## 1. Introduction and Motivation

## Source details

**Canonical URL:** [1. Introduction and Motivation](https://www.imf.org/-/media/files/publications/wp/2017/wp17193.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17193.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17193.pdf.json)

---

### Motivation and research question
- The paper seeks to quantify the importance of a Global Financial Cycle (hereafter GFCy), particularly for the variation in international capital flows.
- Motivation stems from literature arguing that a GFCy co-moves asset prices, credit growth, and capital flows, and is associated with measures of global risk such as the VIX.
- Empirical importance: if the GFCy explains much of capital flow variation (especially for small and emerging economies), then capital flows become largely exogenous to those countries; if not, domestic policy has greater freedom.

### Representative statements from the literature (verbatim)
- “Risky asset prices around the globe, from stocks to corporate bonds, have a strong common component.  So do capital flows ... Global financial cycles are associated with surges and retrenchments in capital flows, booms and busts in asset prices and crises.  The picture emerging is that of a world with powerful global financial cycles characterized by large common movements in asset prices, gross flows, and leverage ... The global financial cycle can be related to monetary conditions in the centre country and to changes in risk aversion and uncertainty ... capital flows, especially credit flows, are largely driven by a global factor ... “ -- Rey (2013, pp 1-2)
- “There is a global financial cycle in capital flows, asset prices, and in credit growth.  This cycle co-moves with the VIX, a measure of uncertainty and risk aversion of the markets.” -- Rey (2013, abstract)
- “Large gross cross-border flows are moving in tandem across countries regardless of the exchange rate regime, they tend to rise in periods of low volatility and risk aversion and decrease in periods of high volatility and risk aversion, as measured by the VIX ... There is a global financial cycle.” -- Passari and Rey (2015, p 693)
- “Global factors, especially global risk, are significantly associated with extreme capital flow episodes. ... Our analysis indicates that waves of capital flows are primarily associated with global factors.  Global risk, which incorporates both risk aversion and economic uncertainty, is the only variable that consistently predicts each type of capital flow episode; an increase in global risk is associated with more stops and retrenchments and fewer surges and flight. ... most domestic factors only have a limited correlation with capital flow volatility ... global factors, and especially global risk, are key to understanding periods of extreme capital flows by domestic and foreign investors.  Increases in global risk predict sudden stops in capital flows by foreigners and retrenchments in capital flows by domestic investors ...” -- Forbes and Warnock (2012)

### Policy implications highlighted
- If GFCy is dominant:
  - Small and emerging economies face large capital flow fluctuations driven by common/global shocks (including center-country factors).
  - They may need to insulate their economies (capital controls, macro-prudential instruments), accepting some loss of international financial integration benefits.
  - The GFCy can lead to excessive credit growth in boom times and excessive retrenchment in bad times, with associated financial stability issues (Rey (2015, pp 9-10)).
- If GFCy is not dominant:
  - Policy authorities in small and emerging countries have greater degrees of freedom to manage the domestic economy with respect to capital flow fluctuations.

### Methodological stance of the paper
- The authors emphasize a conventional, plain-vanilla approach by design:
  - Data sets, techniques, capital flow models, and statistical metrics used are widely used and standard.
  - Focus is on empirical results rather than novel data or methodology.
- Structure of the paper:
  - Section 2: literature review.
  - Section 3: data and methodology for identifying the GFCy.
  - Section 4: empirical results (the core of the paper).
  - Section 5: brief summary and conclusion.
- Terminology note:
  - The authors use “GFCy” to avoid confusion with “GFC”, which commonly refers to the Global Financial Crisis of 2008-09.
  - A companion paper will separately analyze domestic asset prices and credit growth across countries.

*Source: wp17193 - 1. Introduction and Motivation*

### 1. Introduction and Motivation .........................................................................................

### 1. Introduction and Motivation

### Motivation and research question
- The paper seeks to quantify the importance of a Global Financial Cycle (hereafter GFCy), particularly for the variation in international capital flows.
- Motivation stems from literature arguing that a GFCy co-moves asset prices, credit growth, and capital flows, and is associated with measures of global risk such as the VIX.
- Empirical importance: if the GFCy explains much of capital flow variation (especially for small and emerging economies), then capital flows become largely exogenous to those countries; if not, domestic policy has greater freedom.

### Representative statements from the literature (verbatim)
- “Risky asset prices around the globe, from stocks to corporate bonds, have a strong common component.  So do capital flows ... Global financial cycles are associated with surges and retrenchments in capital flows, booms and busts in asset prices and crises.  The picture emerging is that of a world with powerful global financial cycles characterized by large common movements in asset prices, gross flows, and leverage ... The global financial cycle can be related to monetary conditions in the centre country and to changes in risk aversion and uncertainty ... capital flows, especially credit flows, are largely driven by a global factor ... “ -- Rey (2013, pp 1-2)
- “There is a global financial cycle in capital flows, asset prices, and in credit growth.  This cycle co-moves with the VIX, a measure of uncertainty and risk aversion of the markets.” -- Rey (2013, abstract)
- “Large gross cross-border flows are moving in tandem across countries regardless of the exchange rate regime, they tend to rise in periods of low volatility and risk aversion and decrease in periods of high volatility and risk aversion, as measured by the VIX ... There is a global financial cycle.” -- Passari and Rey (2015, p 693)
- “Global factors, especially global risk, are significantly associated with extreme capital flow episodes. ... Our analysis indicates that waves of capital flows are primarily associated with global factors.  Global risk, which incorporates both risk aversion and economic uncertainty, is the only variable that consistently predicts each type of capital flow episode; an increase in global risk is associated with more stops and retrenchments and fewer surges and flight. ... most domestic factors only have a limited correlation with capital flow volatility ... global factors, and especially global risk, are key to understanding periods of extreme capital flows by domestic and foreign investors.  Increases in global risk predict sudden stops in capital flows by foreigners and retrenchments in capital flows by domestic investors ...” -- Forbes and Warnock (2012)

### Policy implications highlighted
- If GFCy is dominant:
  - Small and emerging economies face large capital flow fluctuations driven by common/global shocks (including center-country factors).
  - They may need to insulate their economies (capital controls, macro-prudential instruments), accepting some loss of international financial integration benefits.
  - The GFCy can lead to excessive credit growth in boom times and excessive retrenchment in bad times, with associated financial stability issues (Rey (2015, pp 9-10)).
- If GFCy is not dominant:
  - Policy authorities in small and emerging countries have greater degrees of freedom to manage the domestic economy with respect to capital flow fluctuations.

