## 10. Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit (Monthly)

## Source details

**Canonical URL:** [10. Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit (Monthly)](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024158-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024158-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024158-print-pdf.pdf.json)

---

### Introduction and motivation and scope
- Research context:
  - Builds on Global Financial Cycle (GFCy) literature following Forbes and Warnock (2012), Rey (2013), Bruno and Shin (2015); recent review Miranda-Agrippino and Rey (2022).
  - Policy relevance spans macroprudential policies, capital flow management tools, quantitative easing/tightening, and market interventions.
- Objective:
  - Quantify the proportion of variation in asset prices, domestic credit, and capital flows explained by the GFCy for small advanced and emerging market countries.
- Data and period:
  - Series covered: short- and long-term interest rates, sovereign spreads, equity prices (local currency and USD), house prices, local (real) credit, and capital flows (by type and direction), subject to data availability.
  - Period reported in main summary: 2000Q1-2021Q4.
  - Monthly-frequency analysis performed (2000M1-2019M4 for monthly regressions including MAR factor; house prices drop out at monthly frequency).

### Empirical strategy, factors, and estimation
- Definition of GFCy:
  - Captured by dynamic common factors estimated separately for AEs and EMEs and by eight US "center-country" variables (USFUND_jt).
- US GFCy-drivers (USFUND_jt), end-of-period quarter (eight fundamentals):
  - (i) the VIX;
  - (ii) the change in the dollar real effective exchange rate (REER, quarter over quarter percentage change in IMF’s CPI-based REER);
  - (iii) the nominal policy interest rate (the Federal Funds rate);
  - (iii) the ex post real policy interest rate (nominal rate minus ex post year over year realized CPI inflation rate);
  - (iv) the TED spread (three-month LIBOR minus the US Treasury bill rate);
  - (v) the yield curve slope (US 10-year rate minus the three-month government rate);
  - (vi) US real GDP growth (from the IMF WEO);
  - (viii) US M2 growth (year over year growth).
- Country-level regression:
  - VAR_{i,t} = sum_k γ_{k,i} FAC_{k,t} + sum_j β_{j,i} USFUND_{j,t} + φ_i + ε_{c,t}
  - VAR_{i,t} covers policy rates, short-term rates, gov’t spreads, MSCI growth/returns, residential real estate growth, domestic credit flow/GDP, real credit growth, and capital flows by type/direction.
- Goodness-of-fit metric:
  - Focus on adjusted R^2 (AR2) distributions across 76 small AEs and EMEs (sample listed in Appendix Table 1).
  - AR2s summarized via box-plots for three specifications:
    - AllFactors (dynamic AE and EM factors + 8 US vars);
    - AllFactors only (factors only);
    - U.S. variables only (US vars only).

### Baseline empirical findings (period 2000Q1-2021Q4)
- High-level explanatory power of GFCy (combination of series’ common factor and US GFCy-variables):
  - more than 75% of the variation explained for policy rates, short-term interest rates, and government spreads;
  - about 60% of the variation explained for house prices;
  - up to 40% of the variation explained for stock market returns (measured either in US dollars or local currency);
  - about 30% of the variation explained for domestic credit;
  - up to just 25% of the variations in capital flows (dis-aggregated by four types and by direction).
- Asset prices and credit ranking (median AR2s using all variables):
  - Policy and short-term interest rates: median AR2 above 75%.
  - Government spreads: median AR2 about 75%.
  - House prices: median AR2 around 60%; common factors only ≈ 20%; US variables only ≈ 40%.
  - Stock market returns (USD): median AR2 about 40%; US variables only similar; common factors only often < 10%.
  - Credit flow/GDP and real credit growth: median AR2 slightly below 30%; US variables only yield similar AR2s; common factors only somewhat lower.
- Capital flows:
  - Eight types analyzed: FDI in/out, other investment in/out, portfolio debt in/out, portfolio equity in/out.
  - AR2 distributions for capital flows: rarely do the 75th percentiles exceed the 25% level; median AR2s generally low (up to just 25%).
  - Findings align with Cerutti et al. (2019b) reporting ~25% adjusted R^2 for capital flows for 1990Q1-2015Q4.

