## The countries using preemptive measures (treatment group) are therefore defined differently

## Source details

**Canonical URL:** [The countries using preemptive measures (treatment group) are therefore defined differently](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022003-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022003-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022003-print-pdf.pdf.json)

---

### Definition of preemptive policy measures
- For each risk-off episode (Taper Tantrum and COVID shocks) five measures of preemptive policy implementation are created: Domestic MPMs, CFMs on inflows, CFMs on outflows, CFMs (without distinguishing inflows from outflows) and CFM/MPMs.
- A preemptive policy measure is defined by both:
  - (a) there is a net tightening action in at least one instrument; and
  - (b) the total number of tightening actions (across instruments) exceeds the total number of easing actions.
- Preemptive policies are identified looking at the period preceding the shock (the “preemptive” period). Dummies for policies “during the shock” are analogously defined but referenced to the quarter in which the shock takes place rather than the preceding period.
- Treatment group membership changes across shocks because the preemptive period is defined relative to each shock (e.g., five years before each shock in the main specification).

### Descriptive statistics and patterns of usage (1996–2019)
- Sample: 56 emerging markets.
- Usage (percent of the 56 Emerging Markets that used policies at least once in the sample period):
  - More than 80 percent implemented domestically-oriented prudential measures (MPM_domestic) at least once.
  - Over 90 percent implemented CFM/MPMs at least once.
  - CFMs (inflows and outflows) were used by approximately one-third of the sample for both inflows and outflows.
- Time-series coding: in iMAPP each tightening action in a country-quarter = "1", each easing action = "-1", no action = "0". Average country-quarter values: positive = net tightening, negative = net easing.
- Time-series observations:
  - Domestic MPMs and CFM/MPMs generally move in the same direction.
  - Average usage of MPMs has trended up since the global financial crisis.
  - Since 2008, EMs on average have largely undertaken net tightening MPMs.
- Notable comparisons:
  - The actual FXI variable (percent of GDP) is materially smaller than the stock of FX reserves (percent of GDP).
- Chart- and sample-related specifics:
  - Taper Tantrum occurred in the second quarter of 2013; illustrative preemptive window considered first quarter of 2008 through first quarter of 2013 for that shock.

### Identification and the UIP outcome variable
- Main outcome: the UIP wedge (UIP deviation) over a 12-month horizon.
- UIP deviation formulation (log form preserved):
  - λt ≡ ̃it − ̃i∗t + ̃st − ̃st+k e
  - ̃it = log(1 + it), ̃i∗t = log(1 + i∗t), ̃st = log(St), ̃st+k e = log(Et(St+k))
- Interpretation:
  - λt = 0 implies UIP holds in expectations.
  - λt > 0 implies expected profitable returns from shorting the U.S. dollar and going long the domestic currency.
- Empirical strategy:
  - Difference-in-differences regression with country fixed effects αc and time (month) fixed effects ωt:
    - λc,t = αc + ωt + β Preemptive Policyc × Risk-Off Shockt + εc,t
  - β captures the differential effect of the shock on countries with versus without preemptive policies.
- Empirical observation:
  - Persistent UIP premia in EMs with average expected excess return of 3 percent for the average country.

