The Determinants of Cross-Border Lending in the Euro Zone
The 14th
Dubrovnik Economic ConferenceJune 26-27, 2008
Sylvia Heuchemer Cologne University of Applied Sciences
Stefanie Kleimeier (not present)Maastricht University and METEOR Fellow
Harald SanderCologne University of Applied Sciences, and METEOR Fellow
2 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Introduction (1)
European financial market integration has made a big leap forward.
However, retail banking is lagging far behind, though recently gaining some momentum.
We explore the geography of cross-border lending in the euro zone by means of a gravity approach in order to: identify trade-theoretic determinants of cross-border lending identify important drivers of and barriers to integration special focus on exploring the role of cultural and political
differences
3 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Introduction (2) Gravity models have recently been used in explaining
cross-border finance…(e.g. Guiso,Sapienza & Zingales 2005, Blank & Buch 2006)
…and highlighted the limiting role of cultural and political factors in economic exchange
(Guiso, Sapienza & Zingales 2005, Heuchemer & Sander 2007, Kalemli-Ozcan & Sørensen 2007)
Our approach innovates on: using a non publicly available bilateral data on cross-border
loans for the euro zone adapting a trade-theoretically-augmented gravity model to
cross-border lending modeling cultural and political differences in a gravity approach
4 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Data
Bilateral cross-border loans Annual outstanding volume from 1999-2006 Austria, Belgium, Finland, France, Germany,
Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain except loans to customers in Lux and Por
More than 800 countrypair-year specific observations
5 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
- Increase in banking market integration- Lending is right-skewed, indicating that lending
activities are concentrated in a few countries
Cross-border loansmore than doubled
Table 1: The growth of bilateral cross-border lending in the euro zone over time
total mean median
yearin millions of
euroin millions
of euro
in % of GDP of
bank country
in % of GDP of
customer country
in millions of euro
in % of GDP of
bank country
in % of GDP of
customer country
number of country-
pairs1999 151.976,3 1.746,9 5,1 10,6 854,0 1,5 2,2 872000 174.443,0 1.982,3 5,3 11,2 796,0 1,5 2,2 882001 205.113,3 1.991,4 4,8 11,1 647,0 1,2 2,3 1032002 212.000,5 2.058,3 4,4 10,6 761,3 1,2 2,4 1032003 228.407,3 2.175,3 4,3 11,8 850,0 1,3 2,1 1052004 251.659,8 2.352,0 4,7 11,7 861,8 1,2 2,1 1072005 302.286,3 2.851,8 5,3 15,0 1.165,0 1,7 2,0 1062006 360.896,0 3.341,6 5,7 16,6 1.304,7 1,8 2,4 108
1999 to 2006 2.338,0 4,9 12,4 920,5 1,4 2,2 807
bilateral cross-border loan volume
Data
6 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Methodology (1) Gravity models have been successfully applied for explaining
transactions over space, mainly trade in goods The canonical gravity model explains trade flows between two
countries as a function of their respective economic masses (GDP) and the geographical distance separating them - as a proxy for all
information and transaction costs
2ij
jtitijt
DISTANCE
GDPGDPGX
ij3jt2it10ijt lnDISTANCEβlnGDPβlnGDPβα)ln(X
7 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Methodology (2) Augmented gravity models
are micro-founded by adapting them to trade theories(e.g. Anderson 1979, Bergstrand 1985 & 1989 Deardorf, 1989, Anderson & van Wincoop 2003, Baltagi et al. 2003) Size (sum of GDPs) “new trade theory” Similarity (of GDP) “new trade theory” Relative per-capita income Ricardo-HOS-theory vs. Linder
include additional trade impeding or trade promoting factors as proxy for information and transaction cost distance, common border, language alternative measures of cultural distances (“Hofstede”, trust
to/trust from, differences in trust levels, legal system origin) Differences in governance quality
8 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Methodology (3)Augmented gravity model adjusted to banking market analysis:
ijt
K
6kijtkij5
ij4ijt3ijt2ijt10ijt
ulnYβBORDERβ
lnDISTANCEβlnSIMILARβlnRELβlnSIZEβαlnX
With• Xijt: cross-border loans from bank country to customer country at time t
• Size: sum of the GDPs of the trading partner (“new” trade theory)
)GDPGDPln(SIZE jtitijt
9 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Methodology (3)
