Risk and International Capital Flows Linda S. Goldberg

Transcription

1 Risk and International Capital Flows Linda S. Goldberg EMG Workshop on Global Liquidity and its International Implications April 22, 2016 London Views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System.

2 Risks and International Capital Flows Globally integrated capital markets developed over recent decades. Rapid growth and large volumes of gross capital flows Compositional changes in participation Less banks, more nonbank and bond financing Large price changes. Post - financial crisis Regulatory regimes evolved Deleveraging in some locations Extra-ordinary monetary and credit interventions Financial deglobalization? 2

3 Risks and International Capital Flows Where are we now? In a risk event, are flows likely to be more stable? Are crises are likely to be less frequent, and less severe? Post-crisis period, altered strength of international capital flow drivers US monetary policy stronger for bank and nonbank issuers VIX weaker for banks, stronger for nonbank issuers global GDP growth weaker local GDP growth stronger sovereign ratings weaker o Bank and nonbank issuers behavior more similarly o Bottom line: aggregate capital flows remain risk sensitive, and more sensitive to center country policy changes. 3

4 Organization of talk I. Patterns of international capital flows II. Relation to risk conditions and macro variables Conceptual mechanisms Evidence What has changed? III. Is it regulation? Evidence on changes in prudential policies Intended and unintended consequences IV. Comments on potential safety of the new system V. Research and policy challenges 4

5 I. Patterns of international capital flows and prices for internal use only

6 A changing international financial landscape Globally integrated capital markets developed over recent decades. Rapid growth and large volumes of gross capital flows Broad evolution of banks with different counterparties o sovereign / private, bank/nonbank, AE/EM International price co-movement After decades of rapid growth and compositional changes, sharply different patterns of gross capital flows after Great Recession International Financial Flows in the New Normal: Key Patterns (and Why We Should Care), Matthieu Bussiere, Julia Schmidt and Natacha Valla. Banque de France. [stylized facts using data from 40 countries] 6

7 After pre-crisis flow acceleration, persistent Great Retrenchment in gross flows. (% GDP) Gross inflows in advanced and emerging market economies 25,0 20,0 15,0 10,0 5,0 0,0-5,0-10, Gross inflows to Advanced Economies Gross inflows to Emerging Market Economies Source: Bluedorn et al. (2013). Bussiere et al (2016). 7

8 International bank credit grew sharply, collapsed, slowly recovering. Bank-to-bank flows most volatile. VIX correlation.

9 Within bank-to-bank international lending, unrelated most volatile. International debt securities flows different patterns. Cross-border loans and international debt securities Annual growth rates, in per cent Graph 1 Cross-border loans (by residence) Cross-border loans (by nationality) International debt securities Sources: BIS locational banking statistics; BIS International Debt Securities. 9

10 Bussiere et al (2016) list the usual suspects as explanatory factors for international financial retreat Weaker economic activity Capital Controls / slower pace of liberalization may affect the overall dynamics of gross flows Deleveraging by private agents, especially banks (in line with sectoral composition) Risk aversion (cf. CGFS, 2011) Lower magnitude of global imbalances Effect of regulation and macroprudential measures? [Various research efforts exploring specifics more later] 10

11 Observing declines in capital flow quantities does not necessarily imply financial market fragmentation (1) Rolling Stock Market Correlation with S&P500 Correlation of WoW returns over a 2Q period Correlation m Rolling Average MSCI World SX5E UKX NKY Sources: National authorities, Haver. Data as of Apr

12 Observing declines in capital flow quantities does not necessarily imply financial market fragmentation (2) Exchange Rate Volatility Index 0.10 Index USDJPY VIX (6m rolling average, RHS) GBPUSD USDEUR Trend 1/ Note: Series are calculated as the standard deviation over a 2Q rolling window, normalized against the average level over that period. 1/ Trend line calculated as a rolling 6 month average of all series. Sources: Bloomberg. Data as of Apr-2016

