Financial Fragmentation and Economic Growth in Europe

Transcription

1 Financial Fragmentation and Economic Growth in Europe Isabel Schnabel University of Bonn, CEPR, CESifo, and MPI Bonn Christian Seckinger LBBW International Financial Integration in a Changing Policy Context - the End of an Era? European Commission Brussels, March 1, / 31

2 Motivation: Disintegration of European banking market Foreign Claims by Home Country of Banks (Trillions of U.S. dollars) Euro area Other Europe United States Other Source: IMF (Global Financial Stability Report, April 2015) Substantial deepening of banking sector integration in Europe before the crisis Sharp disintegration since the crisis 2 / 31

3 Disintegration reinforced by economic policy In Europe, the reduction in foreign claims was reinforced by policy measures Multinational banks that had been bailed out by national governments were forced to deleverage and concentrate on their core business (financial protectionism, Rose/Wieladek, 2014) Ring-fencing of foreign banks affiliates, making financial integration less attractive What effects did the financial fragmentation process have on banks host countries? 3 / 31

4 Financial integration and economic growth (pre-crisis) Long-standing debate about the costs and benefits of financial integration Mixed evidence on the relationship between financial integration and economic growth (Kose/Prasad/Rogoff/Wei, 2009) But: Europe is different It benefits significantly from financial integration (Guiso/Jappelli/Padula/Pagano, 2004, Prasad/Rajan/Subramanian, 2007, Abiad/Leigh/Mody, 2009, Friedrich/Schnabel/Zettelmeyer, 2013) Particularly strong benefits of foreign banks: They improve domestic credit access, alleviate financial constraints, and raise economic growth (Gianetti/Ongena, 2012, Bruno/Hauswald, 2013) Foreign bank presence is the most important type of financial integration in Europe 4 / 31

5 Financial integration and crises Financial integration in times of crises: Foreign subsidiaries have a stabilizing effect on bank lending in a crisis in the host country (De Haas/van Lelyveld, 2010) Banks tend to withdraw from foreign markets during crises in the home country: Flight home effect (Cetorelli/Goldberg, 2010, Gianetti/Laeven, 2012) No general run to the exit but differentiated withdrawals, depending on factors such as geographical closeness (De Haas/van Horen, 2013) Effects of financial integration in times of crises: Financial integration has no beneficial growth effects in decades with financial crises in a broad country sample (Eichengreen/Gullapalli/Panizza, 2011) Is Europe again different? 5 / 31

6 Contribution of this paper This paper reassesses the growth effect of banking sector integration (physical bank presence) in Europe from Result: Significant growth effect 6 / 31

7 Contribution of this paper This paper reassesses the growth effect of banking sector integration (physical bank presence) in Europe from Result: Significant growth effect... investigates whether the effect is different during the crisis Result: Foreign bank presence had a more than 4 times stronger growth effect than during normal times 6 / 31

8 Contribution of this paper This paper reassesses the growth effect of banking sector integration (physical bank presence) in Europe from Result: Significant growth effect... investigates whether the effect is different during the crisis Result: Foreign bank presence had a more than 4 times stronger growth effect than during normal times... analyzes whether the effect is different in times of domestic bank deleveraging and disintegration Result: Effect is much larger in times of domestic bank deleveraging, but not in phases of disintegration 6 / 31

9 Contribution of this paper This paper reassesses the growth effect of banking sector integration (physical bank presence) in Europe from Result: Significant growth effect... investigates whether the effect is different during the crisis Result: Foreign bank presence had a more than 4 times stronger growth effect than during normal times... analyzes whether the effect is different in times of domestic bank deleveraging and disintegration Result: Effect is much larger in times of domestic bank deleveraging, but not in phases of disintegration... asks whether similar growth effects can be obtained by cross-border lending Result: Results are less consistent, results on foreign bank presence remain robust when controlling for credit inflows 6 / 31

