## CHAPTER 3 FINANCIAL CONdITIONS ANd GROwTh AT RISk

## Source details

**Canonical URL:** [CHAPTER 3 FINANCIAL CONdITIONS ANd GROwTh AT RISk](https://www.imf.org/-/media/files/publications/gfsr/2017/october/chapter-3/documents/c3.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfsr/2017/october/chapter-3/documents/c3.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfsr/2017/october/chapter-3/documents/c3.pdf.json)

---

### Overview and purpose
- Develops a new analytical tool that maps financial conditions into the probability distribution of future GDP growth for 21 major advanced and emerging market economies over horizons up to three years.
- Defines “financial conditions” as combinations of:
  - key domestic financial market asset returns, funding spreads, and volatility;
  - domestic credit aggregates; and
  - external conditions such as measures of global risk sentiment (including VIX and MOVE).
- Constructs both a univariate financial conditions index (FCI) and a partitioned approach with three subindices: domestic price of risk; credit aggregates (leverage); and external conditions.
- Uses conditional density forecasting via quantile regressions to capture tails and state dependence in the mapping from financial conditions to future GDP growth.

### Methodological approach
- Empirical framework:
  - Baseline quantile regression: y_{t + h,q} = β_{f,q}^{h} FC_{t} + β_{y,q}^{h} y_{t} + ε_{t,q}^{h}.
  - Extended (partitioned) specification: y_{t + h,q} = α_{p,q}^{h} p_{t} + β_{a,q}^{h} Agg_{t} + γ_{y,q}^{h} y_{t} + φ_{f,q}^{h} f_{t} + ε_{t,q}^{h}, where p, Agg, and f are principal components of price-of-risk, credit aggregates, and foreign variables.
- FCIs estimation:
  - FCIs reestimated for 11 advanced economies starting in 1973 and for 10 emerging market economies starting in 1991 using a Koop and Korobilis 2014 implementation building on Primiceri 2005 and Doz, Giannone, and Reichlin 2011.
  - A set of 19 financial indicators is used (term spreads, interbank spreads, corporate spreads, equity returns, house price returns, credit growth, sovereign spreads, VIX, MOVE, commodity prices, exchange rates, etc.).
- From quantiles to density:
  - Estimated conditional quantiles (5th, 25th, 50th, 75th, 95th) are smoothed by fitting a parametric skewed t distribution with parameters {μ_{t + h}, s_{t + h}, v_{t + h}, ξ_{t + h}} obtained by minimizing squared distance to theoretical quantiles.

### Main findings — signals from financial conditions
- General:
  - Changes in a country’s financial conditions shift the distribution of future GDP growth in both advanced and emerging market economies.
  - Asset prices and credit aggregates provide complementary information: fast-moving asset prices signal near-term risks; slowly changing balance-sheet aggregates indicate risks over longer horizons.
- Horizon- and variable-specific findings:
  - Over one to four quarters, tighter financial conditions (higher univariate FCIs) predict increased downside risks to GDP growth in most advanced economies and more uncertain growth in several emerging market economies.
  - A tightening domestic price of risk (decompression in spreads or increase in asset price volatility) is a significant predictor of large macroeconomic downturns within a one-year horizon.
  - Rising leverage is a significant predictor of elevated downside risk over the medium term (one to three years).
  - Asset prices are most informative about risks to growth in the short term; credit aggregates provide more information over longer time horizons.
  - A souring of global risk sentiment (higher VIX) increases downside risks to growth at short time horizons of one quarter.
- Distributional effects:
  - Movements in the FCI are especially powerful signals of changes in downside tail risk (e.g., 5th percentile) but are less informative about baseline growth (median) and the strength of booms (right tail) except for very large FCI changes.
  - Forecasts of worst-case outcomes (5th percentile) are between 3 times (United States) and more than 10 times (Australia) more sensitive to changes in FCIs than forecasts of the central tendency.

### Drivers of FCIs and cross-country heterogeneity
- Common drivers:
  - Higher corporate funding costs and worsening global risk sentiment (VIX and MOVE) tighten FCIs across countries.
  - Exchange rate appreciation uniformly loosens financial conditions.
- Cross-country differences:
  - Sovereign spreads are important in emerging market economies but rarely in advanced economies.
  - Commodity price changes loosen FCIs in commodity exporters (Australia, Brazil, Canada, Chile, Russia) and tighten them in commodity importers.
  - Asset price shocks generally appear more important than credit aggregates in driving changes in FCIs, possibly reflecting slower credit adjustment relative to GDP at turning points.
- Country patterns:
  - Brazil, Korea, and Mexico: higher FCI levels portend a more uncertain one-year growth outlook (fatter tails at both ends).
  - Chile (commodity exporter): tightening FCIs signal risk of stronger recessions and booms of lower intensity.
  - Advanced economies: price of risk predicts downside growth risks up to one year; price-of-risk signals become less informative over longer horizons.