13 Big picture from data Gross flows to AEs accelerated late 1990s, to EMs early 2000s. Steep AE collapse post crisis. Bank-to-bank credit had most dramatic changes, with GDP and VIX correlations. International debt securities have trend declined, but with positive growth. Flows to nonbanks held up in recent years. Despite reduced volumes of international flows, comovements in market indices have increased. Exchange rates, reflecting pressures from international asset allocation desires, move with VIX. Other asset prices closely co-move. 13

14 II. Relations to risk conditions and macro variables for internal use only

15 Explaining macro economic and risk effects on flows Recent contributions Forbes and Warnock (2012) Fratzscher (2012) Cerutti, Claessens and Ratnovski (2014) Bruno and Shin (2015) Miranda-Agrippino and Rey (2015) Correa, Paligorova, Sapriza and Zlate (2015) add to long literature concentrated on EM capital flows 15

16 New research paper examines patterns and drivers and structural changes, using rich BIS databases Global factors, international capital flows, and stabilizing regulatory reforms? Avdjiev, Gambacorta, Goldberg, Schiaffi What are the drivers of different international capital flows? Have these drivers changed since the crisis? Do regulatory reforms explain the changes? Are international capital flows less flighty and more responsive to fundamentals? Detailed country flow data for 64 countries, 2000Q1-2013Q4. 16

17 Who is involved in which flows? XB loans to banks XB loans to nonbanks IDS issued by banks Typical Lenders Typical Borrowers Notes Internationally-active Banks (all sizes) Interbank market banks (unsecured and repo) Internationally-active banks Pension funds; Insurance companies; Money Market Mutual Funds; Hedge funds Large non-financial corporates; exporting/importing firms; Leveraged non-bank financials Large and mid-sized banks Syndicated loan market; trade credit; project financing Smaller investor base than for IDS issued by non-banks IDS issued by nonbanks Pension funds; Insurance companies; Mutual Funds; Hedge funds Non-financial corporates; governments; Insurance companies Broader investor base than for IDS issued by banks

18 Table 1. Descriptive statistics of the main variables used in the estimation Variables Obs. Mean Std. Dev. A) Dependent variables: cross-border bank flows ΔCross-border loans (to all sectors) 3, ΔCross-border loans (to banks) 3, ΔCross-border loans (to non banks) 3, ΔInternational debt securities (issued by all sectors) 3, l16.18 ΔInternational debt securities (issued by banks) 3, ΔInternational debt securities (issued by non banks) 3, Notes: The sample includes quarterly data for 64 recipient countries over the period 2000Q1-2014Q4. 18

19 Structure of empirical analysis logy = α + α FFR + α logvix + α log GDP j t 0 1 t 2 t 3 + α SovRating + α ChinnIto + α log GlobalGDP + µ + ε j j t t t j t j t Cross-border loans: Quarterly Growth Rate (j,t) = Adjusted Flows (j,t) / Outstanding Stock (j,t-1) International Debt Securities: Quarterly Growth Rate (j,t) = Net Issuance (j,t) / Outstanding Stock (j,t-1) quarterly growth rates in stocks constructed using the official BIS definition. Exactly the same interpretation as log changes in stocks of outstanding loans/ids positions, with advantage of being adjusted for FX movements and breaks in series 19

20 Structure of empirical analysis logy = α + α FFR + α logvix + α log GDP j t 0 1 t 2 t 3 + α SovRating + α ChinnIto + α log GlobalGDP + µ + ε j j t t t j t j t 1. Baseline specification 2. Add range of tests for o structural breaks in fit of model, and timing of break o examine which of the coefficients exhibit breaks o explore time variation in coefficients on FFR and VIX o changes in sensitivity with prudential policy changes in j 3. Conclude with comparisons for total international capital flows 20

21 International loans show familiar responses to global and local conditions; interesting intl debt security patterns. Table 2 Baseline Model Dependent variable: Dependent variable: Cross-border loans International debt securities Explanatory variables All to banks o non-banks All by banksby non-banks Fed funds rate (1) -1.87*** -2.07*** -2.11*** -1.43* Log(VIX) -4.45*** -4.26*** -4.93*** -3.56*** -7.55*** -2.65*** Real GDP 0.57*** 0.6*** 0.52*** Sovereign rating (2) 2.49** 4.22*** * Chinn-Ito index (3) * 0.28 Real global GDP * Country FE yes yes yes yes yes yes Observations 2,903 2,903 2,903 2,903 2,572 2,902 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. (2) LT foreign currency, average across 3 agencies. (3) Chinn and Ito (2006) measure of financial openness. 21