### Robustness tests and sensitivity
- Overview:
  - Robustness exercises include FX regime splits (pegged vs flexible), capital account openness splits (below/above median Chinn and Ito 2015 index), addition of Miranda-Agrippino and Rey (2022) (MAR) global factor, and different frequencies and sample periods.
- FX regime and capital account openness:
  - Pegged vs flexible regimes: very similar degrees of comovement in asset price and credit measures; only small differences in AR2s.
  - Stock market returns: medians marginally higher under flexible regimes.
  - Combining FX regime with capital account openness: some differences consistent with traditional models (e.g., higher median AR2 for credit under open + pegged), but differences are small and not consistently meaningful.
  - Capital flows: a few median AR2s lowest for pegged + closed capital account and highest for flexible + open capital account.
  - Interpretation: exchange rate regime choice may matter less for capital flow exposure to the GFCy than some literature suggests.
- Adding MAR factor — Quarterly (2000Q1-2019Q2):
  - Specification: 8 US vars + Adv+EM dynamic factors + MAR factor (three-month average).
  - Little material change: marginal increases in median AR2s and slightly lower dispersion; MAR shows low AR2s for asset prices and credit.
  - Capital flows: few differences in AR2s and ranking compared to baseline.
  - Note: some AR2s become negative when data span is short and many regressors included.
- Adding MAR factor — Monthly (2000M1-2019M4; house prices excluded):
  - Monthly regressions show expected drop in AR2 magnitudes but preserve ranking of variables.
  - Ranking preserved: highest explanatory power for interest rates; next equity returns in USD; lower for credit flows and equity returns in local currency.
  - Conclusion: baseline results robust to monthly frequency and inclusion of MAR factor.

### Policy implications and research priorities
- Policy implications:
  - GFCy explains a much larger share of variation in asset prices than in quantities (capital flows). Quantity-based tools often used to regulate financial systems may affect local asset prices but may not directly or fully control them.
  - Tool selection and calibration should account for the different magnitudes and speeds of GFCy effects across prices and quantities:
    - High-frequency-adjustable tools (e.g., FX interventions) better suited for rapid asset price changes.
    - Lower-frequency tools (e.g., macroprudential limits like LTV restrictions) better suited for slower-moving cycles and vulnerabilities.
- Research agenda priorities:
  - Better understand drivers of the GFCy and why advanced economies play a large role.
  - Clarify the role of the US dollar and AEs’ monetary policy in driving the GFCy.
  - Assess whether source-country macroprudential regulation amplifies or reduces the GFCy.
  - Identify exact transmission channels: common balance sheets/exposures vs contagion; drivers of risk-on vs risk-off; differences between local exchange rate vs broad dollar effects; roles of specific capital flow types.
  - Evaluate how domestic policy responses alter adjustment of asset prices and local financial conditions when quantity tools target capital flows vs local variables.

### Key numeric markers and sample notes
- Sample size: 76 economies (Appendix Table 1).
- Primary period for baseline summary: 2000Q1-2021Q4.
- Quarterly MAR-included regressions sample: 2000Q1-2019Q2.
- Monthly MAR-included regressions sample: 2000M1-2019M4.
- Adjusted R^2 (AR2) benchmarks reported exactly as:
  - >75% for policy rates, short-term rates, government spreads;
  - ≈60% for house prices;
  - up to 40% for stock market returns;
  - ≈30% for domestic credit;
  - up to just 25% for capital flows.

