## 5. Binary Recursive Tree: Defend the Currency Intervention

## Source details

**Canonical URL:** [5. Binary Recursive Tree: Defend the Currency Intervention](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1411.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1411.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1411.pdf.json)

---

### I. Introduction — scope and questions
- Research question: Which exchange rate regimes are most vulnerable to crisis, and why; and where to draw the line between safe floats and risky intermediate regimes.
- Sample: 50 major EMEs over 1980-2011.
- Methodology highlights:
  - Use of IMF’s detailed de facto classification, supplemented with other classifications (IMF de jure; Reinhart and Rogoff, RR).
  - Examination of underlying vulnerabilities (macroeconomic and financial) and crisis frequencies (banking, currency, sovereign debt) plus growth collapses defined relative to historical norms.
  - Decision-theoretic technique: binary recursive tree (BRT) analysis to allow arbitrary thresholds and interactive effects (e.g., exchange rate flexibility; degree/direction/circumstances of FX intervention).

### II. Trends in exchange rate regimes in EMEs
- Three phases identified:
  - Post-1997-98: “hollowing out of the middle” with moves toward free floats.
  - 2004 runup to GFC: increased adoption of intermediate regimes, especially managed floats.
  - GFC and beyond: accelerated move toward intermediate regimes, especially managed floats.
- Transition/steady-state findings:
  - Full-sample steady-state (aggregate): intermediate regimes ~0.697, hard pegs ~0.194, floats ~0.108.
  - Full-sample finer classification steady-state (fine): managed floats 0.307; hard peg 0.195; crawling peg/band 0.169; single currency peg 0.142; independent float 0.114; basket peg 0.040; horizontal band 0.034.
  - Recent sample (2000-11) steady-state: Fixed 0.311, Intermediate 0.580, Float 0.109; regime distribution in 2011 (de facto aggregate): Fixed 0.151, Intermediate 0.774, Float 0.075.
  - Empirical rejection of bipolar hypothesis: regimes are persistent but off-diagonal transition probabilities are non-zero; floats are the least persistent with 0.195 transitions from floats to intermediate regimes and 0.018/0.008 transitions to hard pegs depending on matrix.

### III. Financial and macroeconomic vulnerabilities by regime
- Stylized vulnerabilities under less flexible regimes:
  - Loss of exchange rate as adjustment tool → real exchange rate overvaluation → larger imbalances → higher currency crisis risk.
  - Exchange rate guarantees encourage excessive foreign borrowing and FX-denominated lending.
  - Intervention (if not sterilized) can fuel credit expansion; less flexible regimes may reduce fiscal discipline.
- Key empirical observations (selected exact figures):
  - Domestic credit change (3-year cumulative change in private credit-to-GDP):
    - Almost twice as large under hard pegs as under intermediate regimes, and almost four times as large as under floats (Table 3, col. [1]).
    - Change in credit more than twice as large under basket pegs than under single currency pegs, and almost eight times as large as under managed floats.
    - Regression coefficients for hard peg include: 4.332*, 4.345, 2.614, 4.086*, 4.023*, 0.736 (reference: free float); observations: 1,010 and 646; No. of countries: 51; R-squared up to 0.391.
  - Bank foreign borrowing and FX-denominated domestic lending:
    - Both measures roughly twice as large under hard pegs as under floats (Table 3, cols. [2]-[3]).
    - Hard peg bank foreign liabilities coefficients: 7.795**, 7.828**, 6.628**, 5.376*, 4.881*, 3.034; observations up to 1,237; No. of countries: 50; R-squared up to 0.437.
    - FX lending (percent of total bank lending) hard peg coefficients: 20.379**, 20.505**, 9.645, 4.154, 2.360, 2.351; observations: 571; No. of countries up to 44; R-squared up to 0.571.
    - Net financial flows/GDP: 0.671*** in FX lending regressions; bank foreign liabilities/GDP: 0.716*** to 0.668*** across specs.
  - Policy-relevant controls:
    - Controls on capital inflows associated with significantly lower banking system external liabilities (Open FX position limits coefficients: -5.223**, -4.058*).
    - Restrictions on FX-denominated lending reduce the proportion of such loans (Restrictions on FX lending: -13.805**, -11.362*).
  - Macroeconomic vulnerabilities:
    - Fiscal deficits: hard pegs fiscal balance coefficients include 1.328, 1.285, 0.984, 1.743**, 1.744**, 1.637** (varied specifications).
    - Real exchange rate overvaluation: hard pegs and intermediate regimes associated with greater overvaluation than pure floats; REER deviation coefficients include 3.584*** and 4.157*** (strong positive association).
    - Current account imbalances: Fixed average deficit -8 percent of GDP; Intermediate average deficit -6 percent of GDP; Float average deficit -4 percent of GDP. Reversal probabilities larger for fixed and intermediate regimes versus floats in many comparisons.

### IV. Crisis propensity by regime (banking, currency, sovereign debt, growth)
- Crisis definitions:
  - Banking crisis: systemic banking distress requiring significant policy intervention (Laeven and Valencia, 2012).
  - Currency crisis: nominal depreciation against the US dollar ≥ 30 percent and at least 10 percentage points greater than prior year (Frankel and Rose, 1996).
  - External debt crisis: sovereign default and/or restructuring.
  - Growth collapse: bottom fifth percentile of growth declines (current year vs. average of previous three years) ≈ fall in growth rate of real GDP of about 7.5 percentage points.
- Frequencies (total crisis counts):
  - Banking crises: 58 observations.
  - Currency crises: 64 observations.
  - Debt crises: 25 observations.
  - Growth collapses: 61 observations.
- Regression/probit highlights (selected exact marginal effects and coefficients):
  - Banking crisis propensity:
    - Intermediate aggregate marginal effect ~0.803** in one specification (banking crisis).
    - Within intermediate regimes, crawling arrangements and horizontal bands most crisis-prone (horizontal band coefficients 1.366***, 1.197***; crawling peg/band 0.985**, 0.750*, 0.494**).
    - Managed floats are the least likely and no more likely than pure floats.
    - REER deviation: 3.584*** and 4.157*** (strong positive association).
    - Reserves/GDP protective: -0.026** and -0.082***.
    - Bank foreign liabilities/GDP: 0.013** and 0.015*.
    - Domestic credit expansion: 0.013*.
    - Net financial flows/GDP: 0.023*.
    - Current account balance/GDP: -0.029** (negative association).
    - Observations: 1,063 for banking crises; No. of countries: 50.
  - Currency crisis propensity:
    - Almost five times as likely under intermediate regime than under hard peg, and twice as likely as under a pure float (Table 3, col. [7]); statistical significance varies by specification.
    - Crawling pegs show higher frequency than pure floats; adding controls can render some coefficients insignificant and make hard peg coefficient negative and significant at 10 percent in some specs.
    - Reserves/GDP protective for currency crises as well.
    - Observations: 1,258 for currency crises.
  - Sovereign debt crises:
    - Unconditional likelihood similar under hard pegs and intermediate regimes (~2 percent) and about four times that under pure floats (Table 3, col. [8]).
    - After controlling for overvaluation, reserves, fiscal balance, growth, inflation, no statistically significant differences across finer regimes (Table 11).
    - Fiscal balance/GDP: -0.055*** (associated with lower sovereign crisis probability).
    - Banking crisis indicator strongly associated with sovereign crisis: 0.910***.
    - Observations: up to 1,193; No. of countries: 51.
  - Growth collapses:
    - Unconditionally, growth collapses are far more common under hard pegs: >10 percent of hard peg observations experience sharp decline in real GDP growth (Table 3, col. [9]).
    - Hard pegs, single currency pegs, and basket pegs significantly more prone to growth collapses than managed or pure floats (Table 11, cols. [4]-[5]).
    - Current account balance/GDP: -0.045*** (negative and significant for growth collapse).
    - Reserves/GDP: -0.061**, -0.045*** (protective).
    - Controlling for banking/currency/debt crises does not eliminate association between hard pegs and growth collapses.