22 Model exhibits structural break based on Sum of square residual (SSR) test on structural breaks for all parameters (baseline model) Notes: For every date t in the x-axis, we have created a time dummy that takes value 1 if the date is greater than t and 0 elsewhere. Then we have run the regression of cross-border loans (to all, to banks, to non banks) and international debt securities (issued by all, by banks, by non banks) on Fed fund rates, log(vix) Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP, the time dummy and the interaction of each explanatory variables with the time dummy. Each graph reports the sum of squared residuals (SSR) of each of these regressions and for each different time dummy. The graphs shows the SSR for each different time dummy and for each different dependent variable. When the sequence of SSR's attains its minimum, then the fit of the model is the greatest and this is due to the presence of a specific time dummy. Therefore, the date when the SSR is at its minimum is the most likely candidate for a structural break. 22

23 A clear candidate for when this world changes... Table 3: Chow tests for the presence of structural breaks All Banks Non banks Cross-border loans Break date: 2009:Q1 Break date: 2009:Q1 Break date: 2009:Q1 F(7,2834) = F(7,2834) = F(7,2834) = p-value = p-value = p-value = International debt securities Break date: 2009:Q1 Break date: 2009:Q2 Break date: 2009:Q1 F(7,2834) = 3.43 F(7, 2505) = 0.84 F(7, 2833) = 2.70 p-value = p-value = p-value = Notes: For each bank flow, the table contains the break date identified in Graph 1, together with the F-statistic and the corresponding p-value of a Chow test. The Chow test is conducted on the significance of a time dummy that takes value 1 after the break date and on the interactions of this time dummy with all the explanatory variables, i.e. Fed fund rates, log(vix) Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP. 23

24 Series of specifications test for evidence of parameter breaks. Table 4 - Baseline model with structural breaks Dependent variable: Dependent variable: Cross-border loans International debt securities Explanatory variables All to banks to non-banks All by banks by non-banks Fed funds rate (1) -3.19*** -3.37*** -3.47*** Log(VIX) -3.99*** -4.4*** -4.47*** ** Real GDP 0.34*** 0.35** 0.39*** Sovereign rating (2) 4.17** 7.18*** Chinn-Ito index (3) * ** 0.47 Real global GDP 1.74*** 1.89*** 1.5*** Break dummy (4) Fed funds rate*break -3.19** -4.97** * ** Log(VIX)*Break Real GDP*Break 0.37*** 0.38** *** ** Sovereign rating*break -4.12** -6.64** Chinn-Ito index*break ** Real global GDP*Break -2.22*** -2.11*** -1.97*** -1.97*** ** Country fixed effects yes yes yes yes yes yes Break date 2009:Q1 2009:Q1 2009:Q1 2009:Q1 2009:Q2 2009:Q1 Observations 2,903 2,903 2,903 2,903 2,572 2,902 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. (2) LT foreign currency, average across 3 agencies. (3) Chinn and Ito (2006) measure of financial openness. (4) The break dummy equals 1 for every period after (and including) the break date. 24

25 Summed coefficients show evolution of sensitivities Table 4 - Baseline model with structural breaks Dependent variable: Dependent variable: Explanatory variables All to banks to non-banks All by banks by non-banks Pre-break Cross-border loans International debt securities Log(VIX) -3.99*** -4.4*** -4.47*** ** Fed funds rate (1) -3.19*** -3.37*** -3.47*** Post-break Log(VIX) -1.58* ** * Fed funds rate (1) -6.38*** -8.34*** -5.28*** -6.51*** *** Break date 2009:Q1 2009:Q1 2009:Q1 2009:Q1 2009:Q2 2009:Q1 Observations 2,903 2,903 2,903 2,903 2,572 2,902 R-squared Notes:The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. The regressions include Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP and their interaction with a break dummy that takes value 1 after the break date.the regressions also include a full set of country fixed effects.