• Size: sum of the GDPs of the trading partner (“new” trade theory)
Augmented gravity model adjusted to banking market analysis:
ijt
K
6kijtkij5
ij4ijt3ijt2ijt10ijt
ulnYβBORDERβ
lnDISTANCEβlnSIMILARβlnRELβlnSIZEβαlnX
With• Xijt: cross-border loans from bank country to customer country at time t
• Similar: similarity of the size of the financial sector (“new” trade theory)
22
1jtit
jt
jtit
itijt CreditCredit
Credit
CreditCredit
CreditlnSIMILAR
10 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Methodology (3)
• Size: sum of the GDPs of the trading partner (“new” trade theory)
Augmented gravity model adjusted to banking market analysis:
ijt
K
6kijtkij5
ij4ijt3ijt2ijt10ijt
ulnYβBORDERβ
lnDISTANCEβlnSIMILARβlnRELβlnSIZEβαlnX
With• Xijt: cross-border loans from bank country to customer country at time t
• Similar: similarity of the size of the financial sector (“new” trade theory)
• REL: indicator for relative financial development (“old” or “new” trade theory)
jt
jt
it
itijt GDP
Creditln
GDP
CreditlnREL
11 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Methodology (3)Augmented gravity model adjusted to banking market analysis:
ijt
K
6kijtkij5
ij4ijt3ijt2ijt10ijt
ulnYβBORDERβ
lnDISTANCEβlnSIMILARβlnRELβlnSIZEβαlnX
With• Xijt: cross-border loans from bank country to customer country at time t• Size: sum of the GDPs of the trading partner (“new” trade theory)
• REL: indicator for relative financial development (“old” or “new” trade theory)
• Similar: similarity of the size of the financial sector (“new” trade theory)
• Distance: geographical distance between two countries
• Border: dummy variable that is 1 if two countries share a common border and 0 otherwise
• Yijt: additional factors possibly influencing cross-border lending
12 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
1. Static Fixed Vs Random Effects Estimation
2. Static Vs Dynamic Fixed Effects Estimation habit persistence may be likely in banking
(switching and information costs)
3. Least Square Dummy Variables (LSDV) Estimation
Fixed effects estimations eliminate all time-invariant country-pair specific variables we are interested in
LSDV can remedy this, but we have to model the country-pair specific effects
Estimation Procedure
13 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Estimation Procedure 1: Fixed Vs Random Effects in Loan Markets
None of the explanatory variables are statistically significant.
Variation in cross-border loans can be explained by country-pair and time specific factors.
Credit market proxies for
SIMILAR and REL
SIZE 0,83 1,56 *1,18 6,43
SIMILAR -0,20 0,02-0,92 0,09
REL -0,07 -0,13-0,33 -0,63
countrybank dummies No
countrycustomer dummies Noyear dummies Yes
Hausman test statstics (HT) 48,60HT (p-value) 0,00
adjusted R2 0,938number of observations 807 807Note: In each regression, the dependent variable is the log of bilateralcross-border loan volume. The two-way fixed effects model is estimatedwith White-robust standard errors. For each coefficient, the first rowshows the estimated coefficient and the second row the t-statistic. *indicates significance at least at 10%.
One-way random effects with time
dummies
Table 2: Determinants of cross border lendingFixed effects
0,983
14 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Hausman tests substantiate our choice of fixed effects model.
Strong evidence for country-pair fixed effects Fixed time effects, capturing economic
shocks and account for important regulatory and behavioral changes in the first years of the currency union
Estimation Procedure 1: Fixed Vs Random Effects Estimation
15 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Estimation Procedure 2: Dynamic Fixed Effects
Habit persistence is likely in retail banking (information and transaction costs)
Arellano-Bond Estimator
16 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Data and Estimation Procedure
Credit market proxies for
SIMILAR and REL
Credit market proxies for
SIMILAR and REL
SIZE 0,70 0,301,67 0,29
SIMILAR 0,52 * 0,062,20 0,18
REL -0,08 -0,10-0,24 -0,35
ln(BILOANSt-1) 0,62 * 0,40 *6,56 3,69
countrybank dummies No No
countrycustomer dummies No Noyear dummies No Yes
Sargan test 28,62 19,86p-value 0,10 0,47m2 -0,89 -0,83p-value 0,37 0,41
number of observations 603 603Note: In each regression, the dependent variable is the log of bilateral cross-border loan volume. The dynamic model uses the Arellano-Bond (1991) 2-step generalised method of moments (GMM) estimator. * indicatessignificance at least at 10%.
