Determinants of domestic credit levels in

emerging markets: The role of external factors

Giray Gozgor
Dogus University, Department of International Trade and Business, Acibadem, Kadikoy, Istanbul 34722, Turkey
a r t i c l e i n f o a b s t r a c t
Article history:
Received 23 August 2013
Accepted 6 November 2013
Available online 15 November 2013
This paper empirically examines the determinants of domestic credit
expansion across a wide range of 24 emerging market economies. We
use a dynamic panel data estimation technique to investigate the
short-run and long-run effects of internal demand and external supply
factors, external balance, different measures of trade openness and
global uncertainty on domestic credit. We nd that loose monetary
policy in the domestic market, differences between domestic and global
lending rates and real trade openness positively contribute to domestic
credit levels. The ndings also show that external balance and
perceptions of global tail risk negatively affect domestic credit levels.
2013 Elsevier B.V. All rights reserved.
JEL classication:
E51
F44
Keywords:
Emerging market economies
Domestic credit
Loose monetary policy
External balance
Perception of global tail risk
1. Introduction
What are the main factors that determine domestic credit levels in developing economies?
1
Answering
this question is an increasingly noteworthy issue, particularly given that most emerging market economies
(henceforth EMEs) experienced severe credit boombust cycles in the last decade. Indeed, the effects of
domestic credit on economic activity and nancial stability are well documented, along with the experiences
of the banking and nancial crises in the Latin American and Caribbean countries in 1980s and 1990s
(Montoro and Rojas-Suarez, 2012). Nevertheless, some recent studies have shown that sustainable domestic
credit growth in a developing economy would signicantly improve welfare (Obstfeld, 2012). The aimof this
paper is to empirically investigate domestic credit across a wide range of 24 EMEs over the period 20002011
within dynamic panel data estimation framework.
Emerging Markets Review 18 (2014) 118
Tel.: +90 216 544 55 55/1554; fax: +90 216 544 55 34.
E-mail address: ggozgor@dogus.edu.tr.
1
Throughout the entire paper, domestic credit refers to the credit provided by the banking sector.
1566-0141/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.ememar.2013.11.003
Contents lists available at ScienceDirect
Emerging Markets Review
j our nal homepage: www. el sevi er . com/ l ocat e/ emr
In fact, a growing number of empirical studies, such as Mendoza and Terrones (2008), Obstfeld and
Rogoff (2010), Jorda et al. (2011), Schularick and Taylor (2012), and Gourinchas and Obstfeld (2012),
indicate that a rapid increase in domestic credit availability could play a statistically and economically
signicant role in predicting subsequent nancial or economic crises. As with an upsurge in domestic
credit, a deep decline in domestic credit can cause declining economic activity and nancial instability. For
instance, Gourinchas and Obstfeld (2012) show that the relative strength of EMEs during the great global
recession of 200809 was related to their avoidance of domestic credit booms in the preceding years. They
document that this was not the case for EMEs in Central and Eastern Europe, where the effects of the
global crisis were notably severe. The great global recession of 200809 also reected one type of asset
price bubble, referred to as a credit-driven bubble, which can be notably hazardous (Mishkin, 2010).
Domestic credit is also found to be a signicant determinant of domestic private investment in developing
economies in general therefore understanding the determinants of domestic credit is important (Luca and
Spatafora, 2012). In addition, healthy availability of domestic credit instead of external nancing would
help ease the pressure coming from the exchange rate risk on domestic rms, particularly in an economic
or a political downturn (Fernandez-Villaverde et al., 2013). Hence, the possible causes of domestic credit
cycles in EMEs deserve further investigation for both theoretical and macroeconomic policy reasons.
As discussed in Section 2 below, there is a growing empirical literature examining the determinants of
domestic credit in EMEs, but possible decisive factors remain unclear. Moreover, the determinants of
domestic credit levels appear to be complex (Elekdag and Han, 2012). The ability to borrow from abroad
allows a developing country to borrow at lower interest rates and with less pressure on the nominal
exchange rate and thus on ination than if the country is economically and nancially closed (Obstfeld,
2012). Thus, domestic credit may not only be driven by internal demand factors, such as stronger
economic growth, falling real interest rates, rising ination and rising money stocks, but also by external
supply factors, such as global liquidity, global market conditions, and external imbalances. In view of the
deep decline in domestic credit in EMEs during the period of the great global recession, the literature also
suggests that exchange rate exibility the nominal value of the exchange rate capital ows, and degree
of trade openness might also affect domestic credit availability in EMEs.
This paper investigates not only the systematic relationship between domestic credit and fundamental
internal demand and external supply factors but also takes into account the value of nominal exchange
rates, domestic and global liquidity conditions, private capital ows, and differences between domestic
and global lending rates. Following Edwards (2012) and Lane and McQuade (2013), we also separately
control for the effects on domestic credit availability of external balance indicators, such as the external
balance on goods and services (balance of trade) and the current account balance as well as nominal and
real trade openness measures. On the other hand, we re-examine recent empirical ndings by considering
the period effect of the great global recession and the effect on EME credit levels of perceived global risk
and global uncertainty.
This study seeks to contribute to the existing literature in several ways. To the best of our knowledge,
this paper is the rst to examine the effects on domestic credit expansion in EMEs of both internal
macroeconomic demand and external nancial supply factors while controlling for external balances and
real trade openness. Second, this is the rst study to analyze the persistence of domestic credit. For this
purpose, we initially determine the cross-sectional dependence of domestic credit availability across 24
EMEs using the formal test of cross-sectional dependence of Pesaran (2004) and applying the second
generation panel unit root test of Pesaran (2007). Third, following ndings of signicant persistence in
domestic credit expansion, this paper is the rst to utilize the robust and efcient dynamic panel
estimation technique of Arellano and Bond (1991) to examine the determinants of domestic credit
expansion in EMEs. This method allows us to obtain both short-run and long-run coefcients, a point of
distinction from other studies, which have generally run xed effects panel data estimations. Fourth, we
systematically re-assess our main empirical ndings with a special focus on global uncertainty and
perceived global risk indicators as well as the period effect of the great global recession. Within this
context, this is the rst study to examine Standard and Poor's (S&P 500) Volatility (VIX) and the S&P 500
Skewness (SKEW) indices of the Chicago Board Options Exchange (CBOE) as factors in domestic credit
levels in EMEs.
In this paper we nd that loose monetary policy in the domestic market, differences in domestic and
global lending rates and the degree of real trade openness contribute positively to domestic credit
2 G. Gozgor / Emerging Markets Review 18 (2014) 118
expansion in EMEs. Our results also indicate that external balances and perceptions of global tail risk are
negatively associated with domestic credit expansion in EMEs.
The remainder of this paper is organized as follows. Section 2 briey reviews the related literature.
Section 3 summarizes our data and methodology. Section 4 explains and discusses our empirical results.
Section 5 checks the robustness of our empirical results. Section 6 discusses the implications of our
empirical results. Section 7 presents the concluding remarks.
2. Literature review
Before the collapse of the Lehman Brothers in September 2008, domestic credit availability had been
commonly identied as a basic indicator of nancial development (Barajas et al., 2009). A growing
literature is now examining the determinants of domestic credit, particularly in EMEs. For instance, Takats
(2010) focuses on domestic credit in a panel framework of 21 EMEs for the period from 1995 to 2009,
using the normalized quarterly volatility of the S&P 500 index as the global supply factor. Using gross
domestic product (GDP) as the most important demand factor in his analysis, he concludes that both
demand and supply factors contributed to the decline in cross-border lending to EMEs but that the impact
of the supply shock was stronger. Additionally, these two factor types appear to have had more balanced
effects before the great global recession of 200809. Bakker and Gulde (2010) indicate that external
factors, such as bad luck, were the main causes of domestic credit boombust cycles in new European
Union (EU) countries over the past decade. In their estimation, higher ination caused real interest rates to
fall, which encouraged growth of domestic credit. Aisen and Franken (2010) use data from 80 countries to
investigate the main determinants of domestic credit expansion during the great global recession, nding
that the boom before the recession and the rapid slowdown in economic activity among major trading
partners during the recession were the main determinants of domestic credit levels in EMEs and that the
roles of counter-cyclical monetary policy and domestic liquidity were also important. Kamil and Rai
(2010) consider external and internal sources of domestic credit growth in 13 LAC borrower countries and
14 non-LAC lender countries during the global recession, mainly concluding that the countries reliant on
foreign funding for domestic credit growth suffered more heavily from the great global recession than
those that were less reliant. Guo and Stepanyan (2011) examine the determinants of domestic credit in
EMEs during the last decade, covering both the pre-crisis and post-crisis periods. They nd that domestic
funding and foreign funding were among the most important determinants of domestic credit. They also
indicate that stronger economic growth, higher ination and loose domestic and global monetary policies
tended to increase domestic credit availability in EMEs. Borio et al. (2011) investigate volumes of global
credit in major currencies and their contribution to overall credit growth in particular countries. They
emphasize the role of external factors on domestic credit growth and draw policy implications with
respect to the associated risks, mainly suggesting the importance of international coordination. Similar
empirical ndings are obtained by Avdjiev et al. (2012) for several developed and developing Asian
economies. Similarly, Elekdag and Wu (2011) examine 99 credit booms and conclude that loose monetary
policy and global liquidity conditions are the main drivers of real domestic credit growth across EMEs.
Mendoza and Terrones (2012) focus on 61 developed and developing countries in the period from 1960 to
2010, identifying 70 credit boom events, half of them in EMEs. They nd that there is a systematic and
robust relationship between domestic credit booms (or boombust cycles) and increasing capital inows,
rising house and equity prices, increased economic activity, real exchange rate appreciation, and growing
decits in the external balance on goods and services and the current account balance.
On the other hand, capital ows could also inuence domestic credit levels in EMEs. In the literature,
there is empirical evidence that net capital inows in emerging market economies might generate
domestic credit booms, particularly given a fragile domestic nancial system. For instance, Mendoza and
Terrones (2008) nd that most nancial crises and domestic credit booms in EMEs have occurred as a
result of excessive capital inows. A study of the International Monetary Fund (IMF) (2011) investigates
19 developed countries and 28 EMEs for the period 1960 to 2010, concluding that capital inows
signicantly predicted domestic credit booms. Ostry et al. (2012), using specic denitions of domestic
credit booms and crises in EMEs in a panel data framework covering the period from 1995 to 2008,
conclude that domestic credit booms have been associated with capital inow surges. Lane and McQuade
(2013) investigate the interrelationship between domestic credit growth and international capital ows
3 G. Gozgor / Emerging Markets Review 18 (2014) 118
for the period 1993 to 2008 in 29 European countries, focusing on the period from 2003 to 2008 as an
observation period of a domestic credit boom. They nd a strong asymmetric relationship between
domestic credit growth, net debt ows, and net equity ows, concluding that macroeconomic and
nancial variables could affect both the supply and demand factors in domestic credit growth and that
external balances should also be added to estimation framework. Magud et al. (2012) suggest that net
capital inows are the main determinant of domestic credit growth in EMEs, particularly for countries
with oating exchange rates. They show that increasing exchange rate exibility in emerging market
economies could signicantly decrease the growth rate of domestic credit and the share of foreign
currencies in total domestic credit. They also nd that ination has an insignicant effect, while the
domestic money stock, dened as broad money, has a signicant effect on domestic credit levels in EMEs.
Contrary to that empirical evidence, Hume and Sentence (2009) observe that several large EMEs have
experienced domestic credit booms without net capital inows. In other words, capital ows and domestic
credit might be unrelated in EMEs. They also observe that, as a result of the global recession of 200809,
net capital ows from emerging markets to developed economies was quantitatively far less than the
amount of domestic credit those economies generated in the preceding years. In this line, Furceri et al.
(2012) nd that foreign direct investment (FDI) and portfolio equity inows have no statistically
signicant effect on domestic credit, while only debt inows were a statistically signicant factor in
determining domestic credit levels in EMEs. They indicate that the short-term effect of capital inows on
domestic credit depends on countries' macroeconomic policy stances.
3. Data and methodology
3.1. Data
This paper empirically examines the determinants of domestic credit levels across a wide range of 24
EMEs: Argentina, Brazil, Bulgaria, Chile, China, Estonia, Hungary, India, Indonesia, Latvia, Lithuania,
Malaysia, Mexico, Pakistan, Peru, the Philippines, Poland, Romania, the Russian Federation, South Africa,
Thailand, Turkey, Ukraine, and Venezuela. Our denition of EMEs is based on the classication of the IMF.
Table 1
Descriptive summary statistics for the domestic credit levels in 24 EMEs.
Domestic credit (percentage of GDP) Mean Std. dev. Minimum Maximum Difference (max.min.)
Argentina 36.7 11.1 24.4 62.4 38
Brazil 83.8 11.3 71.9 98.3 26.4
Bulgaria 44.7 20.8 17.4 71.4 54
Chile 80.8 7.9 66.0 89.4 23.4
China 136 11.1 119.7 151.9 32.2
Estonia 71.1 25.2 34.6 105.4 70.8
Hungary 66.6 12.0 49.6 81.7 32.1
India 61.8 7.9 51.4 75.1 23.7
Indonesia 45.3 8.0 36.5 60.7 24.2
Latvia 65.8 26.9 23.3 94.3 71
Lithuania 42.5 21.1 15.2 70.0 54.8
Malaysia 130.6 11.2 113.4 146.5 33.1
Mexico 37 5.0 31.8 45.5 13.7
Pakistan 44.1 4.9 37.2 53.2 16