### V. Sensitivity and endogeneity checks
- Alternate specifications and robustness:
  - Pooling crisis types: almost all less flexible regimes (except managed floats and basket pegs) significantly more susceptible to crisis than pure floats (Table B2).
  - Results robust to inclusion of trade openness, institutional quality, capital account openness, contagion variables, year effects, and different proxies for bank foreign borrowing.
  - For currency crises, controlling for hyperinflation (annual inflation rate ≥ 40 percent) or lagged banking crisis variable does not materially change results.
- Endogeneity:
  - Use of one-year lag for regime classification to mitigate reverse causality.
  - Empirical observation: in less than a quarter of crisis cases does the de facto regime switch between t-2 and t-1; among those switches, near-equal split between moves toward less flexibility and toward greater flexibility.
  - No robust evidence that vulnerabilities systematically prompt regime switches, except that more overvalued countries tend to switch toward more flexibility.
  - Conclusion: reverse causality is unlikely to drive main findings.

### VI. Binary Recursive Tree (BRT) analysis — where to draw the line for managed floats
- Rationale and primitives used:
  - Managed float classification is ambiguous across schemes; BRT identifies crisis-prone intermediate regimes based on primitives: real exchange rate overvaluation, credit expansion, banking foreign liabilities, share of FX credit, nominal exchange rate flexibility, degree/direction/circumstances of intervention.
- Implementation details:
  - Algorithm: Improved CHAID; variables lagged one period.
  - Dependent variable: 1 if country experiences banking or currency crisis; 0 otherwise.
  - Intervention metric: I = |∆R|/(|∆R|+|∆E|) where ∆R = annual % change in reserves, ∆E = annual % change in NEER; I ranges 0 to 1.
  - Exchange rate flexibility measured by monthly NEER volatility via rolling standard deviation over 6, 12, or 36 months.
- Key tree splits, thresholds, and conditional probabilities:
  - First split: REER overvaluation at 5 percent.
    - REER overvaluation > 5 percent → conditional crisis probability 30 percent.
    - REER overvaluation ≤ 5 percent → conditional crisis probability 4 percent.
  - For overvaluation > 5 percent branch:
    - 3-year cumulative change in domestic credit-to-GDP > 30 percentage points → crisis probability increases more than threefold vs. those without such a boom.
    - Overvaluation > 12 percent → crisis probability 55 percent vs 20 percent for overvaluation ≤ 12 percent.
    - Among highly overvalued (>12 percent): heavy intervention increases crisis likelihood; when overvaluation ≤ 12 percent, nominal exchange rate flexibility matters—less flexible exchange rates ~ four times more likely to have a crisis than more flexible ones.
  - For overvaluation ≤ 5 percent branch:
    - Credit expansion > ~32 percentage points (3-year cumulative) → crisis probability ~15 times higher.
    - Among lower credit expansion countries, more flexible exchange rates show higher crisis likelihood if accompanied by overvaluation: overvalued currencies in this subset ~ six times more likely to experience crisis.
  - Classification accuracy: tree correctly classifies ~94 percent of the sample; correctly classifies 29 percent of crisis observations and 99 percent of noncrisis observations.
- Role of IMF and RR classifications:
  - Neither IMF nor RR classification enters the full tree when overvaluation, intervention, and flexibility are included—these primitives better discriminate crisis vs noncrisis.
  - If overvaluation, intervention, and flexibility are dropped, IMF classification can enter the tree (managed/pure floats vs less flexible), but is a weaker discriminator; RR classification does not enter even with restricted variables.
- Directional intervention refinements:
  - Distinguish intervention that buys FX (countering appreciation) vs intervention that sells FX (defending overvaluation).
  - For currencies >12 percent overvalued:
    - Intervention against the wind (central bank buys FX to limit appreciation) → conditional crisis probability 4 percent.
    - No intervention against the wind → conditional crisis probability 24 percent.
    - Central bank selling FX to defend an overvalued exchange rate → conditional crisis probability 83 percent vs 44 percent when not defending.

### VII. Policy implications and conclusions
- Aggregate conclusions:
  - Free floats are the least vulnerable to crises.
  - Hard pegs display the largest financial and macroeconomic vulnerabilities (external imbalances, overvaluation, foreign liabilities, credit expansion, FX lending) but these tend to materialize as growth collapses rather than banking or currency crises.
  - Intermediate regimes are the most susceptible to banking and currency crises as a class; however, managed floats (IMF de facto) behave much like pure floats with significantly lower risks and fewer crises.
- Practical policy guidance on managed floats:
  - No uni-dimensional rule (e.g., a single volatility threshold) separates safe from risky managed floats.
  - Safer managed floats are characterized by:
    - Monitoring and preventing real exchange rate overvaluation (key thresholds highlighted: 5 percent and 12 percent).
    - Intervening to limit appreciation (buying FX) when overvaluation pressures emerge.
    - Refraining from intervening to defend an overvalued exchange rate (i.e., not selling FX to support an overvalued parity).
  - Complementary policies that reduce vulnerabilities:
    - Controls on capital inflows can lower banking external liabilities and the share of FX lending.
    - Restrictions on FX lending and appropriate FX limits affect bank behavior in ways that can reduce systemic risks.
- Research implication: managed floats can be relatively resilient, but central banks face real-time challenges distinguishing temporary from persistent capital flows and assessing real exchange rate misalignment; more research needed to fully define contours of “safe” managed floats.

*Source: IMF working paper section “5. Binary Recursive Tree: Defend the Currency Intervention,” dataset and analysis covering 50 EMEs over 1980-2011.*

### References..............................................................................................................

