## wpiea2025216-source-pdf

## Source details

**Canonical URL:** [wpiea2025216-source-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025216-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025216-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025216-source-pdf.pdf.json)

---

### Introduction and overview
- Purpose: Integrates two strands of research—interaction of business cycles with financial conditions, and macroeconomic consequences of crowding out—to analyze shock transmission in emerging market economies.
- Sample:
  - 18 emerging market economies that collectively account for over a third of global GDP in nominal terms, equivalent to 46 percent in purchasing power parity (PPP) terms.
  - Sample period: 2000 through 2019:Q4.
- Main novelty: A multivariate state-space (MARSS) representation (inspired by predator–prey models) uncovers two opposing short-term forces between domestic financial conditions and growth: a growth-enhancing effect and a growth-inhibiting effect.

### Key quantitative regularities (aggregate EM-level)
- Near-term growth-enhancing effect:
  - A one standard deviation (sd) easing in emerging markets’ financial conditions (EMFCI) is followed by a 0.46sd increase in their annual median GDP growth (EM-GDP) the next quarter, equivalent to 82 bps.
- Growth-inhibiting (feedback) effect:
  - A one-sd increase in emerging markets’ median GDP growth is followed by a 0.2sd tightening in their financial conditions in the next quarter.
- Amplification by global financial cycle:
  - Incorporating global financial conditions into the state-space model raises the GDP→FCI response: a one sd increase in EM-GDP results in a 0.27sd tightening in EMFCI in the next quarter.
- Augmented MARSS (with global financial cycle covariate):
  - GDP→FCI linkage increases by over 37 percent in median-growth specification (example: from 0.35 to 0.48).

### Methodology (model structure and estimation)
- Conceptual approach:
  - Capture bidirectional feedback between financial conditions (f) and output growth (g) and the role of global financial conditions as an exogenous driver via a predator–prey analogy.
- Econometric specification:
  - State-space (MARSS) with latent state vector x_t = (F_t, G_t)′ and observables y_t = (EMFCI, EM-GDP) treated as noisy one-for-one readings of states (Z = I_2).
  - Process: x_t = B x_{t−1} + C c_t + w_t ; observation: y_t = Z x_t + a + v_t.
  - Interpretation of B: diagonal elements (b_ff, b_gg) capture self-interaction/density-dependence; off-diagonals (b_fg, b_gf) capture cross-effects (EMFCI→EMGDP and EMGDP→EMFCI).
  - Covariate C captures contemporaneous influence of inverted global financial conditions (higher values denote tighter global conditions).
- Estimation: Maximum likelihood with Kalman filter/smoother to extract latent states and estimate R, Q.
- Advantages highlighted:
  - Accommodates bi-directional feedback between latent financial conditions and GDP growth.
  - Filters measurement noise via the Kalman filter/smoother to reduce attenuation bias.
  - Incorporates an exogenous global financial cycle covariate.

### Data and Financial Conditions Index (FCI) construction
- Countries: Argentina, Brazil, Chile, China, Colombia, the Czech Republic, India, Indonesia, Korea, Malaysia, Mexico, Philippines, Poland, Russia, Slovakia, South Africa, Thailand, and Türkiye.
- Time coverage for MARSS: 2000:Q1–2019:Q4.
- Output measure: year-on-year growth rate of real GDP (percent).
- Global financial conditions: Miranda-Agrippino and Rey index inverted so higher values denote tighter global conditions.
- Country-level FCI construction:
  - Dynamic factor model (DFM) combining fast-moving stress indicators (e.g., EqReturn, EqVol30, OptionVol3m, term/corporate/interbank spreads, EMPI) and slow-moving vulnerability indicators (e.g., Credit, SriskUT, Debt-to-GDP, Primary Balance, real-estate prices, Industrial Production).
  - Sectoral blocks: banking, fiscal, real sector, external trade and finance.
  - Aggregate EMFCI: median across country-level FCIs each quarter; alternative EMFCI3 excludes real-activity variables for robustness.
- Data sources: Bloomberg L.P., NYU V-Lab, Bank for International Settlements (BIS), IMF-IFS.