### Predictive performance and historical illustration (Global Financial Crisis)
- FCIs add predictive power during crises:
  - Conditioning on FCIs significantly increases the conditional likelihood of a GDP growth outcome less than or equal to the actual growth outturn one quarter ahead for countries that experienced significant growth downturns during the crisis.
  - Example (2008:Q4 one-quarter-ahead densities for 2009:Q1): incorporating lagged FCI produces a more pessimistic (worse) red density versus a blue density based only on lagged GDP.
- Quantitative comparisons (cumulative probability of actual 2009:Q1 growth outturn, percent) — examples from models estimated four quarters earlier (2008:Q1):
  - Selected Advanced Economies — One quarter ahead for 2009:Q1 (Real-time FCI | FCI Augmented | Autoregressive):
    - Germany: 5.4, 2.4, 0.0
    - Sweden: 6.5, 5.9, 4.8
    - United Kingdom: 29.8, 29.5, 5.8
    - United States: 46.7, 30.3, 8.5
  - Selected Advanced Economies — Four quarters ahead for 2009:Q1:
    - Germany: 0.1, 0.4, 0.0
    - Sweden: 0.0, 0.8, 0.5
    - United Kingdom: 0.8, 2.8, 1.5
    - United States: 2.6, 4.0, 4.2
  - Selected Emerging Market Economies — One quarter ahead for 2009:Q1:
    - Brazil: 35.5, 39.6, 7.5
    - Chile: 6.4, 8.0, 2.6
    - South Africa: 7.2, 4.6, 0.8
    - Turkey: 31.5, 27.1, 5.3
  - Selected Emerging Market Economies — Four quarters ahead for 2009:Q1:
    - Brazil: 4.2, 5.0, 5.5
    - Chile: 4.0, 1.7, 2.0
    - South Africa: 5.3, 6.2, 1.6
    - Turkey: 3.5, 2.3, 2.8
- Partitioned financial variables:
  - Partitioning FCIs into subindices (price of risk, credit aggregates, foreign shocks) improves four-quarters-ahead cumulative probability forecasts relative to univariate FCI and autoregressions in many cases (examples shown for Germany, Sweden, United Kingdom, United States, Brazil, Chile, South Africa, Turkey).
- Robustness and real-time estimation:
  - Results robust across a broader cross section of crisis-affected countries.
  - For countries that did not experience an economic contraction, the model augmented with FCIs does not generate false alarms at one- and four-quarter horizons.
  - Real-time conditional density forecasts closely match in-sample forecasts; recursive estimation of quantiles produces similar conditional likelihoods.

### Policy implications and recommendations
- Usage:
  - Financial-condition-based density forecasts provide a close measure of financial vulnerability and a complete depiction of risks to economic activity, allowing policymakers to define risk tolerance in terms of GDP growth and make precise probability statements (for example, probability of GDP growth < –3 percent one year ahead under a given financial-state).
- Monitoring guidance:
  - Short-term (imminent signals): watch domestic asset prices and global risk sentiment (VIX) for downside risk signals one quarter ahead.
  - Medium-term: monitor rising leverage and credit aggregates as predictors of elevated downside risk at horizons of one year and longer.
  - Country-specific calibration recommended due to heterogeneity, though directional signals are robust across countries.
- Policy actions:
  - Signals of an imminent near-term dire economic outcome may warrant crisis-management-type discretionary policy actions encompassing a range of monetary and macroprudential tools.
  - The approach can inform the design of policy rules and the calibration of countercyclical macroprudential tools (for example, bank capital buffers and limits on loan-to-value ratios).
  - Forecasting-model outputs could be used to calibrate parameters of structural macro-financial models used to guide policy.
- Warning:
  - A rapid decompression in spreads and increase in financial market volatility could significantly worsen the global growth risk outlook given prevailing low funding costs and rising leverage.

### Annex 3.1 — Structural simulation of financial vulnerabilities and macroprudential calibration
- Model features:
  - Embeds an occasionally binding collateral constraint (OBCC) in a New Keynesian open-economy structural model with nominal frictions and collateral-based credit limits.
- Simulation key results:
  - The simulated unconditional distribution of future output is negatively skewed; skewness measure = –1.51 (statistically significant).
  - Risk premium threshold for conditional analysis: 30 basis points (conditional densities shown for risk premium in period t − 1 less than 30 basis points and more than 30 basis points).
  - When the OBCC binds, output and asset prices decline significantly due to asset fire sales and tighter credit conditions.
  - Financial crises occur only when the credit-to-GDP ratio is historically high; risk premiums and credit-to-output ratios are significantly higher than steady-state values for several periods before a crisis.
- Macroprudential calibration:
  - Baseline (no optimal macroprudential policy): probability of a recession driven by a financial crisis = 1.3 percent; skewness of future GDP growth density = –1.51.
  - Implementing state-contingent debt tax or state-contingent LTV regulation reduces these to: probability = 0.5 percent; skewness = –0.66.
  - A simple rules-based macroprudential policy (debt taxes as a linear function of risk premiums) delivers almost the same performance as the optimal policy in the model.
- Caveats:
  - Model limitations include reliance on a single OBCC channel for crises and assumed immediate policymaker response.

### Annex 3.2 — Estimating FCIs, data, partitioning, and LDA
- Data coverage and variables:
  - FCIs use 19 financial indicators; advanced economies from 1973; emerging markets from 1991.
  - Global-level variables include VIX and MOVE.
- Data definitions (selected):
  - Term Spreads: yield on 10-year government bonds minus yield on three-month Treasury bills.
  - Corporate Spreads: corporate yield of the country minus yield of the benchmark country.
  - Equity Returns (local currency): log difference of the equity indices.
  - Credit Growth: percent change in the depository corporations’ claims on private sector.
  - Change in Credit to GDP: change in credit provided by domestic banks, all other sectors, and nonresidents (in percent of GDP).
- Partitioning and LDA:
  - Financial indicators are grouped via linear discriminant analysis (LDA) into Price of Risk, Leverage, and Foreign Shocks to maximize discrimination between periods when one-year-ahead GDP growth is below the 20th percentile and normal periods.
  - Price of Risk group includes: term spread; corporate spread; sovereign spread; interbank spread; equity returns; equity historical volatility; house price returns; VIX (except for the United States, where VIX is a price-of-risk variable).
  - Leverage group includes: credit to GDP; credit growth (quarterly); credit growth (percent change in depository corporations’ claims on private sector); short-term rate; real long-term rate.
  - Foreign Shocks group includes: bilateral exchange rate (US dollar to local currency); commodity prices.
- Limitations noted:
  - Variable weights in univariate FCIs need not reflect economic considerations or country-specific characteristics; asset prices may be better short-horizon indicators while credit aggregates may be more informative at longer horizons.