Table 3: Determinants of cross-border lendingDynamic
0,70 0,30 Model performs better without time dummies capturing regulatory changes.
no yesYear dummies In order to capture regulatory changes
17 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Data and Estimation Procedure
Credit market proxies for
SIMILAR and REL
Credit market proxies for
SIMILAR and REL
SIZE 0,70 0,301,67 0,29
SIMILAR 0,52 * 0,062,20 0,18
REL -0,08 -0,10-0,24 -0,35
ln(BILOANSt-1) 0,62 * 0,40 *6,56 3,69
countrybank dummies No No
countrycustomer dummies No Noyear dummies No Yes
Sargan test 28,62 19,86p-value 0,10 0,47m2 -0,89 -0,83p-value 0,37 0,41
number of observations 603 603Note: In each regression, the dependent variable is the log of bilateral cross-border loan volume. The dynamic model uses the Arellano-Bond (1991) 2-step generalised method of moments (GMM) estimator. * indicatessignificance at least at 10%.
Table 3: Determinants of cross-border lendingDynamic
no yesYear dummies
Cross-border lending is habit persistent.
0,62 0,40
0,52
Cross-border lending is promoted by similarity of financial systems.
18 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Estimation Procedure 3: LSDV
Summary of the Fixed Effects Models Strong evidence for country-pair FE Habit persistence in loan markets
Least Square Dummy Variables (LSDV) Estimation
Fixed effects estimations eliminate all time-invariant country-pair specific variables which we are interested in
LSDV can remedy this, but we have to model the country-pair specific effects
19 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Estimation Procedure 3: LSDV Country-pair fixed effects are replaced by country-pair specific
economic, cultural and political variables measured as Euclidean distance or as dummy variables
K
kjtkitkijt VVED
1
2
with V as possible influencing factors on cross-border lending
Country specific effects (country dummies) are included, capturing the effects of “multilateral resistance”
Time dummies to account (at least partly) for habit persistence
20 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Cultural proxies Common language Legal family (La Porta et. al) Trust_from/Trust_to (Eurobarometer and Guiso et. al) Trust levels (World Value Survey) Overall cultural proxy based on Hofstede’s cultural
dimensions (power distance, individualism, masculinity and uncertainty avoidance)
Political proxies Dimension of governance as defined by the World Bank
(voice and accountability, political stability, government effectiveness, regulatory quality, rule of law, control of corruption)
Overall political proxy based on all six dimensions
Data (extended): Cultural and Political Proxies
21 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Table 4: Determinants of cross-border lendingconstant SIZE SIMILAR REL ln(DISTANCE) BORDERD ln(TRUSTED) LEGALFAMD ln(FRGBNKED) ln(VOICEED) adj. R2
obs.
LSDV 3,07 1,32* 0,54** 0,43** -0,59** 0,54* 0,830 8071,15 2,90 2,38 2,00 -6,44 5,79
LSDV 5,09** 0,84** 0,51** 0,54** -0,56* 0,48* -0,24* 0,55* 0,13* 0,14* 0,844 8072,00 1,97 2,31 2,45 -6,10 5,46 -5,11 5,37 3,40 3,73
Note: In each regression, the dependent variable is the log of bilateral cross-border loan volumne. LSDV estimation wie country and year dummies. White-robust t-values in second row. The subscript D indicates a dummy variable and the subscript ED a variable measured as an euclidian distance.*, **, and *** indicate significance at 1%, 5%, and 10% level, respectively.
