Financial Integration, Productivity and Capital Accumulation Alessandra Bon glioli Instituto de Análisis Económico IAE - CSIC ECB-CFS, 13 February 2008 A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 1 / 25
Aim and Motivation The usual question Does International Financial Liberalization (IFL) a ect Growth? Ambiguous theoretical predictions and empirical evidence on the e ects of IFL on GDP growth A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 2 / 25
Aim and Motivation The usual question Does International Financial Liberalization (IFL) a ect Growth? Ambiguous theoretical predictions and empirical evidence on the e ects of IFL on GDP growth A new question Does IFL a ect Productivity (TFP), Capital (K)-Accumulation or both? A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 2 / 25
Aim and Motivation Why a new question? TFP is the main contributor to long-term GDP Growth A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 3 / 25
Aim and Motivation Why a new question? TFP is the main contributor to long-term GDP Growth Test di erent theories on IFL and economic growth A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 3 / 25
Aim and Motivation Why a new question? TFP is the main contributor to long-term GDP Growth Test di erent theories on IFL and economic growth Welfare implications (see Gourinchas and Jeanne, 2006) A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 3 / 25
This Paper Estimate the impact of IFL on: TFP Capital accumulation Account for indirect e ects through: Financial instability: banking and currency crises Financial development Account for heterogeneous e ects depending on economic development A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 4 / 25
Anticipating the Main Results IFL has positive e ect on TFP A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 5 / 25
Anticipating the Main Results IFL has positive e ect on TFP This e ect is stronger in developed countries A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 5 / 25
Anticipating the Main Results IFL has positive e ect on TFP This e ect is stronger in developed countries IFL has no direct e ect on Capital accumulation A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 5 / 25
Anticipating the Main Results IFL has positive e ect on TFP This e ect is stronger in developed countries IFL has no direct e ect on Capital accumulation Financial crises (mainly banking) are bad for both TFP and Capital accumulation A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 5 / 25
Anticipating the Main Results IFL has positive e ect on TFP This e ect is stronger in developed countries IFL has no direct e ect on Capital accumulation Financial crises (mainly banking) are bad for both TFP and Capital accumulation IFL triggers only minor banking crises in developed countries A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 5 / 25
Related Literature Theoretical predictions on IFL and Growth (Obstfeld, 1994, Acemoglu and Zilibotti, 1997, Stiglitz, 2000) Empirical evidence on IFL and GDP Growth (Quinn, 1997, Bekaert et al., 2005, Eichengreen and Leblang, 2003, Grilli and Milesi-Ferretti, 1995, many more) Growth and Development accounting: di erences in TFP account for much of the variation in GDP and Growth (survey by Caselli, 2005) Empirical evidence IFL and investments (Chari and Henry, 2002) No evidence on IFL and TFP (call for it in Kose et al., 2006 and Chari and Henry, 2002) A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 6 / 25
Data and Empirical Strategy Sample: 70 countries between 1975 and 1999 Data frequency: annual (use annual, 5-year, 25-year) Estimate P it = β 0 + β 0 1 X it + γifl it + u it P it 2 flog K it, d log K it, log TFP it, d log TFP it g, computed as in Hall and Jones (1999) with data from PWT 6.1, Barro and Lee (2001) and Psacharopulos (1994) X it = Control Variables IFL it 2 fcal it, Quinn it, IFIGDP it g indicator of IFL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 7 / 25
IFL indicators De jure indicators: A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 8 / 25
IFL indicators De jure indicators: 1 0-1 de nition of restrictions on residents capital account transactions, from the AREAER of the IMF (212 countries, 1967-96) 0 if restrictions CAL it = 1 if no restrictions A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 8 / 25
IFL indicators De jure indicators: 1 0-1 de nition of restrictions on residents capital account transactions, from the AREAER of the IMF (212 countries, 1967-96) 0 if restrictions CAL it = 1 if no restrictions 2 Quinn s scoring of restrictions on residents and non-residents capital account transactions (142 countries, 1958, 1973, 1982, 1988, 1997) Quinn it 2 [0, 100] with 0 = max restrictions, 100 = no restrictions A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 8 / 25
IFL indicators De jure indicators: 1 0-1 de nition of restrictions on residents capital account transactions, from the AREAER of the IMF (212 countries, 1967-96) 0 if restrictions CAL it = 1 if no restrictions 2 Quinn s scoring of restrictions on residents and non-residents capital account transactions (142 countries, 1958, 1973, 1982, 1988, 1997) Quinn it 2 [0, 100] with 0 = max restrictions, 100 = no restrictions De facto measure: A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 8 / 25