### Methodological stance of the paper
- The authors emphasize a conventional, plain-vanilla approach by design:
  - Data sets, techniques, capital flow models, and statistical metrics used are widely used and standard.
  - Focus is on empirical results rather than novel data or methodology.
- Structure of the paper:
  - Section 2: literature review.
  - Section 3: data and methodology for identifying the GFCy.
  - Section 4: empirical results (the core of the paper).
  - Section 5: brief summary and conclusion.
- Terminology note:
  - The authors use “GFCy” to avoid confusion with “GFC”, which commonly refers to the Global Financial Crisis of 2008-09.
  - A companion paper will separately analyze domestic asset prices and credit growth across countries.

*Source: wp17193 - 1. Introduction and Motivation*

### 2.  Literature Review

### 2.  Literature Review

### Overview of two strands
- Two main strands inform the investigation:
  - Traditional "push-pull" analyses explaining capital flows with global “push” (source country) and “pull” (recipient country) factors.
  - Factor models and structural time-series models (often variants of VARs) documenting commonality in global financial developments (capital flows, credit creation, domestic asset prices).
- The paper combines elements of both strands to favor assessing the quantitative importance of the GFCy for international capital flows.
- Robustness is explored across types, sources, destinations of capital flows, and time periods.

### Push-Pull literature: methods and typical findings
- Origins and methods:
  - Rooted in Calvo et al. (1993, 1996); mainly uses panel regressions.
  - Recent contributors include Fratzscher (2011), Forbes and Warnock (2012), Broner et al. (2013), Bruno and Shin (2015a), Cerutti, Claessens, and Ratnovski (2017); Koepke (2015) reviews ~40 papers; see also IMF (2014).
- Typical empirical regularities:
  - Worsening global risk conditions (measured by the VIX) typically lower capital flows, especially to emerging markets.
  - Evidence on US monetary policy effects is mixed:
    - Some papers find capital flows respond negatively to tightening US monetary policy, more so for portfolio bond and equity flows.
    - Other studies find monetary policy variables not always significant or with inconsistent sign; some find opposite signs (e.g., Correa et al. 2016 find banking flows increase for some countries when US interest rates rise).
- Explanatory power limitations:
  - Push variables often have limited explanatory power in absolute terms and sometimes relative to pull variables.
  - Bruno and Shin (2015b) report overall R2s of .1 or less when local factors account for only modest variation and global factors account for an overwhelming part.
  - Forbes and Warnock (2012) note low pseudo-R2s: .04 for flight episodes, increasing to .07, .13, and .15 for surges, retrenchments, and stops, respectively.
  - Cerutti et al. (2017) report R2s less than .1 when regressing banking flows on domestic factors, spreads, and global measures; R2s exceed .5 only when including pull variables and country fixed effects.
- Conclusion: capital flows are difficult to model empirically, especially using only push variables.

### Factor models and VAR approaches
- Rey (2013) and extensions:
  - Documents negative correlations between various capital flows and the VIX, varying between -.06 and -.28; conditional correlations peak at -.36 (debt flows to Central and Eastern Europe) and -.34 (equity flows to East Asia).
  - Principal components analysis finds a large single common factor among various asset prices (about 850 series); the first common factor has a relatively high, negative correlation (about .25) with the VIX.
  - VAR/BVAR evidence: US monetary policy shocks explain 4-17% of the variation in VIX; in the BVAR, about 14% of variation in capital flows to banks and 6% of debt flows to non-banks are explained by US monetary policy.
- Bruno and Shin (2015a):
  - Use a small-scale VAR over 1995Q4-2007Q4 with ordering: 1) Federal Funds rate; 2) broker dealer leverage; 3) BIS banking flows; 4) VIX; 5) US dollar REER.
  - Show US monetary policy affects US dollar exchange rate and capital flows funded by the US dollar.
  - Variance decomposition (in a smaller four-variable VAR) indicates US monetary policy shocks account for almost 30% of the variance of the VIX and between 10% and 20% of the variance of leverage of US broker dealers at horizons longer than 10 quarters; monetary policy shocks are less important for the variance of the US dollar REER.
  - Emphasize interactions among variables and importance of the leverage cycle of global banks for transmission.

### Heterogeneity by flow type, country, source, and time
- Flow direction and net vs. gross:
  - Sensitivities can differ between inflows and outflows, and between gross and net flows.
  - Outflows tend to increase when global interest rates rise; increases in global risk reduce the probability of outflow-driven stops.
  - Net flows can be less volatile if gross inflows and outflows offset each other; such offsetting is more common for advanced countries than for emerging markets.
- Flow type and maturity/currency:
  - Portfolio flows and credit typically react more to global factors than FDI.
  - Sensitivities vary by maturity (short vs. long) and currency (US dollar vs. others; foreign vs. local currency).
  - Changes in flow composition lead aggregate exposures to global factors to vary over time and across countries.
- Post-GFC changes and evidence:
  - Avdjiev et al. (2017a): impact of global risk increased post-GFC for international bond flows and declined for cross-border loan flows; greater sensitivities to US monetary policy driven mainly by behavioral shifts; better-capitalized banking systems show smaller rises in sensitivities and larger increases in international lending shares.
  - McCauley et al. (2015): unconventional monetary policy helped partial shift from borrowing away from global banks toward bonds for US-dollar-denominated credit.
  - Cerutti et al. (2017): changes in banking system conditions since the GFC affected importance of US and European drivers of bank flows.
- Financial openness and country characteristics:
  - More open countries (real and financial) experience greater effects from global factors.
  - Bruno and Shin (2015b): global factors have larger impact than local factors in more financially open countries with bigger banking flows; for countries subject to large bank inflows, R2 (0.21) is 3.6 times higher than for a model with just local variables.
  - Differences in explanatory power are small between developing and developed countries but larger between countries with high and low measures of law and order.
  - Calderón and Kubota (2013) document only small differences in sensitivities between emerging markets and advanced countries.
- Recipient type and source of financing:
  - Avdjiev et al. (2017b): gross capital inflows (1996–2014, 85 countries) show flows into banks and corporations decline in both advanced economies and emerging markets when the VIX rises; flows to emerging markets’ sovereigns decline with VIX rises, but not to advanced economies’ sovereigns.
  - Baskayay et al. (2017): in one country, global factors affect local banks through capital flow fluctuations; a lower VIX leads to lower local borrowing rates and greater credit supply, explaining up to 40% of cyclical credit growth.
  - Raddatz and Schmukler (2012) and Puy (2016): fund flows to/from emerging markets are pro-cyclical with financial conditions at fund domiciles, often independent of borrowing countries’ variables.
  - Jotikasthira et al. (2012): funding shocks in funds’ domiciles can trigger fire sales/purchases in global mutual fund portfolios, affecting emerging markets.
  - Cerutti et al. (2015): capital flows to emerging markets that rely more on mutual funds or international banks are more sensitive (short-term) to global factors.
- Composition shifts since the GFC:
  - Shin (2013) and others note substitutive movement from bank-intermediated flows toward corporate bond financing; Cerutti and Hong (2017) note post-GFC substitution of bank lending for debt securities occurred more among advanced economies’ borrowers, while emerging countries increased borrowing in both bank loans and debt securities.