LSDV Results: LoansBasic Version
Size1,32
SIM0,54
REL0,43
R2
0,830Dist.-0,59
Border0,54
Economic Geography
Common borders increase cross-border banking by 71,6%
Intra-industry trade- Product heterogeneity- Economies of scale
„New“ Trade Theory „Old“ Trade Theory
Comparative Advantages
22 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Table 5: LSDV model selectionCross-border loans
Reg 1 Reg 2 Reg 3 Reg 4 Reg 5 Reg 6 Reg 7 Reg 8 Reg 9 Reg 10 Reg 11 Reg 12constant 3,07 3,46 2,88 3,47 2,38 3,00 2,99 3,27 2,63 1,95 4,27 3,08
1,15 1,32 1,07 1,32 0,90 1,09 1,11 1,20 0,95 0,71 1,60 1,15SIZE 1,32 1,18 1,32 1,34 1,20 1,32 1,32 1,31 1,39 1,49 1,15 1,32
2,90 2,66 2,90 3,00 2,71 2,87 2,90 2,87 2,95 3,17 2,52 2,91SIMILAR_CRE 0,54 0,56 0,55 0,62 0,52 0,54 0,54 0,54 0,57 0,61 0,46 0,54
2,38 2,46 2,42 2,76 2,32 2,36 2,38 2,37 2,43 2,60 2,05 2,38REL_CRE 0,43 0,54 0,43 0,39 0,43 0,43 0,43 0,44 0,44 0,43 0,45 0,43
2,00 2,40 2,01 1,85 2,00 2,00 2,01 2,05 2,02 2,01 2,08 1,98ln(DISTANCE) -0,59 -0,58 -0,56 -0,66 -0,47 -0,59 -0,58 -0,61 -0,59 -0,58 -0,60 -0,60
-6,44 -6,27 -5,44 -7,27 -4,90 -6,42 -6,45 -6,65 -6,41 -6,32 -6,74 -6,42BORDERD 0,54 0,53 0,52 0,48 0,52 0,54 0,54 0,53 0,53 0,53 0,56 0,54
5,79 5,63 5,53 5,24 5,95 5,82 5,75 5,68 5,61 5,62 5,95 5,75ln(FRGBNKED) 0,11
2,79LANGUAGED 0,11
0,71ln(TRUSTED) -0,24
-5,08LEGALFAMD 0,66
6,27ln(POLED) -0,01
-0,16ln(CORRUPED) -0,01
-0,47ln(GOVEFFED) 0,03
0,81ln(POLSTABED) -0,03
-0,88ln(REGQALED) -0,06
-1,75ln(VOICEED) 0,09
2,50ln(LAWED) 0,00
0,07
countrylender dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yescountryborrower dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yesyear dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
adjusted R2 0,830 0,831 0,830 0,835 0,836 0,830 0,830 0,830 0,830 0,830 0,831 0,830number of observations 807 807 807 807 807 807 807 807 807 807 807 807Note: In each regression, the dependent variable is the log of bilateral cross-border loan volume. The Least Square Dummy Variable (LSDV) model isestimated as OLS with White-robust standard errors. For each coefficient, the first row shows the estimated coefficient and the second row the t-statistic.The subscript D indicates a dummy variable and the subscript ED a variable measured as an Euclidean distance where larger values indicate largerdifferences between borrower and lender country.
LSDV Results: LoansImpact of Cultural and Political Variables
0,11Language
Corruption -0,01
Gov. Effectivness
Political Stability
0,03
-0,03
Reg. Quality
Rule of Law
-0,06
0,00
Statisticallyinsignificant
23 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Table 5: LSDV model selectionCross-border loans
Reg 1 Reg 2 Reg 3 Reg 4 Reg 5 Reg 6 Reg 7 Reg 8 Reg 9 Reg 10 Reg 11 Reg 12constant 3,07 3,46 2,88 3,47 2,38 3,00 2,99 3,27 2,63 1,95 4,27 3,08
1,15 1,32 1,07 1,32 0,90 1,09 1,11 1,20 0,95 0,71 1,60 1,15SIZE 1,32 1,18 1,32 1,34 1,20 1,32 1,32 1,31 1,39 1,49 1,15 1,32
2,90 2,66 2,90 3,00 2,71 2,87 2,90 2,87 2,95 3,17 2,52 2,91SIMILAR_CRE 0,54 0,56 0,55 0,62 0,52 0,54 0,54 0,54 0,57 0,61 0,46 0,54
2,38 2,46 2,42 2,76 2,32 2,36 2,38 2,37 2,43 2,60 2,05 2,38REL_CRE 0,43 0,54 0,43 0,39 0,43 0,43 0,43 0,44 0,44 0,43 0,45 0,43
2,00 2,40 2,01 1,85 2,00 2,00 2,01 2,05 2,02 2,01 2,08 1,98ln(DISTANCE) -0,59 -0,58 -0,56 -0,66 -0,47 -0,59 -0,58 -0,61 -0,59 -0,58 -0,60 -0,60
-6,44 -6,27 -5,44 -7,27 -4,90 -6,42 -6,45 -6,65 -6,41 -6,32 -6,74 -6,42BORDERD 0,54 0,53 0,52 0,48 0,52 0,54 0,54 0,53 0,53 0,53 0,56 0,54
5,79 5,63 5,53 5,24 5,95 5,82 5,75 5,68 5,61 5,62 5,95 5,75ln(FRGBNKED) 0,11