### _wp1411 - References...................................................................................................................................... 22

### Appendix Sections
- Appendix A: Exchange Rate Regimes: De Jure and RR Classifications....................................... 38
- Appendix B: Data and Additional Estimation Results .................................................................. 41

### Tables
- 1. Transition Probabilities Matrix for EMEs: IMF’s Aggregate Classification............................. 29
- 2. Transition Probabilities Matrix for EMEs: IMF’s Fine Classification ...................................... 30
- 3. Vulnerabilities and Crisis in EMEs, 19802011 ....................................................................... 30
- 4. Domestic Credit: IMF’s De Facto Classification, 19802011 .................................................. 31
- 5. Bank Foreign Borrowing: IMF’s De Facto Classification, 19802011 .................................... 32
- 6. FX Lending: IMF’s De Facto Classification, 19802011 ......................................................... 33
- 7. Fiscal Balance anad REER Deviation: IMF’s De Facto Classification, 19802011 ................. 34
- 8. Current Account Reversals: IMF’s De Facto Classification, 19802011 ................................. 35
- 9. Crisis Occurrence in EMEs, 19802011 ................................................................................... 35
- 10. Banking and Currency Crises in EMEs, 19802011 ............................................................... 36
- 11. Sovereign Debt Crisis and Growth Collapses in EMEs, 19802011 ...................................... 37
- A1. List of Countries in the Sample .............................................................................................. 39
- A2. Transition Probability Matrix: De Jure Classification, 19802011 ........................................ 39
- A3. Transition Probability Matrix: RR’s Classification, 19802010 ............................................ 40
- B1. Data Description and Sources ................................................................................................. 41
- B2. Robustness Analysis, 19802011 ........................................................................................... 42
- B3. Economic Characteristics and Exchange Rate Regime Switches ........................................... 43
- B4. Crisis Susceptibility: De Jure Classification ........................................................................... 44
- B5. Crisis Susceptibility: RR’s Classification ............................................................................... 45

### Figures
- 1. Distribution of Exchange Rate Regimes in EMEs, 19802011 ................................................ 25
- 2. Current Account Balance in EMEs, 19802011 ....................................................................... 25
- 3. Binary Recursive Tree: Banking or Currency Crisis ................................................................. 26
- 4. Binary Recursive Tree: Against the Wind Intervention ............................................................ 27

*Source: _wp1411 - References..............................................................................................................*

### 5. Binary Recursive Tree: Defend the Currency Intervention ....................................................... 28

### 5. Binary Recursive Tree: Defend the Currency Intervention

### I. Introduction — scope and questions
- Research question: Which exchange rate regimes are most vulnerable to crisis, and why; and where to draw the line between safe floats and risky intermediate regimes.
- Sample: 50 major EMEs over 1980-2011.
- Methodology highlights:
  - Use of IMF’s detailed de facto classification, supplemented with other classifications (IMF de jure; Reinhart and Rogoff, RR).
  - Examination of underlying vulnerabilities (macroeconomic and financial) and crisis frequencies (banking, currency, sovereign debt) plus growth collapses defined relative to historical norms.
  - Decision-theoretic technique: binary recursive tree (BRT) analysis to allow arbitrary thresholds and interactive effects (e.g., exchange rate flexibility; degree/direction/circumstances of FX intervention).

### II. Trends in exchange rate regimes in EMEs
- Three phases identified:
  - Post-1997-98: “hollowing out of the middle” with moves toward free floats.
  - 2004 runup to GFC: increased adoption of intermediate regimes, especially managed floats.
  - GFC and beyond: accelerated move toward intermediate regimes, especially managed floats.
- Transition/steady-state findings (three-way and seven-way classifications):
  - Full-sample steady-state (assuming historical transition probabilities persist): intermediate regimes ~70 percent of EMEs in the long-run; hard pegs ~20 percent.
  - Full-sample finer classification steady-state: managed floats ~31 percent; hard pegs ~20 percent; crawling pegs/bands ~17 percent; single currency pegs ~14 percent; floats ~11 percent.
  - More recent sample (2000-11): roughly equal split between hard pegs and managed floats of about 30 percent each; single currency pegs ~15 percent; pure floats ~10 percent.
- Empirical rejection of bipolar hypothesis as a positive prediction: regimes are persistent but off-diagonal transition probabilities are non-zero; floats are the least persistent with about 20 percent transitions from floats to intermediate regimes and 2 percent to hard pegs.

### III. Financial and macroeconomic vulnerabilities by regime
- Summary of stylized vulnerabilities under less flexible regimes:
  - Loss of exchange rate as adjustment tool → real exchange rate overvaluation → larger imbalances → higher currency crisis risk.
  - Exchange rate guarantees encourage excessive foreign borrowing and FX-denominated lending.
  - Intervention (if not sterilized) can fuel credit expansion; less flexible regimes may reduce fiscal discipline.
- Key empirical observations (raw and regression results):
  - Domestic credit change (3-year cumulative change in private credit-to-GDP):
    - Almost twice as large under hard pegs as under intermediate regimes, and almost four times as large as under floats (Table 3, col. [1]).
    - Change in credit more than twice as large under basket pegs than under single currency pegs, and almost eight times as large as under managed floats.
  - Foreign borrowing by banking system and FX-denominated domestic lending:
    - Both measures are roughly twice as large under hard pegs as under floats (Table 3, cols. [2]-[3]).
    - Regression evidence: foreign borrowing significantly greater under less flexible regimes; hard pegs and single currency pegs significant versus managed floats (Table 5). Hard and basket pegs associated with higher share of FX-denominated lending (Table 6), though effects weaken when controlling for net capital flows and bank foreign liabilities.
  - Policy-relevant controls:
    - Controls on capital inflows associated with significantly lower banking system external liabilities (Table 5, col. [4]) and lower share of FX-denominated domestic bank lending (Table 6, col. [4]).
    - Restrictions on FX-denominated lending reduce the proportion of such loans; open FX-limits increase foreign borrowing by banks.
  - Macroeconomic vulnerabilities:
    - Fiscal deficits are lower under hard pegs than most other regimes (except basket pegs), but differences vs. floats are generally not statistically significant (Table 3, col. [4]; Table 7).
    - Real exchange rate overvaluation: hard pegs and intermediate regimes are associated with significantly greater overvaluation than pure floats; managed floats are not significantly more overvalued than pure floats (Table 7, cols. [4]-[6]).
    - Current account imbalances tend to be larger under hard pegs and intermediate regimes; unconditional reversal probability significantly greater for hard pegs and most intermediate regimes (Table 8).

### IV. Crisis propensity by regime (banking, currency, sovereign debt, growth)
- Crisis definitions:
  - Banking crisis: systemic banking distress requiring significant policy intervention (Laeven and Valencia, 2012).
  - Currency crisis: nominal depreciation against the US dollar ≥ 30 percent and at least 10 percentage points greater than prior year (Frankel and Rose, 1996).
  - External debt crisis: sovereign default and/or restructuring.
  - Growth collapse: bottom fifth percentile of growth declines (current year vs. average of previous three years) ≈ fall in growth rate of real GDP of about 7.5 percentage points.
- Frequencies and relationships:
  - Currency and banking crises most common; sovereign debt crises least prevalent (Table 9).
  - Only about one-third of growth collapses occur with or within three years of a banking or currency crisis; fewer than 10 percent occur in context of a debt crisis.
  - Banks and currency crises show co-movement: roughly one-half of currency or debt crises occur within three years of a banking crisis; 15-30 percent of banking crises occur within three years of debt or currency crisis.
- Regression/probit results:
  - Banking crisis propensity:
    - Intermediate regimes ~ twice as likely as hard pegs and ~ four times as likely as floats (Table 3, col. [6]).
    - Within intermediate regimes, crawling arrangements and horizontal bands most crisis-prone (~7 percent experience a banking crisis); basket pegs also elevated; managed floats are the least likely and no more likely than pure floats.
    - Controlling for vulnerabilities (overvaluation, banking foreign liabilities, credit expansion, net inflows) reduces or eliminates significance of some intermediate regime coefficients (Table 10, col. [3]); vulnerabilities are important channels.
  - Currency crisis propensity:
    - Almost five times as likely under intermediate regime than under hard peg, and twice as likely as under a pure float (Table 3, col. [7]); differences not always statistically significant.
    - Crawling pegs show statistically higher frequency than pure floats; adding controls (overvaluation, reserves, liabilities) makes crawling peg coefficient insignificant and hard peg coefficient negative and significant at 10 percent (Table 10, col. [6]).
    - Interpretation: hard pegs face fewer currency crises than their vulnerabilities imply due to policy discipline and deterrence against speculative attacks.
  - Sovereign debt crises:
    - Unconditional likelihood similar under hard pegs and intermediate regimes (2 percent) and about four times that under pure floats (Table 3, col. [8]).
    - No statistically significant differences across finer regimes once controls (overvaluation, reserves, fiscal balance, growth, inflation) are included (Table 11, cols. [1]-[3]).
  - Growth collapses:
    - Unconditionally, growth collapses are far more common under hard pegs than under intermediate or floating regimes: >10 percent of hard peg observations experience sharp decline in real GDP growth (Table 3, col. [9]).
    - Hard pegs, single currency pegs, and basket pegs significantly more prone to growth collapses than managed or pure floats (Table 11, cols. [4]-[5]).
    - Association holds controlling for trading partner growth, current account, reserves (Table 11, col. [6]) and controlling for banking/currency/debt crises.