### Exchange rate regime, monetary autonomy, and transmission
- Research on "dilemma vs. trilemma" and monetary policy independence:
  - Rey (2013) foregrounded the issue of domestic monetary policy independence under global financial conditions.
  - Han and Wei (2016): a flexible exchange rate by itself does not fully confer monetary policy autonomy; flexible regimes confer autonomy when the center raises its interest rate but not when it lowers it (“fear of floating” mainly takes the form of “fear of appreciation”); capital controls provide insulation when the center lowers its rate.
  - Klein and Shambaugh (2015): a moderately flexible exchange rate confers monetary policy autonomy; partial capital controls do not.
  - Obstfeld et al. (2017): transmission of global financial shocks is magnified under fixed exchange rates relative to more flexible regimes.
  - Aizenman et al. (2016): economies pursuing greater exchange rate stability and financial openness face stronger links with center economies’ monetary policies.
- Role of local vs. global determinants:
  - Ghosh et al. (2014): global factors act as “gatekeepers” for timing of capital flow surges, but local macroeconomic variables determine magnitudes.
  - Case studies show cross-country variation in flows during global stress, but differences are harder to fully explain.

### Gaps and motivation for current paper
- Ongoing research areas and debates:
  - Properties of different types of flows, relationships between global factors and domestic financial/economic developments, and cross-country differences in sensitivities remain active research topics.
- Specific gap addressed:
  - Few papers in either strand have analyzed the relative importance of global factors — the main objective of the present paper.

*Source: wp17193 - 2.  Literature Review*

### 3.  Identifying the GFCy: Data, Methodology and a First Look

### 3.  Identifying the GFCy: Data, Methodology and a First Look

### Strategy
- Objective: quantify the importance of the GFCy for capital flows by measuring what proportion of variation in capital flows is explained by the GFCy.
- Working definition: the GFCy is (high) commonality in financial conditions, manifest in capital flows, driven by observable global determinants.
- Two complementary approaches:
  - Direct: examine observable “fundamental” GFCy drivers — center-country macroeconomic and financial determinants of capital flows.
  - Indirect: examine observable manifestations of the GFCy via commonality of capital flows; attribute all common movements to the GFCy to obtain an upper bound on its importance.
- Rationale for conservatism:
  - Use a wide set of candidate center-country measures (not just the VIX) and include other center economies (EMU, UK) for robustness.
  - Use factor-analytic methods to extract common factors in capital flows and also combine both direct and indirect methods.

### Data: Center Country Variables
- Primary consensus proxy: the VIX (CBOE volatility index); used as the last VIX reading of each quarter, downloaded from Bloomberg.
- Additional volatility analogues used:
  - VDAX (German analogue; began trading in 1992; now replaced by the VDAX-NEW).
  - VSTOXX (began in 1999 with EMU).
  - IVI (British analogue; measures volatility of the FTSE 100 index; used after 2000).
- Other center-country “fundamentals” (seven variables a–g), all end-of-period or period-growth measures as specified:
  - a) nominal policy interest rate: Federal Funds rate (US), UK Base rate, Euro Area deposit facility rate; obtained from Haver.
  - b) ex post real policy interest rate: nominal rate minus the ex post year over year realized CPI inflation rate (inflation from IFSTSUB and GDS).
  - c) TED spread: end of period three month LIBOR minus government rate (Treasury bill rate for US, Gilt rate for UK, Government AAA bill rate for EMU).
  - d) yield curve slope: end of period ten-year minus three-month government rates.
  - e) GDP growth: obtained from IMF WEO.
  - f) growth in the real effective exchange rate (REER): quarter over quarter percentage change in the IMF’s CPI-based REER.
  - g) M2 growth: year over year growth in local currency, obtained from Haver.
- Center countries considered primarily: United States (default), with robustness checks including EMU and the UK. Note: EMU began in 1999Q1, reducing span for some variables.
- Additional candidate measures considered (tested but with little substantive effect): Shiller cyclically-adjusted P/E, US corporate Baa bond yield relative to 10-year T-bond yield, credit growth for US/UK/EMU.

### Data: Capital Flows
- Capital flow series (from IMF BOP Statistics), expressed as percentages of GDP, disaggregated into four types:
  - Foreign Direct Investment (“FDI”)
  - Portfolio Equity Investment (“Equity”)
  - Portfolio Debt Investments (“Debt”)
  - Bank Credit (“Credit”) — subset of other investment liabilities, “other investment liabilities, of which borrowers are banks.”
- In some analyses, portfolio equity and portfolio debt are summed for an overall measure of portfolio flows.
- Panel characteristics:
  - Unbalanced panel running from 1990Q1 through 2015Q4.
  - Data on 85 countries (listed in appendix Table A1).
  - Focus subset: 63 “small” countries by excluding plausible center-countries (USA, UK, EMU members, Japan).
  - Sample selection criteria: countries reporting 2015 data by type of capital flow and with at least a decade of quarterly series; two versions of the data set matched separately; standard checks and outlier inspections performed.
- Caution: raw scatterplots of capital flows against the VIX (Figure 1) show no indication of strong relationships; Figure A1 (flows < five percent of GDP) delivers a similar message, though raw plots gloss over other controllable sources of variation.