Peru 19.2 3.5 14.6 26.0 11.4
the Philippines 51.6 3.8 47.2 58.3 11.1
Poland 46.8 12.3 34.4 66.2 31.8
Romania 30.3 17.5 13.0 55.0 42
the Russian Federation 28.0 6.0 22.1 39.6 17.5
South Africa 175.3 13.3 152.5 195.3 42.8
Thailand 130.3 10.4 109.0 150.8 41.8
Turkey 51.5 10.6 37.9 69.6 31.7
Ukraine 50.4 24.9 23.8 80.6 56.8
Venezuela 18.0 5.9 10.5 29.3 18.8
Panel (average) 64.5 12.2 48.2 82.4 34.1
4 G. Gozgor / Emerging Markets Review 18 (2014) 118
In addition, related to our aim and estimation technique; and also following the recent studies of Guo and
Stepanyan (2011) and Magud et al. (2012), we focus on relatively homogenous observations for
developing economies and include only 24 EMEs in the panel data set. Thus, the homogenous panel data
set would give us the opportunity to compare our results with their previous ndings.
We employ the dynamic panel data estimation procedure of Arellano and Bond (1991) for the period
from 2000 to 2011, using annual frequency data. We focus on this period because it covers and reects the
full boombust cycle rather than only booms or busts in domestic credit levels in EMEs. We obtained all
our data from the World Development Indicators and the Global Financial Development database of the
World Bank.
Related to our purpose and estimation procedure, we focus on both the lagged and current ratios of
domestic credit to GDP. Using the lagged ratio of domestic credit provided by banking sector to GDP, we
aim to determine whether the persistence of the dependent variable is signicant and obtain both the
short-run and long-run coefcients. We report a summary of the descriptive statistics for the domestic
credit level in each country in Table 1. Moreover, we report descriptive summary statistics in Table 2.
We explain and discuss the explanatory variables as follows.
3.2. Internal demand factors
Following the seminal paper of Frankel and Romer (1999), we consider (lagged) real GDP per capita in
U.S. dollars (USD) in logarithmic form, instead of nominal or real GDP, as the benchmark measure of
domestic income in our panel estimations. A higher growth rate of domestic income leads to stronger
domestic demand and thus higher levels of domestic credit (Takats, 2010).
We use the (lagged) ination rate (%) based on consumer price indices (2005 = 100) in the dynamic
panel data estimations. If the coefcient on the ination rate is signicant and less than one, ination will
lead to a decline in real domestic credit. We also use the (lagged) deposit rate (%), the real interest rate
(%), and the domestic money supply (dened as broad money as a percentage of GDP) as proxy variables
for the overall monetary policy stance. Higher deposit and real interest rates and a lower money supply
signal tighter monetary conditions and thus less domestic credit growth. We also use the ratio of
non-performing loans to gross total loans as the fundamental measure of the soundness of EME banking
sectors. Fragile banking systems in developing economies tend to mitigate the growth of domestic credit
in these countries. On the other hand, high net incurrence of the domestic and foreign liabilities (dened
as a percentage of GDP) may generate additional domestic credit growth in EMEs (Bakker and Gulde,
Table 2
Descriptive summary statistics.
Variables Denition Mean Std. dev. Min. Max. Obs.
Domestic credit Percentage of GDP 64.51 41.81 10.51 195.33 288
Ination rate Percentage 7.43 7.93 1.38 54.92 288
GDP per capita In logarithmic form 7.94 0.77 6.11 9.35 288
Nominal exchange rate LCU per USD 433.9 1860.7 0.48 10390 287
Deposit rate Percentage 7.86 8.31 1.00 74.70 250
Broad money Percentage of GDP 57.84 33.88 18.16 180.78 287
The US broad money Percentage of GDP 78.90 5.98 70.95 88.84 288
Non-performing loans to gross total loans Percentage 7.46 7.06 0.2 34.40 281
Net incurrence of liabilities, domestic Percentage of GDP 1.76 2.21 4.14 9.01 186
Net incurrence of liabilities, foreign Percentage of GDP 0.90 2.30 5.39 16.01 190
Lending interest rate differences Percentage 8.59 11.16 3.38 62.96 266
Real interest rate Percentage 5.39 10.09 18.9 47.7 266
Private capital ows Percentage of GDP 2.82 3.94 14.27 28.65 285
Current Account balance Percentage of GDP 0.54 6.96 27.16 18.04 285
Balance of trade Percentage of GDP 0.55 7.92 21.47 23.15 287
Nominal openness Percentage 80.89 44.54 21.72 220.41 286
Real openness Percentage 90.87 54.60 19.35 266.06 283
The 3 months LIBOROIS spread Percentage 0.27 0.28 0.08 1.08 264
The VIX index In logarithmic form 3.05 0.30 2.53 3.46 288
The SKEW index In logarithmic form 4.76 0.03 4.73 4.80 288
5 G. Gozgor / Emerging Markets Review 18 (2014) 118
2010; Guo and Stepanyan, 2011; Magud et al., 2012). We therefore include those variables in the panel
estimation framework.
However, we use the lagged logarithm of real GPD per capita and the lagged ination rate in the
dynamic panel data estimations to avoid the problem of reverse causality, arising from the fact that
domestic credit growth tends to pioneer stronger real income growth and higher ination. A lagged
deposit rate is also used, as the deposit rate will affect domestic credit only with a lag.
3.3. External supply factors
On the other hand, we include in the dynamic panel data estimations the money supply (broad money
as a percentage of GDP) in the US as the benchmark indicator of global liquidity conditions. An increasing
US money supply is expected to raise the level of domestic credit in EMEs (Elekdag and Wu, 2011). We
also consider the nominal (ofcial) exchange rate (local currency per USD, period average) as a candidate
explanatory variable in determining the volume of domestic credit in EMEs. Furthermore, it is important
to note that a fall in the value of ofcial exchange rate denotes appreciation of the domestic currency,
which will lead to an increase in domestic credit (Borio et al., 2011).
Following Magud et al. (2012), we suggest that the difference between the domestic lending rate (%)
and the US (global) lending rate (%) could also be of interest in an empirical investigation of the
determinants of domestic credit levels across a wide range of EMEs. As Obstfeld (2012) indicates, the
overall positions of banks in emerging market economies strongly depend on their ability to borrow from
abroad. This facility could allow them to borrow at lower global interest rates and lend at higher domestic
interest rates. Thus, an increased difference between global and domestic lending rates tends to increase
domestic credit levels in EMEs. Similarly to this effect, high prots in the semi-oligopolistic banking
systems of most EMEs, and banks' strong appetites for risk, would cause a substantial increase in the
domestic credit supply in EMEs (Fungacova et al., 2013; Mamatzakis et al., 2005).
Importantly, we control for the impact of private capital ows (as a percentage of GDP) on domestic
credit. Private capital ows consist of net foreign direct investment and portfolio investment. Within this
context, increasing private capital ows to a developing economy tend to increase the volume of domestic
credit (Kim and Wu, 2008; Lane and McQuade, 2013). On the other hand, the degree of nancial
integration, the sizes of budget decits, and external debt levels are viewed as the main determinants of
capital ows in the literature and thus do not directly impact the volume of domestic credit (Magud et al.,
2012). Hence, we do not consider the degree of nancial integration or the sizes of budget decits and
external debt levels in our panel estimations.
3.4. External balances and degree of trade openness
Following Edwards (2012) and Lane and McQuade (2013), we consider the (lagged) external balance
on goods and services (as a percentage of GDP) and the (lagged) current account balance (as a percentage
of GDP) in the panel estimations. External balances and domestic credit are negatively related in literature
(Lane and McQuade, 2013). We consider the lagged indicator of external balances in the panel data
estimations to avoid the problem of reverse causality, arising from the fact that increased domestic credit
tends to be associated with larger external imbalances (Cheung et al., 2012).
On the other hand, degree of openness to trade may provide additional data regarding the
determinants of domestic credit levels in EMEs. Trade openness can relate to both internal demand
factors and external supply factors. Indeed, the great global recession of 200809 was associated with a
dramatic collapse in global trade, in excess of the fall in production and signicantly larger than the fall in
GDP (Gopinath et al., 2012). In addition, according to Mishkin (2010), the great global recession of
200809 reected the effects of the credit-driven bubble in asset prices. Thus, trade openness and
domestic credit may be related. Generally, we expect a positive relationship between trade openness and
domestic credit expansion. Indeed, if a developing country is more open to trade, that is, more globalized
and more integrated into the global nancial system, a growing ability to borrow from abroad would
enable the country to borrow at lower interest rates from foreign markets (Obstfeld, 2012), providing an
additional boost to domestic credit expansion (Kamil and Rai, 2010). Additionally, credit demand related
to imports and exports of goods and services may introduce an additional channel of domestic credit
6 G. Gozgor / Emerging Markets Review 18 (2014) 118
growth in EMEs. On the other hand, more open economies are more vulnerable to shocks originating in
their main trading partners; hence, their domestic credit cycles can be strongly affected by such shocks
(Aisen and Franken, 2010).
In this paper we dene nominal (trade) openness as nominal imports plus nominal exports relative
to nominal GDP, a measure used in many seminal papers, including Frankel and Romer (1999). Another
measure, real openness, is dened by Alcala and Ciccone (2004) as real imports plus real exports relative
to Purchasing Power Parity GDP, in USD. Alcala and Ciccone (2004) show that because of inaccuracies
caused by cross-country differences in the relative prices of non-tradable goods a nominal measure of
openness might induce a biased estimate of productivity. Hence, we also use real trade openness in this
paper.
To avoid the problem of reverse causality in panel estimations, we use lagged nominal and real
openness measures. Additionally, however, trade openness can be related to domestic income and
economic growth (Alcala and Ciccone, 2004; Frankel and Romer, 1999) and to the ination rate (Romer,
1993). Using lagged openness measures in our estimations also avoids simultaneity bias and possible
multicollinearity in a dynamic panel data framework.
3.5. Global nancial market conditions
Along with the fundamental explanatory variables discussed above, we focus on global uncertainty and
perceived global risk indicators, as well as the period effect of the great global recession of 200809, to
control for the impact of global economic and nancial market conditions on domestic credit levels in
EMEs. We control for the impact of the great global recession of 200809 on domestic credit by using a
time dummy variable for the year 2009. We also use other key indicators for global uncertainty and the
nancial environment. For instance, Aisen and Franken (2010) and Kamil and Rai (2010) both use the
spread between the London Interbank Offer Rate (LIBOR) and the Overnight Indexed Swap (OIS) as an
indicator of global uncertainty and the overall nancial environment. Following them, we consider the
3-month LIBOROIS spread as a possible determinant of domestic credit levels in EMEs.
As already discussed, Takats (2010) also cites the normalized quarterly volatility of the S&P 500
nancial index as a global supply factor affecting the availability of domestic credit. In this line, we use the
logarithms of the S&P 500 VIX and the S&P 500 SKEW indices of the CBOE. As volatility generally signies
nancial turmoil, the VIX index is often referred to as the investor fear gauge (Hartelius et al., 2008). On
the other hand, the SKEW index is a strong benchmark measure of the perceived risk of extreme negative
movementsoften referred to as tail risk or a black swan event in literaturein US equity markets and
global nancial markets generally. To the best of our knowledge, this paper is the rst in literature to use
the VIX and the SKEW indices of the CBOE as benchmark measures of uncertainty in global nancial
markets and perceived tail risk as global supply factors in domestic credit availability in EMEs.
3.6. Preliminary tests and panel data estimation procedure
Lagged values of domestic credit are automatically added into the dynamic panel data estimations
below. This variable denes and measures the magnitude of the persistence of domestic credit levels in
EMEs. However, as a preliminary, we check the robustness of the coefcient on lagged domestic credit. In
doing so, we begin by testing for cross-sectional dependence in domestic credit levels among the 24 EMEs,
using the cross-sectional dependence (CD) test of Pesaran (2004). Following the results fromthe CD test of
Pesaran (2004), we apply second generation panel unit root (PUR) tests, such as that proposed by
Pesaran (2007), to account for cross-sectional dependence. This procedure is crucial in overcoming the
shortfall of rst generation PUR tests that assume cross-sectional independence by default. To the best of
our knowledge, this is the rst study in the literature to apply a second generation PUR test in a large
panel framework for the study of the determinants of domestic credit levels.
The results of the PUR tests suggest signicant persistence of domestic credit in EMEs. In light of this
nding, we employ the differentiated dynamic panel data estimation techniques of Arellano and Bond
(1991) to eliminate not only problems of autocorrelation and heteroskedasticity but also the possible
presence of different degrees of serial integration in the balanced panel data framework used in this paper.
Moreover, the bootstrapped covariance matrix can be used in the estimation procedure and consistent
7 G. Gozgor / Emerging Markets Review 18 (2014) 118
estimators can be obtained using the bias-correction methods of Anderson and Hsiao (1982), Arellano and
Bond (1991), and Blundell and Bond (1998). Thus, we utilize the two-stage estimation procedure of Arellano
and Bond (1991) with consistent estimators to obtain differentiated dynamic panel data estimates and avoid
possible multicollinearity among the explanatory variables. Under this procedure, explanatory variables are
automatically dropped in cases of multicollinearity.
Furthermore, we address potential endogeneity of the explanatory variables and of lagged domestic
credit by instrumenting them with related lagged values. In the Generalized Method of Moments (GMM)
approach of Arellano and Bond (1991), all variables are differenced and differenced endogenous variables
are instrumented by their lags. To use this dynamic panel GMM estimation technique, two assumptions
must be hold. First, instruments must be uncorrelated with error terms, and second, instruments must be
correlated with the instrumented variables. Thus, we must nd empirical evidence in favor of rst-order
autocorrelation but against second-order autocorrelation in the residuals.
We also control for country xed effects in all dynamic panel data estimations because there could be
other heterogeneities across countries that affect domestic credit levels. On the other hand, we use
dummies for time xed effects to control for potential cyclical effects not captured by variables in our
estimations. We do not, however, consider a xed or random effects panel data estimation because of the
signicant persistence of domestic credit in EMEs, a phenomenon not fully captured in xed or random
effects estimations.
3.7. Empirical model
In this paper, we estimate the following panel data regression:
DC
i;t
DC
i;t1
INF
i;t1
lnGPC
i;t
NER
i;t
DR
i;t1
CONTROL
i;t
CONTROL
i;t1
v
l
v
t