*Source: wpiea2024158-print-pdf — "10. Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit (Monthly)"*

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

### 1. Introduction and Motivation

### Chapter structure and major sections
- 1. Introduction and Motivation ........................................................................................................................... 3
- 2. Literature Review ............................................................................................................................................ 7
- 3. Empirical Strategy, Data, and Country Sample .......................................................................................... 11
- 4. Empirical Results .......................................................................................................................................... 14
  - 4.1 Baseline Results ..................................................................................................................................... 14
  - 4.2 Robustness Tests ................................................................................................................................... 16
- 5. Summary and Policy Implications and Issues ........................................................................................... 22
- Appendix ............................................................................................................................................................ 31

### Appendix contents and figures referenced
- Table 1: List of Small Advanced and Emerging Market Countries ....................................................................... 31
- Figure A.1: Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit ......... 32
- Figure A.2: Goodness of Fit Measures for GFCy in Capital Flows. .............................................................. 32
- Figure A.3: Goodness of Fit Measures for GFCy (including MAR Factor at eop) in Asset Prices and Credit
  ..................................................................................................................................................................... 33
- Figure A.4: Goodness of Fit Measures for GFCy (including MAR Factor) in Capital Flows ......................... 33

### Figures listed in the chapter
- Figure 1. Common Factor for AEs and EMEs in Asset Prices and Credit ...................................................................... 12
- Figure 2. Common Factor for AEs and EMEs in Capital Flows ...................................................................................... 13
- Figure 3. Goodness of Fit Measures for GFCy in Asset Prices and Credit .................................................................... 15
- Figure 4. Goodness of Fit Measure for GFCy in Captial Flows ...................................................................................... 16
- Figure 5. Goodness of Fit Measures for GFCy in Asset Prices by FX Regime .............................................................. 17
- Figure 6. Goodness of Fit Measures for GFCy in Capital Flows by FX Regime ............................................................ 18
- Figure 7. Goodness of Fit Measures for GFCy in Asset Prices by FX Regime & Captial Account ................................ 19
- Figure 8. Goodness of Fit Measure for GFCy in Captial Flows by FX Regime & Captial Account ................................ 19
- Figure 9. Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit (Quarterly) ............ 20

*Source: wpiea2024158-print-pdf - 1. Introduction and Motivation (PDF chapter). Canonical URL: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024158-print-pdf.pdf*

### 10. Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit (Monthly) ............ 21

### 10. Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit (Monthly)

### Introduction and motivation
- Research context:
  - The Global Financial Cycle (GFCy) literature expanded after Forbes and Warnock (2012), Rey (2013), and Bruno and Shin (2015); recent reviews include Miranda-Agrippino and Rey (2022).
  - Policy relevance spans capital-receiving and capital-sending countries and tools such as macroprudential policies, capital flow management tools, quantitative easing/tightening, and market interventions.
- Prior findings and motivation for current analysis:
  - Cerutti et al. (2019b) measured GFCy relevance for capital flows over 1990Q1-2015Q4 and found it rarely explained more than 25 percent of variation in capital flows.
  - The GFCy may differ in importance between capital flows and local financial conditions (asset prices and credit); theory suggests comovement in prices need not mirror comovement in quantities.
  - Policy tools are often quantity-based but aim to influence asset prices; understanding differences in GFCy importance across quantities and prices matters for tool choice and calibration.
- Scope of current study:
  - Focus on comovements in short-term and long-term interest rates, sovereign spreads, equity prices (local currency and USD), house prices, and local (real) credit, subject to data availability for the period 2000Q1-2021Q4.
  - Empirical approach is model-free and empirical-based, capturing GFCy via common factors in series and conventional center-country (US) variables.
- High-level overall finding summary (period 2000Q1-2021Q4):
  - A combination of each series’ common factor and conventional US GFCy-variables typically explains:
    - more than 75% of the variation for policy rates, short-term interest rates and government spreads,
    - about 60% for house prices,
    - up to 40% for stock market returns (measured either in US dollars or local currency),
    - about 30% of domestic credit variation,
    - up to just 25% of the variations in capital flows (dis-aggregated by four types and by direction).
- Contribution to literature:
  - Explains discrepancy between high comovement found in papers using prices and lower comovement in papers using capital flows.
  - Uses a more general proxying strategy for the GFCy (common factors and key push drivers) and reports robustness across factor and country samples.
  - Aligns with findings that common factors explain much more variance in asset prices than in credit or quantities (examples cited: Monnet and Puy (2019); Fernandez and Vicondoa (2024)).