### Factors (Indirect measurement via commonality)
- Approach: estimate common factors in capital flows using factor analysis; construct many variants to test sensitivity.
- Implementation details:
  - Generate factors using two time samples: 1990Q1-2015Q4 and 1996Q1-2015Q4.
  - Use three country sets: advanced economies; emerging markets; mix of advanced and emerging economies (members listed in Table A1).
  - Construct factors for each direction (inflows, outflows) and five types of flows: FDI, debt, equity, credit, and portfolio debt+equity.
  - Estimate common factors in three ways for each data set:
    - Dynamic factor model with a single lag; extract factor from the largest eigenvalue (default, following Miranda Agrippino and Rey (2015)).
    - Dynamic factor model with two lags; extract factor from the largest eigenvalue (sensitivity to lag length).
    - Static factor model; extract factor from the largest eigenvalue.
  - Total factors generated: 180 = (2 samples of time) × (3 country sets) × (2 directions) × (5 flow types) × (3 estimation methods).
  - Additional robustness: use global factors provided by Miranda Agrippino and Rey (2015).
  - Estimation note: dynamic factor models estimated with the BHHH technique for a maximum of 20 iterations.

### A First Look at the Factors — empirical patterns and correlations
- Expectation: a ubiquitous GFCy should produce factors that are positively correlated across most capital flow types/directions and correlate with direct measures like the VIX.
- Figure 2 (dynamic factors for advanced economies, 1990Q1-2015Q4, single lag) findings:
  - Many scatterplots are clouds without a clear shape; some positive correlations, particularly among outflow factors, but overall limited commonality.
  - VIX correlations with the eight capital-flow factors:
    - All have correlations smaller than .5.
    - Three of the eight correlations are negative.
    - Only three correlations are statistically different from zero:
      - FDI inflows: .41 (positive)
      - FDI outflows: .34 (positive)
      - Debt inflows: -.29 (negative)
- Figure 3 (factors for capital inflows to advanced vs. emerging economies) findings:
  - FDI common factors: correlation coefficient .66 between advanced- and emerging-economy FDI inflow factors.
  - Other inflow types show modest or negligible correlation coefficients:
    - Debt inflows: -.01
    - Equity inflows: .18
    - Credit inflows: -.01
- Figure 4 (factors from stock returns, commodity prices, and VIX analogues) findings:
  - Two stock-market factors provided by Miranda Agrippino and Rey (MAR) are strongly correlated: .78.
  - Few strong positive correlations between stock/commodity price factors and other price factors or the capital-flow factors.
- Supplementary notes:
  - Time-series correlation between the VDAX and the VSTOXX is .9.
  - Figures A2 and A3 (not reproduced here) indicate limited commonality in capital flows across small advanced economies and a dozen emerging markets, respectively — an issue for a strong GFCy interpretation.

*Source: IMF Working Paper — 3. Identifying the GFCy: Data, Methodology and a First Look*

### Appendix Figure A4 examines the effects of different modeling assumptions for the

### Appendix Figure A4 examines the effects of different modeling assumptions for the

### Factor-modeling assumptions and preliminary evidence
- Dynamic factor models were estimated with one and two lags; a static factor model was also compared.
- Finding on lag choice:
  - "The exact lag length used in a dynamic factor model does not seem to make a large impact on the results, but using a static factor model delivers factors only loosely correlated with those of dynamic factor models. It seems reasonable to set the one-lag dynamic factor as our default."
- Observed relationships:
  - Factors extracted from different types and directions of capital flows are not strongly related with the VIX or with each other.
  - Limited relationships between capital-flow factors and the VIX, and limited relationships among capital flows, are difficult to reconcile with a substantive Global Financial Cycle (GFCy).

### Panel regression framework (equations and regressors)
- Baseline pooled panel (equation (1)):
  - CAPFLOW_{d,e,i,t} = {φ_i} + {θ_t} + ε_{d,e,i,t} across i,t
  - CAPFLOW_{d,e,i,t} = capital flow (direction d, type e, country i, quarter t) as percent of country i GDP.
  - {φ_i} and {θ_t} are country- and time-specific fixed effects.
  - Sample: pooled across countries with data available for the entire period 1990Q1-2015Q4.
- Time-varying "push" regressors (equation (2)):
  - CAPFLOW_{d,e,i,t} = Σβ_j USFUND_{j t} + Σγ_k FAC_{d,e k t} + {φ_i} + ε_{d,e,i,t}
  - USFUND: contemporary US values of eight variables: a) VIX; b) real GDP growth rate; c) nominal policy rate; d) real policy rate; e) TED spread; f) yield curve slope; g) REER change; and h) M2 growth.
  - FAC_{d,e k t}: two dynamic capital flow factors (one for advanced economies, one for emerging markets), estimated with one lag and matched to direction/type of regressand.
  - Estimation approach: least squares, pooling across countries.
- Country-specific slopes (equation (3)):
  - CAPFLOW_{d,e,i,t} = Σβ_{i j s} USFUND_{j t-s} + Σγ_k FAC_{d,e k t} + {φ_i} + ε_{d,e,i,t} across i,t
  - Allows country-specific contemporary and lagged responses to US variables.

### Panel-regression empirical findings (goodness of fit)
- Table 1 (equation (1)):
  - Eight rows (two directions × four types). All sixteen goodness-of-fit measures (within and overall R^2) are low across samples (1990Q1-2015Q4; extended sample to 1996Q1; restricted to small economies).
  - Conclusion: global time-specific phenomena (e.g., GFCy) do not have an important effect on capital flows if assumed to have identical impact across countries.
- Table 2 (equation (2)):
  - Standard push factors (eight US variables) plus two capital-flow factors explain only a small fraction of variation in capital flows.
  - Substituting Miranda Agrippino-Rey factors for default factors leaves results essentially unchanged.
  - Adding one lag or four lags of US variables improves goodness of fit only marginally.
- Table 3 (equation (3)):
  - Allowing country-specific slopes for the VIX alone does not materially improve fit; maximum R^2 remains less than .15 when only VIX has country-specific slopes.
  - Allowing country-specific slopes for all eight US variables (contemporary or with one lag) improves fit but remains limited; no R^2 approaches .5.
  - Best-fitting specification: FDI inflows with country-specific slopes, within R^2 = .42, but estimated with over 300 regressors and only 5.8 (≈2016/349) observations per coefficient; overall R^2 = .08.
  - Improvement in fit comes with a large increase in free parameters and potential overfitting.