i;t
1
where DC
i,t
and DC
i,t 1
are the current and lagged ratios of domestic credit provided by banking sector to
GDP in country i at times t and t 1; INF
i,t 1
is the lagged ination rate in country i at time t 1; lnGPC
i,t
is
the log of real per capita GDP in country i at time t; NER
i,t
is the nominal exchange rate in units of local
currency per USD for county i at time t; and DR
i,t 1
is the lagged deposit rate (%) in country i at time t 1.
Country effects, period effects, and the error term are denoted by v
i
, v
t
and
i,t
, respectively.
We also use several CONTROL
i,t
variables in country i at time t as follows: The domestic money supply
(dened as broad money) as a percentage of GDP in country i at time t; the foreign money supply (dened
as broad money) as a percentage of US GDP at time t; the ratio of non-performing loans to gross total loans
in country i at time t; the net domestic incurrence and foreign incurrence of liabilities as a percentage of
GDP in country i at time t; the difference between the domestic lending interest rate (%) and the US
(global) lending interest rate (%) in country i at time t; the real interest rate (%) in country i at time t; and
private capital ows as a percentage of GDP in country i at time t. We also consider CONTROL
i,t 1
variables
in country i at time t 1 as follows: the current account balance as a percentage of GDP in country i at
time t 1; the external balance on goods and services as a percentage of GDP in country i at time t 1,
and nominal and the real openness measures in country i at time t 1. In the next section, we represent
and discuss the empirical results of the panel data estimation technique of Arellano and Bond (1991) for
Eq. (1).
Table 3
Results of the CD test of Pesaran (2004) for the domestic credit (DC) (% of GDP).
Cross-sectional dependence DC
The CD statistics and probability 6.63 (0.00)
Average absolute value of the off-diagonal elements 0.309
Notes: The CD test of Pesaran (2004) is dened under the null hypothesis of no cross-sectional
dependence in domestic credit in EMEs. p-Values are in parentheses.
8 G. Gozgor / Emerging Markets Review 18 (2014) 118
4. Empirical results
4.1. Results of the persistence in domestic credit
As a preliminary, we test for persistence of domestic credit in 24 EMEs, using a second-generation
cross-sectional dependence panel unit root test. First, we test for cross-sectional dependence of domestic
credit by applying the cross-section independence (CD) test of Pesaran (2004). We report the results in
Table 3.
In the CD test of Pesaran (2004), the null hypothesis of no cross-sectional dependence of domestic
credit among EMEs is rejected. Thus, domestic credit is cross-sectionally dependent. We then apply the
second-generation panel unit root test, assuming cross-sectional dependence; in particular, we perform
the cross-sectional augmented panel unit root (CIPS) test of Pesaran (2007). We report the results of the
CIPS test, using only a constant term as well as both constant and trend terms in Table 4.
Results from the CIPS test of Pesaran (2007) show that domestic credit in EMEs can be described as
non-stationary processes. Therefore, the ndings of dynamic panel data estimation demonstrating
persistence in the lagged variables of domestic credit are correct and robust, and we do not consider the
xed or random effects panel estimation techniques.
4.2. Results of the dynamic panel data estimations of Arellano and Bond (1991)
The results of our dynamic panel data estimation, using the technique of Arellano and Bond (1991),
are reported in Tables 5, 6, and 7. All the results of the Sargan test indicate no over-identication problem
in the estimation specications. Additionally, results of the LM tests for AR(1) and AR(2) indicate
rst-order but no second-order autocorrelation. Thus, the empirical results in Tables 5, 6, and 7 satisfy
the necessary conditions for application of the dynamic panel data estimation technique of Arellano and
Bond (1991).
The empirical ndings in Tables 5, 6, and 7 show statistically signicant persistence of domestic credit
in EMEs, with an average coefcient 0.464. This result signals medium-level persistence of domestic credit
in EMEs.
The results in Table 5 report estimates using various fundamental framework specications of
internal demand and external supply factors as determinants of domestic credit levels in EMEs. In
Column I, a relatively basic framework in the literature is employed. We then control for domestic and
global money supplies in Column II. Additionally, we add several indicators of the soundness of the
banking and nancial sectors of EMEs into the regressions, and report the results in Columns III and IV.
The empirical results in Table 5 indicate that the logarithm of lagged per capita GDP and the quantity of
money (both domestic and US, broadly dened) positively affect domestic credit availability in EMEs.
Among all variables considered, these are the only ones whose coefcients are statistically signicant in
all frameworks.
Table 6 reports on the explanatory variables used in the model of Column III, various other indicators of
internal demand and external supply factors, and private capital ows and the lagged current account
balance as determinants of domestic credit volumes in EMEs. In Column V, along with the variables in the
model of Column III, our regression also includes differences in the lending rates between global and
domestic markets. We also control for the real interest rate, private capital ows, and the lagged current
Table 4
Results of the CIPS tests for the domestic credit (DC) (% of GDP).
Heterogeneous unit root test: the CIPS Constant Constant and trend
Zt-bar statistics 0.759 (0.776) 1.298 (0.105)
Notes: The CIPS test assumes cross-sectional dependence in the form of a single unobserved common factor. The test is dened
under the null hypothesis that there is a unit root in domestic credit in EME. The optimal lag length is selected by the Akaike
Information Criterion (AIC). The p-values are in parentheses.
9 G. Gozgor / Emerging Markets Review 18 (2014) 118
Table 5
Results of the dynamic panel data estimation of Arellano and Bond (1991).
Regressors I II III IV
Lagged domestic credit (% of GDP) 0.614 (0.051)