### Literature review (asset prices and credit comovements)
- Asset prices:
  - Interest rate arbitrage and exchange rate regime issues imply partial comovement of interest rates; debate on whether monetary autonomy is constrained by capital mobility (Rey (2013) vs. critiques).
  - International asset-pricing (ICAPM) literature has generally found limited guidance from ICAPM models for cross-country asset prices.
  - Empirical approaches shifted toward documenting global factor importance (e.g., principal components, correlations); findings of increased comovement over decades (Bekaert et al. (2016)).
  - Examples: Longstaff et al. (2011) find one principal component accounts for 64 percent of sovereign CDS variation; Adrian et al. (2016) link common movements to the VIX.
- Credit:
  - Literature on commonality in credit is smaller and more eclectic; identification of financial cycles often pragmatic (e.g., HP and band-pass filters).
  - Gross external credit flows, rather than net flows or current account positions, identified as important for disruptive credit booms (Borio et al. (2011); Avdjiev et al. (2020)).
  - Some studies link global credit shocks to macro fluctuations but less often model local credit directly (Helbling et al. (2011); Eickmeier and Ng (2015)).
  - Aldasoro et al. (2023) find both domestic financial cycles and GFCy in capital flows and asset prices rise before crises; their correlation between capital flows and asset price GFCy is 97 percent.
- Relationship between prices and quantities:
  - Classic models allow price changes without quantity changes; real-world frictions imply prices and quantities often correlate.
  - Empirical examples show global factors can affect local borrowing rates and credit supply (Baskaya et al. (2022)); FX interventions can affect domestic credit through borrowers’ creditworthiness (Hofmann et al. (2021)).
  - Recent studies document portfolio and exchange rate amplification effects in EMEs when foreign investors adjust positions (Hofmann et al. (2020); Bruno et al. (2022)).

### Empirical strategy, data, and country sample
- Definition and objective:
  - GFCy defined as high commonality in financial conditions captured by common factors in each series and observable US (“center-country”) determinants.
  - Goal: quantify proportion of variations in asset prices, domestic credit, and capital flows explained by the GFCy for small countries (center-country variables treated as plausibly exogenous).
- Country sample and period:
  - Sample comprises 76 AEs and EMEs (listed in Table1 of the source) for period 2000Q1-2021Q4.
  - Largest AEs (US, euro area, Japan, UK) are excluded from regressions to avoid confounding cause and effect.
- Variables analyzed:
  - Policy rates and short-term rates (I(0)); government spreads constructed as local currency sovereign 10-year rate minus dollar 10-year rate to address trending in long-term rates.
  - MSCI growth rate (local currency and USD), MSCI end-of-period returns, residential real estate price growth rate, domestic credit flow as percentage of GDP, real domestic credit growth.
  - Capital flows by type and direction (as percentage of GDP).
- Common factors:
  - Dynamic common factors estimated separately for AEs and EMEs (one lag); factor patterns described for policy rates, spreads, stock returns, house prices, and credit.
  - AE factor countries used to generate AE factors: Australia, Canada, Iceland, New Zealand, Norway, and Sweden.
  - EME factor countries used to generate EME factors: Brazil, Chile, China, Indonesia, Korea, Mexico, Philippines, Russia, South Africa, Thailand, and Türkiye.
- US GFCy-drivers (USFUND j t), eight fundamentals (end-of-period, quarter):
  - (i) the VIX;
  - (ii) the change in the dollar real effective exchange rate (REER, quarter over quarter percentage change in IMF’s CPI-based REER);
  - (iii) the nominal policy interest rate (the Federal Funds rate);
  - (iii) the ex post real policy interest rate (nominal rate minus ex post year over year realized CPI inflation rate);
  - (iv) the TED spread (three-month LIBOR minus the US Treasury bill rate);
  - (v) the yield curve slope (US 10-year rate minus the three-month government rate);
  - (vi) US real GDP growth (from the IMF WEO);
  - (viii) US M2 growth (year over year growth).
- Country-level regression framework:
  - Estimated equation for country i, quarter t:
    - VAR_{i,t} = sum_k γ_{k,i} FAC_{k,t} + sum_j β_{j,i} USFUND_{j,t} + φ_i + ε_{c,t}
  - Where:
    - VAR_{i,t} is the country-quarter series under analysis (policy rates, short-term rates, government spread, MSCI growth rates, residential real estate growth rate, domestic credit flow/GDP, real credit growth, and capital flows by type/direction).
    - FAC_{k,t} denotes the dynamic common factor for AEs or EMEs (indexed by k).
    - USFUND_{j,t} denotes one of the eight US GFCy-drivers (indexed by j).
    - φ_i captures country-specific constant.
- Estimation details:
  - Up to 18 individual country time-series equations per country (8 asset prices, 8 capital flow types, 2 credit series).
  - Focus on adjusted R^2 as the metric for quantitative importance of GFCy (penalizes over-parameterization).
  - The distribution of adjusted R^2s is presented across the 76 small AEs and EMEs.