### Main empirical findings and robustness
- Baseline MARSS (no global covariate):
  - EMFCI→EMGDP: one-sd easing → 0.46sd increase in EM-GDP next quarter (82 bps).
  - EMGDP→EMFCI: one-sd increase → 0.2sd tightening in EMFCI next quarter.
- Augmented MARSS (with global financial cycle covariate):
  - GDP→FCI linkage is amplified: example coefficient linking GDP growth to EMFCI increases by over 37 percent (from 0.35 to 0.48 in a median-growth specification).
  - Model diagnostics: improved fit (lower corrected AIC, higher log-likelihood).
- Robustness checks:
  - Results hold using alternative EMFCI definitions (EMFCI1 vs EMFCI3) and alternative aggregate EM GDP measures (median, simple average, weighted averages with WB/IMF weights), though precision varies by weighting.
  - Quantile analysis: both b_gf and b_fg are stable in the bottom half of growth distribution; both attenuate toward the right-tail (example: b_fg from −0.41 at Q50 to −0.12 at Q95, remaining statistically significant).
- Demand decomposition (Table 5 highlights):
  - Growth-enhancing effect transmits primarily through private consumption and investment.
  - Government spending is largely decoupled in the short run (government Bfg 0.13 (ns), Bgf 0.16 (ns) in reported table).

### Country-level and subgroup findings
- Country-level MARSS results (2000–2019):
  - In 15 of 18 countries, an easing of domestic financial conditions is associated with higher annual GDP growth next quarter (statistically significant in 11 countries).
  - In 14 countries, higher real GDP growth is associated with tightening financial conditions next quarter (statistically significant in 8 countries).
  - Global financial conditions generally behave as expected: tighter global conditions → tighter domestic conditions and lower near-term GDP growth in most countries.
- Twin-balance-sheets (sample-split) analysis and subgroup medians (Table 8 highlights):
  - Classification: countries grouped using long-term benchmarks of debt-to-GDP and current account balance-to-GDP into HD-CAS, LD-CAS, HD-CAD, and LD-CAD.
  - Findings:
    - HD–CAD (twin-deficit) countries exhibit coexistence of growth-enhancing and growth-inhibiting effects; growth reductions after a tightening last about four quarters for this group.
    - HD–CAS countries also show coexistence of both effects.
    - LD–CAS countries (ample fiscal and external space) are the only group where neither effect is statistically significant (LD–CAS Bfg 0.02 (ns), Bgf 0.12 (ns)).
  - Quantitative amplification:
    - Median-GDP regressions controlling for public debt/current account increase absolute vulnerability of growth to financial tightening (EMFCI3 coefficient from −0.17 to −0.25 in select models).
  - Representative subgroup coefficients reported:
    - HD Bfg −0.47***, Bgf 0.30***.
    - LD Bfg −0.27***, Bgf 0.22***.
    - CAD Bfg −0.37***, Bgf 0.31***.
    - CAS Bfg −0.44***, Bgf 0.30***.
    - HD–CAS Bfg −0.44***, Bgf 0.22***.
    - LD–CAD Bfg −0.24***, Bgf 0.26***.
    - HD–CAD Bfg −0.31***, Bgf 0.33***.
    - LD–CAS Bfg 0.02 (ns), Bgf 0.12 (ns).

### Role of global factors and amplification
- Inclusion of inverted global financial conditions as a covariate:
  - Improves model fit and raises the estimated strength/persistence of the growth-inhibiting feedback (GDP→EMFCI).
  - Example: reported b_gf rises from 0.35 to 0.48 (median GDP Growth specification) when global financial cycle is included.
  - Global financial cycle mainly amplifies the growth-inhibiting link rather than the growth-enhancing link.

### Model comparison, simulations, and diagnostics
- VAR/PVAR vs state-space:
  - VAR/PVAR models estimated on observables often find a significant EMFCI→GDP response but a statistically insignificant GDP→EMFCI response.
  - Attenuation explanation: measurement error in noisy quarterly aggregates mutes lag coefficients in observable-based VARs; state-space approach estimates measurement noise and filters it out.
- Simulation exercises on interaction matrix B:
  - Fully unconstrained B (allowing bi-directional feedback) provides the best model fit.
  - Constraining B (to unequal diagonal, equal diagonal, or identity/no interaction) progressively worsens model fit.
  - Treating EMGDP as exogenous (removing it from feedback) makes GDP→EMFCI coefficient statistically insignificant, reproducing VAR-like results.
- Example simulation coefficients and diagnostics (Appendix/Table D examples):
  - B11 0.82*** (0.12), B21 −0.15** (0.09), B12 0.32** (0.09), B22 0.77*** (0.07); C11 −0.18 (0.13), C21 0.06 (0.09); AICc 252.38; Log-likelihood −113.13.

