and 3) controlling for inflation and the expected real GDP
growth—as projected by the WEO—rather than the actual one, to better measure expectations
and demand for credit.
5 Conclusions
This paper revisits the classic question of the impact of terms of trade movements on real activ-
ity in low-income countries and explores the role of the bank lending channel in the transmis-
11
sion mechanism. We add to the existing literature on commodity price movements and growth
in low-income countries on two fronts. First, we quantify the role of the credit supply chan-
nel in amplifying the effects of commodity prices on real activity, something which has been
overlooked in the existing literature, but can be quite important for LICs, many of which are
dependent on commodity exports and are potentially vulnerable to commodity price move-
ments, given that the traditional banking sector is the main source of firm financing. Sec-
ond, we integrate the literature on banking structure and the transmission of shocks through
the bank lending channel by looking at the differential effect of changes in commodity prices
across bank characteristics.
Our findings suggest that there is a positive, unconditional, association between commod-
ity prices and bank lending in LICs, which becomes smaller once we control for country- and
bank-specific unobserved heterogeneity. However, there is no evidence of any effect of com-
modity prices on loan growth once we include bank characteristics and macroeconomic fac-
tors, which may affect bank lending through the demand channel. Conditional on the intensity
of banks’ exposure to commodity exports, we do find a positive and significant association be-
tween commodity prices and lending which is not very large in terms of economic significance.
We also find evidence which suggests that domestic banks are more likely to curtail lend-
ing than foreign banks in the event of a commodity price bust and the effect is larger for
weakly capitalized banks and banks with low retail funding ratios. For a typical commodity-
dependent LIC, with commodity exports to total exports around 15 percent, a drop in com-
modity prices in the order of what registered in 2015 reduces bank lending by weakly capital-
ized banks by an additional 2 percentage points, which is not trivial, given that average loan
growth of weakly capitalized banks is 7.4%
Overall, our results suggest that the credit supply channel does not matter for the entire
sample of low-income countries but this channel can be potent for domestic banks which are
more exposed to commodity exports and have weak capital and funding positions. Given that
domestic banks account for two-thirds of the total number of banks in our sample, these find-
ings underscore the need to put macro-prudential polices in place which can make the banking
system more resilient to movements in commodity prices and prevent the amplification of the
credit cycle.
12
References
ABIAD, A., BLUEDORN, J., GUAJARDO, J. and TOPALOVA, P. (2015). The Rising Resilience of
Emerging Market and Developing Economies. World Development, 72, 1 – 26.
ABUKA, C., ALINDA, R. K., MINOIU, C., PRESBITERO, A. et al. (2015). Monetary Policy in a De-
veloping Country; Loan Applications and Real Effects. IMF Working Paper 15/270, International
Monetary Fund, Washington DC.
AIYAR, S., CALOMIRIS, C. W. and WIELADEK, T. (2016). How does credit supply respond to
monetary policy and bank minimum capital requirements? European Economic Review, 82,
142–165.
BAZZI, S. and BLATTMAN, C. (2014). Economic shocks and conflict: Evidence from commodity
prices. American Economic Journal: Macroeconomics, 6 (4), 1–38.
BECK, T., DEMIRGUC-KUNT, A. and LEVINE, R. (2010). Financial institutions and markets
across countries and over time: The updated financial development and structure database.
The World Bank Economic Review, 24 (1), 77–92.
BERGER, A. N. and UDELL, G. F. (2002). Small business credit availability and relationship
lending: The importance of bank organisational structure. The Economic Journal, 112 (477),
F32–F53.
BLEANEY, M. and GREENAWAY, D. (2001). The impact of terms of trade and real exchange rate
volatility on investment and growth in sub-Saharan Africa. Journal of Development Economics,
65 (2), 491–500.
BRÜCKNER, M. and CICCONE, A. (2010). International Commodity Prices, Growth and the
Outbreak of Civil War in Sub-Saharan Africa. The Economic Journal, 120 (544), 519–534.
CLAESSENS, S. and VAN HOREN, N. (2014). Foreign Banks: Trends and Impact. Journal of
Money, Credit and Banking, 46 (s1), 295–326.