### Baseline empirical findings (summary highlighted in the source)
- Relative explanatory power of GFCy (period 2000Q1-2021Q4, combination of series’ common factor and US GFCy-variables):
  - more than 75% of the variation explained for policy rates, short-term interest rates, and government spreads;
  - about 60% of the variation explained for house prices;
  - up to 40% of the variation explained for stock market returns (USD or local currency);
  - about 30% of the variation explained for domestic credit;
  - up to just 25% of the variation explained for capital flows (dis-aggregated by four types and by direction).
- Comparative literature figures cited:
  - Cerutti et al. (2019b) reported a similar 25 percent adjusted R^2 for capital flows for 1990Q1-2015Q4.
  - Miranda-Agrippino and Rey (2020) estimate first factors that explain about 24.1 percent (prices) and 20.7 percent (quantities) of total variance (frequency and sample differences noted).
  - Monnet and Puy (2019) show GFCy is much weaker for credit than asset prices in long-run quarterly data.
  - Fernandez and Vicondoa (2024) find the common factor explains 64 percent of government bond spreads fluctuations but only 16 percent of variability in net capital flows.
  - Aldasoro et al. (2023) report a 97 percent correlation between their capital-flow GFCy factor and asset-price GFCy factor.
- Policy implication emphasized:
  - The greater importance of the GFCy for prices than for quantities implies that quantity-based policy tools (macroprudential, capital flow management, microprudential) often used to regulate financial systems affect local asset prices but may not directly or fully control them; understanding differences in GFCy importance across prices and quantities is important for policy design, calibration, and assessment.
- Methodological notes and robustness:
  - Quarterly frequency chosen as most appropriate for capital flows comparisons; monthly estimations for several asset prices and inclusion of more factors (including MAR’s global price factor) in robustness checks yield similar results to quarterly estimates.

### Next sections in the source (outline)
- The source indicates Section 4 presents empirical results and robustness tests and Section 5 concludes with policy discussion and future research agenda. The provided content ends at the start of Section 4.

*Source: IMF working paper chapter: "10. Goodness of Fit Measures for GFCy (including MAR Factor) in Asset Prices and Credit (Monthly)" (content from the supplied PDF chapter).*

### 4.1  Baseline Results

### 4.1  Baseline Results

### Methodology and Goodness-of-Fit Measure
- Comovement measured using box-plots of (adjusted) R2s (AR2s) across three specifications:
  - using the first dynamic factor loadings for both AEs and EMEs and eight core country (US) variables ("AllFactors").
  - using only the first factor loadings ("AllFactors only").
  - using only the core country variables ("U.S. variables only").
- Box-plot construction details:
  - Each box runs from the 25th to the 75th percentiles of individual country results; median marked by a vertical bar.
  - Whiskers extend to the most extreme values within 150 percent of the interquartile range of the nearest quartile.
  - Outliers are individually marked.
- Sample and model specification noted on figures: "8 contemporary US vars, Adv + EM dynamic factors, intercept; small countries, 2000Q1-2021Q4".
- Results robust to using the raw, non-adjusted R2s.