### Annex 3.3 — Quantile regressions and density construction
- Estimation approach:
  - Quantile regressions estimate conditional quantiles ŷ_{t + h,q} used to derive a full conditional density.
  - Percentiles matched for parametric smoothing: 5th, 25th, 50th, 75th, and 95th.
  - Parametric form: skewed t distribution with parameters {μ, s, v, ξ} obtained by minimizing squared distance between estimated and theoretical quantiles.
- Advantages:
  - Quantile regression is robust to extreme outliers and violations of normality and homoscedasticity, allows time-varying structural parameters, and avoids overfitting compared with more complex models.
- Implementation references and technical notes preserved from source.

### Country coverage (Annex Table 3.2.1)
- Australia
- Brazil
- Canada
- Chile
- China
- France
- Germany
- India
- Indonesia
- Italy
- Japan
- Korea
- Mexico
- Russia
- South Africa
- Spain
- Sweden
- Switzerland
- Turkey
- United Kingdom
- United States

*Sources: Bloomberg Finance L.P.; Haver Analytics; IMF, Global Data Source and World Economic Outlook databases; Thomson Reuters Datastream; and IMF staff estimates.*

### Introduction

### c3 - Introduction

### Overview
- The global financial crisis highlighted that financial vulnerabilities can increase both the duration and severity of economic recessions.
- Financial vulnerabilities are defined as the extent to which the adverse impact of shocks on economic activity may be amplified by financial frictions; they usually grow in buoyant economic conditions when investment opportunities seem ample, funding conditions are easy, and risk appetite is high.
- Financial indicators can provide intelligence on risks to the economic outlook by linking the state of the financial system to the probability of a financial crisis or bank capital shortage.

### Prepared by
- Prepared by a staff team consisting of Jay Surti (team leader), Mitsuru Katagiri, Romain Lafarguette, Sheheryar Malik, and Dulani Seneviratne, with contributions from Vladimir Pillonca, Aquiles Farias, André Leitão Botelho, Kei Moriya, and Changchun Wang, under the general guidance of Claudio Raddatz and Dong He.
- The chapter team benefited from discussions with Norman Swanson, Nellie Liang, and Domenico Giannone. Claudia Cohen and Breanne Rajkumar provided editorial assistance.

### Methodological approach
- Develops a new analytical tool that maps financial conditions into the probability distribution of future GDP growth.
- Financial conditions correspond to combinations of:
  - key domestic financial market asset returns, funding spreads, and volatility;
  - domestic credit aggregates; and
  - external conditions such as measures of global risk sentiment.
- Extends a nascent literature deriving a direct empirical link between financial conditions and risks to the real economy and applies it to 21 major advanced and emerging market economies over the near and medium term.

### Key questions addressed
- Do changes in financial conditions signal risks to future GDP growth? Are they equally informative for advanced and emerging market economies, about the intensity of recessions and the strength of booms, and over different time horizons?
- What types of financial variables are more informative regarding the risks to growth at different time horizons and in different countries?
- Could we have used financial conditions to shed light on the likelihood of extremely negative growth outcomes of the past, such as the global recession following the bankruptcy of Lehman Brothers?
- How can policymakers make use of this new tool of macro-financial surveillance?

### Main findings
- Changes in a country’s financial conditions shift the distribution of future GDP growth in both advanced and emerging market economies.
  - A tightening of financial conditions, reflected in a decompression in spreads or an increase in asset price volatility, is a significant predictor of large macroeconomic downturns within a one-year horizon.
  - In emerging market economies, tighter financial conditions could also portend stronger booms over the subsequent four quarters, possibly because of procyclical capital flows.
- Asset prices are most informative about risks to growth in the short term, whereas credit aggregates provide more information over longer time horizons.
  - A rising cost of funding and falling asset prices signal a greater threat of severe recession at time horizons of up to four quarters.
  - Higher leverage signals increased downside risk to growth at horizons between one and three years.
- Movements in commodity prices and exchange rates affect the real economy in a significant, albeit complex, manner, complicating simple economic interpretation of their predictive content.
- A souring of global risk sentiment increases downside risks to growth at short time horizons of one quarter.
- Heterogeneity exists across countries in the information content of financial conditions for growth risks (e.g., asset prices are no longer informative over horizons longer than a year for advanced economies, but remain informative for emerging markets).
- A retrospective real-time analysis of the global financial crisis shows that forecasting models augmented by financial conditions would have assigned a much higher likelihood to the post-Lehman economic contraction than models based on recent growth performance alone.

### Policy implications
- Policymakers should maintain heightened vigilance regarding risks to growth during periods of benign financial conditions that may foster accumulation of financial vulnerabilities.
- Signals of an imminent near-term dire economic outcome may warrant crisis-management-type discretionary policy actions encompassing a range of monetary and macroprudential tools.
- The approach can inform the design of policy rules and the calibration of countercyclical macroprudential tools as financial vulnerabilities develop.
- The forecasting-model outputs could be used to calibrate parameters of structural macro-financial models used to guide policy.
- The cross-country richness of results suggests scope for authorities to adapt the broad approach to specific country conditions and structural changes in financial markets and the real economy.