### V. Sensitivity and endogeneity checks
- Alternate specifications:
  - Pooling crisis types shows almost all less flexible regimes (except managed floats and basket pegs) significantly more susceptible to crisis than pure floats (Table B2).
  - Results robust to inclusion of additional controls: trade openness, institutional quality, capital account openness, contagion variables, year effects, different proxies for bank foreign borrowing.
  - For currency crises, controlling for hyperinflation (annual inflation rate ≥ 40 percent) or lagged banking crisis variable does not materially change results.
- Endogeneity:
  - Use of one-year lag for regime classification to mitigate reverse causality.
  - Empirical observation: in less than a quarter of crisis cases does the de facto regime switch between t-2 and t-1; among those switches, near-equal split between moves toward less flexibility and toward greater flexibility.
  - No robust evidence that vulnerabilities systematically prompt regime switches, except that more overvalued countries tend to switch toward more flexibility.
  - Conclusion: reverse causality is unlikely to drive main findings.

### VI. Binary Recursive Tree (BRT) analysis — where to draw the line for managed floats
- Rationale: managed float classification is ambiguous across schemes (IMF vs RR); BRT used to identify crisis-prone intermediate regimes based on primitives: real exchange rate overvaluation, credit expansion, banking foreign liabilities, share of FX credit, nominal exchange rate flexibility, degree/direction/circumstances of intervention.
- BRT implementation details:
  - Algorithm: Improved CHAID (chi-squared test for best split); variables lagged one period.
  - Dependent variable: 1 if country experiences banking or currency crisis; 0 otherwise.
  - Intervention metric: I = |∆R|/(|∆R|+|∆E|) where ∆R = annual % change in reserves (reserve flows from BoP), ∆E = annual % change in NEER; I ranges 0 (no intervention) to 1 (full intervention).
  - Exchange rate flexibility: monthly NEER volatility via rolling standard deviation over 6, 12, or 36 months.
- Key BRT thresholds and conditional probabilities (Figure 3):
  - First split: real exchange rate overvaluation at 5 percent.
    - Overvaluation > 5 percent → conditional crisis probability 30 percent.
    - Overvaluation ≤ 5 percent → conditional crisis probability 4 percent.
  - For overvaluation > 5 percent branch:
    - Second split: 3-year cumulative change in domestic credit-to-GDP > 30 percentage points → crisis probability increases more than threefold vs. those without such a boom.
    - Further split: overvaluation > 12 percent → crisis probability 55 percent vs 20 percent for less overvalued.
    - Among highly overvalued (>12 percent): heavy intervention increases crisis likelihood; when overvaluation ≤ 12 percent, nominal exchange rate flexibility matters—less flexible exchange rates ~ four times more likely to have a crisis than more flexible ones.
  - For overvaluation ≤ 5 percent branch:
    - Credit expansion > ~32 percentage points (3-year cumulative) → crisis probability ~15 times higher.
    - Among lower credit expansion countries, more flexible exchange rates show higher crisis likelihood if accompanied by overvaluation: overvalued currencies in this subset ~ six times more likely to experience crisis.
  - Classification accuracy: tree correctly classifies ~94 percent of the sample; 29 percent of crisis observations and 99 percent of noncrisis observations correctly classified.
- Role of IMF and RR classifications in the tree:
  - Neither IMF nor RR classification enters the full tree when overvaluation, intervention, and flexibility are included—these primitives better discriminate crisis vs noncrisis.
  - If overvaluation, intervention, and flexibility are dropped, IMF classification can enter the tree (managed/pure floats vs less flexible), but is a weaker discriminator; RR classification does not enter even with restricted variables.
- Directional intervention refinements (Figures 4 and 5):
  - Distinguish intervention that buys FX (countering overvaluation) vs intervention that sells FX (defending overvaluation).
  - For currencies >12 percent overvalued:
    - Intervention against the wind (central bank buys FX to limit appreciation) → conditional crisis probability 4 percent.
    - No intervention against the wind → conditional crisis probability 24 percent (Figure 4).
    - Central bank selling FX to defend an overvalued exchange rate → conditional crisis probability 83 percent vs 44 percent when not defending (Figure 5).
- Main BRT insight: no simple single metric (e.g., nominal exchange rate flexibility) cleanly separates safe vs risky intermediate regimes. The critical determinants are:
  - Real exchange rate overvaluation (primary discriminator; thresholds 5 percent and 12 percent noted).
  - Credit expansion (3-year cumulative thresholds ~30 and ~32 percentage points).
  - Nature and direction of FX intervention: intervening to limit overvaluation lowers crisis risk; intervening to defend an overvalued parity markedly raises crisis risk.
  - Nominal exchange rate flexibility interacts with the above factors.

### VII. Policy implications and conclusions
- Aggregate conclusions:
  - Free floats are the least vulnerable to crises.
  - Hard pegs display the largest financial and macroeconomic vulnerabilities (external imbalances, overvaluation, foreign liabilities, credit expansion, FX lending) but these tend to materialize as growth collapses rather than banking or currency crises—making the hard end of the bipolar prescription largely illusory.
  - Intermediate regimes are the most susceptible to banking and currency crises as a class; however, managed floats (as defined by the IMF de facto classification) behave much like pure floats with significantly lower risks and fewer crises.
- Practical policy guidance on managed floats:
  - There is no uni-dimensional rule (e.g., a single volatility threshold) for safe managed floating.
  - Safer managed floats are characterized by:
    - Monitoring and preventing real exchange rate overvaluation.
    - Intervening to limit appreciation (buying FX) when overvaluation pressures emerge.
    - Refraining from intervening to defend an overvalued exchange rate (i.e., not selling FX to support an overvalued parity).
  - Complementary policies that reduce vulnerabilities:
    - Controls on capital inflows can lower banking external liabilities and the share of FX lending.
    - Restrictions on FX lending and appropriate FX limits affect bank behavior in ways that can reduce systemic risks.
- Research implication: managed floats can be relatively resilient, but central banks face real-time challenges distinguishing temporary from persistent capital flows and assessing real exchange rate misalignment; more research needed to fully define contours of “safe” managed floats.