0.442 (0.048)

0.436 (0.058)

0.368 (0.076)

Lagged ination rate (%) 0.086 (0.132) 0.016 (0.113) 0.004 (0.018) 0.016 (0.173)
Lagged GDP per capita (ln) 8.303 (3.646)

13.54 (4.673)

12.67 (6.343)

5.025 (2.348)

Nominal exchange rate (LCU per USD) 0.0003 (0.003) 0.0008 (0.002) 0.0007 (0.002) 0.0007 (0.004)
Lagged deposit rate (%) 0.205 (0.157) 0.085 (0.013) 0.089 (0.131) 0.329 (0.222)
Broad money (% of GDP) 0.641 (0.079)

0.647 (0.081)

0.768 (0.108)

The US broad money (% of GDP) 0.301 (0.101)

0.290 (0.104)

0.256 (0.149)

Non-performing loans to gross total loans (%) 0.029 (0.120) 0.014 (0.186)
Net incurrence of liabilities, domestic (% of GDP) 0.030 (0.315)
Net incurrence of liabilities, foreign (% of GDP) 0.175 (0.349)
Lending interest rate differences (%)
Real interest rate (%)
Private capital ows (% of GDP)
Lagged current account balance (% of GDP)
Lagged balance of trade (% of GDP)
Lagged nominal openness
Lagged real openness
Observations 220 220 216 137
Sargan test [0.00] [0.00] [0.00] [0.00]
AR(1) [0.00] [0.00] [0.00] [0.00]
AR(2) [0.35] [0.39] [0.32] [0.17]
Notes: The dependent variable is the ratio of domestic credit provided by banking sector to GDP. The constant term is also estimated
but not reported. The Sargan test shows the results of the over-identifying restrictions (null hypothesis: there is over-identication
in the estimation specication). AR(1) and AR(2) showresults of the LM statistics for autocorrelation (null hypothesis: no rst-order
autocorrelation and no second-order autocorrelation, respectively). We report robust standard errors. Standard errors are in
parentheses, and the p-values are in brackets.