10 Implication of results Concerns of policymakers about increasing fragmentation of the European banking sector are justified Process of fragmentation generated significant losses in economic growth, especially in times of scarce domestic financing (during crises and in time of domestic deleveraging) Policy measures enforcing financial disintegration are very harmful for host countries 7 / 31

11 Methodology Application of Rajan-Zingales methodology (Rajan/Zingales, AER 1998) based on industry data Exploit that industries differ in their needs for external capital to finance their capital expenditures (rather than internal financing through retained earnings) Theoretical prediction: If foreign bank assets reduce financial constraints for industrial firms, then industries with a strong need for external financing should benefit more in terms of higher production growth. 8 / 31

12 Empirical model Production growth i,j,t =α i,j + β j,t + γ i,t + δ 0 Industry share i,j,t 1 + δ 1 Ext. dependence j Foreign bank assets i,t 1 + δ 2 Ext. dependence j Domestic bank assets i,t 1 + ɛ i,j,t i denotes countries, j denotes industries, t denotes years Industry share controls for convergence effects Model contains full sets of industry-country fixed effects α i,j (country-specific industry characteristics), industry-year fixed effects β j,t (industry-specific developments), and country-year fixed effects γ i,t (country-specific developments) to control for a large variety of unobserved heterogeneity 9 / 31

13 Empirical model Empirical model is designed to estimate short-term (annual) industry growth effects Foreign bank assets are measured relative to GDP to obtain growth effects of additional foreign capital, holding domestic capital constant (not: composition effects) Note: The overall effect of foreign bank assets is not identified because the base effect is absorbed by the country-time fixed effect Model is only able to identify differential effects on different industries 10 / 31

14 Empirical model Sample is split in order to analyze growth effects of banking sector integration for... normal times versus times of crisis times of deleveraging of domestic banks versus other times times of disintegration versus integration Table 1: Definition of crises, deleveraging, and disintegration Variable Value Definition Crisis 1 Beginning of crisis: Significant distress in banking sector and rescue measures by government (Laeven and Valencia, 2013) End of crisis: Real GDP growth positive and decrease in unemployment rate 0 otherwise Deleveraging i,t 1 Change in domestic bank assets / GDP from year t 1 to year t 35th percentile ( 25th, 15th) 0 otherwise Disintegration i,t 1 Change in foreign bank assets / GDP from year t 1 to year t 35th percentile ( 25th, 15th) 0 otherwise 11 / 31

15 Data - Industry production Dependent variable: Annual real growth of industry production Industry production data from Structural business statistics by Eurostat for all member countries of the European Union Industries are classified according to European business classification NACE Rev. 2 (since 2008, before NACE Rev. 1.1) on 4-digit-level and matched to industry classification ISIC Rev. 2 (original classification in Rajan and Zingales, 1998) Data cover the period from 2000 to 2012 including the financial and sovereign debt crisis Final dataset: 24 EU countries (excluding Croatia, Cyprus, Luxembourg, and Malta) 35 industrial sectors Panel is unbalanced 12 / 31

16 Data - External dependence Reliance on observed use of external finance problematic since observed financing behavior reflects financial frictions Idea of Rajan/Zingales (1998): Use the observed external financing of U. S. industries to assess the external dependence of industries in other countries, assuming that the U. S. capital market is subject to fewer financial frictions External dependence as an industry characteristic External dependence = ( capital expenditures - cash-flow from operations capital expenditures This study: Data on external dependence from Laeven/Valencia (2013) who updated the values of Rajan/Zingales (1998) by using firm-level data from ) 13 / 31

17 Data - External dependence Industrial Sector nobs External dependence Tobacco Leather Footwear Pottery Other chemicals Furniture Petroleum refineries Apparel Printing and publishing Basic excluding fertilizers Beverages Spinning Nonmetal products Pulp, paper Synthetic resins Transportation equipment Paper and products Wood products Food products Textile Metal products Iron and steel Plastic products Glass Petroleum and coal products Ship Nonferrous metal Rubber products Motor vehicle Electric machinery Machinery Other industries Office and computing Drugs Professional goods Radio Note: External dependence ratios are provided by Laeven and Valencia (2013) and estimated for the United States from / 31