### Policy-relevant implications and recommendations
- Vulnerability management:
  - Countries with high public debt and large current account deficits face faster re-tightening of financial conditions following growth spurts; strengthening fiscal and external positions can lengthen and amplify beneficial effects of easier financial conditions.
- Early-warning and monitoring:
  - Composite FCIs combining fast-moving stress indicators and slow-moving vulnerability indicators can provide timely signals of macro-financial vulnerability and the likely persistence of growth impulses.
- Macroprudential design and buffers:
  - Tightening financial conditions “bites hardest in the lower half of the growth distribution (Table 4),” making the payoff to countercyclical macroprudential buffers state-dependent: buffers built in easing phases should be available for release when growth is weak and conditions tighten.
- Modeling and forecasting:
  - State-space approaches that extract latent macro-financial states and include global covariates can improve near-term predictive content and early-warning utility compared with observables-based VARs.
- Suggested next methodological steps:
  - Structural or semi-structural designs that jointly identify fiscal and private-credit transmission—allowing multipliers to vary over the cycle—and that nest balance-sheet linkages and twin balance sheets heterogeneity.

### Conclusions (summary)
- The MARSS framework identifies two distinct drivers in emerging markets:
  - A growth-enhancing effect: easier financial conditions are associated with stronger near-term growth.
  - A growth-inhibiting effect: stronger growth is associated with tighter near-term financial conditions.
- Heterogeneity:
  - Emerging market economies with high public debt and current account deficits display greater macro-financial feedback more consistently and significantly.
  - Countries with low public debt and strong external balances primarily display a standard growth-enhancing effect.
- Methodological note:
  - Estimating dynamics in a state-space framework helps counter measurement-error attenuation in observables-based VAR/PVAR models by filtering noisy aggregates via the Kalman filter.
  - The estimates remain reduced form: the feedback coefficients capture predictive linkages, not identified fiscal or credit multipliers.
- Recommended priorities:
  - Jointly track domestic FCIs and global financial conditions indicators, while addressing public-debt and external-balance risks, to stabilize growth and reduce adverse tail outcomes.

*Source: wpiea2025216-source-pdf - Introduction (IMF Working Paper).*

### Introduction ...........................................................................................................

### wpiea2025216-source-pdf - Introduction ...........................................................................................................

### Introduction
- Located on page 4.

### Methodology
- Located on page 8.
- Subsections:
  - Conceptual Framework (page 8)
  - Econometric Model (page 10)
  - Data (page 13)

### Data and Sample
- Located on page 13.
- Financial Conditions Index (page 13)
  - Concept (page 13)
  - Input Variables (page 14)
  - Index Construction (page 15)

### Results and Discussion
- Located on page 16.
- Regional Analysis (page 16)
- Country-level Analysis (page 18)
- Balance Sheet Conditions and Finance-Growth Interdependence (page 19)
- Role of Global Factors (page 20)
- Model Simulation Exercise (page 21)

### Conclusions
- Located on page 23.

### References
- Located on page 24.

### Tables and Figures (located starting page 29)
- Figures
  - 1. Macro-financial Associations in Emerging Markets - Analytical Framework (page 29)
  - 2. Financial Conditions Index: Framework and Estimation (page 29)
  - 3. Financial Conditions Index (FCI) for Emerging Economies (page 30)
  - 4. Fiscal and Current Account Balance (page 31)
- Tables
  - 1. Financial Conditions Index: Data and Variable Construction (page 32)
  - 2 (a). Results: Estimated Model Coefficients with and without Covariates (page 33)
  - 2 (b). Results: Estimated Model Coefficients with Covariates (page 33)
  - 3 (a). Results: Estimated Model Coefficients with Different EM Growth Measures (page 34)
  - 3 (b). Results: Estimated Model Coefficients with Different EM Growth Measures (page 34)
  - 4 (a). Results: Model Coefficients for Different EM Growth Quantiles (page 35)
  - 4 (b). Results: Model Coefficients for Different EM Growth Quantiles (page 35)
  - 5. Results: Finance-Growth Feedback by Demand Component (page 36)
  - 6. Summary Results: Sign Test (page 36)
  - 7. Median Quantile Regressions: Financial Conditions, Growth, and Macro-imbalances (page 36)
  - 8. Median Finance-Growth Feedback by Debt and External Balance Subgroups (page 37)