### Empirical findings — Taper Tantrum (May 2013) and COVID-19 shocks
- Treatment definition timing:
  - Main preemptive window: 5 years prior to the shock (used in main tables).
  - For COVID-19 robustness, a 12-year window starting in 2008 is also used.
- Taper Tantrum benchmark results (Table 3 summary):
  - Preemptive CFM/MPMs and preemptive CFMs on inflows have negative effects on the UIP premium during the shock (columns (2) and (4)).
  - Preemptive domestic MPMs and preemptive CFMs on outflows have positive effects on the UIP premium during the shock.
  - When using all policies together (column (6)), the above pattern holds.
  - Economic magnitude: the UIP premium is 0.03-0.06 percentage points lower in countries with preemptive policies (CFM/MPM and CFM on inflows). This represents a 30 percent lower external finance premium in those countries relative to the average premium.
- Interpretation of signs:
  - Negative effect of preemptive inflow CFMs and CFM/MPMs: these measures slow down capital inflows by nonresidents, lower FX debt accumulation, and thereby reduce external finance premia during shocks.
  - Positive effect of preemptive outflow CFMs: may reflect a “reputation effect”—preemptive outflow restrictions (often applied to residents) signal potential future constraints and lead nonresident investors to demand higher premia.
- Robustness to country×year fixed effects:
  - Adding country×year effects weakens significance on preemptive CFM/MPM but the negative effect of preemptive CFMs on inflows remains.
- COVID-19 shock (Table 4 summary):
  - Using the five-year window: no effect from domestic MPMs.
  - Using the 12-year window: all preemptive policies negatively affect external finance premia during COVID-19, except preemptive CFMs on outflows which again show a positive impact.
  - Interpretation: since COVID-19 is both an external and domestic shock, preemptive domestic MPMs that rein in credit growth before the shock helped reduce external finance premia during the shock.
  - Economic impact magnitudes for COVID-19 are similar to those in the Taper Tantrum results.

### Evidence on FX-denominated debt and portfolio debt flows (treatment group dynamics)
- For the Taper Tantrum treatment group (median treatment country):
  - FX/LC debt ratio declined from 15 to 5.
  - FX / total debt declined from almost 90 percent to 76 percent.
  - Size of FX debt averages around 12 percent of GDP in the treatment group.
- Example comparison:
  - Brazil (preemptive CFMs on inflows such as IOF taxes) versus Mexico (no CFMs at time, but domestic MPMs): during the Taper Tantrum, Brazil experienced a much lower UIP deviation relative to Mexico despite the same exogenous shock.

### Granular preemptive measures (disaggregation of CFMs by instrument)
- CFMs on inflows and CFMs on outflows are disaggregated into nine instruments (Binici et al. (2020) disaggregation).
- Preemptive CFMs on inflows (and outflows) are defined as a net tightening in each individual instrument in a quarter.
- Limitation: certain instruments have limited observations and are dropped where too few observations exist.

Main granular empirical findings (Table 6)
- Across instruments used as CFMs on inflows:
  - Estimated coefficients on all but one instrument are negative, reinforcing that CFMs on inflows on average lower UIP deviations.
  - Statistically significant negative coefficients observed for:
    - reserve requirements (column 6)
    - Other instruments (column 9)
  - Suggestive interpretation: aggregate inflow-CFM results may be driven by a subset of instruments.
- Across instruments used as CFMs on outflows:
  - Every instrument shown bears both a positive and statistically significant coefficient, including approval requirements, bans, limits, surrender and repatriation requirements, and Other instruments.
  - Interpretation: consistent evidence across instruments that preemptive outflow CFMs raise UIP premia, indicative of a reputational cost from restricting outflows.
- Consistency:
  - Estimated coefficients on Domestic MPMs and CFM/MPMs are virtually identical across granular specifications and match Table 3 estimates.

### Threats to identification and robustness checks
- Main identification threat: omitted variables at the country-month level correlated with the key interaction (preemptive policy × global shock), notably contemporaneous policies used in response to the shock.
- Primary contemporaneous policies considered:
  - FX intervention (FXI)
  - Monetary policy actions
  - Easing of existing inflow CFMs and tightening of outflow CFMs during the shock
- Controls implemented:
  - FXI reserves to GDP and manually collected actual FX intervention data (focused on the TT shock)
  - Monetary policy during the shock
  - Easing/tightening of CFMs during the shock
- Key robustness results (Table 7):
  - The negative effect of preemptive CFMs (and CFM/MPMs) on inflows on UIP premia remains after controlling for FXI and monetary policy.
  - FX reserves and actual FXI employed during the shock have a negative effect on UIP premia.
  - Monetary policy tightening has a positive impact on UIP premia (consistent with Kalemli-Ozcan (2019)): tightening can be ineffective or counterproductive by raising external finance premia.
  - Only around 10 percent of countries tightened monetary policy during Taper Tantrum, but large tightenings in some countries could drive the observed positive effect.
  - Loosening of preexisting CFMs on inflows and/or tightening CFMs on outflows during the shock has a positive impact on UIP premia, consistent with reputational costs from tightening outflow CFMs.
  - Policy implication: employing preemptive measures to slow foreign capital inflows and reduce FX debt accumulation is more effective at reducing financial stress than tightening outflow CFMs during the shock.