2,79LANGUAGED 0,11
0,71ln(TRUSTED) -0,24
-5,08LEGALFAMD 0,66
6,27ln(POLED) -0,01
-0,16ln(CORRUPED) -0,01
-0,47ln(GOVEFFED) 0,03
0,81ln(POLSTABED) -0,03
-0,88ln(REGQALED) -0,06
-1,75ln(VOICEED) 0,09
2,50ln(LAWED) 0,00
0,07
countrylender dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yescountryborrower dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yesyear dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
adjusted R2 0,830 0,831 0,830 0,835 0,836 0,830 0,830 0,830 0,830 0,830 0,831 0,830number of observations 807 807 807 807 807 807 807 807 807 807 807 807Note: In each regression, the dependent variable is the log of bilateral cross-border loan volume. The Least Square Dummy Variable (LSDV) model isestimated as OLS with White-robust standard errors. For each coefficient, the first row shows the estimated coefficient and the second row the t-statistic.The subscript D indicates a dummy variable and the subscript ED a variable measured as an Euclidean distance where larger values indicate largerdifferences between borrower and lender country.
LSDV Results: LoansImpact of Cultural and Political Variables
-0,24Trust
Legal Family
Foreign Bank 0,11
0,66
Voice + Account. 0,09
Culture -0,34
Statistically significant
24 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Table 4: Determinants of cross-border lendingconstant SIZE SIMILAR REL ln(DISTANCE) BORDERD ln(TRUSTED) LEGALFAMD ln(FRGBNKED) ln(VOICEED) adj. R2
obs.
LSDV 3,07 1,32* 0,54** 0,43** -0,59** 0,54* 0,830 8071,15 2,90 2,38 2,00 -6,44 5,79
LSDV 5,09** 0,84** 0,51** 0,54** -0,56* 0,48* -0,24* 0,55* 0,13* 0,14* 0,844 8072,00 1,97 2,31 2,45 -6,10 5,46 -5,11 5,37 3,40 3,73
Note: In each regression, the dependent variable is the log of bilateral cross-border loan volumne. LSDV estimation wie country and year dummies. White-robust t-values in second row. The subscript D indicates a dummy variable and the subscript ED a variable measured as an euclidian distance.*, **, and *** indicate significance at 1%, 5%, and 10% level, respectively.
LSDV Results: LoansPreferred Estimation
R2
0,844Size0,84
SIM0,51
REL0,54
Dist.-0,56
Border0,48
Legal0,55
Bank0,13
Voice0,14
Trust-0,24
Same legal system increases cross-border
loans by 73,3%
Cultural (rather than governance) factors play an important limiting role in banking market integration
Influence of economic geography is slightly
smaller
Intra-industry trade Comparative advantages
25 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Conclusion Cross-border lending is promoted by
similarity of financial systems but… …differences in factor endowments and
financial development still play a role. Cross-border lending is habit persistent. Strong geographic and cultural limits to
full(er) integration Distance and border effects Differences in legal system origin Cultural Differences (Trust)
26 Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Research Challenges Ahead
Towards a more theory-based measurement of cultural dimensions
Exploring the interaction of the various concepts and proxies for information and transaction costs (also for trade, FDI, international finance)
Efficient estimation of time invariant and rarely changing variables in gravity models with fixed effects (e.g. Hausman/Taylor, or more recently: Plümper/Troeger 2007)
Thank you for your attention!
Sylvia Heuchemer, Stefanie Kleimeier & Harald Sander
Cultural and geographical factors remain for quite some time a barrier to banking market integration in the euro zone.
Top Related