### Country-level time-series regressions and distribution of fit (equation (4))
- Country-by-country specification:
  - CAPFLOW_{d,e,i,t} = Σβ_{i j} USFUND_{j t} + Σγ_k FAC_{d,e k t} + φ_i + ε_{d,e,i,t} across t
  - Estimated for small countries, up to ten time-series equations per country (two directions × five types).
  - Focus metric: adjusted Rଶ (to penalize over-parameterization); results robust to using R^2.
- Aggregate results (Figure 5):
  - Number of estimated Rଶ values: 598 (all combinations of countries, directions, types).
  - Summary statistics:
    - Mean Rଶ = .12.
    - Over a quarter of Rଶ values are negative.
    - Only a few Rଶ exceed .5.
    - FDI inflows: mean Rଶ = .25 (FDI inflows are relatively easiest to model, but still weak).
  - Conclusion: conventional time-series models using eight center-country variables and two capital-flow factors typically fit small countries poorly.

### Sensitivity analyses (Appendix Figures A6–A13)
- Variations explored:
  - Figure A6: use conventional R^2 (not adjusted); mean R^2 < .25 — similar poor fit.
  - Figure A7: remove US variables, retain two capital-flow factors.
  - Figure A8: restrict sample to small countries with complete series 1990Q1–2015Q4.
  - Figure A9: restrict to countries with above-median capital mobility (Chinn-Ito index).
  - Figure A10: restrict to countries with per capita annual real GDP ≥ $5,000.
  - Figure A11: add four quarterly lags of the VIX.
  - Figure A12: add a quarterly lag of all eight US variables.
  - Figure A13: add contemporaneous values of eight British and Eurozone variables to the eight US variables (sample restricted to 2000Q1–2015Q4).
- Overall sensitivity result:
  - "None of the nine appendix variations on Figure 5 alters our view substantively; as with the straightforward estimates of equation (4), national capital flow equations typically fit the time-series data poorly, even when incorporating eight center-country variables and two capital flow factors."
  - Exceptions exist but are rare and do not indicate a systemic GFCy presence.

### Concise interpretation and implications
- Repeated evidence across pooled panel, country-specific slope, and country-level time-series regressions indicates:
  - Center-country variables (including the VIX) and common capital-flow factors explain little of the cross-country and time-series variation in capital flows for most countries, especially small countries.
  - Allowing country-specific responses improves fit only by substantially increasing parameters, raising concerns about overfitting (e.g., best FDI specification: within R^2 = .42 but overall R^2 = .08 with >300 regressors).
  - The weak and inconsistent relationships between capital-flow factors and the VIX, and between capital flows across countries, are difficult to reconcile with a dominant, systemic Global Financial Cycle (GFCy) that uniformly drives capital flows.
- Evidence robustness:
  - Results are robust to alternative factor choices (e.g., Miranda Agrippino-Rey factors), additional lags, inclusion/exclusion of US variables, sample restrictions by capital mobility and income, and augmenting with British and Eurozone variables.

*Source: wp17193 - Appendix Figure A4 examines the effects of different modeling assumptions for the (IMF working paper material provided).*

### Appendix Figure A14 substitutes the natural logarithm of the VIX for its level; the results are inconsequentially

### wp17193 - Appendix Figure A14 substitutes the natural logarithm of the VIX for its level; the results are inconsequentially different from Figure 5, so we stick to the level of the VIX. Appendix Figure A15 is the same as Figure 5 but with a top row of histograms for net capital flows, in the same order as the gross flows below; the absence of novel results for net flows is the reason we stick to gross flows below.

### Robustness of VIX specification and choice of gross vs net flows
- Appendix Figure A14: substituting the natural logarithm of the VIX for its level yields results "inconsequentially different" from Figure 5; authors therefore stick to the level of the VIX.
- Appendix Figure A15: repeats Figure 5 but adds a top row of histograms for net capital flows; absence of novel results for net flows motivates focus on gross flows.

### Distribution of goodness-of-fit across country time-series regressions (Figure 6 and related)
- Sample construction and estimation:
  - Separate time-series regression estimated for each small economy with sufficiently long series of capital flow data (not necessarily for all of 1990Q1-2015Q4).
  - Sample contains 63 such countries, delivering a distribution of goodness-of-fit measures.
- Box-plot presentation:
  - Box extends from the 25th to 75th percentiles of the Rଶ statistics, median marked by a vertical bar.
  - Whiskers extend to adjacent values defined as the most extreme values within 150% of the interquartile range of the nearest quartile; outliers individually marked.
  - For continuity, a vertical line at .25 is marked for each set of box plots as a notional benchmark.
- Key findings from Figure 6:
  - The boxes in the top-left panel are located well to the left; seven of the eight interquartile boxes lie below the .25 marker, implying majority of equations fit poorly.
  - Occasional outliers to the right indicate countries where a particular capital flow is well modelled by the eleven coefficients in equation (4), but these are exceptions.
  - Panels using R2 instead of Rଶ show little difference; exact choice of statistic makes little difference.
  - Middle panels restrict sample to countries with complete time series over the entire 26 years; right panels drop all eight US variables from estimation. All six panels look similar.
  - Conclusion: the GFCy does not seem to explain much variation in capital flows when modelled with the ten time-varying regressors of (4).