### Exchange rate volatility and depreciation
- Standardization and measures:
  - Each bilateral exchange rate standardized to set its value in 2008 January at 100.
  - Volatility computed over 8-month rolling windows on this indexed series; log change in the index used as realized depreciation.
- Volatility results (Table 8 column (2)):
  - Preemptive CFM on inflows: estimated coefficient about -2, which is half of the sample mean volatility of the indexed exchange rate.
  - Preemptive CFM/MPMs: statistically insignificant on volatility.
  - Preemptive domestic MPMs: contributed to lowering volatility but impacts are economically small and sometimes statistically insignificant.
  - Preemptive CFMs on outflows: associated with an increase in exchange rate volatility of about half the mean volatility (consistent with foreigners perceiving higher risk).
- Effects of policies used during the shock:
  - Use of CFMs during the shock increases exchange rate volatility.
  - Use of monetary policy during the shock raises exchange rate volatility.
  - FXI during the shock has no significant effect on exchange rate volatility.
- Exchange rate level (depreciation/appreciation):
  - FXI has a statistically significant impact in appreciating the local currency during the shock, but the estimated impact is extremely small.
  - Preemptive outflow CFMs lead to depreciation during the shock (reflecting reputation effects and measurement issues).
- Overall inference:
  - Preemptive CFM/MPMs and inflow CFMs reduce exchange rate-related stress during risk-off shocks and thereby reduce external finance premia.

### Placebo shocks
- Purpose: rule out that countries with preemptive policies are intrinsically different in ways that lower UIP deviations in normal periods.
- Method:
  - Constructed dummy variables for “placebo shocks” (periods not global risk-off shocks) between 2009 and 2017.
  - Retained preemptive period as the preceding 5 years before the placebo shocks.
  - For earlier placebo shocks used Aizenman-Pasricha CFM data when other CFM data coverage was incomplete.
- Results (Table 9):
  - Preemptive CFMs on inflows have no negative impact on UIP deviations during placebo shocks other than the 2015 shock.
  - Interpretation: the 2015 episode likely functioned as a global risk-off shock in practice (e.g., Chinese stock market turbulence).
  - Preemptive CFMs on outflows are positive and significant in certain placebo cases, consistent with the finding that such measures raise investors’ demanded risk premia.
  - Conclusion: placebo tests support that main results are not spurious and that preemptive policies lower UIP deviations during genuine risk-off shocks.

### Heterogeneity by pre-shock FX-denominated debt
- Theoretical prediction: higher pre-shock FX-denominated debt amplifies the beneficial impact of preemptive CFMs/CFM+MPMs in lowering UIP deviations after a shock.
- Empirical setup:
  - Triple difference-in-differences regression interacting preemptive policy indicators with average FX-denominated debt of households and the non-financial sector (percent of GDP) in the 5 years prior to the shock.
- Data limitations:
  - FX-denominated debt from BIS available for only 8 countries (vs 43 in baseline), reducing sample size by about three-fourths.
  - FX-denominated data available only since 2013, so test applies to the Covid-19 shock but not Taper Tantrum.
- Findings (Table 10):
  - Sum of estimated coefficients of the triple interaction with shock × FX-denominated debt is negative.
  - Evaluated at the sample mean of FX-denominated debt (0.15):
    - For preemptive MPM policies the impact is -0.006.
    - For preemptive CFM/MPM policies the impact is -0.07.
  - Interpretation: countries with higher than average FX-denominated debt would experience even more negative UIP deviations; countries with lower than average FX-denominated debt would experience smaller, potentially positive, deviations for very low levels of FX-denominated debt.