DABLA-NORRIS, E. and BAL GUNDUZ, Y. (2014). Exogenous Shocks and Growth Crises in
Low-Income Countries: A Vulnerability Index. World Development, 59, 360 – 378.
DEATON, A. (1999). Commodity prices and growth in africa. Journal of Economic Perspectives,
13 (3), 23–40.
— and MILLER, R. I. (1995). International commodity prices, macroeconomic performance, and poli-
tics in Sub-Saharan Africa. International Finance Section, Department of Economics, Princeton
University.
GAMBACORTA, L. (2005). Inside the bank lending channel. European Economic Review, 49 (7),
1737–1759.
— and MARQUES-IBANEZ, D. (2011). The bank lending channel: lessons from the crisis. Eco-
13
nomic Policy, 26 (66), 135–182.
GHOSH, A. (2016). Banking sector globalization and bank performance: A comparative anal-
ysis of low income countries with emerging markets and advanced economies. Review of
Development Finance, forthcoming, –.
GRUSS, B. (2014). After the boom–commodity prices and economic growth in Latin America and the
Caribbean. IMF Working Paper 14/154, International Monetary Fund, Washington DC.
INTERNATIONAL MONETARY FUND (2015). World Economic Outlook – Adjusting to Lower Com-
modity Prices. Washington DC: International Monetary Fund.
KASHYAP, A. K. and STEIN, J. C. (1995). The impact of monetary policy on bank balance
sheets. In Carnegie-Rochester Conference Series on Public Policy, Elsevier, vol. 42, pp. 151–195.
KINDA, T., MLACHILA, M. et al. (2016). Commodity Price Shocks and Financial Sector Fragility.
IMF Working Paper 16/12, International Monetary Fund, Washington DC.
LOAYZA, N. V., RANCIÈRE, R., SERVÉN, L. and VENTURA, J. (2007). Macroeconomic Volatility
and Welfare in Developing Countries: An Introduction. The World Bank Economic Review,
21 (3), 343–357.
MENDOZA, E. G. (1997). Terms-of-trade uncertainty and economic growth. Journal of Develop-
ment Economics, 54 (2), 323–356.
RADDATZ, C. (2007). Are external shocks responsible for the instability of output in low-
income countries? Journal of Development Economics, 84 (1), 155–187.
RANKIN, N., SÖDERBOM, M. and TEAL, F. (2006). Exporting from Manufacturing Firms in
Sub-Saharan Africa. Journal of African Economies, 15 (4), 671–687.
14
Figures
Figure 1: Changes in commodity prices, real GDP growth, and credit
-.4
-.2
0
.2
Com
mod
ity p
rice
inde
x (%
gro
wth
)
2
4
6
8
Rea
l GD
P gr
owth
(%)
1995 1998 2001 2004 2007 2010 2013 2016
Growth: Commodity exporters Growth: Diversified exportersCommodity prices
(a) Commodity prices and growth
3.5
4
4.5
5
5.5
Com
mod
ity p
rice
inde
x
10
20
30
40
50
60
Cre
dit t
o th
e pr
ivat
e se
ctor
ove
r GD
P (%
)
1995 1998 2001 2004 2007 2010 2013 2016
Growth: Commodity exporters Growth: Diversified exportersCommodity prices
(b) Commodity prices and credit
Notes: Data on Real GDP growth is from the IMF World Economic Outlook, 2016, data on credit to private sector is from theWorld Bank’s World Development Indicators and data on commodity prices is from the IMF commodity price system. All dataare GDP-PPP weighted averages.
Figure 2: Commodity prices and bank lending: transmission channels
Motivation Research Question Stylized Facts Empirical Analysis Conclusion and next steps
Transmission channels
Commodity prices fall
Government rev-enues/Firm
profits go down
Bank lendinggoes down
Deteriorationof bank health
Lower demand
supply channeldemand channel
15
Figure 3: Commodity prices, loan growth, and bank health
1020
3040
50Lo
an G
row
th (%
)
2004 2006 2008 2010 2012 2014
Commodity exportersDiversified exporters
(a) Loan growth
11.