### Extensive robustness checks across specifications and samples (Figures 7–9)
- Figure 7 specification variations and results:
  - Top-left: drops factors from RHS of (4).
  - Top-middle: substitutes static for dynamic factors.
  - Top-right: substitutes dynamic factors estimated by Miranda Agrippin and Rey for authors' dynamic factors.
  - Bottom-left: adds 4 lags of the VIX to RHS of (4).
  - Bottom-middle: adds a single lag of all eight US variables as regressors.
  - Bottom-right: adds contemporary values of all eight variables for both the UK and the EMU (restricts time span).
  - None of these changes leads to a dramatic improvement in fit.
  - Highest Rଶ statistics are in bottom-right when 24 US, British, and European variables plus two factors extracted from capital flows are used to model 64 observations available since 2000Q1. Even then:
    - A reasonable median Rଶ is .25.
    - There are plenty of negative statistics.
- Figure 8 sample perturbations:
  - Top-left: allows each of the 8 US variables to have a different coefficient from 2009Q1 onwards (following Avdjiev et al. (2017a)); breaking at 2008Q1 instead of 2009Q1 changes results insubstantially.
  - Bottom-left: drops outlier observations where residual lies more than 2.5 standard errors from the mean.
  - Top-middle: drops observations where capital flows exceed 8% of GDP.
  - Bottom-middle: winsorizes all observations at 5% of GDP.
  - Top-right: drops advanced economies leaving only developing and emerging markets.
  - Bottom-right: drops observations between 1990Q1 and 1995Q4.
  - None of these perturbations alters results substantially.
- Figure 9 sample splits by country characteristics:
  - Top-left: drops observations with below-median financial openness (Chinn-Ito measure).
  - Bottom-left: drops observations for countries with annual real GDP per capita below $5,000.
  - Top-middle and panel below: split by region (East/Central Asia, the Pacific and Europe vs other countries).
  - Right-side panels: split by exchange rate regime using Reinhart-Rogoff annual data (flexible regimes above; more rigid regimes below).
  - Results: not particularly sensitive to region or exchange rate regime; GFCy is quantitatively not so important.

### Event study on periods of high VIX (Figure 10 and appendices)
- Event definition and sample:
  - Periods when the VIX ended the quarter above 30 occurred on eight occasions during the 26-year sample; these are treated as events depicted in Figure 10.
  - VIX closed above 30 on 1998Q3, 2001Q3, 2002Q3, 2008Q3, 2008Q4, 2009Q1, 2010Q2, and 2011Q3.
  - Authors plot movements of capital flows for 63 small countries during the twelve quarters on either side of these high-VIX quarters.
- Event-study findings (Figure 10 and appendices A17–A20):
  - Top-left graph: average FDI inflows (relative to GDP) traced from twelve quarters before to twelve quarters after events; (5, 95) confidence intervals plotted.
  - Little action in FDI inflows around high-VIX quarters; analogous patterns for portfolio debt, portfolio equity, and credit inflows, and for outflows.
  - All eight graphs in Figure 10 deliver same message: capital flows do not change systematically and significantly during periods around high VIX values; some small changes but no statistically significant movements.
  - Robustness to alternative event definitions:
    - Appendix Figure A17: nineteen events when VIX ended quarter exceeding 25 produce little change.
    - Figure A18: dozen periods when VIX rose by more than 5.
    - Figure A19: seven quarters when average (not just end-of-quarter) VIX exceeded 30.
    - Figure A20: seven quarters when the US$ real effective exchange rate appreciated more than 3.5%.
  - Conclusion: periods of financial stress (high and/or rapidly rising VIX or dollar) do not seem systematically associated with unusual capital flow movements at quarterly frequency across sample.

### Coefficient estimates and explanatory power
- Focus of paper:
  - Emphasis on quantitative importance of the GFCy for determining capital flows; focus on goodness-of-fit rather than sign/size/significance of coefficients.
  - Thousands of capital flow equations estimated; coefficients vary dramatically across countries and flow types.
- Appendix Table A3 and Figure A21:
  - Tabulated coefficient estimates from (4) for capital flows into five emerging markets; coefficients and significance levels vary dramatically.
  - General confirmation that portfolio flows, especially debt flows, tend to co-move negatively with the VIX more so than other types of capital flows.
  - Scatter plots of actual vs fitted capital flows confirm limited explanatory power.
- Empirical emphasis implies:
  - Even when models fit reasonably for particular country-flows, general distribution indicates limited explanatory power of GFCy across many countries and flow types.

### Conclusion and policy implications
- Main quantitative conclusions:
  - Difficult to find consistent manifestations of the GFCy in capital flows across a broad sample using conventional models and data.
  - Little evidence that the GFCy explains systematically as much as a quarter of the variation in capital flows.
  - Where real credit growth replaces capital flows as regressand, median R2 is about 0.3, suggesting credit may be slightly more linked to the GFCy than most capital flows (Figure A16).
- Caveats and directions for future work:
  - Empirics based on conventional capital-flow models estimated in straightforward ways; future work could use out-of-sample techniques, consider non-linear effects, include richer dynamics (e.g., VARs), examine higher-frequency data, and incorporate endogenous domestic responses.
  - Inclusion of domestic determinants is likely to reduce apparent effects of GFCy further.
  - GFCy may matter more in other variables (domestic asset prices, credit) or at higher frequencies for short periods.
- Policy implications (two contrasted scenarios):
  - If most variation in capital flows were explained by center-country GFCy:
    - Peripheral countries might view benefits of financial integration as less than risks of volatile exogenous flows, and could choose insulation measures (capital controls, macro-prudential policies, self-insurance).
  - Given results suggesting GFCy explains only small fraction of variation:
    - More idiosyncratic phenomena explain capital flows, making systematic policies to screen "good" from "bad" flows difficult.
    - These idiosyncratic influences may be domestic, implying policy authorities should take ownership and responsibility for managing capital flows.
  - Because GFCy effects can show up in other variables (asset prices, credit), more analysis needed before firm policy conclusions.

*Source: wp17193 (appendix extract).*

### References

### References (wp17193 - References)

### Bibliographic sources
- Extensive list of working papers, journal articles, IMF World Economic Outlook chapters, IMF Staff Discussion Note, NBER and CEPR papers, and central bank/BIS mimeos cited. Notable authors and works include:
  - Avdjiev, Stefan; Gambacorta, Leonardo; Goldberg, Linda S.; Schiaffi, Stefano (2017a) “The shifting drivers of international capital flows,” Bank for International Settlements (mimeo).
  - Avdjiev, Stefan; Hardy, Bryan; Kalemli-Ozcan, Sebnem; Serven, Luis (2017b) “Gross Capital Flows into Banks, Corporates and Sovereigns” CEPR Discussion Paper DP11,806.
  - Aizenman, Joshua; Binici, Mahir; Hutchison, Michael M. (2014) “The Transmission of Federal Reserve Tapering News to Emerging Financial Markets,” NBER Working Paper No. 19,980.
  - Blanchard, Olivier; Ostry, Jonathan D.; Ghosh, Atish R.; Chamon, Marcos (2015) “Are Capital Inflows Expansionary or Contractionary? Theory, Policy Implications, and Some Evidence," NBER Working Paper No. 21,619.
  - Bruno, Valentina and Hyun Song Shin (2015a, 2015b) on capital flows, risk-taking, and cross-border banking.
  - IMF pieces: IMF (2013) “The Yin and Yang of Capital Flow Management: Balancing Capital Inflows with Capital Outflows,” WEO, Chapter 4, Fall; IMF (2014) “Global Liquidity: Issues for Surveillance”; IMF (2016) "Understanding The Slowdown In Capital Flows To Emerging Markets," WEO, Chapter 2, Spring.
- Research topics covered by references: global liquidity, push vs. pull factors for capital flows, gross capital flows dynamics, international spillovers and local credit cycles, transmission of monetary policy (tapering), capital flow surges/stops, portfolio versus banking flows, and the global financial cycle.