### Asset Prices and Credit: Ranking of Comovement with GFCy
- Policy and short-term interest rates:
  - Most closely correlated with the GFCy.
  - Both have a median AR2 above 75% for regressions using all variables.
  - AR2s are somewhat lower when using only common factors or only US variables, but median AR2s remain above 50% in those specifications.
- Government spread:
  - Displays a median AR2 of about 75% when using all variables.
  - Median AR2s are lower when using either common factors only or US variables only.
- House prices:
  - Display median AR2s around 60% when using all three sets of variables.
  - Using common factors only yields median AR2s of some 20%.
  - Using US variables only yields median AR2s of some 40%.
  - Interpretation: house prices are considerably driven by global factors with an important role for US conditions, despite real estate being only partially owned by non-residents and immovable.
- Stock market returns:
  - Rates of return measured in USD have the highest median AR2s among equity measures, about 40% for both USD series.
  - Results with US variables only are very similar to those with all variables, while common factors only produce much lower median AR2s, often less than 10%.
  - Local currency returns display lower median AR2s than USD-based returns, reflecting the important role of the US dollar in global finance and for equity prices.
- Credit variables:
  - Credit flow to GDP and real credit growth display the lowest comovements.
  - Median AR2s are slightly below 30% when using both two sets of variables.
  - Common factors only show somewhat lower median AR2s than the two-set specifications, while US variables only yield very similar AR2s, confirming the importance of the US in global finance.

### Capital Flows: Lower Comovement than Asset Prices
- Eight capital flow types analyzed: FDI in- and outflows, other investment in- and outflows, portfolio debt in- and outflows, and portfolio equity in- and outflows.
- Results for the more recent and slightly enlarged sample are very similar to earlier work covering 1990Q1-2015Q4.
- Observation:
  - Rarely do the 75 percentiles of the AR2s (the right side of the boxes) go above the 25% mark.
  - Capital flows "quantities" comove much less than asset prices with the GFCy and somewhat less than local credit comoves.

*Source: 4.1  Baseline Results (wpiea2024158-print-pdf)*

### 4.2  Robustness tests

### 4.2 Robustness tests

### Overview of robustness strategy
- Conducted multiple robustness exercises to assess sensitivity of results on the importance of the global financial cycle (GFCy) for local financial conditions:
  - Analyzed distribution of the AR2s across countries to detect country characteristics driving results.
  - Tested robustness to the specific GFCy factors by adding the Miranda-Agrippino and Rey (2022) (MAR) global factor, at both quarterly and monthly frequencies.
  - Considered different time-periods and sample frequencies.

### Breaking down results by FX regime and capital account openness
- Sample splits:
  - Exchange rate regime: pegged versus flexible (based on Ilzetzki et al. (2019)’s exchange rate classification).
  - Capital account openness: below or above median openness (2015 Chinn and Ito (2008) Index).
  - Baseline sample: small countries, 8 contemporary US variables, Adv + EM dynamic factors, intercept; 2000Q1-2021Q4.
- Main findings:
  - Pegged and flexible exchange rate regimes show very similar degrees of comovements in asset price and credit measures; only small differences in AR2s.
  - Medians for AR2s of stock market rates of return (measured in US dollars or local currency) are marginally higher for flexible exchange rate regime countries.
  - Results contrast with some literature (e.g., Obstfeld et al. (2019)) that predicts magnified transmission under fixed exchange rates.
  - For capital flows across FX regimes, no noticeable differences in median AR2s; some flows have slightly lower medians under flexible regimes and others higher.
  - Combining FX regime with capital account openness:
    - Asset prices and credit: some differences consistent with traditional models (e.g., credit flows and growth have a higher median AR2 under open capital account + pegged exchange rate), but differences are small and not consistently meaningful.
    - Example counterintuitive case: stock market returns have a lower median AR2 under open capital account + pegged exchange rate than under open capital account + flexible exchange rate.
    - Capital flows: a few median AR2s are lowest for economies with pegged exchange rate + closed capital account, and highest for flexible exchange rate + open capital account.
- Interpretation:
  - Choice of exchange rate regime may matter less for countries’ capital flow exposure to the GFCy than traditional literature suggests.
  - A flexible exchange rate could, in some cases, accentuate the presence of a GFCy (noted in “original sin redux” literature).