### Conceptual issues: financial conditions, vulnerabilities, and growth risks
- Economic growth has a complex and nonlinear relationship with shocks and financial vulnerabilities; financial vulnerabilities increase risks to growth.
- In buoyant macro-financial environments, ease of borrowing and high asset prices reduce incentives to manage liquidity and solvency risks and can lead to increased leverage by households and firms.
- Booming asset prices boost capital adequacy, lending capacity, and risk appetite of financial intermediaries, encouraging maturity transformation and short-term wholesale funding for long-term credit exposures; this can accumulate maturity mismatches and balance sheet weaknesses.
- When vulnerabilities are elevated, small negative shocks can trigger credit tightening, defaults, asset liquidation, and further rounds of contraction in credit, investment, and growth; rising volatility and risk spreads increase capital buffer requirements and constrain risk-bearing capacity, also through funding liquidity concerns.
- Asset prices and credit aggregates provide complementary signals: fast-moving asset prices signal near-term risks, while gradually changing balance sheet aggregates indicate risks over longer horizons.

*Source: https://www.imf.org/-/media/files/publications/gfsr/2017/october/chapter-3/documents/c3.pdf*

### CHAPTER 3 FINANCIAL CONdITIONS ANd GROwTh AT RISk

### CHAPTER 3 FINANCIAL CONdITIONS ANd GROwTh AT RISk

### Overview
- Defines “financial conditions” as a combination of a broad set of financial variables (including nominal exchange rate and commodity prices) that influence economic behavior and thereby the future of the economy.
- Examines two approaches to constructing measures of financial conditions: a univariate financial conditions index (FCI) and a partitioned approach with three subindices (domestic price of risk; credit aggregates; external conditions).
- Empirical framework: forecasts the probability distribution of future GDP growth for major advanced and emerging market economies for horizons of up to three years through quantile projections.

### Construction of Financial Conditions Measures
- Univariate FCI:
  - Aggregates multiple financial indicators into a single index (parsimony reduces parameter uncertainty but can let volatile indicators dominate).
- Three subindices (partitioned approach):
  - Domestic price of risk: risk spreads, asset returns, price volatility.
  - Credit aggregates: leverage and credit growth.
  - External conditions: global risk sentiment, commodity prices, exchange rates.
- The global FCI is defined as the first principal component of the country-level FCIs.

### Empirical Framework and Rationale
- Uses conditional density forecasting and quantile regressions to map current financial conditions into the distribution of future GDP growth.
- Focuses on full density (tails and central tendency) to capture nonlinearities and state dependence.
- Example application: quantifying the probability of GDP growth being less than –3 percent one year ahead given a state of the financial system.

### Key Findings — How Changes in Financial Conditions Indicate Risks to Growth
- Horizons and effects:
  - Over a horizon of one to four quarters, tighter financial conditions (higher univariate FCIs) predict increased downside risks to GDP growth in most advanced economies and more uncertain growth in several emerging market economies.
  - An increasing domestic price of risk signals an elevated threat of imminent, severe recession in advanced and emerging market economies.
  - Rising leverage is a significant predictor of elevated downside risk over the medium term.
- Distributional effects:
  - Movements in the FCI are especially powerful signals of changes in downside tail risk to the global economy but are less informative about baseline growth (median) and strength of booms (right tail), except for very large FCI changes.
  - Forecasts of worst-case outcomes (5th percentile) are between 3 times (United States) and more than 10 times (Australia) more sensitive to changes in FCIs than forecasts of the central tendency.
- Short-term easing example:
  - Easing of global financial conditions through 2016 signaled reduced tail risk to global growth for 2017, reflected in upward movement in the bottom tail (5th percentile) of the GDP growth density forecast.

### Drivers of FCIs and Cross-Country Variation
- Common drivers:
  - Higher corporate funding costs and worsening global risk sentiment (VIX and MOVE) tighten FCIs across countries.
  - Exchange rate appreciation uniformly loosens financial conditions.
- Cross-country differences:
  - Sovereign spreads are important in emerging market economies but rarely in advanced economies.
  - Commodity price changes loosen FCIs in commodity exporters (Australia, Brazil, Canada, Chile, Russia) and tighten them in commodity importers.
  - Asset price shocks generally appear more important than credit aggregates in driving changes in FCIs, possibly reflecting slower credit adjustment relative to GDP at turning points.

### Asset Prices, Aggregates, and Time Horizons
- Domestic asset prices:
  - Term and interbank spreads, followed by corporate and sovereign spreads, are the most important risk indicators for the investment and growth outlook in advanced economies.
  - House price dynamics matter where homeownership and floating-rate mortgages are high (United Kingdom) or where mortgages underpin systemic funding markets (United States).
  - For many emerging markets with limited data, sovereign spreads and equity returns are most significant.
  - Domestic asset prices dominate short-term drivers of the domestic price of risk.
- Credit aggregates:
  - Credit growth and credit-to-GDP signal greater downside risk at horizons of one year and longer; effects are stronger at lower quantiles and in advanced economies.
  - Over one-quarter horizons, rising leverage signals higher downside risks in emerging market and large advanced economies, but lower downside risks in small open advanced economies.
- External conditions:
  - Changes in external variables are complex to interpret due to mixed channels (real and financial) and commodity exporter/importer status.
  - Isolating global risk sentiment (VIX) gives clearer interpretation: higher VIX signals greater downside risks in the short term but may signal lower downside risks at one- to two-year horizons, possibly through slower leverage growth.

### Country Examples and Patterns
- Emerging market specifics:
  - In Brazil, Korea, and Mexico, higher FCI levels portend a more uncertain one-year growth outlook (fatter tails at both ends).
  - In some commodity exporters (Chile), tightening FCIs signal risk of stronger recessions and booms of lower intensity.
- Advanced economies:
  - The price of risk is a significant predictor of downside growth risks up to one year, particularly in the left tail; price-of-risk signals become less informative over longer horizons.

### Predictive Performance and Historical Illustration
- Augmenting growth forecast models with financial conditions significantly improves the ability to forecast downside outcomes (greater likelihood assigned to actual negative growth outcomes during severe episodes).
- Historical episode: global financial crisis forecasting
  - At a one-quarter horizon (fourth quarter of 2008), conditioning risk forecasts on financial conditions added significantly to capturing subsequent negative outcomes.
  - The model was used to predict the distribution of growth for the first quarter of 2009, broadly corresponding to the peak of the global financial crisis.