### Tabular evidence and regression fit statistics
- Table 1: Importance of Common Phenomena for Goodness of Fit in Panel Regressions (panel regressions with time- and country-specific fixed effects; data pooled across countries with series continuously available over relevant sample)
  - Time span columns: 1990Q1-2015Q4; 1996Q1-2015Q4; 1990Q1-2015Q4 (Country span: All)
  - Selected Within / Overall R2 values (exact as reported):
    - Inflow FDI: .06, .05, .06, .05, .08, .07
    - Inflow Portfolio Debt: .10, .09, .10, .09, .07, .07
    - Inflow Portfolio Equity: .05, .03, .05, .03, .11, .10
    - Inflow Credit: .10, .09, .10, .09, .07, .06
    - Outflow FDI: .08, .07, .08, .07, .08, .07
    - Outflow Portfolio Debt: .07, .06, .07, .06, .09, .09
    - Outflow Portfolio Equity: .08, .06, .08, .06, .07, .07
    - Outflow Credit: .11, .11, .11, .11, .07, .06
- Table 2: Importance of Common Phenomena for Goodness of Fit in Panel Regressions (panel regressions with country-specific fixed effects; data pooled across small countries 1990Q1-2015Q4)
  - Regressors variations: MAR factors; Lags of 8 US vars: None / None / Add 1 lag / Add 4 lags
  - Selected Within/Overall R2 values (exact):
    - Inflow FDI: .05/.04 ; .07/.06 ; .05/.04 ; .06/.06
    - Inflow Portfolio Debt: .04/.04 ; .05/.04 ; .05/.04 ; .06/.05
    - Inflow Portfolio Equity: .05/.05 ; .05/.05 ; .06/.06 ; .08/.08
    - Inflow Credit: .03/.03 ; .04/.04 ; .04/.04 ; .06/.06
    - Outflow FDI: .03/.03 ; .04/.04 ; .03/.03 ; .05/.05
    - Outflow Portfolio Debt: .05/.05 ; .06/.06 ; .05/.05 ; .07/.07
    - Outflow Portfolio Equity: .01/.01 ; .01/.01 ; .02/.02 ; .05/.05
    - Outflow Credit: .03/.03 ; .04/.03 ; .04/.03 ; .05/.04
  - Common phenomena include contemporary (and lagged) values of: eight US variables (VIX; real GDP growth rate; nominal policy rate; real policy rate; TED (2m LIBOR/Tbill) spread; yield curve (10y/3m treasury) spread; REER change; M2 growth) and two dynamic capital flow factors (advanced and emerging), estimated with one lag and specific to direction/type of capital flow.
- Table 3: Country-Heterogeneous Responses to Common Phenomena in Panel Regressions (panel regressions with country-specific fixed effects and dynamic factors; pooled across all small countries 1990Q1-2015Q4)
  - Add country-specific interactions with various sets of US variables (Current VIX; Current, 1 lag of VIX; Current, 4 lags of VIX; Current of 8 US vars; Current, lag of 8 US vars)
  - Selected Within/Overall R2 values for Inflow FDI and others (exact):
    - Inflow FDI: .05/.05 ; .05/.04 ; .11/.02 ; .35/.08 ; .42/.08
    - Inflow Portfolio Debt: .09/.02 ; .11/.02 ; .12/.01 ; .25/.00 ; .42/.05
    - Inflow Portfolio Equity: .04/.02 ; .05/.03 ; .09/.05 ; .26/.07 ; .35/.11
    - Inflow Credit: .03/.00 ; .05/.00 ; .11/.00 ; .32/.00 ; .38/.00
    - Outflow FDI: .05/.01 ; .06/.01 ; .11/.02 ; .27/.01 ; .36/.01
    - Outflow Portfolio Debt: .08/.00 ; .11/.00 ; .13/.01 ; .27/.02 ; .37/.04
    - Outflow Portfolio Equity: .03/.03 ; .05/.02 ; .08/.00 ; .15/.00 ; .33/.13
    - Outflow Credit: .07/.02 ; .07/.02 ; .10/.03 ; .27/.02 ; .35/.01
  - Up to eight current and lagged US variables included as indicated: a) VIX; b) nominal policy rate; c) real policy rate; d) TED (2m LIBOR/Tbill) spread; e) yield curve (10y/3m treasury) spread; f) growth; g) REER change; and h) M2 growth.

### Event regression and goodness-of-fit summaries
- Table A2: Goodness of Fit for Panel Forbes-Warnock Capital Flow Event Equations (53 countries, 1990Q1-2009Q4)
  - Reported goodness-of-fit measures (Quasi R2 for probit/logit or within R2 for LS):
    - Probit, random Time-Effects: Flight .01 ; Retrenchment .00 ; Stop .00 ; Surge .01
    - Logit, fixed Time-Effects: Flight .07 ; Retrenchment .17 ; Stop .21 ; Surge .12
    - LS, fixed Time-Effects: Flight .06 ; Retrenchment .16 ; Stop .21 ; Surge .08
    - Probit, random 8 US Variables: Flight .01 ; Retrenchment .00 ; Stop .00 ; Surge .01
    - Logit, fixed 8 US Variables: Flight .04 ; Retrenchment .12 ; Stop .15 ; Surge .07
    - LS, fixed 8 US Variables: Flight .03 ; Retrenchment .12 ; Stop .16 ; Surge .04