26 Key message from extensive analysis of international capital flows International flows that involve banks either as a borrower or as a lender (i.e. B-to-B flows, B-to-NB flows, NB-to-B flows) are: less flighty in response to global uncertainty, but more flighty in response to global capital costs. International flows that do not involve banks either as a borrower or as a lender (i.e. NB-to-NB flows in the form of international debt securities issued by non-banks) are: more flighty both with respect to global uncertainty (the VIX), more flighty with respect to global capital costs (the fed funds rate). 26

27 Aggregate flows are more sensitive to Fed Funds; Pattern of sensitivity to global risk depends on borrower type. Table 8a Baseline with structural breaks, aggregate flows Dependent variable: Total cross-border flows (loans and debt securities) Explanatory variables All to banks to non-banks Pre-break Log(VIX) -3.29*** -3.54*** -2.85*** Fed funds rate (1) -2.18*** -2.85*** -2.19*** Post-break Log(VIX) -2.24** *** Fed funds rate (1) -6.68*** -7.67*** -5.77*** Break date 2009:Q1 2009:Q1 2009:Q1 Observations 2,903 2,572 2,902 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu- Xia Shadow rate for the period 2009:Q1 2014:Q4. The regressions include Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP and their interaction with a break dummy that takes value 1 after the break date. The regressions also include a full set of country fixed effects.

28 Do regulatory explain changed sensitivities? for internal use only

29 Regulatory and prudential instrument environment internationally IBRN Changes in Prudential Policy Instruments A New Cross- Country Database Cerutti, Correa, Fiorentino and Segalla (2015) Covers 2000Q1 through 2014Q4. 64 countries. Contains 5 types of prudential instruments: Countercyclical capital buffers General capital requirements Sector-specific capital requirements (split into real estate credit, consumer credit, and other) Interbank exposure limits Concentration limits Loan-to-value ratio limits, and Reserve requirements (local and foreign currency) 29

30 Incidence of prudential changes #distinct countries #episodes #tightening episodes #loosening episodes General capital requirements Sector specific capital requirements Loan to value ratio limits Reserve requirement: foreign Reserve requirement: local Interbank exposure limits Concentration limits

31 Across countries, most changes in general capital requirements are recent. Other instruments have changes spread out over time. 31

32 Do instrument changes at the country level explain results? 32

33 Table 5a LTV with structural breaks Dependent variable: Dependent variable: Cross-border loans International debt securities Explanatory variables All to banks to non-banks All by banks by non-banks Pre-break Log(VIX) -4.11** -5.08* -4.14** -3.29** -8.37** -2.66* Fed funds rate (1) -3.39*** -3.22** -4.26*** Post-break Log(VIX) -2.98** *** -4.3*** -5.26** -3.63*** Fed funds rate (1) -5.21*** -6.47* -5.04*** -4.27** ** Break date 2009:Q1 2009:Q1 2009:Q1 2009:Q1 2009:Q2 2009:Q1 Observations 1,031 1,031 1,031 1, ,031 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. The regressions include Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP and their interaction with a break dummy that takes value 1 after the break date. The regressions also include a full set of country fixed effects and the following macro-prudential index: loan to value ratio caps. 33

34 Table 5b Capital requirements with structural breaks Dependent variable: Dependent variable: Cross-border loans International debt securities Explanatory variables All to banks to non-banks All by banks by non-banks Pre-break Log(VIX) -4.12*** -4.58*** -4.62*** ** Fed funds rate (1) -3.16*** -3.16*** -3.46*** Post-break Log(VIX) *** Fed funds rate (1) -6.63*** -8.59*** -5.38*** -6.81*** -11.4** -6.72*** Break date 2009:Q1 2009:Q1 2009:Q1 2009:Q1 2009:Q2 2009:Q1 Observations 2,847 2,847 2,847 2,847 2,516 2,847 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. The regressions include Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP and their interaction with a break dummy that takes value 1 after the break date. The regressions also include a full set of country fixed effects and the following macro-prudential index: capital requirements. 34