### Policy-Relevant Implications
- Financial-condition-based density forecasts provide:
  - A close measure of financial vulnerability (how the financial system amplifies shocks).
  - A complete depiction of risks to economic activity that allows policymakers to define risk tolerance in terms of GDP growth.
  - Precise probability statements (for example, probability of GDP growth < –3 percent one year ahead under a given financial-state).
- Monitoring implications:
  - Short-term: watch domestic asset prices and global risk sentiment for imminent downside risk signals.
  - Medium-term: monitor rising leverage and credit aggregates as predictors of elevated downside risk.
  - Country-specific calibration may improve magnitude estimates, though directional signals are robust.
- Stress scenario warning:
  - A rapid decompression in spreads and increase in financial market volatility could significantly worsen the global growth risk outlook, given prevailing low funding costs and rising leverage.

*International Monetary Fund | October 2017 — CHAPTER 3 FINANCIAL CONdITIONS ANd GROwTh AT RISk*

### CHAPTER 3 FINANCIAL CONdITIONS ANd GROwTh AT RISk

### CHAPTER 3 FINANCIAL CONdITIONS ANd GROwTh AT RISk

### Price of risk, leverage, and market volatility as predictors of downside growth risks
- Figure 3.4 (quantile regressions) — Higher price of risk is a significant predictor of downside growth risks:
  - Advanced economies: one quarter ahead — "Economic significance is highest over one quarter ..."
  - Emerging market economies: one quarter ahead — "... albeit less so in emerging market economies."
  - Advanced economies: one year ahead — "It remains so over one year in advanced economies ..."
  - Emerging market economies: one year ahead — "... and in emerging market economies."
  - Advanced economies: two years ahead — "Price of risk becomes uninformative over longer horizons in advanced economies ..."
  - Emerging market economies: two years ahead — "... but, in emerging market economies, higher funding costs signal lower risk over longer horizons."
- Figure 3.5 (quantile regressions) — Rising leverage signals higher downside growth risks at longer time horizons (three years ahead) for both advanced and emerging market economies.
- Figure 3.6 (quantile regressions; VIX = Chicago Board Options Exchange Volatility Index) — Waning global risk appetite (higher VIX) signals imminent downside risks to growth one quarter ahead in advanced and emerging market economies.

### Financial conditions indices (FCIs) add predictive power during crises
- Conditioning on FCIs significantly increases the conditional likelihood of a GDP growth outcome less than or equal to the actual growth outturn one quarter ahead for countries that experienced a significant growth downturn during the crisis.
- Figure 3.7 (probability densities) — For the United States and Chile at 2008:Q4:
  - Model with single regressor (one-quarter-lagged GDP growth) = blue density.
  - Model with two regressors (one-quarter-lagged GDP growth and one-quarter-lagged FCI) = red density.
  - "Accounting for financial conditions generates a more pessimistic outlook for risks to growth one quarter before 2009:Q1."
  - The likelihood attached to poor growth outcomes around the actual realization is significantly higher when rapidly tightening financial conditions are incorporated into the growth forecast (red density) versus using only past growth (blue density).

### Quantitative comparisons for the Global Financial Crisis (2009:Q1) — cumulative probabilities of actual growth outturn
- Table 3.1 (cumulative probability of actual 2009:Q1 growth outturn, percent) — probabilities from models estimated four quarters earlier (in 2008:Q1):
  - Selected Advanced Economies — One quarter ahead for 2009:Q1 (Real-time FCI | FCI Augmented | Autoregressive):
    - Germany: 5.4, 2.4, 0.0
    - Sweden: 6.5, 5.9, 4.8
    - United Kingdom: 29.8, 29.5, 5.8
    - United States: 46.7, 30.3, 8.5
  - Selected Advanced Economies — Four quarters ahead for 2009:Q1:
    - Germany: 0.1, 0.4, 0.0
    - Sweden: 0.0, 0.8, 0.5
    - United Kingdom: 0.8, 2.8, 1.5
    - United States: 2.6, 4.0, 4.2
  - Selected Emerging Market Economies — One quarter ahead for 2009:Q1:
    - Brazil: 35.5, 39.6, 7.5
    - Chile: 6.4, 8.0, 2.6
    - South Africa: 7.2, 4.6, 0.8
    - Turkey: 31.5, 27.1, 5.3
  - Selected Emerging Market Economies — Four quarters ahead for 2009:Q1:
    - Brazil: 4.2, 5.0, 5.5
    - Chile: 4.0, 1.7, 2.0
    - South Africa: 5.3, 6.2, 1.6
    - Turkey: 3.5, 2.3, 2.8
- Table 3.2 — Market consensus forecasts for 2009:Q1 were considerably more optimistic than forecasts conditional on lagged GDP and FCI (examples shown as presented):
  - Brazil: 3.1 −4.3 4.6 2.1 −6.9
  - Canada: 1.7 −5.3 1.7 −0.1 −8.8
  - France: 1.9 −1.2 1.6 −0.6 −6.4
  - Mexico: 2.6 −3.6 2.8 −0.1 −14.7
  - South Africa: 2.7 −2.0 4.7 2.7 −6.1
  - Switzerland: 1.9 −2.0 2.8 −1.6 −5.5
  - Turkey: 3.4 −7.4 4.8 0.8 −15.2
  - United States: 1.9 −3.8 1.6 −1.3 −5.4
  - Note: Columns denote conditional mean forecasts based on lagged FCI and GDP (one quarter and one year earlier), market consensus forecasts (one quarter and four quarters earlier), and the actual growth outturn.
- Table 3.3 — Partitioned financial variables (subindices) improve four-quarters-ahead cumulative probability forecasts relative to univariate FCI and autoregressions (examples):
  - Advanced economies (four quarters ahead for 2009:Q1; Real-time partitioned | Partitioned financial variables | FCI Augmented | Autoregressive):
    - Germany: 0.8, 0.7, 0.4, 0.0
    - Sweden: 7.1, 5.7, 0.8, 0.5
    - United Kingdom: 6.4, 5.0, 2.8, 1.5
    - United States: 24.7, 19.1, 4.0, 4.2
  - Emerging market economies (four quarters ahead for 2009:Q1):
    - Brazil: 14.0, 6.7, 5.0, 5.5
    - Chile: 12.7, 10.4, 1.7, 2.0
    - South Africa: 5.4, 7.3, 6.2, 1.6
    - Turkey: 7.4, 4.4, 2.3, 2.8
- Interpretation:
  - Adding FCIs to autoregressive growth forecasts increases the conditional likelihood assigned to the realized severe contractions in many crisis-hit countries one quarter ahead.
  - Partitioning FCIs into subindices (for example, separating credit aggregates from asset prices) allows their signals to regain predictive gains at horizons beyond one quarter, improving one-year-ahead likelihoods of poor growth outcomes.