,

and

indicate statistical signicance at the 1%, 5% and 10% levels, respectively.
Table 6
Results of the dynamic panel data estimation of Arellano and Bond (1991).
Regressors V VI VII VIII
Lagged domestic credit (% of GDP) 0.471 (0.059)

0.490 (0.061)

0.466 (0.059)

0.442 (0.059)

Lagged ination rate (%) 0.021 (0.141) 0.004 (0.140) 0.021 (0.014) 0.109 (0.143)
Lagged GDP per capita (ln) 8.354 (6.994) 6.422 (7.164) 6.923 (7.031) 9.278 (6.878)
Nominal exchange rate (LCU per USD) 0.0004 (0.003) 0.001 (0.003) 0.0006 (0.003) 0.0007 (0.002)
Lagged deposit rate (%) 0.018 (0.175) 0.009 (0.176) 0.021 (0.177) 0.114 (0.177)
Broad money (% of GDP) 0.686 (0.084)

0.652 (0.089)

0.686 (0.086)

0.718 (0.084)

The US broad money (% of GDP) 0.133 (0.121) 0.103 (0.124) 0.090 (0.126) 0.186 (0.120)
Non-performing loans to gross total loans (%) 0.090 (0.135) 0.123 (0.138) 0.083 (0.135) 0.014 (0.136)
Net incurrence of liabilities, domestic (% of GDP)
Net incurrence of liabilities, foreign (% of GDP)
Lending interest rate differences (%) 0.329 (0.112)

0.289 (0.117)

0.322 (0.112)

0.220 (0.114)

Real interest rate (%) 0.122 (0.101)

Private capital ows (% of GDP) 0.164 (0.128)
Lagged current account balance (% of GDP) 0.394 (0.098)

Lagged balance of trade (% of GDP)

Lagged nominal openness
Lagged real openness
Observations 205 205 203 205
Sargan test [0.00] [0.00] [0.00] [0.00]
AR(1) [0.00] [0.00] [0.00] [0.00]
AR(2) [0.55] [0.47] [0.57] [0.48]
Notes: The dependent variable is the ratio of domestic credit provided by banking sector to GDP. The constant term is also estimated
but not reported. The Sargan test shows the results of the over-identifying restrictions (null hypothesis: there is over-identication
in the estimation specication). AR(1) and AR(2) showresults of the LM statistics for autocorrelation (null hypothesis: no rst-order
autocorrelation and no second-order autocorrelation, respectively). We report robust standard errors. Standard errors are in
parentheses, and the p-values are in brackets.

,

and

indicate statistical signicance at the 1%, 5% and 10% levels, respectively.
10 G. Gozgor / Emerging Markets Review 18 (2014) 118
account balance in Columns VI, VII, and VIII, respectively. The results in Table 6 indicate that the domestic
money supply and differences between global and domestic lending rates positively contribute to levels of
domestic credit. Additionally, the lagged current account balance negatively contributes to present
domestic credit levels. However, in contrast to the rst four cases, the coefcients of the logarithm of
lagged per capita GDP and the current US money supply are not statistically signicant. Thus, they are not
statistically robust explanatory factors of domestic credit expansion in EMEs, according to the evidence in
Table 6.
In addition, the empirical results presented in Table 7 report the benchmark model in Column V,
controlling for the lagged balance of trade and lagged nominal and real trade openness measures as
possible determinants of domestic credit expansion in EMEs. In Column IX, along with the variables in the
benchmark model in Column V, we control for the lagged external balance on goods and services, nding
that its coefcient is statistically signicant and that the impact on domestic credit is negative. We again
obtain signicant and positive effects for the domestic money supply (broadly dened) and for differences
in global and domestic lending rates, in the results reported in Column XI. These empirical results are in
line with the results presented in Column VIII. In addition, we nd a statistically insignicant effect of
lagged nominal openness on domestic credit in Column X, whereas the coefcient on lagged real openness
is statistically signicant in the regression presented in Column XI. Finally, in Column XII, we control for
both the lagged current account balance and lagged real openness, nding a negative coefcient in the
case of the lagged current account balance. Additionally, similarly to framework XI, more real trade
openness in the previous period has a positive effect on domestic credit in the current period. Thus, we
obtain four statistically signicant determinants of domestic credit.
On the other hand, the average short-run coefcients of the domestic money supply and of the
differences in global and domestic lending rates are 0.693 and 0.295, respectively. The long-run impact,
namely, the short-run effect together with the effect through the lagged domestic credit to GDP ratio, is
Table 7
Results of the dynamic panel data estimation of Arellano and Bond (1991).
Regressors IX X XI XII
Lagged domestic credit (% of GDP) 0.461 (0.059)

0.472 (0.059)

0.467 (0.059)

0.440 (0.059)

Lagged ination rate (%) 0.042 (0.144) 0.052 (0.149) 0.028 (0.140) 0.147 (0.145)
Lagged GDP per capita (ln) 9.116 (6.969) 9.484 (7.007) 9.917 (6.991) 10.59 (6.941)
Nominal exchange rate (LCU per USD) 0.0005 (0.003) 0.0006 (0.003) 0.0002 (0.003) 0.001 (0.003)
Lagged deposit rate (%) 0.079 (0.180) 0.006 (0.177) 0.012 (0.176) 0.135 (0.178)
Broad money (% of GDP) 0.702 (0.085)

0.706 (0.086)

0.694 (0.084)

0.722 (0.084)

The US broad money (% of GDP) 0.157 (0.121) 0.121 (0.122) 0.041 (0.126) 0.101 (0.126)
Non-performing loans
to gross total loans (%)
0.028 (0.137) 0.110 (0.135) 0.058 (0.135) 0.039 (0.137)
Net incurrence of liabilities,
domestic (% of GDP)

Net incurrence of liabilities,
foreign (% of GDP)

Lending interest rate differences (%) 0.241 (0.117)

0.364 (0.115)

0.354 (0.112)

0.247 (0.116)

Real interest rate (%)

Private capital ows (% of GDP)
Lagged current account balance (% of GDP) 0.374 (0.099)

Lagged balance of trade (% of GDP) 0.308 (0.118)

Lagged nominal openness 0.084 (0.055)
Lagged real openness 0.122 (0.047)

0.109 (0.047)

Observations 205 205 205 205

Sargan test [0.00] [0.00] [0.00] [0.00]
AR(1) [0.00] [0.00] [0.00] [0.00]
AR(2) [0.52] [0.38] [0.42] [0.44]
Notes: The dependent variable is the ratio of domestic credit provided by banking sector to GDP. The constant term is also estimated
but not reported. The Sargan test shows the results of the over-identifying restrictions (null hypothesis: there is over-identication
in the estimation specication). AR(1) and AR(2) showresults of the LM statistics for autocorrelation (null hypothesis: no rst-order
autocorrelation and no second-order autocorrelation, respectively). We report robust standard errors. Standard errors are in
parentheses, and the p-values are in brackets.