18 Data - Bank assets over GDP Data on the size and composition of national banking sectors are obtained from the ECB Foreign bank assets: Total assets of branches and subsidiaries in the reference country that are owned by banks of foreign EU countries Domestic bank assets are the difference between total and foreign bank assets (including foreign assets from countries from outside the EU, but: minor role) 15 / 31

19 National banking sectors in Europe Table 2: Country-specific overview over banking sector integration Country Mean TA banking Mean TA foreign Mean TA domestic Mean share sector / GDP affiliates / GDP banks / GDP foreign affiliates Austria Belgium Bulgaria Czech Republic Denmark Estonia Finland France Germany Greece Hungary Ireland Italy Latvia Lithuania Netherlands Poland Portugal Romania Slovakia Slovenia Spain Sweden United Kingdom / 31

20 Financial fragmentation in Europe Year Foreign Assets Domestic Assets Share of foreign banks Year (a) Foreign/domestic bank assets / GDP (b) Foreign bank assets / total assets Figure 1: Source: Own calculation 17 / 31

21 Results - Growth effect over the cycle Table 3: Estimation results over the cycle VARIABLES Over the cycle Production growth Industry share *** (0.644) Industry growth effect of foreign bank assets, 0.116*** depending on external dependence (0.0275) Industry growth effect of domestic bank assets, depending on external dependence (0.0219) Constant 0.127*** (0.0196) Differential in real 1.83*** growth rates foreign bank assets (0.433) Differential in real growth rates domestic bank assets (1.515) Observations 5,414 Number of industries 713 Country-time fixed-effects yes Industry-time fixed-effects yes Country-industry fixed-effects yes Standard errors clustered on industry level in parentheses *** p<0.01, ** p<0.05, * p< / 31

22 Summary of results - Growth effect over the cycle Over the entire period, foreign bank assets have a significantly positive growth effect No such effect is found for domestic bank assets Note: Results are robust to focussing on loans to non-financial companies Methodology only allows for identification of growth effects stemming from differences across two dimensions Consider two industries with different external dependence in two countries with different degrees of banking sector integration Table 4: Percentiles of foreign bank assets, domestic bank assets and external dependence Variable 10% 25% Median (50%) 75% 90% Foreign bank assets Domestic bank assets External dependence Note: Percentiles for foreign and domestic bank assets are on the country-year-level and for external dependence on the industry level and therefore, are unweighted by the number of industry-country-year observations. 19 / 31

23 Differential in real growth rates In a highly integrated country, the difference of production growth rates between an industry that is highly dependent and an industry that has a low dependence is: growth differential 75% int. = ( ) = 2.95 pp In a less integrated country, the difference of production growth rates between an industry that is highly dependent and an industry that has a low dependence is: growth differential 25% int. = ( ) = 1.12 pp Hence, the difference of the growth differentials is 1.83 percentage points per year. This is an economically important magnitude. 20 / 31

24 Normal times versus times of crisis Hypothesis Ceteris paribus, the effect of banking sector integration on industry production growth differentials is stronger during the crisis than in normal times. During the financial crisis, financing conditions deteriorated for industrial firms (Campello/Graham/Harvey, 2010) and corporate investment decreased (Duchin/Ozbas/Sensoy, 2010) Financial crises have negative effects on industry growth of externally dependent firms (Kroszner/Laeven/Klingebiel, 2007, Dell Arricia/Detragiache/Rajan, 2008) We expect foreign bank presence to be more beneficial in terms of production growth during a domestic crisis 21 / 31