*Source: IMF working paper section “5. Binary Recursive Tree: Defend the Currency Intervention,” dataset and analysis covering 50 EMEs over 1980-2011.*

### REFERENCES

### REFERENCES

### Exchange rate regimes — surveys, classifications, and theory
- Anderson, H., 2008, “Exchange Policies before Widespread Floating (1945–89),” mimeo (Washington DC: International Monetary Fund).
- Bubula, A., and I. Ötker, 2002, “The Evolution of Exchange Rate Regimes since 1990: Evidence from De Facto Policies?” IMF Working Paper WP/02/155 (Washington DC: IMF).
- Bubula, A., and I. Ötker, 2003, “Are Pegged and Intermediate Regimes More Crisis-prone?” IMF Working Paper WP/03/223 (Washington DC: International Monetary Fund).
- Calvo, G., and C. Reinhart, 2002, “Fear of Floating,” Quarterly Journal of Economics, 117(2), pp. 379-408.
- Eichengreen, B., 1994, International Monetary Arrangements for the 21st Century (Washington, DC: Brookings Institution).
- Fischer, S., 2001, “Exchange Rate Regimes: Is the Bipolar View Correct?” Journal of Economic Perspectives, 15(2), 3-24.
- Fischer, S., 2008, “Mundell-Fleming Lecture: Exchange Rate Systems, Surveillance, and Advice,” IMF Staff Papers, 55(3), 367-383.
- Frankel, J., 1999, “No Single Currency Regime is Right for All Countries or at All Times,” NBER Working Paper 7338 (Cambridge, MA: National Bureau of Economic Research).
- Frankel, J., S. Schmukler, and L. Serven, 2000, “Verifiability and the Vanishing Intermediate Exchange Rate Regime,” NBER Working Paper 7901 (Cambridge, MA: NBER).
- Ghosh, A., A. Gulde, and Holger Wolf, 2003, Exchange Rate Regimes: Choices and Consequences (Cambridge, MA: MIT Press).
- Ghosh, A., J. Ostry, and C. Tsangarides, 2010, “Exchange Rate Regime and the Stability of the International Monetary System,” IMF Occasional Paper No. 270 (Washington DC: IMF).
- Ghosh, A., M. Qureshi, and C. Tsangarides, 2013, “Is Exchange Rate Regime Really Irrelevant for External Adjustment?” Economic Letters, 118(1), 104-109.
- Klein, M., and J. Shambaugh, 2010, Exchange Rate Regimes in the Modern Era (Cambridge, MA: MIT Press).
- Masson, P., 2000, “Exchange Rate Regime Transitions,” IMF Working Paper WP/00/134 (Washington DC: International Monetary Fund).
- Obstfeld, M., and K. Rogoff, 1995, “The Mirage of Fixed Exchange Rates,” Journal of Economic Perspectives, 9(4), pp. 73-96.
- Reinhart, C., and K. Rogoff, 2004, “The Modern History of Exchange Rate Arrangements: A Reinterpretation,” Quarterly Journal of Economics, 119(1), pp. 1-48.
- Rogoff, R., A. Husain, A. Mody, R. Brooks, and N. Oomes, 2004, “Evolution and Performance of Exchange Rate Regimes,” IMF Occasional Paper 229 (Washington DC: IMF).
- Rose, A., 2011, “Exchange Rate Regimes in the Modern Era: Fixed, Floating, and Flaky,” Journal of Economic Literature, 49(3), pp. 652-672.
- Shambaugh, J., 2004, “The Effects of Fixed Exchange Rates on Monetary Policy,” Quarterly Journal of Economics, 119(1), pp. 301-352.

### Capital flows, adjustment, and currency behavior
- Chinn, M., and S.-J. Wei, 2013, “A Faith-based Initiative Meets the Evidence: Does a Flexible Exchange Rate Regime Really Facilitate Current Account Adjustment?” Review of Economics and Statistics, 95(1), 168-184.
- Goldfajn, I., and R. Valdes, 1999, “The Aftermath of Appreciations,” Quarterly Journal of Economics, 114(1), pp. 229-262.
- Magud, N., C. Reinhart, and E. Vesperoni, 2011, “Capital Inflows, Exchange Rate Flexibility, and Credit Booms,” NBER Working Paper No. 17670 (Cambridge, MA: NBER).
- Montiel, P. and C. Reinhart, 2001, “The Dynamics of Capital Movements to Emerging Economies during the 1990s,” in S. Griffith-Jones, M. Montes, and A. Nasution (eds.), Short-term Capital Flows and Economic Crises (Oxford: Oxford University Press, 2001).
- Rose, A., 2011, “Exchange Rate Regimes in the Modern Era: Fixed, Floating, and Flaky,” Journal of Economic Literature, 49(3), pp. 652-672.
- Herrmann, S., 2009, “Do We Really Know that Flexible Exchange Rates Facilitate Current Account Adjustment? Some New Empirical Evidence for CEE Countries,” Applied Economics Quarterly, 55, pp. 295–312.

### Banking crises, credit booms, and macrofinancial stability
- Angkinand, A., and T. Willett, 2011, “Exchange Rate Regimes and Banking Crises: The Channels of Influence Investigated,” International Journal of Finance and Economics, 16(3), pp. 256-274.
- Backe, P. and C. Wojcik, 2008, “Credit Booms, Monetary Integration, and the Neoclassical Synthesis,” Journal of Banking and Finance, 32(3), pp. 458–470.
- Demirgüç-Kunt, A., and E. Detragiache,1998, “The Determinants of Banking Crises in Developing and Developed Countries,” IMF Staff Papers, 45(1), pp. 81-109.
- Domaç, I., and M. Peria, 2003, “Banking Crises and Exchange Rate Regimes: Is There a Link?” Journal of International Economics, 61(1), pp. 41-72.
- Kaminsky, G., and C. Reinhart, 1999, “The Twin Crises: The Causes of Banking and Balance of Payments Problems,” American Economic Review, 89(3), pp. 473–500.
- Laeven, L., and F. Valencia, 2012, “Systemic Banking Crises Database: An Update,” IMF Working Paper No. WP/12/163 (Washington DC: International Monetary Fund).
- Dell’Ariccia, G., D. Igan, L. Laeven, H. Tong, B. Bakker, and J. Vandenbussche, 2012, “Policies for Macrofinancial Stability: How to Deal with Credit Booms,” IMF Staff Discussion Note SDN/12/06 (Washington DC: International Monetary Fund).
- Ostry, J., A. Ghosh, M. Chamon, and M. Qureshi, 2012, “Tools for Managing Financial Stability Risks,” Journal of International Economics, 88(2), 407-421.

### IMF publications, datasets, and related applied work
- IMF, 2008, Annual Report on Exchange Rate Arrangements and Exchange Rate Restrictions (Washington DC: International Monetary Fund).
- Ghosh, A., B. Joshi, J. Kim, U. Ramakrishnan, A. Thomas, J. Zalduendo, 2008, “IMF Support and Crisis Prevention,” IMF Occasional Paper 262 (Washington DC: IMF).
- Rosenberg, C. and M. Tirpák, 2008, “Determinants of Foreign Currency Borrowing in the New Member States of the EU,” IMF Working Paper WP/08/173 (Washington DC: IMF).

### Empirical methods and categorical analysis
- Kass, G., 1980, “An Exploratory Technique for Investigating Large Quantities of Categorical Data,” Applied Statistics, 29(2), pp. 119–127.