and

indicate statistical signicance at the 1% and 5% levels, respectively.
11 G. Gozgor / Emerging Markets Review 18 (2014) 118
found to be 1.292
2
for the domestic money supply and 0.550
3
for the lending rate differences. In addition,
the empirical evidence shows that a one percent increase in the lagged current account balance reduces
the current domestic credit to GDP ratio by approximately 0.384% (on average) in the short run and by
approximately 0.716
4
percent (on average) in the long run.
5
Additionally, a 10 percent increase in lagged
real openness leads to an approximately 1.15 percent increase (on average) in the current domestic credit
to GDP ratio in the short run and an approximately 2.14
6
percent increase (on average) in the long run.
5. Robustness check
5.1. Inclusion and exclusion of explanatory variables
In this section, we check the robustness of our empirical results by also controlling for the impact of
three indicators of global economic and nancial market conditions: the 3-month LIBOROIS spread, the
log of the VIX index and the log of the SKEW index. We also include a dummy variable for the great global
recession of 200809 (2009 = 1; other years = 0). We again run the dynamic panel data estimation
procedure of Arellano and Bond (1991), with results reported in Tables 8 and 9.
All results of the Sargan test show no over-identication problem in the estimation specications.
Additionally, the results of the LM test indicate rst-order, but not second-order, autocorrelation. As
shown shortly, the regressions meet the necessary conditions of the dynamic panel data estimation of
Arellano and Bond (1991).
We consider eight frameworks for the panel estimation of the four control variables (the dummy
variable for the great global recession of 200809; the 3-month LIBOROIS spread; the log of the VIX index
and the log of the SKEW index), the lagged balance of trade (rst columns) and the lagged current account
balance (second columns). Tables 8 and 9 present the results of the dynamic panel data regression of the
four control variables and of the nal estimation framework in Table 7. We nd that the effects of the great
global recession of 200809, the 3-month LIBOROIS spread, and the log of VIX index are not statistically
signicant. We nd the log of the SKEW index to be the only signicant determinant of domestic credit in
EMEs, among the global economic and nancial market condition indicators. Indeed, the SKEW index
mainly describes the tail-risk of the distribution. Thus, our results show that the SKEW index, which is a
strong benchmark measure of the perceived risk of extreme negative movements in US equities markets
and global nancial markets, is negatively related to domestic credit levels in EMEs.
On the other hand, similarly to the previous results in Tables 5, 6, and 7, the results in Tables 8 and 9
demonstrate that lagged domestic credit as a percentage of GDP, the domestic money supply as a
percentage of GDP, the difference between global and domestic lending rates, the lagged current account
balance as a percentage of GDPor the lagged external balance on goods and services as a percentage of
GDPand lagged real openness are statistically signicant determinants of domestic credit levels in EMEs.
According to the empirical ndings in Tables 8 and 9, the signicant persistence in domestic credit now
has a coefcient of 0.445, on average. The short-run coefcients on the domestic money supply and
differences in global and domestic lending rates are, on average, 0.74 and 0.253, while the long-run
coefcients are 1.333
7
and 0.455
8
, respectively. Furthermore, the empirical ndings indicate that a one
percent rise in the lagged current account balance reduces the current domestic credit to GDP ratio by
approximately 0.353% (on average) in the short run and by approximately 0.636
9
percent (on average) in
the long run.
10
Additionally, a 10 percent increase in lagged real openness causes an approximately 1.11
2
Note that 0.693 / (1 0.464).
3
Note that 0.295 / (1 0.464).
4
Note that 0.384 / (1 0.464).
5
Similarly, the short-run and long-run coefcients of the external balance on goods and services are found to be 0.308 and
0.308 / (1 0.464) = 0.574, respectively.
6
Note that 1.15 / (1 0.464).
7
Note that 0.74 / (1 0.445).
8
Note that 0.253 / (1 0.445).
9
Note that 0.353 / (1 0.445).
10
In addition, the short-run and long-run coefcients of the external balance on goods and services are obtained as 0.282 and
0.282 / (1 0.445) = 0.508, respectively.
12 G. Gozgor / Emerging Markets Review 18 (2014) 118
percent increase (on average) in the current domestic credit to GDP ratio in the short run and an
approximately 2.01
11
percent increase (on average) in the long run.
5.2. Cross-section data estimation and leave-one-out (LOO) type of analysis
In this section we conduct some further robustness tests with respect to the countries in the sample
and evaluate whether or not there are inuential countries in the dataset. This is an important issue since
we use a relatively homogenous sample of 24 countries, leaving doubts on the generalizability of our
conclusions. At this point, we initially average the data over time and run the Ordinary Least Squares (OLS)
cross-section data estimation. Then, we performthe leave-one-out (LOO) type of analysis. We nowuse the
statistically signicant explanatory variables in the dynamic panel data estimations: broad money (% of
GDP), the lending interest rate differences (%), real openness, current account balance (% of GDP) and the
SKEW index (in logarithmic form) in the regressions and report their coefcients in Table 10.
The rst regression covers whole observations of 24 EMEs. In the second regression, we exclude South
Africa and Thailand in the analysis. In the exclusion, we focus on the ratios of domestic credit provided by
banking sector to GDP: According to the descriptive statistics in Table 1, South Africa, China, Malaysia, and
Thailand have the highest ratios of domestic credit provided by banking sector to GDP on average. We also
consider the difference between maximum and minimum observationswhich is a strong an indicator of
boom patternand select South Africa and Thailand as contradictory observations. In similar, the third
regression does not include Peru and Venezuela in the analysis due to they have the lowest ratios of
domestic credit provided by banking sector to GDP on average. In the fourth regression, we exclude South
Africa, Thailand, Peru and Venezuela in the estimation. Furthermore, the fth regression does not consider
Estonia and Latvia in the estimation. This exclusion is based on the volatility values (standard deviations)
11
Note that 1.11 / (1 0.445).
Table 8
Results of the dynamic panel data estimation of Arellano and Bond (1991).
Regressors I II III IV
Lagged domestic credit (% of GDP) 0.436 (0.060)

0.448 (0.060)

0.435 (0.059)

0.445 (0.060)

Lagged ination rate (%) 0.135 (0.148) 0.069 (0.147) 0.124 (0.148) 0.059 (0.147)
Lagged GDP per capita (ln) 10.35 (6.928) 10.25 (7.009) 8.621 (7.067) 7.893 (7.143)
Nominal exchange rate (LCU per USD) 0.0004 (0.003) 0.0008 (0.003) 0.001 (0.003) 0.001 (0.003)
Lagged deposit rate (%) 0.129 (0.179) 0.096 (0.128) 0.118 (0.180) 0.085 (0.182)
Broad money (% of GDP) 0.735 (0.090)

0.733 (0.091)

0.750 (0.089)

0.751 (0.090)

The US broad money (% of GDP) 0.081 (0.134) 0.030 (0.135) 0.006 (0.159) 0.085 (0.157)
Non-performing loans to gross total
loans (%)
0.050 (0.139) 0.025 (0.141) 0.069 (0.139) 0.052 (0.140)
Lending interest rate differences (%) 0.245 (0.116)

0.260 (0.119)

0.244 (0.116)

0.258 (0.119)

Lagged real openness 0.109 (0.047)

0.120 (0.047)

0.105 (0.047)

0.114 (0.047)

Lagged current account balance

(% of GDP)
0.369 (0.100)

0.351 (0.103)

Lagged balance of trade (% of GDP) 0.299 (0.119)

0.280 (0.120)

Dummy for the great global recession 0.577 (1.369) 1.085 (1.373)
The 3 months LIBOROIS spread (%) 1.977 (1.845) 2.898 (1.829)
The VIX index (ln)
The SKEW index (ln)
Observations 205 205 205 205
Sargan test [0.00] [0.00] [0.00] [0.00]
AR(1) [0.00] [0.00] [0.00] [0.00]
AR(2) [0.75] [0.69] [0.62] [0.57]
Notes: The dependent variable is the ratio of domestic credit provided by the banking sector to GDP. The constant term is also
estimated but not reported. The Sargan test shows the results of the over-identifying restrictions (null hypothesis: there is
over-identication in the estimation specication). AR(1) and AR(2) show results of the LM statistics for autocorrelation (null
hypothesis: no rst-order and no second-order autocorrelation, respectively). We report robust standard errors. Standard errors are
in parentheses, and the p-values are in brackets.

and

indicate statistical signicance at the 1% and 5% levels, respectively.
13 G. Gozgor / Emerging Markets Review 18 (2014) 118
of the ratios of domestic credit provided by banking sector to GDP in Table 1. The domestic credit cycles in
Latvia and Estonia present a volatile structure compare to other observations in the data set; therefore,
they may also introduce a break down in the main results. In the sixth regression, we exclude Latvia,
Estonia, Peru, and Venezuela. Similarly, the seventh regression excludes Latvia, Estonia, South Africa, and
Thailand; while the eighth regression leaves out all possible extreme cases: Latvia, Estonia, South Africa,
Thailand, Peru, and Venezuela in the cross-section data estimations.
The results in Table 10 indicate that the coefcients for the explanatory variables in the cross-section
data estimations are statistically signicant in all cases. In addition, they have expected same signs with
the dynamic panel data estimations. Therefore, the results of the OLS cross-section data estimations
strongly support the main results of the ArellanoBond regressions. In short, we run the LOO type of
analysis in the study and our main conclusions stick up when some of the observations are not included in
the sample; thus the results are robust.
6. Discussion and implications
There are several possible economic and nancial explanations for our empirical results. First, we nd
that increases in the domestic money supply, or more generally, loose domestic monetary policy,
positively affect EME credit levels. According to both the short-run and long-run coefcients, this is the
most important (internal demand) factor in domestic credit expansion in EMEs, especially in the long
term. In fact, loose monetary policy is generally designed to encourage investment-led economic growth
in a developing economy, an objective that necessarily involves enabling economic agents to obtain credit
more easily and cheaply. Therefore, businesses, investors and consumers seize the opportunity of cheap
nance and domestic credit demand increases. This nding is in line with previous ndings of Aisen and
Franken (2010), Guo, and Stepanyan (2011), Elekdag and Wu (2011), Elekdag and Han (2012), and Magud
et al. (2012). However, we nd statistically insignicant effects of another internal demand factor
domestic incomeon domestic credit expansion; in other words, there is no signicant relationship that
runs from income to domestic credit. Contrary to the early ideas by Knut Wicksell, Ludwig von Mises,
Table 9
Results of the dynamic panel data estimation of Arellano and Bond (1991).
Regressors V VI VII VIII
Lagged domestic credit (% of GDP) 0.445 (0.058)

0.462 (0.058)

0.438 (0.059)

0.449 (0.059)

Lagged ination rate (%) 0.139 (0.145) 0.083 (0.145) 0.087 (0.150) 0.022 (0.149)
Lagged GDP per capita (ln) 10.91 (7.001) 10.37 (7.076) 5.772 (7.429) 4.576 (7.510)
Nominal exchange rate (LCU per USD) 0.001 (0.003) 0.001 (0.003) 0.001 (0.003) 0.0008 (0.003)
Lagged deposit rate (%) 0.123 (0.078) 0.098 (0.110) 0.063 (0.184) 0.211 (0.186)
Broad money (% of GDP) 0.729 (0.085)