### Robustness, real-time estimation, and forecasting implications
- Results remain robust across a broader cross section of countries that experienced significant growth downturns during the crisis.
- For countries that did not experience an economic contraction, the model augmented with FCIs does not generate false alarms—that is, it does not produce a significantly lower conditional probability of a recession at one- and four-quarter horizons.
- The quantile autoregression model of GDP growth is a conservative benchmark due to high persistence in GDP growth; augmenting this benchmark with FCIs demonstrates measurable gains in assigning higher probability to severe downside outcomes.
- Real-time conditional density forecasts are almost identical to in-sample forecasts (Figures 3.8 and 3.9); recursive (real-time) estimation of quantiles produces similar conditional likelihoods to in-sample estimation.
- Timing explanation for forecasting gains:
  - Financial conditions (risk spreads, market volatility, asset prices) deteriorated sharply in Q4 2008 immediately after the Lehman Brothers bankruptcy, signaling potential negative fallout for economic activity that economic indicators caught up with only later.
  - Hence, conditioning on FCIs improves short-horizon (one-quarter) forecasts late in 2008; at longer horizons (one year), univariate FCIs may be less informative unless partitioned into subindices.

*Sources: Bloomberg Finance L.P.; Haver Analytics; IMF, Global Data Source and World Economic Outlook databases; Thomson Reuters Datastream; and IMF staff estimates.*

### 1. Germany2. Brazil

### c3 - 1. Germany2. Brazil

### Policy implications and overarching findings
- The chapter develops a new macroeconomic measure of financial stability by linking financial conditions to the probability distribution of future GDP growth, enabling policymakers to assess the whole distribution of future GDP growth and specify bad outcomes according to risk preference or tolerance.
- Financial conditions contain useful information to help forecast risks to economic growth at short- and medium-term horizons.
- Elevated leverage signals downside risks to growth in the medium term, while the short-term risk can be mitigated by a low price of risk; however, a scenario of rapid decompression in spreads and increased financial market volatility would significantly worsen the growth outlook.
- Changes in the domestic price of risk are potent signals of imminent threats to growth and can justify swift deployment of monetary easing and crisis-management policy actions.
- Countercyclical macroprudential tools (for example, bank capital buffers and limits on loan-to-value ratios) could be designed and calibrated to contain the growth of financial vulnerabilities during loose financial conditions.
- Practical implementation requires closing data gaps, continuous calibration as data availability improves, and incorporation of country-level information and relevant financial indicators.

*Italic line: Sources: Bloomberg Finance L.P.; Haver Analytics; IMF, Global Data Source and World Economic Outlook databases; Thomson Reuters Datastream; and IMF staff estimates.*

### Annex 3.1 — Financial vulnerabilities, growth hysteresis, and a structural simulation
- Purpose: Illustrate the nonlinear response of output growth to shocks depending on the level of financial vulnerabilities by embedding an occasionally binding collateral constraint (OBCC) in a New Keynesian open-economy structural model.
- Model features:
  - OBCC modeled as in Kiyotaki and Moore 1997.
  - Nominal frictions and open-economy NK features in the spirit of Galí and Monacelli 2005.
  - Households endowed with tradable goods as in Bianchi 2011; nontradables produced with capital and labor.
  - Borrowing limited to a fixed fraction of capital value (collateral constraint).
  - Asset prices determined under fixed supply of capital; nominal interest rates set by a standard Taylor rule; exchange rate from uncovered interest parity.
  - Parameters calibrated to standard values in OBCC and open-economy NK literature.
- Key simulation results:
  - The simulated unconditional distribution of future output is negatively skewed; skewness measure = –1.51 (statistically significant).
  - Risk premium threshold used for conditional analysis: 30 basis points (conditional densities shown for risk premium in period t − 1 less than 30 basis points and more than 30 basis points).
  - When the OBCC binds (rare event), output and asset prices decline significantly due to asset fire sales and tighter credit conditions.
  - Relationship findings:
    - Rising risk premiums shift the conditional density of one-period-ahead output leftward and increase negative skewness (fatter left tail).
    - The lower quantile of one-period-ahead output declines significantly with rising risk premiums while the upper quantile is less sensitive.
    - Financial crises occur only when the credit-to-GDP ratio is historically high.
    - Risk premiums and credit-to-output ratios are significantly higher than steady-state values for several periods before a crisis.
- Calibration of macroprudential policy:
  - Baseline (no optimal macroprudential policy): probability of a recession driven by a financial crisis = 1.3 percent; skewness of future GDP growth density = –1.51.
  - Implementing state-contingent debt tax or state-contingent LTV regulation reduces these to: probability = 0.5 percent; skewness = –0.66.
  - A simple rules-based macroprudential policy (debt taxes as a linear function of risk premiums) delivers almost the same performance as the optimal policy in the model.
- Caveats:
  - All crises in the OBCC model are caused by a simple collateral constraint, whereas real-world crises may have many contributing factors.
  - The model assumes policymakers can respond immediately; delays in policy reactions or transmission may alter implications.