52
2.5
Ret
urn
on A
sset
s (%
)
2004 2006 2008 2010 2012 2014
Commodity exportersDiversified exporters
(b) Bank profitability
8590
9510
0R
etai
l Fun
ding
(%)
2004 2006 2008 2010 2012 2014
Commodity exportersDiversified exporters
(c) Retail funding
1011
1213
1415
Equ
ity to
Ass
ets
(%)
2004 2006 2008 2010 2012 2014
Commodity exportersDiversified exporters
(d) Bank capital
Notes: Data on loan growth, return on assets, retail funding and equity-to-assets ratio are from Bankscope. The chart reports themedian for each year for banks in commodity and diversified exporters. The shaded area represents episodes of commodity pricebusts.
16
Figure 4: Share of commodity exports over GDP, sample distribution
0.0
5.1
.15
.2E
xpor
t sha
re o
f com
mod
ities
in G
DP
(%)
2004 2006 2008 2010 2012 2014
Lower quartile MedianUpper quartile
17
Tables
Table 1: Summary Statistics: Loan growth by bank characteristics
Loan growth Obs Mean S.D. 25th Median 75th
All banks 3,559 27.4 30.0 7.9 20.3 39.5Domestic banks 2,243 27.5 29.8 8.4 20.2 39.1
Foreign banks 1,316 27.3 30.3 6.8 20.5 40.7Small banks (lowest quartile) 889 39.0 39.6 8.2 30.7 62.5
Other banks 2,670 23.6 24.8 7.8 18.8 33.3Low equity banks (lowest quartile) 889 24.1 27.1 7.4 18.7 33.7
Other banks 2,670 28.5 30.8 8.1 21.2 41.5Low retail funding banks (lowest quartile) 889 29.5 32.6 7.9 21.3 42.5
Other banks 2,670 26.7 29.0 7.8 20.2 38.4
Notes: Dummies for low bank characteristics are constructed considering the banks in the bottom quartile of the sample distribu-tion. All other banks are grouped in the high characteristic dummy.
Table 2: Commodity prices and bank health
Dep. Var.: EQUITY ROA NPL
PRICE INDEXt−1 0.0015 -0.0024 -0.0071(0.009) (0.003) (0.010)
PRICE INDEX × EXPOSURE 0.0760** 0.0365*** -0.0823**(0.030) (0.009) (0.039)
GROWTHt -0.0067 0.0304** -0.1138***(0.037) (0.012) (0.044)
IRt 0.4271 0.2534** -0.0616(0.370) (0.108) (0.513)
Observations 4,821 4,791 3,055Number of banks 751 749 524R2 0.801 0.595 0.681Bank FE Yes Yes YesCountry FE Yes Yes YesYear FE Yes Yes Yes
Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
18
Table 3: Baseline results
Dep. Var.: LOAN GROWTHt (1) (2) (3) (4) (5) (6)
PRICE INDEXt−1 0.1756*** 0.1020** 0.0884* -0.0069 -0.0427 -0.0781*(0.039) (0.046) (0.045) (0.043) (0.044) (0.045)
LIQUIDITYt−1 0.3037*** 0.3025*** 0.2968***(0.046) (0.046) (0.046)
SIZEt−1 -22.1065*** -21.9804*** -22.1047***(2.191) (2.196) (2.200)
EQUITYt−1 0.1060 0.1325 0.1224(0.150) (0.152) (0.152)
RETAIL FUNDINGt−1 0.1149* 0.0991 0.1053*(0.063) (0.062) (0.062)
GROWTHt 0.8290*** 0.8504***(0.205) (0.203)
IRt -7.4338*** -7.2030***(2.145) (2.144)
PRICE INDEX × EXPOSURE 0.3609**(0.163)
Observations 3,569 3,569 3,569 3,569 3,569 3,569Number of banks 584 584 584 584 584 584R2 0.051 0.122 0.382 0.454 0.462 0.463Bank FE No No Yes Yes Yes YesCountry FE No Yes . . . .Year FE Yes Yes Yes Yes Yes Yes
Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
19
Table 4: Heterogeneous effects across bank characteristics
Dep. Var.: LOAN GROWTHt (1) (2) (3) (4) (5)
PRICE INDEXt−1 -0.0768* -0.0779* -0.0266 -0.0763* -0.0716(0.046) (0.046) (0.047) (0.046) (0.046)
PRICE INDEX × EXPOSURE 0.3609**(0.163)
PRICE INDEX × EXPOSURE, domestic banks 0.4555**(0.205)
PRICE INDEX × EXPOSURE, foreign banks 0.2462(0.219)
PRICE INDEX × EXPOSURE, high size 0.2376(0.168)
PRICE INDEX × EXPOSURE, low size 0.3697*(0.218)
PRICE INDEX × EXPOSURE, high equity 0.3118*(0.170)
PRICE INDEX × EXPOSURE, low equity 0.5096***(0.188)
PRICE INDEX × EXPOSURE, high retail funding 0.2369(0.170)
PRICE INDEX × EXPOSURE, low retail funding 0.6093***(0.191)
Observations 3,569 3,569 3,569 3,569 3,569Number of banks 584 584 584 584 584R2 0.463 0.463 0.428 0.463 0.464Bank controls Yes Yes Yes Yes YesMacro controls Yes Yes Yes Yes YesBank FE Yes Yes Yes Yes YesYear FE Yes Yes Yes Yes Yes
Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Dummies for low bank characteristics are constructedconsidering the banks in the bottom quartile of the sample distribution. All other banks are grouped in the high characteristicdummy.