35 Table 5c Reserve requirements with structural breaks Dependent variable: Dependent variable: Cross-border loans International debt securities Explanatory variables All to banks to non-banks All by banks by non-banks Pre-break Log(VIX) -3.81*** -4.23*** -4.27*** ** Fed funds rate (1) -3.13*** -3.31*** -3.41*** Post-break Log(VIX) -1.61* *** -2.85** * Fed funds rate (1) -6.25*** -8.22*** -5.14*** -6.56*** *** Break date 2009:Q1 2009:Q1 2009:Q1 2009:Q1 2009:Q2 2009:Q1 Observations 2,903 2,903 2,903 2,903 2,572 2,902 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. The regressions include Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP and their interaction with a break dummy that takes value 1 after the break date. The regressions also include the following macro-prudential index: reserve requirements. 35

36 Do instrument changes at the country level explain results? LTV ratio caps -- no Capital requirements: higher cumulative capital requirement levels reduce the negative effect of a hike of FFrates (global capital costs) on XB loans to banks and non-banks; Reserve requirements: higher cumulative reserve requirement levels increase the negative effect of an increase in the VIX (global uncertainty) on XB loans to banks 36

37 Table 7b Cumulative capital requirements with structural breaks and interactions with global factors Explanatory variables All to banks to non-banks All by banks by non-banks Pre-break Log(VIX) -4.14*** -4.6*** -4.66*** ** Log(VIX) * CapReq (2) * Fed funds rate (1) -3.16*** -3.2*** -3.43*** Fed funds rate * CapReq ( * ** Post-break Dependent variable: Dependent variable: Cross-border loans International debt securities Log(VIX) -3.08** *** -5.72*** *** Log(VIX) * CapReq (2) Fed funds rate (1) -7.25*** -9.5*** -6.1*** -5.98** ** Fed funds rate * CapReq ( 3.82** 5.44** 3.1** Break date 2009:Q1 2009:Q1 2009:Q1 2009:Q1 2009:Q2 2009:Q1 Observations 2,847 2,847 2,847 2,847 2,516 2,847 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. (2) Capital requirements. The regressions include Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP and their interaction with a break dummy that takes value 1 after the break date. The regressions also include a full set of country fixed effects and the following cumulative macro-prudential index: capital requirements. The cumulative macro-prudential index is obtained in each quarter by adding the non-cumulative macro-prudential index up to that quarter.

38 Table 7c Cumulative reserve requirements with structural breaks and interactions Dependent variable: Dependent variable: Cross-border loans International debt securities Explanatory variables All to banks to non-banks All by banks by non-banks Pre-break Log(VIX) -4.38*** -5.03*** -4.76*** ** Log(VIX) * ResReq (2) -1.09*** -1.65*** -0.71** Fed funds rate (1) -3.4*** -3.64*** -3.58*** -1.66* Fed funds rate * ResReq (2) Post-break Log(VIX) -2.00** *** -2.76** * Log(VIX) * ResReq (2) -0.69** -1.19** Fed funds rate (1) -6.01*** -7.82*** -4.94*** -6.32** *** Fed funds rate * ResReq (2) Break date 2009:Q1 2009:Q1 2009:Q1 2009:Q1 2009:Q2 2009:Q1 Observations 2,903 2,903 2,903 2,903 2,572 2,902 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. to borrowers in country j. issued by borrowers in country j. (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. (2) Reserve requirements. The regressions include Real GDP, Sovereign Ratings, Chinn-Ito Index, Real Global GDP and their interaction with a break dummy that takes value 1 after the break date. The regressions also include a full set of country fixed effects and the following cumulative macroprudential index: reserve requirements. The cumulative macro-prudential index is obtained in each quarter by adding the non-cumulative macroprudential index up to that quarter.