### Representative coefficient evidence (selected from Table A3)
- Table A3 presents signs and coefficient estimates (with robust standard errors in parentheses) from selected capital inflow equations; regressand is capital inflow of given type, as percentage of domestic GDP. LS estimation, 1990Q1-2015Q4 (with some gaps). Intercept and advanced/emerging market factors included but not reported.
- Examples (coefficients and standard errors reported exactly):
  - South Africa, FDI: VIX coefficient -.08 (.05)
  - Brazil, FDI: VIX coefficient .01 (.02)
  - Chile, FDI: VIX coefficient -.03 (.06)
  - Mexico, FDI: VIX coefficient .01 (.02)
  - Indonesia, FDI: VIX coefficient -.05 (.02)
  - South Africa, Debt: VIX coefficient -.09 (.05)
  - Brazil, Debt: VIX coefficient -.05 (.03)
  - Chile, Debt: VIX coefficient .02 (.04)
  - Mexico, Debt: VIX coefficient -.09 (.03)
  - Indonesia, Debt: VIX coefficient -.07 (.02)
  - South Africa, Equity: VIX coefficient -.03 (.05)
  - Brazil, Equity: VIX coefficient -.04 (.02)
  - Chile, Equity: VIX coefficient -.03 (.02)
  - Mexico, Equity: VIX coefficient -.03 (.01)
  - Indonesia, Equity: VIX coefficient -.03 (.02)
  - South Africa, Credit: VIX coefficient -.03 (.02)
  - Brazil, Credit: VIX coefficient -.03 (.03)
  - Chile, Credit: VIX coefficient .02 (.03)
  - Mexico, Credit: VIX coefficient .01 (.05)
  - Indonesia, Credit: VIX coefficient -.01 (.01)

### Figures and visual evidence (selected metadata and captions)
- Sample coverage and dates repeatedly used across figures:
  - Primary sample period: 1990Q1-2015Q4.
  - Many figures use 63 small countries, 1990Q1-2015Q4.
  - Dynamic factors are often estimated with 1 lag.
- Key figure captions and notes (exact wording where provided):
  - Figure 1: "Capital Flows, %GDP (y) against VIX (x); 63 small countries 1990Q1-2015Q4" (shows correlations: Corr=.02 for FDI Inflows; Corr=-.09 for Debt Inflows; Corr=-.03 for Equity Inflows; Corr=0.05 for Credit Inflows; Corr=.01 for FDI Outflows; Corr=-.05 for Debt Outflows; Corr=-.00 for Equity Outflows; Corr=-.06 for Credit Outflows).
  - Figure 2: "Dynamic factors (1 lag) 1990Q1-2015Q4, Advanced Economies" — plots VIX and capital flow factors (FDI, Debt, Equity, Credit inflows/outflows).
  - Figure 3: "Dynamic factors 1990Q1-2015Q4, 1 lag — Capital Inflow Factors: Advanced and Emerging Economies."
  - Figure 4: "Dynamic factors 1990Q1-2015Q4, 1 lag — Fear Measures and Factors from Asset/Commodity Prices."
  - Figure 5: "Histograms of (up to 598) adjusted R2s, small countries 1990Q1-2015Q4 — Regressors: Adv/EM 26-yr dyn factors, 8 US var's (VIX/TED/yield/policy/real/growth/REER chg/M2 growth)." Includes mean values for panels (e.g., Mean=.12 All; Mean=.14 Inflows; Mean=.10 Outflows; Mean=.11 All Portfolio; Mean=.25 FDI Inflows; Mean=.09 Debt Inflows; Mean=.12 Equity Inflows; Mean=.13 Credit Inflows; Mean=.14 FDI Outflows; Mean=.10 Debt Outflows; Mean=.14 Equity Outflows; Mean=.02 Credit Outflows).
  - Figures 6–21: Multiple robustness and event-window plots showing distributions of adjusted/raw R2s and capital flows around stress events (e.g., "Capital Flows (%GDP) around quarters when VIX close>30"; "Capital Flows (%GDP) around quarters when VIX close>25"; "Capital Flows (%GDP) around quarters when final VIX rises >5"; "US REER rises>3.5%, 7 events"; etc.). Many figures report means with (5,95) confidence intervals and sample/event counts (e.g., 8 events, 19 events, 12 events, 7 events).
  - Figure A1 and subsequent appendix figures: variations and robustness checks (e.g., |Values|<5%GDP, 63 small countries 1990Q1-2015Q4; Advanced Economies and Emerging Markets factor extraction plots).
  - Figure A21: "Modeling Capital Inflows: Some Examples" — linear regressions with 8 US fundamental variables, 2 dynamic factors, and intercept; 1990Q1-2015Q4. Reports fitted vs. actual gross flows (% GDP) and R2 values for country/type examples (exact R2s reported, e.g., South Africa FDI R2 = .17; Brazil FDI R2 = .65; Chile FDI R2 = .23; Mexico FDI R2 = .19; Indonesia FDI R2 = .58; and additional R2s for debt/equity/credit per country as displayed).

### Appendix: country coverage and sample annotations
- Table A1: List of Countries (63 small countries and others used in factor construction). Countries listed explicitly include: Argentina; Armenia; Australia; Austria; Azerbaijan; Bangladesh; Belarus; Belgium; Bolivia; Brazil; Bulgaria; Cabo Verde; Canada; Chile; China; Colombia; Costa Rica; Croatia; Cyprus; Czech Republic; Denmark; Ecuador; Estonia; Finland; France; Georgia; Germany; Greece; Hong Kong; Honduras; Hungary; Iceland; India (not listed); Indonesia; Ireland; Israel; Italy; Japan; Jordan; Kazakhstan; Kyrgyzstan; Korea; Laos; Latvia; Lebanon; Lesotho; Lithuania; Luxembourg; Macedonia; Malta; Mauritius; Mexico; Moldova; Mongolia; Namibia; Netherlands; New Zealand; Nicaragua; Norway; Pakistan; Panama; Paraguay; Peru; Philippines; Poland; Portugal; Romania; Russia; Slovakia; Slovenia; South Africa; Spain; Sri Lanka; Suriname; Sweden; Switzerland; Thailand; Turkey; Uganda; Ukraine; UK; USA; Uruguay; Venezuela; Viet Nam; Zambia. (Exact list entries and grouping as presented in Table A1.)
- Footnote keys in Table A1 (exact text):
  - "1 EMU member 2017; 2 large non-Euro economy; 3 used to generate advanced economy factors; 4 used to generate emerging market factors; 5 used to generate advanced/emerging factors."

*Italic: Source — wp17193 - References (PDF chapter/section), canonical PDF: wp17193 - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17193.pdf_