### Appendix
- Located starting page 38.

*Source: wpiea2025216-source-pdf - Introduction (IMF Working Paper).*

### Introduction

### Introduction

### Overview
- Purpose: Integrates two strands of research—interaction of business cycles with financial conditions, and macroeconomic consequences of crowding out—to analyze shock transmission in emerging market economies.
- Sample: 18 emerging market economies that collectively account for over a third of global GDP in nominal terms, equivalent to 46 percent in purchasing power parity (PPP) terms.
- Sample period: 2000 through 2019:Q4.
- Main novelty: A multivariate state-space (MARSS) representation (inspired by predator–prey models) uncovers two opposing short-term forces between domestic financial conditions and growth: a growth-enhancing effect and a growth-inhibiting effect.

### High-level empirical regularities
- Near-term growth-enhancing effect:
  - A one standard deviation (sd) easing in emerging markets’ financial conditions (EMFCI) is followed by a 0.46sd increase in their annual median GDP growth (EM-GDP) the next quarter, equivalent to 82 bps.
- Growth-inhibiting (feedback) effect:
  - A one-sd increase in emerging markets’ median GDP growth is followed by a 0.2sd tightening in their financial conditions in the next quarter.
- Amplification by global financial cycle:
  - Incorporating global financial conditions into the state-space model raises the GDP→FCI response: a one sd increase in EM-GDP results in a 0.27sd tightening in EMFCI in the next quarter.

### Key contributions
- Empirical: Documents coexistence of growth-enhancing and rapid growth-inhibiting linkages in emerging market economies, especially prominent in countries with weak fiscal and external balance-sheets.
- Methodological: Proposes and applies a reduced-form multivariate state-space model (MARSS) that:
  - Accommodates bi-directional feedback between latent financial conditions and GDP growth,
  - Filters measurement noise via the Kalman filter/smoother to reduce attenuation bias,
  - Incorporates an exogenous global financial cycle covariate.

### Core interpretation
- Mechanism: In many emerging markets, shallower domestic financial markets and twin deficits (high public debt and/or large current account deficits) reduce the elasticity of domestic credit supply. As growth rises, competing demands for scarce funding (private credit vs. government and external refinancing) can cause rapid tightening of financial conditions that shortens and limits growth spurts.
- Channels: The growth-enhancing effect operates mainly through private consumption and investment; government spending is largely decoupled in the short run.

---

### Methodology (model structure and estimation)
- Conceptual approach:
  - Capture bidirectional feedback between financial conditions (f) and output growth (g) and the role of global financial conditions as an exogenous driver.
  - Analogy: predator–prey class of models for endogenous macro-financial cycles.
- Econometric specification:
  - State-space (MARSS) with latent state vector x_t = (F_t, G_t)′ and observables y_t = (EMFCI, EM-GDP) treated as noisy one-for-one readings of states (Z = I_2).
  - Process: x_t = B x_{t−1} + C c_t + w_t ; observation: y_t = Z x_t + a + v_t.
  - Interpretation of B: diagonal elements (b_ff, b_gg) capture self-interaction/density-dependence; off-diagonals (b_fg, b_gf) capture cross-effects (EMFCI→EMGDP and EMGDP→EMFCI).
  - Covariate C captures contemporaneous influence of inverted global financial conditions (higher values denote tighter global conditions).
- Estimation: Maximum likelihood with Kalman filter/smoother to extract latent states and estimate R, Q.