### Fiscal and monetary discipline, and other theoretical contributions
- Frankel, J., and A. Rose, 1996, “Currency Crashes in Emerging Markets: An Empirical Treatment,” Journal of International Economics, Vol. 41, pp. 351-366.
- Tornell, A., and A. Velasco, 2000, “Fixed versus Flexible Exchange Rates: Which Provides More Fiscal Discipline?” Journal of Monetary Economics, 45(2), pp. 399-436.

*Source: _wp1411 - REFERENCES*

### 1. Aggregate classification 2. Fine classification

### _wp1411 - 1. Aggregate classification 2. Fine classification

### Exchange rate regime distributions and transition probabilities
- IMF’s de facto aggregate classification (1980-2011) transition matrix (Fixed / Intermediate / Float):
  - Fixed → Fixed: 0.977, Fixed → Intermediate: 0.015, Fixed → Float: 0.008
  - Intermediate → Fixed: 0.004, Intermediate → Intermediate: 0.965, Intermediate → Float: 0.031
  - Float → Fixed: 0.018, Float → Intermediate: 0.195, Float → Float: 0.787
  - Steady-state regime distribution: Fixed 0.194, Intermediate 0.697, Float 0.108
  - Regime distribution in 2011: Fixed 0.151, Intermediate 0.774, Float 0.075
  - Likelihood ratio tests:
    - LR test statistic (fixed as an absorbing state): 5.931 (p-value=0.084)
    - LR test statistic (float as an absorbing state): 61.934 (p-value=0.000)
    - LR test statistic (fixed and float being a closed set): 61.274 (p-value=0.000)

- IMF’s de facto aggregate classification (2000-2011) transition matrix:
  - Fixed → Fixed: 0.988, Fixed → Intermediate: 0.012, Fixed → Float: 0.000
  - Intermediate → Fixed: 0.005, Intermediate → Intermediate: 0.967, Intermediate → Float: 0.028
  - Float → Fixed: 0.008, Float → Intermediate: 0.140, Float → Float: 0.851
  - Steady-state regime distribution: Fixed 0.311, Intermediate 0.580, Float 0.109
  - Regime distribution in 2011: Fixed 0.151, Intermediate 0.774, Float 0.075
  - Likelihood ratio tests:
    - LR test statistic (fixed as an absorbing state): 1.988 (p-value=0.635)
    - LR test statistic (float as an absorbing state): 33.179 (p-value=0.000)
    - LR test statistic (fixed and float being a closed set): 33.479 (p-value=0.000)

- IMF’s de facto fine classification (1980-2011) transition matrix (rows = from; columns = to):
  - Hard peg → Hard peg: 0.977, → Peg to single currency: 0.000, → Basket peg: 0.000, → Horizontal band: 0.000, → Crawling peg/band: 0.000, → Managed float: 0.015, → Independent float: 0.008
  - Peg to single currency → Hard peg: 0.008, → Peg to single currency: 0.751, → Basket peg: 0.016, → Horizontal band: 0.012, → Crawling peg/band: 0.080, → Managed float: 0.120, → Independent float: 0.012
  - Basket peg → Hard peg: 0.000, → Peg to single currency: 0.043, → Basket peg: 0.871, → Horizontal band: 0.032, → Crawling peg/band: 0.022, → Managed float: 0.022, → Independent float: 0.011
  - Horizontal band → Hard peg: 0.014, → Peg to single currency: 0.056, → Basket peg: 0.000, → Horizontal band: 0.732, → Crawling peg/band: 0.070, → Managed float: 0.113, → Independent float: 0.014
  - Crawling peg/band → Hard peg: 0.000, → Peg to single currency: 0.055, → Basket peg: 0.003, → Horizontal band: 0.010, → Crawling peg/band: 0.846, → Managed float: 0.048, → Independent float: 0.039
  - Managed float → Hard peg: 0.002, → Peg to single currency: 0.067, → Basket peg: 0.007, → Horizontal band: 0.015, → Crawling peg/band: 0.037, → Managed float: 0.827, → Independent float: 0.044
  - Independent float → Hard peg: 0.018, → Peg to single currency: 0.018, → Basket peg: 0.000, → Horizontal band: 0.000, → Crawling peg/band: 0.000, → Managed float: 0.177, → Independent float: 0.787
  - Steady-state distribution (fine): Hard peg 0.195, Peg to single currency 0.142, Basket peg 0.040, Horizontal band 0.034, Crawling peg/band 0.169, Managed float 0.307, Independent float 0.114
  - Regime distribution in 2011 (fine): Hard peg 0.137, Peg to single currency 0.196, Basket peg 0.039, Horizontal band 0.000, Crawling peg/band 0.137, Managed float 0.412, Independent float 0.078

- IMF’s de jure classification (1980-2011) aggregate transition matrix:
  - Fixed → Fixed: 0.970, Fixed → Intermediate: 0.008, Fixed → Float: 0.023
  - Intermediate → Fixed: 0.004, Intermediate → Intermediate: 0.959, Intermediate → Float: 0.037
  - Float → Fixed: 0.008, Float → Intermediate: 0.067, Float → Float: 0.925
  - Steady-state regime distribution (de jure aggregate): Fixed 0.157, Intermediate 0.534, Float 0.309
  - Regime distribution in 2011 (de jure): Fixed 0.151, Intermediate 0.585, Float 0.264

- IMF’s de jure fine classification (1980-2011) steady-state distribution and 2011 distribution (selected):
  - Steady-state (fine de jure): Hard peg 0.153, Peg to single currency 0.031, Basket peg 0.025, Horizontal band 0.014, Crawling peg/band 0.030, Managed float 0.424, Independent float 0.323
  - Regime distribution in 2011 (fine de jure): Hard peg 0.137, Peg to single currency 0.039, Basket peg 0.039, Horizontal band 0.000, Crawling peg/band 0.020, Managed float 0.490, Independent float 0.275

### Current account balances by regime (EMEs, 1980-2011)
- Figure note: Panel (a) aggregate shows average current account deficits and surpluses by aggregate regimes:
  - Fixed average deficit: -8 percent of GDP; Intermediate average deficit: -6 percent of GDP; Float average deficit: -4 percent of GDP.
  - Fixed average surplus: 2.5 percent of GDP; Intermediate average surplus: 4 percent of GDP; Float average surplus: 2.5 percent of GDP.