0.720 (0.086)

0.753 (0.087)

0.749 (0.088)

The US broad money (% of GDP) 0.150 (0.138) 0.137 (0.139) 0.004 (0.139) 0.049 (0.138)
Non-performing loans to gross total
loans (%)
0.046 (0.138) 0.019 (0.139) 0.062 (0.137) 0.037 (0.138)
Lending interest rate differences (%) 0.270 (0.123)

0.303 (0.125)

0.216 (0.117)

0.227 (0.119)

Lagged real openness 0.094 (0.050)

0.098 (0.051)

0.118 (0.047)

0.131 (0.048)

Lagged current account balance (% of GDP) 0.364 (0.101)

0.329 (0.103)

Lagged balance of trade (% of GDP) 0.297 (0.119)

0.252 (0.122)

Dummy for the great global recession

The 3 months LIBOROIS spread (%)
The VIX index (ln) 1.132 (1.498) 1.674 (1.504)
The SKEW index (ln) 28.80 (17.11)

36.01 (16.97)

Observations 205 205 205 205

Sargan test [0.00] [0.00] [0.00] [0.00]
AR(1) [0.00] [0.00] [0.00] [0.00]
AR(2) [0.32] [0.39] [0.31] [0.42]
Notes: The dependent variable is the ratio of domestic credit provided by banking sector to GDP. The constant term is also estimated
but not reported. The Sargan test shows the results of the over-identifying restrictions (null hypothesis: there is over-identication
in the estimation specication). AR(1) and AR(2) showresults of the LM statistics for autocorrelation (null hypothesis: no rst-order
autocorrelation and no second-order autocorrelation, respectively). We report robust standard errors. Standard errors are in
parentheses, and the p-values are in brackets.

,

and

indicate statistical signicance at the 1%, 5% and 10% levels, respectively.
14 G. Gozgor / Emerging Markets Review 18 (2014) 118
Joseph A. Schumpeter, and Friedrich August von Hayek, Joan Robinson (1952) argued that an increase in
output (income) increases the demand for nancial services, suggesting a positive inuence of domestic
income on credit expansion. In fact, this argument against the classical Schumpeterian view is not be
veried by our evidence. However, whether relation runs in the opposite direction, from domestic credit
to income, as in the Schumpeterian view, is not a question that lies within the scope in this paper.
Second, we nd a positive effect on domestic credit of domestic and global lending rate differences.
This nding accords with Magud et al. (2012). In fact, the increasing interest rate spread between
domestic and global markets would allow banks in EMEs to borrow at the lower global interest rate and
lend at the higher domestic interest rate. Moreover, semi-oligopolistic banking systems in most EMEs
and banks' superior appetite for risk could help explain higher interest rate spreads that give rise to
domestic credit upsurges. This variable most likely overshadows the impact of capital ows on domestic
credit, as we nd a statistically insignicant effect of private capital ows on domestic credit, a nding that
accords with Hume and Sentence (2009). Indeed, capital ows are mostly associated only with credit
booms (not booms and busts) in the literature.
12
Fortunately, the time period used in this paper enables us
to avoid results that pertain only to credit booms or credit busts. On the other hand, the short-run impact
of capital inows on domestic credit depends on countries' macroeconomic policy stances. This issue calls
for further research.
Third, we nd that real openness is positively related to expansions of domestic credit. In other words,
the empirical results indicate that trade openness, as measured by differences between countries in
relative prices of non-tradable goodswhich in turn reect productivity differences is a signicant
determinant of the volume of domestic credit in EMEs. In the literature, trade openness is generally
viewed as a key factor in capital ows. Therefore, the impact of trade openness on domestic credit is
generally seen as indirect. According to our ndings, however, real openness and the spread between
global and domestic lending rates have a direct effect on domestic credit in EMEs, possibly larger than that
of capital ows. The reason is that increasing trade openness means that a developing economy is more
integrated into the global nancial and economic system. This enables banks in EMEs to borrow at lower
interest rates in global markets, causing an expanded domestic credit supply.
Fourth, another driver of domestic credit expansion is external balances, found to be negatively
associated with the domestic credit expansion in EMEs, ndings that are in line with Mendoza and
Terrones (2012) and Lane and McQuade (2013). External imbalances are certainly a noteworthy problem
in EMEs. Not only do they undermine domestic residents' ability to borrow from abroad at lower interest
12
See, International Monetary Fund (2011), Mendoza and Terrones (2012), and Ostry et al. (2012) for details of the issue.
Table 10
Results of the cross-section data estimations and LOO type of analysis.
Regressors I II III IV V VI VII VIII
Broad
money
1.085
(0.187)

0.907
(0.081)

1.055
(0.210)

0.863
(0.232)

1.104
(0.197)

1.066
(0.221)

0.926
(0.071)

0.872
(0.067)

Interest rate
difference
0.249
(0.104)

0.526
(0.174)

0.281
(0.082)

0.556
(0.124)

0.282
(0.063)

0.318
(0.147)

0.572
(0.218)

0.609
(0.189)

Real
openness
0.006
(0.001)

0.103
(0.046)

0.018
(0.008)

0.093
(0.028)

0.041
(0.016)

0.054
(0.014)

0.063
(0.018)

0.049
(0.021)

Current
account
balance
1.317
(0.346)

0.528
(0.214)

1.186
(0.246)

0.286
(0.141)

1.052
(0.538)

0.838
(0.384)

0.186
(0.042)

0.174
(0.076)

The SKEW 35.22

(15.23)

33.69
(13.11)

36.11
(16.03)

34.24
(16.23)

34.52
(15.14)

34.07
(12.18)

33.91
(15.96)

32.54
(16.81)

Observations 24 22 22 20 22 20 20 18
Adjusted
R-squared
0.67 0.88 0.64 0.88 0.68 0.65 0.91 0.93
Notes: The dependent variable is the ratio of domestic credit provided by banking sector to GDP. The constant term is also estimated
but not reported. We use robust standard errors. Standard errors are in parentheses.