---

### Data and Financial Conditions Index (FCI) construction
- Countries: Argentina, Brazil, Chile, China, Colombia, the Czech Republic, India, Indonesia, Korea, Malaysia, Mexico, Philippines, Poland, Russia, Slovakia, South Africa, Thailand, and Türkiye.
- Time coverage for MARSS: 2000:Q1–2019:Q4.
- Output measure: year-on-year growth rate of real GDP (percent).
- Global financial conditions: Miranda-Agrippino and Rey index inverted so higher values denote tighter global conditions.
- Country-level FCI construction:
  - Dynamic factor model (DFM) combining fast-moving stress indicators (e.g., EqReturn, EqVol30, OptionVol3m, term/corporate/interbank spreads, EMPI) and slow-moving vulnerability indicators (e.g., Credit, SriskUT, Debt-to-GDP, Primary Balance, real-estate prices, Industrial Production).
  - Sectoral blocks: banking, fiscal, real sector, external trade and finance.
  - Aggregate EMFCI: median across country-level FCIs each quarter; alternative EMFCI3 excludes real-activity variables for robustness.
- Data sources: Bloomberg L.P., NYU V-Lab, Bank for International Settlements (BIS), IMF-IFS.

---

### Main empirical findings (aggregate EM-level and robustness)
- Baseline MARSS (no global covariate):
  - EMFCI→EMGDP: one-sd easing → 0.46sd increase in EM-GDP next quarter (82 bps).
  - EMGDP→EMFCI: one-sd increase → 0.2sd tightening in EMFCI next quarter.
- Augmented MARSS (with global financial cycle covariate):
  - GDP→FCI linkage is amplified: coefficient linking GDP growth to EMFCI increases by over 37 percent (example: from 0.35 to 0.48 in median-growth specification).
  - Model diagnostics: improved fit (lower corrected AIC, higher log-likelihood).
- Robustness:
  - Results hold using alternative EMFCI definitions (EMFCI1 vs EMFCI3) and alternative aggregate EM GDP measures (median, simple average, weighted averages with WB/IMF weights), though precision varies by weighting.
  - Quantile analysis: both b_gf and b_fg are stable in the bottom half of growth distribution; both attenuate toward the right-tail (example: b_fg from −0.41 at Q50 to −0.12 at Q95, remaining statistically significant).
- Demand decomposition:
  - Growth-enhancing effect transmits primarily through private consumption and investment; government spending is statistically decoupled (Table 5).

---

### Country-level and subgroup findings
- Country-level MARSS results (2000–2019):
  - In 15 of 18 countries, an easing of domestic financial conditions is associated with higher annual GDP growth next quarter (statistically significant in 11 countries).
  - In 14 countries, higher real GDP growth is associated with tightening financial conditions next quarter (statistically significant in 8 countries).
  - Global financial conditions generally behave as expected: tighter global conditions → tighter domestic conditions and lower near-term GDP growth in most countries.
- Twin-balance-sheets (sample-split) analysis:
  - Classification: countries grouped using long-term benchmarks of debt-to-GDP and current account balance-to-GDP into HD-CAS, LD-CAS, HD-CAD, and LD-CAD.
  - Findings:
    - HD–CAD (twin-deficit) countries exhibit coexistence of growth-enhancing and growth-inhibiting effects; growth reductions after a tightening last about four quarters for this group.
    - HD–CAS countries also show coexistence of both effects.
    - LD–CAS countries (ample fiscal and external space) are the only group where neither effect is statistically significant.
  - Quantitative amplification: median-GDP regressions controlling for public debt/current account increase absolute vulnerability of growth to financial tightening (EMFCI3 coefficient from −0.17 to −0.25 in select models).

---

### Role of global factors and amplification
- Inclusion of inverted global financial conditions as a covariate:
  - Improves model fit and raises the estimated strength/persistence of the growth-inhibiting feedback (GDP→EMFCI).
  - Example magnitudes reported: b_gf rises from 0.35 to 0.48 (median GDP Growth specification) when global financial cycle is included.
  - Global financial cycle mainly amplifies the growth-inhibiting link rather than the growth-enhancing link.