### Vulnerabilities and crisis indicators by regime (selected statistics, 1980-2011)
- Summary table columns (bank, currency, debt, growth) and macro/financial vulnerability indicators. Selected entries:
  - Hard pegs: Credit boom 6.1, Foreign borrowing 14.3, FX lending 8.9, Fiscal balance -2.7, REER deviation 0.3, Bank 3.0, Currency 1.0, Debt Growth 2.0, Total no. of crisis observations 10.5
  - Intermediate (aggregate): Credit boom 2.4, Foreign borrowing 9.4, FX lending 36.1, Fiscal balance -3.6, REER deviation 0.2, Bank 4.7, Currency 5.2, Debt Growth 1.9, Total no. of crisis observations 4.4
  - Peg to single currency: Credit boom 3.5, Foreign borrowing 12.3, FX lending 34.9, Fiscal balance -4.6, REER deviation 0.9, Bank 3.6, Currency 5.2, Debt Growth 2.8, Total no. of crisis observations 6.9
  - Managed float: Credit boom 1.2, Foreign borrowing 8.0, FX lending 35.4, Fiscal balance -3.5, REER deviation -0.7, Bank 2.7, Currency 4.9, Debt Growth 1.5, Total no. of crisis observations 3.3
  - Independent float: Credit boom 0.8, Foreign borrowing 7.3, FX lending 29.4, Fiscal balance -3.2, REER deviation -1.6, Bank 1.2, Currency 2.4, Debt Growth 0.6, Total no. of crisis observations 3.8

### Domestic credit (dependent variable: cumulative change in private sector credit to GDP ratio over 3 years)
- Regression highlights (samples and coefficients reported with clustered standard errors):
  - Hard peg coefficients across specifications include: 4.332*, 4.345, 2.614, 4.086*, 4.023*, 0.736 (standard errors in parentheses).
  - Intermediate aggregate: 2.432 and 3.104** in reported specs.
  - Peg to single currency: 2.588, 2.372, 5.214***, 0.537.
  - Basket peg: 10.263**, 9.201**, 8.843***, 5.367***.
  - Horizontal band: 5.051**, 4.913**, 3.041**, 1.907.
  - Real GDP growth: 0.418*** and 0.288*** in specs.
  - Initial domestic credit/GDP: -0.162*** and 0.111*** reported (note sample restrictions for positive changes).
  - Observations: 1,010 (full sample specs) and 646 (restricted); No. of countries: 51; R-squared up to 0.391.
  - Notes: Reference category is free float. ***, **, * indicate significance at 1, 5 and 10 percent levels, respectively.

### Bank foreign borrowing (dependent variable: bank foreign liabilities to GDP, percent)
- Selected coefficients and diagnostics:
  - Hard peg: 7.795**, 7.828**, 6.628**, 5.376*, 4.881*, 3.034.
  - Peg to single currency: 8.277**, 6.062, 9.176*, 8.206*, 10.001** (various specs).
  - Basket peg and other fine regimes reported with varied coefficients.
  - Domestic credit/GDP: 0.156***, 0.147***, 0.153***, 0.124** across specs.
  - Open FX position limits: -5.223**, -4.058* in some specs.
  - Real GDP per capita (log): 5.708***, 5.964***, 3.542**, 4.239**, 4.492**, 4.928**.
  - Observations: up to 1,237; No. of countries: 50; R-squared up to 0.437.

### FX lending (dependent variable: bank lending in FX to total bank lending, percent; sample 1995-2011)
- Key reported coefficients:
  - Hard peg: 20.379**, 20.505**, 9.645, 4.154, 2.360, 2.351.
  - Basket peg: 14.842**, -5.925, -10.436, -13.978*, -13.371* (different specs).
  - Net financial flows/GDP: 0.671*** consistently.
  - Bank foreign liabilities/GDP: 0.716***, 0.676***, 0.686***, 0.668***.
  - Restrictions on FX lending: -13.805**, -11.362* in some specifications.
  - Real GDP per capita (log): negative and significant in several specs: -9.733***, -10.029***, -10.268***, -9.769***.
  - Observations: 571 (overall), subsets 548–516; No. of countries up to 44; R-squared up to 0.571.

### Fiscal balance and REER deviation
- Fiscal balance regressions (general government fiscal balance to GDP in percent) and REER deviation (in percent):
  - Hard peg fiscal balance coefficients: 1.328, 1.285, 0.984, 1.743**, 1.744**, 1.637** across specs.
  - Peg to single currency: -0.779, -0.946, 2.386**, 2.319** (in REER specs).
  - Basket peg: 3.035**, 2.585*, 1.837**, 1.693**.
  - Crawling peg/band: 0.398, 0.167, 2.741***, 2.295**.
  - Terms of trade change: 0.055*** in one REER specification.
  - Net financial flows/GDP: 0.070* in one fiscal spec.
  - Observations: up to 1,277; No. of countries up to 52; R-squared up to 0.175 (fiscal) and 0.094 (REER).

### Current account reversals (1980-2011)
- Reversals defined per Freund (2005). Prior balance indicates maximum surplus/deficit prior to reversal (percent of GDP); reversal probability is frequency of reversal as proportion of regime observations.
- Surplus prior balances and reversal probabilities (selected):
  - Fixed prior balance: 4.0, Upper quartile 4.0, Lower quartile 4.0, Reversal probability 0.9
  - Intermediate prior balance: 10.3, Upper quartile 11.9, Lower quartile 5.3, Reversal probability 3.2
  - Managed float prior balance: 12.6, Upper quartile 17.3, Lower quartile 5.5, Reversal probability 3.7
  - Float prior balance: 6.9, Upper quartile 11.9, Lower quartile 4.2, Reversal probability 1.8
- Deficit prior balances and reversal probabilities (selected):
  - Fixed prior deficit: -15.9, Upper quartile -7.1, Lower quartile -19.2, Reversal probability 9.3*** (statistical significance vs independent float)
  - Intermediate prior deficit: -11.5, Upper quartile -6.4, Lower quartile -13.1, Reversal probability 7.3***
  - Float prior deficit: -6.7, Upper quartile -4.2, Lower quartile -9.9, Reversal probability 1.8

### Crisis occurrence in EMEs (1980-2011)
- Table reports percentage of crisis observations preceded or accompanied by other crisis types, plus total crisis observation counts for each type. Selected totals:
  - Total number of banking crisis observations: 58
  - Total number of currency crisis observations: 64
  - Total number of debt crisis observations: 25
  - Total number of growth collapse observations: 61
- Cross-occurrence fragments (as presented):
  - Bank... 27.6, 13.8, 19.0, total 58
  - Currency 43.8,... 21.9, 29.7, total 64
  - Debt 56.0, 68.0,... 12.0, total 25
  - Growth 31.1, 32.8, 8.2, 11.5, total 61

### Banking and currency crises (probit regressions, 1980-2011)
- Selected coefficient highlights (marginal effects / coefficients with clustered SEs):
  - Hard peg effects on banking crisis: 0.473, 0.502, 0.243; on currency crisis: -0.390, -0.389, -0.696*.
  - Intermediate aggregate: 0.803** (banking crisis) and 0.351 (other spec).
  - Horizontal band: 1.366***, 1.197*** in banking crisis specs.
  - Crawling peg/band: 0.985**, 0.750*, 0.494** in banking crisis specs.
  - REER deviation: 3.584*** and 4.157*** (strong positive association).
  - Reserves/GDP: -0.026** and -0.082*** (protective effect).
  - Bank foreign liabilities/GDP: 0.013** and 0.015* (positive association).
  - Domestic credit expansion: 0.013* (positive).
  - Net financial flows/GDP: 0.023* (positive).
  - Current account balance/GDP: -0.029** (negative association with banking crises).
  - Observations: 1,063 for banking crises; 1,258 for currency crises; No. of countries: 50.