*Italic line: Source: Prepared by Mitsuru Katagiri (summary of Katagiri, forthcoming); IMF staff estimates.*

### Annex 3.2 — Estimating Financial Conditions Indices (FCIs)
- Main changes to FCI construction in this chapter versus April 2017 GFSR:
  - Expanded coverage to include additional indicators relevant to domestic financial vulnerabilities, global risk sentiment (Chicago Board Options Exchange Volatility Index [VIX], Merrill Lynch Option Volatility Estimate [MOVE] Index), credit aggregates, commodity prices, and exchange rates.
- Data coverage and variables:
  - FCIs reestimated for 11 advanced economies starting in 1973 and for 10 emerging market economies starting in 1991.
  - A set of 19 financial indicators is used to capture domestic and global developments influencing a country’s financial conditions.
- Methodology:
  - FCIs estimated using Koop and Korobilis 2014 implementation; methodology builds on Primiceri 2005 time-varying parameter VAR and Doz, Giannone, and Reichlin 2011 dynamic factor models.
  - Model form (paraphrased notation preserved from source):
    - x_t = λ^y_t Y_t + λ^f_t f_t + u_t
    - [Y_t f_t] = B_{1,t} [Y_{t−1} f_{t−1}] + B_{2,t} [Y_{t−2} f_{t−2}] + . . . + ε_t
    - f_t is the latent factor interpreted as the FCI; Y includes macro variables such as real GDP growth and inflation.
- Rationale and advantages:
  - Univariate FCIs summarize information from multiple financial indicators and can reduce parameter uncertainty in forecasting.
  - The approach controls for current macroeconomic conditions and allows dynamic interaction between FCIs and macroeconomic variables.
- Limitations noted:
  - Variable weights in univariate FCIs are not necessarily driven by economic considerations or country-specific characteristics.
  - Asset prices may be better short-horizon indicators while slower-moving credit aggregates may offer more information at longer horizons, especially in emerging markets.
  - Consequently, indicators need not receive the same weight across time horizons and countries; the chapter also uses approaches that exploit individual indicators’ information content.

*Italic line: Prepared by Romain Lafarguette and Dulani Seneviratne; IMF staff estimates.*

### Annex Table 3.2.1. Country Coverage

### Annex Table 3.2.1. Country Coverage

### Country coverage
- Australia
- Brazil
- Canada
- Chile
- China
- France
- Germany
- India
- Indonesia
- Italy
- Japan
- Korea
- Mexico
- Russia
- South Africa
- Spain
- Sweden
- Switzerland
- Turkey
- United Kingdom
- United States

*Source: IMF staff.*

### Data sources and variable definitions (Annex Table 3.2.2)
- Term Spreads: Yield on 10-year government bonds minus yield on three-month Treasury bills — Bloomberg Finance L.P.; IMF staff
- Interbank Spreads: Interbank interest rate minus yield on three-month Treasury bills — Bloomberg Finance L.P.; IMF staff
- Change in Long-Term Real Interest Rate: Percentage point change in the 10-year government bond yield, adjusted for inflation — Bloomberg Finance L.P.; IMF staff
- Corporate Spreads: Corporate yield of the country minus yield of the benchmark country; JPMorgan CEMBI Broad is used for emerging market economies where available — Bloomberg Finance L.P.; Thomson Reuters Datastream
- Equity Returns (local currency): Log difference of the equity indices — Bloomberg Finance L.P.
- House Price Returns: Log difference of the house price index — Bank for International Settlements; Haver Analytics; IMF staff
- Equity Return Volatility: Exponential weighted moving average of equity price returns — Bloomberg Finance L.P.; IMF staff
- Change in Financial Sector Share: Log difference of the market capitalization of the financial sector to total market capitalization — Bloomberg Finance L.P.
- Credit Growth: Percent change in the depository corporations’ claims on private sector — Bank for International Settlements; Haver Analytics; IMF, International Financial Statistics database
- Sovereign Spreads: Yield on 10-year government bonds minus the benchmark country’s yield on 10-year government bonds — Bloomberg Finance L.P.; IMF staff
- Banking Sector Vulnerability: Expected default frequency of the banking sector — Moody’s Analytics, CreditEdge; IMF staff
- Exchange Rate Movements: Change in US dollar per national currency exchange rate; for the United States, Bloomberg Finance L.P.’s DXY index is used — Bloomberg Finance L.P.; IMF, Global Data Sources and International Financial Statistics databases
- Domestic Commodity Price Inflation: A country-specific commodity export price index constructed following Gruss 2014; change in the estimated country-specific commodity export price index is used — Bloomberg Finance L.P.; IMF, Global Data Sources database; United Nations, COMTRADE database; IMF staff
- Trading Volume (equities): Equity markets’ trading volume, calculated as level to 12-month moving average — Bloomberg Finance L.P.
- Market Capitalization (equities): Market capitalization of the equity markets, calculated as level to 12-month moving average — Bloomberg Finance L.P.; Thomson Reuters Datastream
- Market Capitalization (bonds): Bonds outstanding, calculated as level to 12-month moving average — Dealogic; IMF staff
- Change in Credit to GDP: Change in credit provided by domestic banks, all other sectors of the economy, and nonresidents (in percent of GDP) — Bank for International Settlements; Haver Analytics; IMF staff
- Real GDP Growth: Percent change in GDP at constant prices — IMF, World Economic Outlook database
- Inflation: Percent change in the consumer price index — Haver Analytics; IMF, International Financial Statistics database
- Global-Level Variables:
  - VIX: Chicago Board Options Exchange Market Volatility Index — Bloomberg Finance L.P.; Haver Analytics
  - MOVE: Merrill Lynch Option Volatility Estimate Index — Bloomberg Finance L.P.