### Adding MAR common factor — Quarterly frequency
- Implementation details:
  - Added Miranda-Agrippino and Rey (2022) global factor to regressions; MAR variable used as three month average in quarterly regressions.
  - Quarterly sample shortened because MAR series ends earlier: 2000Q1-2019Q2.
  - Specification: 8 contemporary US vars, Adv+EM dynamic factors, MAR factor, intercept.
- Findings:
  - Little material change from baseline results:
    - Regressions using all three factors show only very marginal increases in median AR2s and slightly lower dispersion in AR2s (relative to prior Figures).
    - MAR variable itself shows only low AR2s for asset prices and credit variables.
  - Using end-of-period MAR rather than three month average produces no major differences.
  - Regressions for capital flows with quarterly MAR show few differences in AR2s and ranking of individual capital flows compared to prior results.
  - Conclusion: Adding MAR makes little change overall; the earlier conclusion that AR2s in capital flows are generally low remains.
- Notable data issue:
  - Some AR2s are negative (example: FDI) when the specific data span is short and many regressors are included (in total 11: 8 US variables + 2 factors + MAR).

### Adding MAR common factor — Monthly frequency
- Implementation details:
  - MAR factor available monthly; monthly regressions run on smaller sample of variables (house prices drop out due to lack of monthly frequency).
  - Monthly sample: 8 contemporary US vars, Adv+EM dynamic factors, MAR factor, intercept; 2000M1-2019M4.
- Findings:
  - Very little change relative to quarterly results:
    - Expected drop in overall AR2s when using higher-frequency data, but rankings of local financial variables preserved.
    - Ranking from highest to lowest explanatory power: interest rates (highest), returns on equity prices in dollars, credit flows and equity returns in local currency (lowest).
    - Using monthly data for single series yields similar rankings to quarterly, though AR2s tend to drop (notably when using US variables only).
  - Overall conclusion: base results are robust to using monthly frequency and to including the MAR factor.

### Key quantitative statements and comparative magnitudes (from broader study context)
- Importance of GFCy in capital flows: generally low (less than 25%).
- Importance of GFCy in local financial variables:
  - Interest rates, house prices, and credit show much higher explanatory power.
  - Quantitative importance: highest for interest rates, second for house prices, third for credit, with explanatory powers between 50% and 75%.
- Implication: GFCy explains a large share of movements in interest rates, house prices, and credit, but a much smaller share of capital flows (including portfolio flows).

### Analysis implications for policy design and research
- Robustness exercises indicate that the core finding — stronger GFCy role in asset prices and credit than in capital flows — is stable across:
  - FX regime splits (pegged vs flexible),
  - Capital account openness splits (closed vs open),
  - Addition of the Miranda-Agrippino and Rey (2022) factor (quarterly and monthly),
  - Different sample frequencies and slightly shorter samples.
- Policy-relevant inferences:
  - Knowing the magnitude and timing (lead-lag) of GFCy effects across quantities and prices is important for selecting and timing tools (FXI, APPs, MaPPs, CFMs).
  - Some quantity-based tools (e.g., FXI, APPs) can affect prices quickly; others (e.g., MaPPs like limits on LTVs) affect prices more slowly.
  - Tools that can be adjusted at high frequency (e.g., FXI) are better suited to address high-frequency shocks with rapid asset price changes; lower-frequency tools better address slower-moving cycles and vulnerabilities.
- Research agenda priorities highlighted by robustness results:
  - Better understanding of what drives the GFCy and why advanced economies play a large role.
  - Role of the US dollar and AEs’ monetary policy in driving the GFCy.
  - Whether source-country macroprudential, regulation, and supervision amplify or reduce the GFCy.
  - Exact transmission channels: common balance sheets/exposures vs contagion; drivers of risk-on vs risk-off behavior; differences between local exchange rate vs broad dollar effects; roles of specific capital flow types.
  - Impacts of domestic policy responses: do asset prices and local financial conditions adjust more (or less) when quantity tools target capital flows vs local financial variables?