---

### Model comparison and simulation exercises
- VAR/PVAR vs state-space:
  - VAR/PVAR models estimated on observables often find a significant EMFCI→GDP response but a statistically insignificant GDP→EMFCI response.
  - Attenuation explanation: measurement error in noisy quarterly aggregates mutes lag coefficients in observable-based VARs; state-space approach estimates measurement noise and filters it out.
- Simulation exercises on interaction matrix B:
  - Fully unconstrained B (allowing bi-directional feedback) provides the best model fit.
  - Constraining B (to unequal diagonal, equal diagonal, or identity/no interaction) progressively worsens model fit.
  - Treating EMGDP as exogenous (removing it from feedback) makes GDP→EMFCI coefficient statistically insignificant, reproducing VAR-like results and highlighting the importance of modeling endogenous feedback.

---

### Policy-relevant implications (inferred from findings)
- Vulnerability management:
  - Countries with high public debt and large current account deficits face faster re-tightening of financial conditions following growth spurts; hence, strengthening fiscal and external positions can lengthen and amplify beneficial effects of easier financial conditions.
- Early-warning and monitoring:
  - Composite FCIs combining fast-moving stress indicators and slow-moving vulnerability indicators can provide timely signals of macro-financial vulnerability and the likely persistence of growth impulses.
- Modeling and forecasting:
  - State-space approaches that extract latent macro-financial states and include global covariates can improve near-term predictive content and early-warning utility compared with observables-based VARs.

*IMF Working Paper — Introduction and associated methodology, data, and results as provided in the supplied content.*

### Conclusions

### Conclusions

### Framework and core findings
- The paper employs a MARSS model to characterize macro-financial dynamics in emerging market economies.
- The framework identifies two distinct drivers:
  - A growth-enhancing effect: easier financial conditions are associated with stronger near-term growth.
  - A growth-inhibiting effect: stronger growth is associated with tighter near-term financial conditions.
- A sample split shows:
  - Emerging market economies with high public debt and current account deficits display greater macro-financial feedback more consistently and significantly.
  - Countries with low public debt and strong external balances only display a standard growth-enhancing effect previously identified in advanced economies.
- In emerging market economies, the growth-enhancing effect operates mainly through private demand, indicating emphasis on private-sector credit conditions when conditions tighten.

### Role of global financial conditions and financial conditions indicators (FCIs)
- Accounting for global financial conditions is associated with better model fit and larger estimates of macro-financial feedback.
- Incorporating information on both stress and vulnerabilities in domestic FCIs aligns with emerging market-specific cyclical features (e.g., higher consumption volatility and exposure to sudden stops) and yields more informative state estimates.

### Methodology, limitations, and next steps
- Estimating dynamics in a state-space framework helps counter measurement-error attenuation in observables-based VAR/PVAR models by filtering noisy emerging market aggregates via the Kalman filter.
- The estimates remain reduced form: the feedback coefficients capture predictive linkages, not identified fiscal or credit multipliers.
- A natural next step is a structural or semi-structural design that:
  - Jointly identifies fiscal and private-credit transmission—allowing multipliers to vary over the cycle.
  - Nests the balance-sheet linkages and twin balance sheets heterogeneity.

### Policy implications
- Tightening financial conditions “bites hardest in the lower half of the growth distribution (Table 4),” making the payoff to countercyclical macroprudential buffers state-dependent:
  - Buffers built in easing phases should be available for release when growth is weak and conditions tighten.
- As emerging market economies’ financial integration deepens, frameworks that jointly track domestic FCIs and global financial conditions indicators, while addressing public-debt and external-balance risks, will be central to stabilizing growth and reducing tail outcomes.