IFL indicators De jure indicators: 1 0-1 de nition of restrictions on residents capital account transactions, from the AREAER of the IMF (212 countries, 1967-96) 0 if restrictions CAL it = 1 if no restrictions 2 Quinn s scoring of restrictions on residents and non-residents capital account transactions (142 countries, 1958, 1973, 1982, 1988, 1997) Quinn it 2 [0, 100] with 0 = max restrictions, 100 = no restrictions De facto measure: 1 Gross external positions from Lane and Milesi-Ferretti (145 countries, 1970-2004) IFIGDP = TotForeignAssets + TotForeignLiabilities GDP A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 8 / 25
Control Variables X BC anecdotal indicator of banking crises (Caprio and Klingebiel, 2003) CC indicator of currency crises (Glick and Hutchison, 2000) Financial depth, privo = private credit/gdp (Beck et al., 2001) Real per capital GDP (PWT 6.1) Goverment expenditure/gdp (PWT 6.1) Trade/GDP (PWT 6.1) IPR s protection (Ginarte and Park, 1997) GADP as in Hall and Jones (1999) with data from ICRG Development indicator (World Development Indicators) A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 9 / 25
Data: summary and statistics A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 10 / 25
Econometric Model (1): yearly Di -in-di Estimate on the yearly panel P it = β 0 + β 0 1 X it 1 + γifl it 1 + η i + ν t + ε it With P 2 flog K, log TFPg Note: η i and ν t country and year xed e ects With IFL 2 f0, 1g: γ estimated o the change in P between pre and post-liberalization in open countries vs the change in closed countries, i.e. γ is a di erence in di erences estimator. A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 11 / 25
Results: yearly D-i-D, P=logTFP, IFL=CAL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 12 / 25
Results: yearly D-i-D, P=logTFP, other IFL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 13 / 25
Results: yearly D-i-D, P=logK, IFL=CAL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 14 / 25
Results: yearly D-i-D, P=logK, other IFL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 15 / 25
Econometric Model (2): 5-year Dynamic Panel Data Estimate on the 5-year panel: dp it = β 0 + θdp it 5 + β 0 1 dx it + γdifl it + dν t + dε it P it = β 0 + θp it 5 + β 0 1 X i(t 5,t) + γifl i(t 5,t) + η i + ν t + ε it, with Two-Step System GMM. Instrument all RHS variables: Di erences with lagged levels Levels with lagged di erences Validate the model with: Sargan test of overidentifying restrictions m2 test of second order serial correlation A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 16 / 25
Results: 5-year DPD, P=logTFP, IFL=CAL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 17 / 25
Results: 5-year DPD, P=logTFP, other IFL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 18 / 25
Results: 5-year DPD, P=logK, IFL=CAL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 19 / 25
Results: 5-year DPD, P=logK, other IFL A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 20 / 25
What about GDP? A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 21 / 25
IFL and Financial Crises Does IFL make Financial Crises more likely? Estimate on the yearly panel Pr (Crisis_type it = 1) = Φ (b o + b 1 X it 1 + gifl it 1 ) with Crisis_type 2 fbc_bl, BC_SYS, BC_ANY, CC g Banking Crises (Caprio and Klingebiel, 2003) BC_BL = 1 0 if borderline crisis otherwise, BC_SYS = 1 0 if systemic crisis otherwise Currency Crises (Glick and Hutchison, 2000) CC = 0 1 if no crisis if crisis A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 22 / 25
IFL and Crises: probit estimates A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 23 / 25
Conclusions New question: Does IFL a ect TFP and/or K? A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 24 / 25
Conclusions New question: Does IFL a ect TFP and/or K? New and robust empirical evidence: A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 24 / 25
Conclusions New question: Does IFL a ect TFP and/or K? New and robust empirical evidence: 1 IFL spurs TFP A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 24 / 25
Conclusions New question: Does IFL a ect TFP and/or K? New and robust empirical evidence: 1 IFL spurs TFP 2 IFL spurs TFP more in developed countries A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 24 / 25
Conclusions New question: Does IFL a ect TFP and/or K? New and robust empirical evidence: 1 IFL spurs TFP 2 IFL spurs TFP more in developed countries 3 IFL has almost no e ect on K A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 24 / 25
Conclusions New question: Does IFL a ect TFP and/or K? New and robust empirical evidence: 1 IFL spurs TFP 2 IFL spurs TFP more in developed countries 3 IFL has almost no e ect on K 4 Crises (mainly banking) harm both TFP and K A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 24 / 25
Conclusions New question: Does IFL a ect TFP and/or K? New and robust empirical evidence: 1 IFL spurs TFP 2 IFL spurs TFP more in developed countries 3 IFL has almost no e ect on K 4 Crises (mainly banking) harm both TFP and K 5 IFL makes minor banking crises slightly more likely A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 24 / 25
What s Next? How to explain theoretically 1. and 3.? Build a model with: Unobservable heterogeneity of projects (good vs bad) Economic volatility: unobservable shocks In Autarky: bad selection (low TFP) With IFL: investment abroad! more information More info! less noise! better selection (higher TFP) What about investment? Ambiguous e ect A. Bon glioli (IAE - CSIC) Financial Integration, TFP and K ECB-CFS, 13 February 2008 25 / 25