*Source: IMF Working Paper — Section 4.2 Robustness tests (figures and text as supplied)*

### References

### References

### Major thematic coverage in the reference list
- Capital flows, global liquidity, and the Global Financial Cycle (GFC)
- Macroprudential policy design, tools, and effects
- Integrated monetary and financial policy frameworks, including the Integrated Policy Framework (IPF)
- International spillovers, credit cycles, and synchronization of business, credit, and housing cycles
- Measurement and decomposition approaches for financial openness, asset pricing, and global linkages
- Empirical methods used across studies: GVAR, factor models, adjusted R2 goodness-of-fit assessments, credit-to-GDP gap comparisons, and high-frequency microdata analyses

### Representative authors, institutions, and outlets cited
- Authors and contributors include Adrian; Gopinath; Pazarbasioglu; Borio; Disyatat; Cerutti; Claessens; Rey; Obstfeld; Ostry; Rey; Miranda-Agrippino; Gopinath; and many others as listed in the source.
- Institutional sources include IMF working papers and policy papers, BIS reports and working papers, World Bank Policy Research Working Papers, NBER Working Papers, ESRB Working Papers, Federal Reserve Bank of New York Staff Reports, and peer-reviewed journals (Journal of International Economics, Journal of Finance, Journal of Monetary Economics, Review of Economic Studies, IMF Economic Review, Journal of International Money and Finance, American Economic Journal: Macroeconomics, Journal of Housing Economics, Economic Policy, and others).

### Key numeric and bibliographic markers preserved from the source
- Working Paper No. WP/2024/158
- Appendix Table 1: List of Small Advanced and Emerging Market Economies — sample size: 76 economies
- Notation in Table 1: 1/ used to generate advanced economy factors; and 2/ used to generate emerging market factors
- Figure captions and associated sample/time coverage and labels as presented in the source:
  - Figure A.1: Goodness of fit measures for GFCy (including MAR factor) in asset prices and credit
    - Note line: "8 contemporary US vars, Adv+EM dynamic factors, MAR factor, intercept; small countries, 2000Q1-2019Q2"
    - Note label: "Box-Plots of Adjusted R2s (25 Countries with all R2s available)"
  - Figure A.2: Goodness of fit measures for GFCy in capital flows
    - Note line: "8 contemporary US vars, Adv + EM dynamic factors, intercept; small countries, 2000Q1-2021Q4"
    - Note label: "Box-Plots of Adjusted R2s (29 Countries with all R2s available)"
  - Figure A.3: Goodness of fit measures for GFCy (including MAR factor at eop) in asset prices and credit
    - Note line: "8 contemporary US vars, Adv+EM dynamic factors, MAR factor (Last), intercept; small countries, 2000Q1-2019Q2"
    - Note label: "Box-Plots of Adjusted R2s (Quarterly)"
  - Figure A.4: Goodness of fit measures for GFCy (including MAR factor) in capital flows
    - Note line: "8 contemporary US vars, Adv+EM dynamic factors, MAR factor, intercept; small countries, 2000Q1-2019Q2"
    - Note label: "Box-Plots of Adjusted R2s"

### Appendix contents highlighted
- Appendix A contains:
  - Table 1: full listing of the 76 small advanced and emerging market economies in the sample (with 1/ and 2/ annotations for factor groupings)
  - Figures A.1–A.4 presenting box-plots of adjusted R2s for GFCy explanatory power across asset prices, credit, and various capital flow categories, under model specifications that include:
    - "AllFactors only"
    - "U.S. variables only"
    - "MAR only"
  - Repeated variable and series labels across figures include: Real Credit Growth; Credit Flow / GDP; Real House Price Growth; MSCI Return USD EOP; MSCI Return LCY EOP; MSCI Return USD; MSCI Return LCY; Govt Spread; Short-term Rate; Policy Rate; Portfolio Equity Outflow; Portfolio Equity Inflow; Portfolio Debt Outflow; Portfolio Debt Inflow; Other Investment Outflow; Other Investment Inflow; FDI Outflow; FDI Inflow.

*References and appendix material as presented in wpiea2024158-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024158-print-pdf.pdf_