,

and

indicate statistical signicance at
the 1%, 5% and 10% levels, respectively.
15 G. Gozgor / Emerging Markets Review 18 (2014) 118
rates, attraction of FDI and foreign portfolio investments. They also reduce investors' and businesses'
appetite for risk and consumers' demands for goods and services. External balances are linked to both
external supply and internal demand factors in creating an environment of credit scarcity or richness in
developing economies.
Fifth, recent empirical evidences such as those obtained by Tang et al. (2008) and Arndt et al. (2010)
show that external funding complements domestic credit in developing economies. On the contrary, if
foreign rms compete against domestic rms in the use of domestic credits, foreign and domestic funding
will be substitutes and this eliminates investment opportunities for domestic rms (Agosin and Machado,
2005). In line with the issue, this paper discusses the role of the external factors in understanding
domestic credit behavior and controls the effects of the downside indicators such as the VIX and the SKEW
indices on domestic credit. We also address a relevant question and test the hypothesis whether domestic
credit and external nancing are substitutes or complements in general and whether this behavior
changes with respect to the business cycle phase or global economic conditions. The ndings show that
the relationship between the domestic credit and the foreign funding is statistically insignicant; thus
they are neither complements nor substitutes in 24 EMEs. These results are in line with the recent ndings
of Furceri et al. (2012). In fact, we nd that the S&P 500 SKEW index, a strong benchmark measure of the
perceived risk of extreme negative movements in US equity markets and global nancial markets, has a
negative effect on domestic credit. This result is in the line with Bakker and Gulde (2010), who suggest the
role of external factors in determining domestic credit levels, and with Takats (2010), who emphasizes the
importance of volatility of the S&P 500 index as a global supply factor in domestic credit in EMEs. Indeed,
perceived risk of extreme negative movements in nancial markets might cause closure of the credit taps.
Our empirical results suggest that the SKEW index could also be a powerful global supply factor in
domestic credit availability in EMEs. We draw out the implications for offsetting policy responses to tail
risk indicated by the SKEW index.
More generally, using the domestic money supply as a policy tool, policy-makers in developing
economies should carefully calibrate their monetary policy stance to counter external factors (i.e., the
external imbalances of the country, the perceived global risks, the benchmark interest rate in global
markets, and economic and nancial conditions at the international level), thus avoiding the transmission
boombust credit cycles into their domestic economies. Economic and nancial dynamics particular to
each emerging market will, of course, modify these general implications.
7. Concluding remarks
This paper empirically investigates the determinants of domestic credit across a wide range of 24 EMEs
over the period 20002011. We use dynamic panel data estimation techniques to examine the short-run
and long-run effects of internal demand and external supply factors, external balances, and different
measures of global economic and nancial market conditions on domestic credit levels. We check the
robustness of our empirical ndings by considering the effects on domestic credit levels in EMEs of
perceived global risk and global uncertainty as well as the period effect of the great global recession of
200809.
We obtain several notable empirical results. First, we nd a strong persistence of domestic credit in
EMEs, a nding we conrm using the second generation PUR test of Pesaran (2007). This result allows us
to distinguish between the short run and the long run in interpreting the coefcients of the explanatory
variables. Second, the empirical results indicate that loose monetary policy conditions in the domestic
market, lending rate differences between domestic and global markets, and real trade openness
systematically and positively contribute to domestic credit expansion. Third, we nd that external
balances contribute negatively to domestic credit expansion. Fourth, we nd that perceived risk of
extreme negative movements in global nancial markets suppresses domestic credit.
Within this context, we investigate the economic and nancial roots of our empirical results and
discuss possible implications in the literature. The main empirical ndings in this paper concern the role of
loose monetary policy in domestic credit expansion, a result in line with Aisen and Franken (2010), Guo,
and Stepanyan (2011), Elekdag and Wu (2011), Elekdag and Han (2012), and Magud et al. (2012). The
remaining results also accord with recent papers, such as those by Takats (2010), Mendoza and Terrones
(2012), and Lane and McQuade (2013). Future studies of this issue might use different panel data
16 G. Gozgor / Emerging Markets Review 18 (2014) 118
estimation techniques, such as the system GMM panel and bias-corrected least square dummy variable
estimators, to examine possible determinants of domestic credit levels in EMEs. On the other hand,
another empirical strategy is that to employ a regime-dependent model for understanding behavior of
domestic credit in EMEs. However, we also leave it to another study.
References
Agosin, M., Machado, R., 2005. Foreign investment in developing countries: does it crowd in domestic investment? Oxf. Dev. Stud. 33
(2), 149162.
Aisen, A., Franken, M., 2010. Bank credit during the 2008 nancial crisis: a cross-country comparison. International Monetary Fund
Working Paper, No. 10/47.
Alcala, F., Ciccone, A., 2004. Trade and productivity. Q. J. Econ. 119 (2), 613646.
Anderson, T.W., Hsiao, C., 1982. Formulation and estimation of dynamic models using panel data. J. Econ. 18 (1), 4782.
Arellano, M., Bond, S., 1991. Some tests of specication for panel data: Monte Carlo evidence and an application to employment
equations. Rev. Econ. Stud. 58 (2), 277297.
Arndt, C., Buch, C.M., Schnitzer, M.E., 2010. FDI and domestic investment: an industry-level view. B.E. J. Econ. Anal. Policy 10 (1),
122.
Avdjiev, S., McCauley, R., McGuire, P., 2012. Rapid credit growth and international credit: challenges for Asia. Bank for International
Settlements Working Paper, No. 377.
Bakker, B., Gulde, A., 2010. The credit boom in the EU new member states: bad luck or bad policies? International Monetary Fund
Working Paper, No. 10/130.
Barajas, A., DellAriccia, G., Levchenko, A., 2009. Credit booms: the good, the bad, and the ugly, Washington, DC: International
Monetary Fund, mimeo.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. J. Econ. 87 (1), 115143.
Borio, C., McCauley, R., McGuire, P., 2011. Global credit and domestic credit booms. Bank Int. Settl. Q. Rev. 3, 4357.
Cheung, Y., Chinn, M.D., Qian, X.W., 2012. Are Chinese trade ows different? J. Int. Money Financ. 31 (8), 21272146.
Edwards, S., 2012. The Federal Reserve, emerging markets, and capital controls: a high frequency empirical investigation. J. Money
Credit Bank. 44 (S2), 151184.
Elekdag, S., Han, F., 2012. What drives credit growth in emerging Asia? International Monetary Fund Working Paper, No. 12/43.
Elekdag, S., Wu, Y., 2011. Rapid credit growth: boon or boombust? International Monetary Fund Working Paper, No. 11/241.
Fernandez-Villaverde, J., Garicano, L., Santos, T., 2013. Political credit cycles: the case of the Eurozone. J. Econ. Perspect. 27 (3),
145166.
Frankel, J., Romer, D., 1999. Does trade cause growth? Am. Econ. Rev. 89 (3), 379399.
Fungacova, Z., Herrala, R., Weill, L., 2013. The inuence of bank ownership on credit supply: evidence from the recent nancial crisis.
Emerg. Mark. Rev. 15, 136147.
Furceri, D., Guichard, S., Rusticelli, E., 2012. The effect of episodes of large capital inows on domestic credit. N. Am. J. Econ. Finance
23 (3), 325344.
Gopinath, G., Itskhoki, O., Neiman, B., 2012. Trade prices and the global trade collapse of 200809. IMF Econ. Rev. 60 (3), 303328.
Gourinchas, P., Obstfeld, M., 2012. Stories of the twentieth century for the twenty-rst. Am. Econ. J. Macroecon. 4 (1), 226265.
Guo, K., Stepanyan, V., 2011. Determinants of bank credit in emerging market economies. International Monetary Fund Working
Paper, No. 11/51.
Hartelius, K., Kashiwase, K., Kodres, L., 2008. Emerging market spread compression: is it real or is it liquidity? International Monetary
Fund Working Paper, No. 08/10.
Hume, M., Sentence, A., 2009. The global credit boom: challenges for macroeconomics and policy. J. Int. Money Financ. 28 (8),
14261461.
International Monetary Fund, 2011. World Economic Outlook: Slowing Growth, Rising Risks. International Monetary Fund,
Washington, DC.
Jorda, O., Schularick, M., Taylor, A.M., 2011. Financial crises, credit booms, and external imbalances: 140 years of lessons. IMF Econ.
Rev. 59 (2), 340378.
Kamil, H., Rai, K., 2010. The global credit crunch and foreign banks' lending to emerging markets: why did Latin America fare better?
International Monetary Fund Working Paper, No. 10/102.
Kim, S.J., Wu, E., 2008. Sovereign credit ratings, capital ows and nancial sector development in emerging markets. Emerg. Mark.
Rev. 9 (1), 1739.
Lane, P.R., McQuade, P., 2013. Domestic credit growth and international capital ows. European Central Bank Working Paper Series,
No. 1566.
Luca, O., Spatafora, N., 2012. Capital inows, nancial development, and domestic investment: determinants and inter-relationships.
International Monetary Fund Working Paper, No. 12/120.
Magud, N.E., Reinhart, C.M., Vesperoni, E.R., 2012. Capital inows, exchange rate exibility, and credit booms. International Monetary
Fund Working Paper, No. 12/41.
Mamatzakis, E., Staikourasa, C., Koutsomanoli-Fillipakia, N., 2005. Competition and concentration in the banking sector of the south
eastern European region. Emerg. Mark. Rev. 6 (2), 192209.
Mendoza, E.G., Terrones, M.E., 2008. An anatomy of credit booms: evidence from macro aggregates and micro data. National Bureau
of Economic Research Working Paper, No. 14049.
Mendoza, E.G., Terrones, M.E., 2012. An anatomy of credits booms and their demise. J. Econ. Chil. (Chil. Econ.) 15 (2), 432.
Mishkin, F.S., 2010. Monetary policy exibility, risk management, and nancial disruptions. J. Asian Econ. 21 (3), 242246.
Montoro, C., Suarez, L.R., 2012. Credit in times of stress: lessons from Latin America. Bank Int. Settl. Q. Rev. 3, 5160.
Obstfeld, M., 2012. Does the current account still matter? Am. Econ. Rev. 102 (3), 123.
Obstfeld, M., Rogoff, K.S., 2010. Global imbalances and the nancial crisis: products of common causes. In: Glick, R., Spiegel, M.M.
(Eds.), Asia and the Global Financial Crisis. Federal Reserve Bank of San Francisco, San Francisco.
17 G. Gozgor / Emerging Markets Review 18 (2014) 118
Ostry, J.D., Ghosh, A.R., Chamon, M., Qureshi, M.S., 2012. Tools for managing nancial-stability risks from capital inows. J. Int. Econ.
88 (2), 407421.
Pesaran, M.H., 2004. General diagnostic tests for cross section dependence in panels. IZA Discussion Paper Series, No. 1240.
Pesaran, M.H., 2007. A simple panel unit root test in the presence of cross-section dependence. J. Appl. Econ. 22 (2), 265312.
Robinson, J., 1952. The generalization of the general theory. In the Rate of Interest and Other Essay.Macmillan, London.
Romer, D., 1993. Openness and ination: theory and evidence. Q. J. Econ. 108 (4), 869903.
Schularick, M., Taylor, A.M., 2012. Credit booms gone bust: monetary policy, leverage cycles and nancial crises, 18702008. Am.
Econ. Rev. 102 (2), 10291061.
Takats, E., 2010. Was it credit supply? Cross-border bank lending to emerging market economies during the nancial crisis. Bank Int.
Settl. Q. Rev. 2, 4956.
Tang, S., Selvanathan, E., Selvanathan, S., 2008. Foreign direct investment, domestic investment and economic growth in China: a
time series analysis. World Econ. 31 (10), 12921309.
18 G. Gozgor / Emerging Markets Review 18 (2014) 118