### Policy implications and synthesis
- Empirical support for theoretical predictions: preemptive CFMs and MPMs lower UIP deviations and exchange rate volatility during global risk-off shocks, particularly when targeted at inflows.
- Preemptive policies help limit buildup of FX debt and currency mismatches, preserving access to international markets during stress.
- Preemptive policies reduce the need for monetary policy to counteract demand shortfalls after shocks and reduce the need to defend the exchange rate, thereby allowing monetary policy to focus on domestic objectives.
- Cautionary findings:
  - Preemptive outflow CFMs tend to raise UIP premia and exchange rate volatility (reputational costs).
  - Using CFMs or monetary policy reactively during shocks can be destabilizing.
- Net takeaway: targeted preemptive measures on inflows and CFM/MPM combinations can be effective ex ante tools to reduce financial vulnerabilities and external finance premia during global risk-off episodes.

### Key summary statistics (selected figures from Table 1)
- UIP premium t: (Mean) 0.03; (Standard Deviation) 0.16; (Minimum) -5.04; (Maximum) 1.9
- FXI actual t: (Mean) 0.0008; (Standard Deviation) 0.004; (Minimum) -0.03; (Maximum) 0.05
- FXI reserves t: (Mean) 17.44; (Standard Deviation) 11.50; (Minimum) 0.74; (Maximum) 85.94
- Preemptive MPM T T: (Mean) 0.75; (Standard Deviation) 0.43; (Minimum) 0; (Maximum) 1
- Preemptive CFM/MPM T T: (Mean) 0.69; (Standard Deviation) 0.46; (Minimum) 0; (Maximum) 1
- Preemptive CFM inflows T T: (Mean) 0.24; (Standard Deviation) 0.46; (Minimum) 0; (Maximum) 1
- Preemptive CFM outflows T T: (Mean) 0.29; (Standard Deviation) 0.46; (Minimum) 0; (Maximum) 1
- Monetary Policy t: (Mean) 0.07; (Standard Deviation) 0.05; (Minimum) 0.001; (Maximum) 0.7
- FX debt: (Mean) 0.16; (Standard Deviation) 0.085; (Minimum) 0.069; (Maximum) 0.32

*Source: wpiea2022003-print-pdf — Section 6.2 and surrounding sections*

### 0.   The  countries  using  preemptive  measures  (treatment  group)  are  therefore  defined  differently

### 0.   The  countries  using  preemptive  measures  (treatment  group)  are  therefore  defined  differently

### Definition of preemptive policy measures
- For each risk-off episode (Taper Tantrum and COVID shocks) the authors create five measures of preemptive policy implementation: Domestic MPMs, CFMs on inflows, CFMs on outflows, CFMs (without distinguishing inflows from outflows) and CFM/MPMs.
- A preemptive policy measure is defined in two complementary ways:
  - (a) there is a net tightening action in at least one instrument; and
  - (b) the total number of tightening actions (across instruments) exceeds the total number of easing actions.
- Preemptive policies are identified looking at the period preceding the shock (the “preemptive” period). Dummies for policies “during the shock” are analogously defined but referenced to the quarter in which the shock takes place rather than the preceding period.
- The treatment group membership changes across shocks because the preemptive period is defined relative to each shock (e.g., five years before each shock in the main specification).

### Descriptive statistics and patterns of usage (1996–2019)
- Sample: 56 emerging markets.
- Usage (percent of the 56 Emerging Markets that used policies at least once in the sample period):
  - More than 80 percent implemented domestically-oriented prudential measures (MPM_domestic) at least once.
  - Over 90 percent implemented CFM/MPMs at least once.
  - CFMs (inflows and outflows) were used by approximately one-third of the sample for both inflows and outflows.
- Time-series coding: in iMAPP each tightening action in a country-quarter = "1", each easing action = "-1", no action = "0". Average country-quarter values: positive = net tightening, negative = net easing.
- Time-series observations:
  - Domestic MPMs and CFM/MPMs generally move in the same direction.
  - Average usage of MPMs has trended up since the global financial crisis.
  - Since 2008, EMs on average have largely undertaken net tightening MPMs.
- Notable comparisons of balance-sheet measures:
  - The actual FXI variable (percent of GDP) is materially smaller than the stock of FX reserves (percent of GDP).
- Chart- and sample-related specifics:
  - Taper Tantrum occurred in the second quarter of 2013; illustrative preemptive window considered first quarter of 2008 through first quarter of 2013 for that shock.