25 Results - Growth effect in times of crisis Table 5: Estimation results for hypothesis Normal times versus times of crisis Normal times Times of crisis Difference VARIABLES Production growth Production growth Production growth Industry share *** ** 4.220** (1.478) (1.598) (1.700) Industry share in times of crisis (2.623) Industry growth effect of foreign bank assets, *** 0.451*** *** depending on external dependence (0.0272) (0.185) (0.0313) Industry growth effect of foreign bank assets 0.354* in times of crisis, depending on external dependence (0.213) Industry growth effect of domestic bank assets, depending on external dependence (0.0227) (0.0838) (0.262)) Industry growth effect of domestic bank assets in times of crisis, depending on external dependence (0.101) Constant *** * * (0.0319) (0.110) (0.0239) Differential in real 1.53*** 7.10*** 5.57* growth rates foreign bank assets (0.428) (2.913) (3.351) Differential in real growth rates domestic bank assets (1.571) (5.800) (7.033) Observations 4,148 1,266 5,414 Number of industries Country-time fixed-effects yes yes yes Industry-time fixed-effects yes yes yes Country-industry fixed-effects yes yes yes Standard errors clustered on industry level in parentheses *** p<0.01, ** p<0.05, * p< / 31

26 Summary of results - Growth effect in times of crisis The growth effect of banking sector integration during times of crisis is more than 4 times as high than in normal times While the differential in real growth rates is about 1.5 percentage points in normal times, it increases to about 7.1 percentage points during times of crisis Difference is economically large and statistically significant (5.6 percentage points) This suggests that financing provided by foreign banks could mitigate financial constraints significantly in times of crisis Banking sector integration generates significant growth effects for industrial firms particularly in times of crisis 23 / 31

27 The role of domestic deleveraging Hypothesis Ceteris paribus, the effect of banking sector integration on industry production growth differentials is stronger in times of deleveraging of domestic banks than at other times. Sharp deleveraging of European banks (Bologna/Caccavaio/Miglietta, 2014) Deleveraging of domestic banks makes obtaining external financing for corporates more difficult, especially for smaller firms (Hoffmann/Sorensen, 2015) Financial constraints in the economy become ceteris paribus more severe The stronger domestic bank deleveraging, the more beneficial should be the presence of foreign banks 24 / 31

28 Results - Growth effect in times of domestic deleveraging Table 6: Estimation results for hypothesis Domestic deleveraging Deleveraging Deleveraging Deleveraging Deleveraging Deleveraging Deleveraging 35th > 35th 25th > 25th 15th > 15th perc. perc. perc. perc. perc. perc. VARIABLES Prod. growth Prod. growth Prod. growth Prod. growth Prod. growth Prod. growth Industry share ** *** *** *** *** *** (1.540) (0.581) (1.495) (0.597) (2.550) (0.655) Industry growth effect of foreign bank ** * 1.682** ** assets, depending on external dependence (0.206) (0.0461) (0.199) (0.0396) (0.803) (0.0384) Industry growth effect of domestic bank * assets, depending on external dependence (0.0358) (0.0560) (0.0654) (0.0480) (0.362) (0.0269) Constant *** *** (0.0525) (0.0312) (0.118) (0.0331) (0.198) (0.0167) Differential in real ** 1.18* 26.49** 1.42** growth rates foreign bank assets (3.244) (0.726) (3.134) (0.624) (12.646) (0.605) Differential in real -4.56* growth rates domestic bank assets (2.477) (3.875) (4.525) (3.321) (25.048) (1.861) Observations 1,941 3,473 1,340 4, ,697 Number of industries Country-time fixed effects yes yes yes yes yes yes Industry-time fixed effects yes yes yes yes yes yes Country-industry fixed effects yes yes yes yes yes yes Standard errors clustered on industry level in parentheses *** p<0.01, ** p<0.05, * p< / 31