39 Key messages from interacting sensitivities to global drivers with cumulative prudential policy changes Controlling for LTV caps increases sensitivity of incoming flows to global uncertainty (VIX). Cumulative caps -- similar results. Controlling for capital requirements (table 5b) Increases sensitivity of XB loans (especially to NB) to VIX decreases sensitivity of IDS (especially issued by NB) to VIX; With cumulative measure, sensitivity to VIX post-crisis considerably higher; Controlling for reserve requirements (table 5c) increased sensitivity of XB loans (especially to NB) to VIX; no effect on sensitivity of IDS to VIX cumulative measures yield similar results. 39

40 What is the new normal? for internal use only

41 International Financial Integration Global financial crisis sparked a critical examination of the ability of banks to withstand shocks, prompting policy change o stress testing o capital and liquidity regulation changes o recovery and resolution regimes o systematic development of macro-prudential tools one of the most significant changes in banking regulation in the past decade. Intent to strengthen the resilience of the banking sector and reduce systemic events, fire sales, adverse feedback loops. 41

42 Evidence on international capital flows, global factors and flightiness Bank sensitivities appear to have evolved o Less flighty in response to global uncertainty o More flighty in response to global capital costs Non-bank flows have increased importance, and o More response to global uncertainty and global capital costs. Country-specific prudential policy changes do not fully account for the post-crisis structural break, but have had a non-negligible impact on the evolution of sensitivities. Exchange rates and asset prices remain tightly correlated. Across international capital flows examined, aggregate sensitivity to risk conditions and global factors remains. 42

43 Thank you. 43

44 Table 8b Baseline with structural breaks, aggregate flows, full specification Dependent variable: Total cross-border flows (loans and debt securities) Explanatory variables All to banks to non-banks Fed funds rate (1) -2.18*** -2.85*** -2.19*** Log(VIX) -3.29*** -3.54*** -2.85*** Real GDP 0.24*** 0.25* 0.18** Sovereign rating (2) ** 0.01 Chinn-Ito index (3) * Real global GDP 1.14*** 1.85*** 0.83*** Break dummy (4) Fed funds rate*break -4.50*** -4.81*** -3.58*** Log(VIX)*Break Real GDP*Break 0.36*** 0.49*** 0.30*** Sovereign rating*break Chinn-Ito index*break *** Real global GDP*Break -1.79*** -2.07*** -1.55*** Country fixed effects yes yes yes Break date 2009:Q1 2009:Q1 2009:Q1 Observations 2,903 2,903 2,903 R-squared Notes: The sample includes quarterly data on cross-border flows (loans and debt securities) for 64 recipient countries over the period 2000Q1-2014Q4. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 (1) Effective federal funds rate for the period 2001:Q1 2008:Q4, Wu-Xia Shadow rate for the period 2009:Q1 2014:Q4. (2) LT foreign currency, average across 3 agencies. (3) Chinn and Ito (2006) measure of financial openness. (4) The break dummy equals 1 for every period after (and including) the break date.

45 Graph 4 - Rolling window estimation for the coefficient on the ΔFED fund rate Notes: The set of graphs is organized over three in three different rows. The first row reports the value of the coefficient of Fed fund rates, together with 90% confidence intervals obtained by rolling-window regression of cross-border loans (to all, to banks, to non banks) using the specification reported in the first column of Table 3. The second row reports the value of the coefficient of Fed fund rates, together with 90% confidence intervals obtained by rolling-window regression of international debt securities (issued by all, by banks, by non banks) using the specification reported in the second column of Table 3. The window size of the rolling regressions is 40 quarters. The dates shown on the x-axis are the last ones included in every window. The last row reports the difference between the impact of Fed fund rates on CBL and IDS for every different sector. 45

46 Graph 4 - Rolling window estimation for the coefficient on the Log(VIX) Notes: The set of graphs is organized over three in three different rows. The first row reports the value of the coefficient of log(vix), together with 90% confidence intervals obtained by rolling-window regression of cross-border loans (to all, to banks, to non banks) using the specification reported in the first column of Table 3. The second row reports the value of the coefficient of log(vix), together with 90% confidence intervals obtained by rolling-window regression of international debt securities (issued by all, by banks, by non banks) using the specification reported in the second column of Table 3. The window size of the rolling regressions is 40 quarters. The dates shown on the x-axis are the last ones included in every window. The last row reports the difference between the impact of Fed fund rates on CBL and IDS for every different sector 46