*Source: wpiea2025216-source-pdf - Conclusions*

### References

### wpiea2025216-source-pdf - References

### Bibliographic themes and emphases
- Core literature on financial conditions, systemic risk, and Growth-at-Risk: Acharya VV; Adrian T; Adrian, Grinberg, Liang, Malik, and Yu; Brownlees C., and Engle R.F.; Surti J., Katagiri M., Lafarguette R., Malik S., and Seneviratne D.
- Macro-financial interaction, credit cycles, sudden stops, and leverage dynamics: Bianchi J.; Mendoza E.G.; Mendoza E.G. (2002, 2010); Korinek A., and Mendoza E.G. (2014); Jeanne O., and Korinek A. (2010); Jordà O., Schularick M., and Taylor A.M. (2013).
- Financial stability, banking risk, and asset prices: Acharya, V. V., and Steffen, S. (2015); Gennaioli, Shleifer, and Vishny (2012); Bruno V., and Shin H.S. (2020); Gorton G., and Ordonez G. (2014); Rajan R.G. (1994); Saunders A., Strock E., and Travlos N.G. (1990).
- Methodological foundations: Dynamic Factor Models and Kalman filtering (Giannone, Reichlin, Small; Doz, Giannone, Reichlin; Forni et al. (2000)); MARSS state-space modeling (interaction matrix estimation); Dumitrescu E.I., and Hurlin C. (2012) panel Granger causality testing.
- Work on fiscal policy, crowding in/out, and public debt interactions: Afonso A., and Sousa R.M. (2012); Furceri D., and Sousa R.M. (2011); Jalles, J. T., Kiendrebeogo, Y., Lam, R., and Piazza, R. (2024); Liaqat Z. (2019); Park J.K., and Meng X. (2024).
- Emerging market–specific analyses and indices: Rey H. (global channels); Miranda-Agrippino S., and Rey H. (2022) “The Global Financial Cycle”; Surti et al. (2016) “The Growing Importance of Financial Spillovers from Emerging Markets”; Central Bank of Chile (2020) Financial Stability Report Chapter IV.

### Tables, figures, and empirical elements documented
- Figures enumerated:
  - Figure 1. Macro-financial Associations in Emerging Markets - Analytical Framework
  - Figure 2. Financial Conditions Index: Framework and Estimation
  - Figure 3. Financial Conditions Index (FCI) for Emerging Economies — period 2000-2021; EMFCI constructed as median of country-wise FCIs; MARSS estimates span 2000Q1 to 2019Q4.
  - Figure 4. Fiscal and Current Account Balance — country 2000–2019 averages (percent of GDP); classification: High Debt (HD) vs. Low Debt (LD) and Current Account Surplus (CAS) vs. Current Account Deficit (CAD).
  - Additional figures: lead-lag correlation between EM GDP and FCI (correlation coefficient −0.79 for one-period lagged FCI); VAR and panel-VAR impulse response analyses (responses of EMGDP to EMFCI shocks and vice versa).
- Tables enumerated and summarized:
  - Table 1. Financial Conditions Index: Data and Variable Construction — variable list includes Equity Return; EqVol30; SriskUT; PLR; OptionVol3m; Term Spread; Corporate Spread; Interbank Spread; Credit; EMPI; Debt-to-GDP; Primary Balance; Real-estate prices; Industrial production. Country codes: ARG, BRA, CHI, CHN, COL, CZE, INDI, INDO, KOR, MAL, MEX, PHI, POL, RUS, SLO, SAF, THA, TUR.
  - Table 2 (a) / 2 (b). Estimated Model Coefficients with and without Covariates — MARSS interaction matrix estimates; baseline and augmented model coefficients reported (e.g., Bff 0.91*** (0.10) baseline; Augmented: Bff 0.85*** (0.10); Cff 0.18** (0.09); Cfg −0.14** (0.06)); AICc and Log-likelihood values given.
  - Table 3 (a) / 3 (b). Estimated Model Coefficients with Different EM Growth Measures — alternative EMFCI measures and weighting schemes; reported coefficients with standard errors in parentheses.
  - Table 4 (a) / 4 (b). Model Coefficients for Different EM Growth Quantiles — coefficients for quantiles Q50, Q5, Q10, Q90, Q95 with AICc and Log-likelihood diagnostics reported.
  - Table 5. Results: Finance-Growth Feedback by Demand Component — reported coefficients (Bfg and Bgf): GDP Bfg −0.20***, Bgf 0.19**; Investment Bfg −0.14**, Bgf 0.20**; Consumption Bfg −0.19**, Bgf 0.18*; Government Bfg 0.13 (ns), Bgf 0.16 (ns).
  - Table 6. Summary Results: Sign Test — counts across 18 countries for expected sign and significance for Growth-enhancing Effect (Bfg), Growth-inhibiting Effect (Bgf), Covariate Effects Cff and Cfg; e.g., “No. of countries with Expected sign and significant: 11; Total # countries 18.”
  - Table 7. Median Quantile Regressions: Financial Conditions, Growth, and Macro-imbalances — model summaries and coefficients (τ=0.50): Model 1 GDP EMFCI3 (−1) −0.17 (0.06); Model 2 −0.23 (0.02) with CAB; Model 3 −0.23 (0.02) with DEBT; Model 4 −0.25 (0.01) Debt + CAB; Models 5–8 show EMFCI on GDP (lags) with coefficients 0.11 (0.05), 0.12 (0.08), 0.12 (0.04), 0.11 (0.15).
  - Table 8. Median Finance-Growth Feedback by Debt and External Balance Subgroups — subgroup Bfg and Bgf estimates: HD Bfg −0.47***, Bgf 0.30***; LD Bfg −0.27***, Bgf 0.22***; CAD Bfg −0.37***, Bgf 0.31***; CAS Bfg −0.44***, Bgf 0.30***; HD–CAS Bfg −0.44***, Bgf 0.22***; LD–CAD Bfg −0.24***, Bgf 0.26***; HD–CAD Bfg −0.31***, Bgf 0.33***; LD–CAS Bfg 0.02 (ns), Bgf 0.12 (ns).
  - Simulation tables (Table D1, Table D2) reporting model diagnostics and parameter estimates for alternative interaction matrix structures; example coefficients: B11 0.82*** (0.12), B21 −0.15** (0.09), B12 0.32** (0.09), B22 0.77*** (0.07); C11 −0.18 (0.13), C21 0.06 (0.09); AICc 252.38; Log-likelihood −113.13.