### Sovereign debt crises and growth collapses (probit regressions, 1980-2011)
- Selected coefficients and significance:
  - Hard peg coefficients for debt crisis: 0.480, 0.487, 0.254; for growth collapse: 0.410**, 0.416**, 0.447*.
  - Peg to single currency: 0.341, -0.057, 0.563**, 0.821*** (some specs).
  - Basket peg: 0.327, 0.022, 0.698**, 0.841**.
  - Current account balance/GDP: -0.006 (debt), -0.045*** (growth collapse) — negative and significant for growth collapse.
  - Reserves/GDP: -0.061**, -0.045*** (protective).
  - Fiscal balance/GDP: -0.055*** (associated with lower sovereign crisis probability).
  - Real GDP growth: -0.062*** (negatively associated with crises).
  - Advanced trading partner growth: -0.279***.
  - Banking crisis indicator: 0.910*** (associated with higher sovereign crisis probability).
  - Observations: up to 1,193; No. of countries: 51.

### Appendix A — de jure and RR classifications; country list (sample)
- Figures and tables provided for distribution of exchange rate regimes under IMF’s de jure classification (1980-2011) and RR’s de facto classification (1980-2010).
- Country sample (selected listing from Table A1): Albania, Algeria, Argentina, Armenia, Belarus, Bosnia & Herzegovina, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Croatia, Czech Republic, Dominican Republic, Ecuador, Egypt, El Salvador, Estonia, Georgia, Guatemala, Hungary, India, Indonesia, Jamaica, Jordan, Kazakhstan, Korea Republic of, Latvia, Lebanon, Lithuania, Macedonia FYR, Malaysia, Mexico, Morocco, Panama, Peru, Philippines, Poland, Romania, Russian Federation, Serbia Republic of, Slovak Republic, South Africa, Sri Lanka, Thailand, Tunisia, Turkey, Ukraine, Uruguay, Venezuela, Vietnam, Pakistan.

_Italic: Source — _wp1411 - 1. Aggregate classification 2. Fine classification (IMF PDF content provided)._

### 3. Aggregate classification

### 3. Aggregate classification

### Aggregate classification: definitions and note
- Fixed includes: no separate legal tender, pre-announced peg and currency board.
- Intermediate includes: pre-announced horizontal band ≤ +/-2%, de facto peg, pre-announced crawling peg, pre-announced crawling ≤ +/-2%, de facto crawling peg, de facto crawling band ≤ +/-2%; pre-announced crawling band ≥ +/-5%, de facto crawling band ≤ +/-5%, moving band ≤ +/-2%, and managed float.
- Float includes: free float.
- The categories of freely falling and dual market where parallel data is missing are excluded from the computations.

### Aggregate classification transition matrix (rows = origin, columns = destination)
- Row labels: Fixed / Intermediate / Float
- Column labels: Fixed / Intermediate / Float
- Transition matrix:
  - Fixed → Fixed: 0.95
  - Fixed → Intermediate: 0.05
  - Fixed → Float: 0.00
  - Intermediate → Fixed: 0.01
  - Intermediate → Intermediate: 0.99
  - Intermediate → Float: 0.00
  - Float → Fixed: 0.04
  - Float → Intermediate: 0.07
  - Float → Float: 0.89

### Aggregate steady-state and 2010 regime distributions (aggregate)
- Steady-state regime distribution:
  - Fixed: 0.17
  - Intermediate: 0.82
  - Float: 0.01
- Regime distribution in 2010:
  - Fixed: 0.19
  - Intermediate: 0.79
  - Float: 0.02

### Fine classification: categories (note)
- The categories of freely falling and dual market where parallel data is missing are excluded from the computations.

### Fine-classification transition matrix (Ilzetzki/Reinhart/Rogoff fine categories)
- Column/Row order (fine categories):
  1. No separate legal tender/Pre announced peg/Currency board
  2. De facto peg
  3. Pre announced crawling peg
  4. Pre announced crawling band ≤ +/-2%
  5. De facto crawling peg
  6. De facto crawling band ≤ +/-2%
  7. Pre announced crawling band ≥ +/-2%
  8. De facto crawling band ≤ +/-5%
  9. Moving band ≤ +/-2%
  10. Managed floating
  11. Free float
- Transition probabilities (rows = origin category 1..11; entries correspond to columns 1..11):
  - Origin 1 (No separate legal tender/Pre announced peg/Currency board):
    - 0.953 0.006 0.000 0.000 0.000 0.012 0.012 0.000 0.012 0.000 0.006 0.000
  - Origin 2 (De facto peg):
    - 0.009 0.889 0.000 0.000 0.000 0.028 0.009 0.009 0.028 0.009 0.019 0.000
  - Origin 3 (Pre announced crawling peg):
    - 0.000 0.000 0.200 0.800 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
  - Origin 4 (Pre announced crawling band ≤ +/-2%):
    - 0.053 0.000 0.000 0.842 0.053 0.000 0.053 0.000 0.053 0.000 0.000 0.000
  - Origin 5 (De facto crawling peg):
    - 0.013 0.019 0.000 0.013 0.906 0.050 0.000 0.000 0.000 0.000 0.000 0.000
  - Origin 6 (De facto crawling band ≤ +/-2%):
    - 0.011 0.011 0.004 0.000 0.040 0.879 0.004 0.029 0.004 0.015 0.004 0.000
  - Origin 7 (Pre announced crawling band ≥ +/-2%):
    - 0.000 0.000 0.000 0.000 0.125 0.125 0.500 0.250 0.000 0.000 0.000 0.000
  - Origin 8 (De facto crawling band ≤ +/-5%):
    - 0.000 0.024 0.000 0.000 0.006 0.047 0.012 0.882 0.000 0.030 0.000 0.000
  - Origin 9 (Moving band ≤ +/-2%):
    - 0.059 0.059 0.000 0.000 0.000 0.000 0.000 0.000 0.882 0.000 0.000 0.000
  - Origin 10 (Managed floating):
    - 0.000 0.008 0.000 0.000 0.000 0.008 0.033 0.000 0.025 0.008 0.917 0.000
  - Origin 11 (Freely floating):
    - 0.037 0.000 0.000 0.000 0.000 0.000 0.000 0.074 0.000 0.000 0.000 0.889

(Note: matrix entries are presented as in the source; trailing zeros shown where present.)

### Fine-classification steady-state and 2010 distributions
- Steady-state distribution (categories 1..11):
  - 0.173 0.119 0.004 0.014 0.182 0.200 0.009 0.143 0.024 0.125 0.007
- Regime distribution in 2010 (categories 1..11):
  - 0.192 0.096 0.000 0.000 0.173 0.135 0.000 0.231 0.019 0.135 0.019

### Switch characteristics: averages in year before regime switch
- Statistics reflect average in year before the exchange rate regime switch.
- Switch to less flexible regimes vs Switch to more flexible regimes (averages):
  - Net capital flows/GDP (in pct.): 2.2 vs 3.3
  - FX loans/total loans (in pct.): 37.5 vs 38.9
  - Domestic credit change (in ppt.): 2.6 vs 1.7
  - Current acct. balance/GDP (in pct.): -3.9 vs -4.4
  - REER overvaluation (in pct.): -2.2 vs 2.0**
  - Real GDP growth (in pct.): 2.0 vs 1.8
  - Institutional quality index: 0.6 vs 0.6
  - Capital account openness index: -0.1 vs -0.4
  - Trade openness (in pct.): 73.1 vs 73.7
- Note: ** indicates that the difference in means between those switching toward more flexible regimes and those toward less flexible regimes is statistically significant at the 5 percent level.

*Source: IMF Working Paper content in "3. Aggregate classification" (Appendix tables and notes).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1411.pdf_