Note: CEMBI = Corporate Emerging Markets Bond Index; DXY = Dollar Index Spot; MOVE = Merrill Lynch Option Volatility Estimate Index; VIX = Chicago Board Options Exchange Volatility Index.

### Data partitioning and Linear Discriminant Analysis (LDA)
- Individual financial indicators are aggregated into groups using linear discriminant analysis (LDA), a data-reduction technique (Annex Table 3.2.3).
- LDA projects the data set onto a lower-dimensional space while ensuring adequate separation of data into categories.
- The categorical variable used in LDA is a dummy variable, defined at the country level, equal to one when future GDP growth at a one-year horizon is below the 20th percentile of historical outcomes and equal to zero otherwise.
- LDA determines loadings to maximize contribution to discriminating between periods of low GDP growth and periods of normal GDP growth, linking financial indicators and GDP growth in the data-reduction process.
- Contrast with PCA: PCA aggregates only information about the common trend among financial indicators.
- LDA assumptions: independence of normally distributed data and homoscedastic variance among each class, although LDA is considered robust when these assumptions are violated (see Duda, Hart, and Stork 2001; Izenman 2013).

### Partitioning of financial indicators into groups (Annex Table 3.2.3)
- Price of Risk:
  - Term spread
  - Corporate spread
  - Sovereign spread
  - Interbank spread
  - Equity returns
  - Equity historical volatility
  - House price returns
  - VIX (except for the United States, for which VIX enters as a price-of-risk variable)
- Leverage:
  - Credit to GDP
  - Credit growth (quarterly)
  - Credit growth (percent change in depository corporations’ claims on private sector — see Annex Table 3.2.2)
  - Short-term rate
  - Real long-term rate
- Foreign Shocks:
  - Bilateral exchange rate (US dollar to local currency)
  - Commodity prices
- Persistence:
  - GDP growth

*Source: IMF staff.*

### Quantile regressions and estimation of the conditional density of future GDP growth (Annex 3.3)
- The conditional density forecast is estimated through quantile projections using quantile regressions.
- Baseline quantile regression specification (equation A3.3.1):
  y_{t + h,q} = β_{f,q}^{h} FC_{t} + β_{y,q}^{h} y_{t} + ε_{t,q}^{h}.
  - In the baseline approach, FC corresponds to a predetermined univariate financial conditions index (FCI) constructed as described in Annex 3.2.
- Extended specification disentangling contributions (equation A3.3.2):
  y_{t + h,q} = α_{p,q}^{h} p_{t} + β_{a,q}^{h} Agg_{t} + γ_{y,q}^{h} y_{t} + φ_{f,q}^{h} f_{t} + ε_{t,q}^{h},
  - where p, Agg, and f correspond to the principal components of the price of risk (asset prices and risk spreads), credit aggregates, and global or foreign variables (commodity prices, exchange rates, and global risk sentiment).
- Quantile regressions advantages (Koenker 2005; Komunjer 2013): desirability of the conditional quantile estimator as a predictor of the true future quantile; robustness to extreme outliers and violations of normality and homoscedasticity; flexibility for time-varying structural parameters and optimal weighting of predictors depending on country, horizon, and portion of the distribution; ability to avoid overfitting compared with more complex models.

### Deriving the density forecast from quantile estimates
- The quantile regression yields estimated conditional quantile function ŷ_{t + h,q} = β̂_{f,q}^{h} FC_{t} + β̂_{y,q}^{h} y_{t}.
- Smoothing the noisy estimated quantile function is achieved by fitting a parametric skewed t distribution.
- For each quarter, the analysis pins down four parameters {μ_{t + h}, s_{t + h}, v_{t + h}, ξ_{t + h}} by minimizing the squared distance between estimated quantiles ŷ_{t + h,q} and theoretical quantile function y_{q}^{f}(μ_{t + h}, s_{t + h}, v_{t + h}, ξ_{t + h}) corresponding to the skewed t distribution.
- The four parameters are, respectively: location (μ), scale (s), degrees of freedom (v), and shape (ξ).
- Percentiles matched: 5th, 25th, 50th, 75th, and 95th.
- Optimization statement (textual form):
  {μ_{t + h}, s_{t + h}, v_{t + h}, ξ_{t + h}} = argmin_{μ_{t + h} ∈ ℝ, s_{t + h} > 0, v_{t + h} ≥ 2, ξ_{t + h} > 0} ∑_{q} { ŷ_{t + h,q} − y_{q}^{f}(μ_{t + h}, s_{t + h}, v_{t + h}, ξ_{t + h}) }^{2}
- Properties and motivations:
  - Choice of skewed t functional form provides flexibility.
  - As v → ∞, f(y; μ, s, v, ξ) is characterized by tail properties resembling a Gaussian distribution.
  - The density is symmetric for ξ = 1.
  - References for skewed t specifications and alternatives include Fernandez and Steel (1998); Giot and Laurent (2003); Lambert and Laurent (2002); Hansen (1994); Azzalini and Capitanio (2003).

*Prepared by Sheheryar Malik and Romain Lafarguette. Source: IMF staff.*

---


_Source: https://www.imf.org/-/media/files/publications/gfsr/2017/october/chapter-3/documents/c3.pdf_