### Identification and the UIP outcome variable
- Main outcome: the UIP wedge (UIP deviation) over a 12-month horizon.
- UIP deviation formulation (log form preserved in source):
  - λt ≡ ̃it − ̃i∗t + ̃st − ̃st+k e
  - ̃it = log(1 + it), ̃i∗t = log(1 + i∗t), ̃st = log(St), ̃st+k e = log(Et(St+k))
- Interpretation:
  - λt = 0 implies UIP holds in expectations.
  - λt > 0 implies expected profitable returns from shorting the U.S. dollar and going long the domestic currency.
- Empirical strategy:
  - Difference-in-differences regression with country fixed effects αc and time (month) fixed effects ωt:
    - λc,t = αc + ωt + β Preemptive Policyc × Risk-Off Shockt + εc,t
  - β captures the differential effect of the shock on countries with versus without preemptive policies.
- Empirical observation:
  - Persistent UIP premia in EMs with average expected excess return of 3 percent for the average country.

### Empirical findings — Taper Tantrum (May 2013) and COVID-19 shocks
- Treatment definition timing:
  - Main preemptive window: 5 years prior to the shock (used in main tables).
  - For COVID-19 robustness, a 12-year window starting in 2008 is also used.
- Taper Tantrum (May 2013) benchmark results (Table 3 summary):
  - Preemptive CFM/MPMs and preemptive CFMs on inflows have negative effects on the UIP premium during the shock (columns (2) and (4)).
  - Preemptive domestic MPMs and preemptive CFMs on outflows have positive effects on the UIP premium during the shock.
  - When using all policies together (column (6)), the above pattern holds.
  - Economic magnitude: the UIP premium is 0.03-0.06 percentage points lower in countries with preemptive policies (CFM/MPM and CFM on inflows). This represents a 30 percent lower external finance premium in those countries relative to the average premium.
- Interpretation of signs:
  - Negative effect of preemptive inflow CFMs and CFM/MPMs: these measures slow down capital inflows by nonresidents, lower FX debt accumulation, and thereby reduce external finance premia during shocks.
  - Positive effect of preemptive outflow CFMs: may reflect a “reputation effect”—preemptive outflow restrictions (often applied to residents) signal potential future constraints and lead nonresident investors to demand higher premia.
- Robustness to country×year fixed effects:
  - Adding country×year effects (column (7)) weakens significance on preemptive CFM/MPM but the negative effect of preemptive CFMs on inflows remains.
- COVID-19 shock (Table 4 summary):
  - Using the five-year window: no effect from domestic MPMs.
  - Using the 12-year window: all preemptive policies negatively affect external finance premia during COVID-19, except preemptive CFMs on outflows which again show a positive impact.
  - Interpretation: since COVID-19 is both an external and domestic shock, preemptive domestic MPMs that rein in credit growth before the shock helped reduce external finance premia during the shock.
  - Economic impact magnitudes for COVID-19 are similar to those in the Taper Tantrum results.

### Evidence on FX-denominated debt and portfolio debt flows (treatment group dynamics)
- For the Taper Tantrum treatment group (median treatment country):
  - FX/LC debt ratio declined from 15 to 5.
  - FX / total debt declined from almost 90 percent to 76 percent.
  - Size of FX debt averages around 12 percent of GDP in the treatment group.
- Example comparison (thought experiment and charts):
  - Brazil (preemptive CFMs on inflows such as IOF taxes) versus Mexico (no CFMs at time, but domestic MPMs): during the Taper Tantrum, Brazil experienced a much lower UIP deviation (lower external finance premia) relative to Mexico despite the same exogenous shock.

### Robustness analyses
- Broader CFM/MPM definition:
  - Expanded CFM/MPM definition includes Liquidity Measures, Taxes and measures applied to SIFIs; these additional measures are removed from the “Domestic MPM” definition in this exercise.
  - Table 5 results: magnitude and significance of CFM/MPM coefficients and other covariates are largely unaffected, supporting robustness of the CFM/MPM measure.
- Alternative shock specification:
  - As an alternative to single-event shocks, monthly fluctuations in the VIX index are used (results available upon request in the source).