29 Summary of results - Growth effect in times of domestic deleveraging The growth effect of banking sector integration is much larger in times of domestic deleveraging than at other times Effect is increasing sharply and monotonically in the extent of domestic bank deleveraging For domestic deleveraging of more than 1.1 pp (25th percentile) the differential in real growth rates is 7.5 percentage points and increases further for more extreme bounds Differentials in real growth rates are much smaller at other times (around 1 percentage point) Banking sector integration generates significant growth effects for industrial firms particularly in times of domestic deleveraging No such results for phases of disintegration (effect even becomes negative for very high levels of disintegration) 26 / 31

30 The role of cross-border lending Can we find similar effects for cross-border lending? Foreign banks with a physical presence may be better able to deal with financial frictions because the distance between lenders and borrowers is smaller (Degryse/Ongena, 2005, Agarwal, 2010) Bruno and Hauswald (2014) indeed find a smaller growth effect of cross-border lending for the pre-crisis period Employed measure for cross-border lending: Credit inflow = credit provided to non-financial institutions in a country by banks hosted in other European countries (taken from ECB) 27 / 31

31 Results - Growth effect of cross-border lending Table 7: Estimation results for hypothesis Cross-border lending Over the cycle Normal times Times of crisis Difference VARIABLES Prod. growth Prod. growth Prod. growth Prod. growth Industry share *** *** ** *** (0.725) (1.425) (1.551) (1.698) Industry share in times of crisis (2.438) Industry growth effect of foreign bank assets, 0.108** 0.113*** 0.452** 0.113** depending on external dependence (0.0399) (0.0374) (0.201) (0.0445) Industry growth effect of foreign bank assets in times of crisis, depending on external dependence (0.246) Industry growth effect of domestic bank assets, depending on external dependence (0.0287) (0.0335) (0.0832) (0.0399) Industry growth effect of domestic bank assets in times of crisis, depending on external dependence (0.0981) Industry growth effect of credit inflow, 0.680** depending on external dependence (0.317) (0.952) (1.392) (1.134) Industry growth effect of credit inflow in times of crisis, depending on external dependence (2.301) Constant *** * * (0.0188) (0.0431) (0.101) (0.0242) Differential in real 1.70** 1.78*** 7.12** 5.34 growth rates foreign bank assets (0.628) (0.589) (3.165) (3.875) Differential in real growth rates domestic bank assets (1.99) (2.318) (5.757) (6.776) Differential in real 1.37** growth rates credit inflow (0.639) (1.918) (2.805) (5.725) Observations 4,386 3,120 1,266 4,386 Number of industries Country-time fixed effects yes yes yes yes Industry-time fixed effects yes yes yes yes Country-industry fixed effects yes yes yes yes Standard errors clustered on industry level in parentheses *** p<0.01, ** p<0.05, * p< / 31

32 Summary of results - The role of cross-border lending Over the cycle there is a positive growth effect of cross-border lending Effect vanishes in both subsamples when distinguishing between crisis and non-crisis periods (same for subsamples according to domestic deleveraging or disintegration) Overall cross-border lending has a less robust effect than the provision of finance through foreign banks Effects of foreign banks assets are robust to including credit inflow This suggests that lending by foreign banks cannot easily be substituted by credit inflows Soft information matters 29 / 31

33 Conclusion - Main results Banking sector integration had sizable differential growth effects on industrial production in the European Union over the entire period Particularly strong growth effects in times of crisis and domestic bank deleveraging Financial fragmentation in times of crisis or domestic deleveraging can cause significant losses in economic growth Less robust effect of cross-border lending Credit by foreign banks with physical presence is particularly important to buffer the adverse effects of domestic financial constraints for firms 30 / 31

34 Conclusion - Policy implications Financial fragmentation in times of crisis and domestic bank deleveraging can be extremely harmful for industry growth Concerns of European policymakers about negative effects of fragmentation in the European banking sector are justified Policy makers should avoid measures fostering disintegration because financial integration plays an important role in buffering country-specific shocks Banking Union and Capital Markets Union are important steps towards re-integrating European financial markets 31 / 31