20
Table 5: Only domestic banks
Dep. Var.: LOAN GROWTHt (1) (2) (3) (4)
PRICE INDEXt−1 -0.0105 0.0553 -0.0062 -0.0034(0.057) (0.061) (0.058) (0.057)
PRICE INDEX × EXPOSURE 0.4058*(0.214)
PRICE INDEX × EXPOSURE, high size 0.2086(0.225)
PRICE INDEX × EXPOSURE, low size 0.4418(0.273)
PRICE INDEX × EXPOSURE, high equity 0.3492(0.223)
PRICE INDEX × EXPOSURE, low equity 0.7308***(0.246)
PRICE INDEX × EXPOSURE, high retail funding 0.3062(0.227)
PRICE INDEX × EXPOSURE, low retail funding 0.5580**(0.239)
Observations 2,243 2,243 2,243 2,243Number of banks 387 387 387 387R2 0.487 0.450 0.487 0.487Bank controls Yes Yes Yes YesMacro controls Yes Yes Yes YesBank FE Yes Yes Yes YesYear FE Yes Yes Yes Yes
Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Dummies for low bank characteristics are constructedconsidering the banks in the bottom quartile of the sample distribution. All other banks are grouped in the high characteristicdummy.
21
Online Appendix
A-I Additional Tables
Table A1: Low-income countries and number of banks in the sample
Country # Banks Country # Banks
Afghanistan 12 Mali 11Bangladesh 53 Mauritania 12Benin 10 Mongolia 13Bolivia 20 Mozambique 19Burkina Faso 10 Myanmar 16Burundi 7 Nepal 34Cambodia 32 Niger 7Cameroon 16 Nigeria 46Congo, Dem. Rep. 7 Moldova 13Cote D’Ivoire 25 Rwanda 10Congo, Rep. 18 Senegal 16Djibouti 5 Sierra Leone 15Ethiopia 16 South Sudan 6Gambia, The 9 Sudan 28Ghana 44 Tajikistan 9Guinea 8 Togo 12Haiti 5 Uganda 32Honduras 25 Tanzania 43Kenya 56 Uzbekistan 21Lao People’s Democratic Republic 10 Vietnam 61Liberia 6 Yemen 11Madagascar 7 Zambia 36Malawi 18 Zimbabwe 38
Table A2: Variable definitions and data sources
Variable Definition Source
Loan Growth Growth of gross loans in US dollars(%) BankscopeSize log of total assets BankscopeNPL Ratio of non-performing loans to gross loans BankscopeRetail Funding Ratio of customer deposits to total funding (excluding
derivatives)Bankscope
Equity Total equity to assets ratio BankscopeLiquidity Liquid assets to deposits and short-term funding BankscopeROA Return on assets BankscopePrice index Country-specific commodity price index Gruss (2014)Exposure Ratio of commodity exports to GDP (three-year moving
average)Gruss (2014) and authors’calculations
Growth Growth rate of real GDP World Economic Outlook(IMF), 2016
IR log of domestic interest rates International FinancialStatistics (IMF) and WDI
22