*Source: wpiea2022003-print-pdf (extracted chapter/section content).*

### 6.2    Granular Preemptive Measures

### 6.2    Granular Preemptive Measures

### Disaggregation of CFMs by instrument
- Extended the baseline results by separating both CFMs on inflows and CFMs on outflows into nine distinct instruments (using the Binici et al. (2020) disaggregation).
- Defined preemptive CFMs on inflows (and on outflows) as a net tightening in each individual instrument in a quarter (as opposed to an overall net tightening).
- Note limitation: for certain instruments there is a limited number of observations; instruments with too few observations are dropped from estimated regressions.

### Main empirical findings (from Table 6)
- Across instruments used as CFMs on inflows:
  - The estimated coefficients on all but one instrument are negative, reinforcing the Table 3 finding that CFMs on inflows on average lower UIP deviations.
  - Statistically significant negative coefficients are observed only for:
    - reserve requirements (column 6)
    - Other instruments (column 9)
  - Suggestive interpretation: aggregate inflow-CFM results may be driven by a subset of instruments.
- Across instruments used as CFMs on outflows:
  - Every instrument shown in Table 6 bears both a positive and statistically significant coefficient.
  - This includes approval requirements (column 1), bans (column 2), limits (column 4), limits used in conjunction with approval requirements (column 5), surrender and repatriation requirements (column 7), and Other instruments (column 9).
  - Interpretation: consistent evidence across instruments that preemptive outflow CFMs raise UIP premia, indicative of a reputational cost from restricting outflows.
- Consistency with prior results:
  - Estimated coefficients on Domestic MPMs and CFM/MPMs are virtually identical across every column in Table 6 and match the estimated coefficients in Table 3.
  - Overall interpretation: results are robust that preemptive inflow CFMs lower UIP deviations, and provide instrument-level evidence of heterogeneous impacts.

### Threats to identification and robustness checks
- Main identification threat: omitted variables at the country-month level correlated with the key interaction (preemptive policy × global shock), notably contemporaneous policies used in response to the shock.
- Primary contemporaneous policies considered:
  - FX intervention (FXI)
  - Monetary policy actions
  - Easing of existing inflow CFMs and tightening of outflow CFMs during the shock
- Controls implemented:
  - FXI reserves to GDP and manually collected actual FX intervention data (focused on the TT shock where intervention data are pinned down)
  - Monetary policy during the shock
  - Easing/tightening of CFMs during the shock
- Key robustness results (from Table 7):
  - The negative effect of preemptive CFMs (and CFM/MPMs) on inflows on UIP premia remains intact after controlling for FXI and monetary policy.
  - Both FX reserves and actual FXI employed during the shock have a negative effect on UIP premia.
  - Monetary policy tightening has a positive impact on UIP premia (consistent with Kalemli-Ozcan (2019)): tightening can be ineffective or counterproductive by raising external finance premia.
  - Only around 10 percent of countries tightened monetary policy during Taper Tantrum, but large tightenings in some countries could drive the observed positive effect.
  - Loosening of preexisting CFMs on inflows and/or tightening CFMs on outflows during the shock has a positive impact on UIP premia (combined variable reflects these concurrent policy moves), consistent with reputational costs from tightening outflow CFMs.
  - Policy implication: employing preemptive measures to slow foreign capital inflows and reduce FX debt accumulation is more effective at reducing financial stress than tightening outflow CFMs during the shock.