### Appendix content and methodological specifics
- Appendix A. Measuring Financial Conditions – Dynamic Factor Model (DFM) Approach:
  - DFM specification: each series λi,t driven by common factors Ft and idiosyncratic errors εi,t; equation (A1) λi,t = γi Ft + εi,t; idiosyncratic AR(1) εi,t = αi εi,t−1 + ei,t; ei,t ∼ iid N(0, σi2) (A2).
  - Matrix form: Xt = Γ Ft + Πt (A3); factor dynamics Ft = A Ft−1 + B ut (A4).
  - Two-step estimation: principal components initialization (S = 1/T ∑ Xt Xt′), extract r largest PCs using V and D, approximate factors F̃ = V′ Xt (A6); estimate Γ̂ and Π̂ via regressions (A7, A8); re-estimate factors via Kalman smoother (Doz et al. 2011).
- Appendix B. Constructing Financial Conditions Index (FCI)— Role of Financial Stress and Vulnerabilities:
  - Stress vs. vulnerabilities taxonomy: fast-moving stress indicators (e.g., asset prices, risk spreads) and slowly accumulating vulnerability indicators (e.g., S-risk, Debt-to-GDP).
  - Defined index variants in the text: DPOR; DPOR-EXT; DPOR-EMPI; DPOR-MF; DPOR-AGG; DPOR-AGG-FISC; DPOR-AGG-BAL; DPOR-ALL. EMFCI3 is based on DPOR-EMPI and EMFCI1 is based on DPOR-ALL.
  - Forecasting exercise: AR(1) model of GDP growth with current-period FCIs; training sample 2000Q1–2015Q4, forecast sample 2016Q1–2020Q4; out-of-sample RMSE comparison shows DPOR-ALL outperforms alternatives for the majority of EMs in the sample.
- Appendix C. Macro-financial Interaction — comparisons with alternative models:
  - Time-series and panel evidence: lead-lag correlation between EM GDP and FCI with correlation −0.79 at one-period lag; VAR and panel-VAR (panel VAR-X) evidence showing significant negative GDP response to FCI tightening and weaker/insignificant reverse response.
  - Panel Granger causality (Dumitrescu and Hurlin (2012)) results: strong Granger causality from EMFCI to EMGDP at the 1 percent significance level; bi-directional feedback between global financial cycle (GFC) and EMFCI confirmed.
  - Group-wise VAR results: Group A (twin-deficits) shows stronger and more prolonged GDP response to EMFCI tightening than Group B.
- Appendix D. Model Simulation Exercises:
  - MARSS interaction-matrix based simulations summarized; alternative input-variable configurations and diagnostics presented; model forms (D1) and (D2) and parameter reporting included.

*Emerging Market Cycles: Twin-Balance Sheet Conditions and Macro-Financial Linkages — Working Paper No. WP/2025/216*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025216-source-pdf.pdf_