### Exchange rate volatility and depreciation (Section 8)
- Standardization and measures:
  - Each bilateral exchange rate standardized to set its value in 2008 January at 100.
  - Volatility computed over 8-month rolling windows on this indexed series; log change in the index used as realized depreciation.
- Volatility results (Table 8 column (2)):
  - Countries with preemptive CFM and CFM/MPM on inflows experienced lower volatility during the shock than countries without such preemptive policies.
  - Estimated coefficient on preemptive inflow CFM is about -2, which is half of the sample mean volatility of the indexed exchange rate (suggesting an economically meaningful impact).
  - Estimated coefficient on preemptive CFM/MPMs is statistically insignificant.
  - Preemptive domestic MPM policies contributed to lowering volatility but have economically small and sometimes statistically insignificant impacts.
  - Preemptive CFMs on outflows are associated with an increase in exchange rate volatility of about half the mean volatility (consistent with foreigners perceiving higher risk).
- Effects of policies used during the shock (rather than preemptively):
  - Use of CFMs during the shock increases exchange rate volatility (consistent with destabilizing effects on UIP premia).
  - Use of monetary policy during the shock raises exchange rate volatility.
  - FXI during the shock has no significant effect on exchange rate volatility.
  - Interpretation: monetary policy used during shocks can be counterproductive in managed floats; flexible exchange rates can absorb risk premia shocks better (consistent with Kalemli-Ozcan (2019)).
- Exchange rate level (depreciation/appreciation):
  - FXI has a statistically significant impact in appreciating the local currency during the shock (column (4)), but the estimated impact is extremely small.
  - Employing preemptive outflow CFMs leads to depreciation during the shock (likely reflecting reputation effects and measurement issues); preemptive outflow CFMs targeting domestic residents do not smooth risk-off shocks linked to foreign inflows.
- Overall inference:
  - Preemptive CFM/MPMs and inflow CFMs reduce exchange rate-related stress during risk-off shocks and thereby reduce external finance premia.

### Placebo shocks (Section 9)
- Purpose: rule out that countries with preemptive policies are intrinsically different in ways that lower UIP deviations in normal periods.
- Method:
  - Constructed dummy variables for “placebo shocks” (periods not global risk-off shocks) between 2009 and 2017.
  - Retained definition of the preemptive period as the preceding 5 years before the placebo shocks.
  - Note: for placebo shocks in earlier periods, used Aizenman-Pasricha CFM data when coverage of other CFM data was incomplete.
- Results (Table 9):
  - Preemptive CFMs on inflows have no negative impact on UIP deviations during placebo shocks other than the 2015 shock (column 2).
  - Interpretation: the 2015 episode likely functioned as a global risk-off shock in practice (e.g., Chinese stock market turbulence), explaining the negative impact there.
  - Preemptive CFMs on outflows are positive and significant in certain placebo cases, consistent with earlier findings that such measures raise investors’ demanded risk premia.
  - Conclusion: placebo tests support that main results are not spurious and that preemptive policies lower UIP deviations during genuine risk-off shocks.

### Heterogeneity by pre-shock FX-denominated debt (Section 10)
- Theoretical prediction: higher pre-shock FX-denominated debt amplifies the beneficial impact of preemptive CFMs/CFM+MPMs in lowering UIP deviations after a shock.
- Empirical setup:
  - Triple difference-in-differences regression interacting preemptive policy indicators with average FX-denominated debt of households and the non-financial sector (in percent of GDP) in the 5 years prior to the shock.
  - Data limitations:
    - FX-denominated debt from BIS available for only 8 countries (vs 43 in baseline), reducing sample size by about three-fourths.
    - FX-denominated data available only since 2013, so test applies to the Covid-19 shock but not Taper Tantrum.
- Findings (Table 10):
  - Sum of estimated coefficients of the triple interaction with shock × FX-denominated debt is negative in both columns (interpreted as higher pre-shock FX-denominated debt associated with larger declines in UIP deviations for countries using preemptive policies).
  - Evaluated at the sample mean of FX-denominated debt (0.15):
    - For preemptive MPM policies the impact is -0.006.
    - For preemptive CFM/MPM policies the impact is -0.07.
  - Interpretation: countries with higher than average FX-denominated debt would experience even more negative UIP deviations; countries with lower than average FX-denominated debt would experience smaller, potentially positive, deviations for very low levels of FX-denominated debt.

### Policy implications and synthesis
- Empirical support for theoretical predictions: preemptive CFMs and MPMs lower UIP deviations and exchange rate volatility during global risk-off shocks, particularly when targeted at inflows.
- Preemptive policies help limit buildup of FX debt and currency mismatches, preserving access to international markets during stress.
- Preemptive policies reduce the need for monetary policy to counteract demand shortfalls after shocks and reduce the need to defend the exchange rate, thereby allowing monetary policy to focus on domestic objectives.
- Cautionary findings:
  - Preemptive outflow CFMs tend to raise UIP premia and exchange rate volatility (reputational costs).
  - Using CFMs or monetary policy reactively during shocks can be destabilizing.
- Net takeaway: targeted preemptive measures on inflows and CFM/MPM combinations can be effective ex ante tools to reduce financial vulnerabilities and external finance premia during global risk-off episodes.

*Italic — Source: wpiea2022003-print-pdf — Section 6.2 and surrounding sections*

### References

### wpiea2022003-print-pdf - References

### Key Summary Statistics (Table 1)
- UIP premium t: (Mean) 0.03; (Standard Deviation) 0.16; (Minimum) -5.04; (Maximum) 1.9
- FXI actual t: (Mean) 0.0008; (Standard Deviation) 0.004; (Minimum) -0.03; (Maximum) 0.05
- FXI reserves t: (Mean) 17.44; (Standard Deviation) 11.50; (Minimum) 0.74; (Maximum) 85.94
- Preemptive MPM T T: (Mean) 0.75; (Standard Deviation) 0.43; (Minimum) 0; (Maximum) 1
- Preemptive CFM/MPM T T: (Mean) 0.69; (Standard Deviation) 0.46; (Minimum) 0; (Maximum) 1
- Preemptive MPM COV ID: (Mean) 0.89; (Standard Deviation) 0.31; (Minimum) 0; (Maximum) 1
- Preemptive CFM/MPM COV ID: (Mean) 0.82; (Standard Deviation) 0.39; (Minimum) 0; (Maximum) 1
- Preemptive CFM inflows T T: (Mean) 0.24; (Standard Deviation) 0.46; (Minimum) 0; (Maximum) 1
- Preemptive CFM outflows T T: (Mean) 0.29; (Standard Deviation) 0.46; (Minimum) 0; (Maximum) 1
- Preemptive CFM inflows COV ID: (Mean) 0.4; (Standard Deviation) 0.49; (Minimum) 0; (Maximum) 1
- Preemptive CFM outflows COV ID: (Mean) 0.41; (Standard Deviation) 0.49; (Minimum) 0; (Maximum) 1
- Monetary Policy t: (Mean) 0.07; (Standard Deviation) 0.05; (Minimum) 0.001; (Maximum) 0.7
- CFM inflows T T: (Mean) 0.036; (Standard Deviation) 0.19; (Minimum) 0; (Maximum) 1
- CFM outflows T T: (Mean) 0.008; (Standard Deviation) 0.09; (Minimum) 0; (Maximum) 1
- FX debt: (Mean) 0.16; (Standard Deviation) 0.085; (Minimum) 0.069; (Maximum) 0.32
- Notes: "See the data section for the definition of variables."

### Preemptive CFMs by Country and Episode (Table 2)
- Preemptive policies defined as policies in place 5 years prior to the shock.
- Taper Tantrum (TT) pre-period: 2008Q1-2013Q1.
- COVID pre-period: 2014Q1-2019Q3.
- Both inflow and outflow policies are net (tightening minus easing).
- Table indicates country-level presence of preemptive inflow/outflow CFMs for Taper Tantrum and COVID-19 (entries marked with Y or multiple Ys).

### Regression Evidence: Taper Tantrum (Table 3)
- Dependent variable: UIP c,t; sample: Emerging Market Economies, 1996m1-2020m6.
- Key coefficient estimates (interaction with TT):
  - Preemptive MPM domestic c × TT t: 0.02*** (columns 1–3 show 0.02***, column 3 shows 0.01*** in one specification). Standard errors: (0.003), (0.004), (0.004).
  - Preemptive CFM/MPM c × TT t: -0.02*** (columns 1–2) and -0.003 (column 3). Standard errors: (0.003), (0.003), (0.004).
  - Preemptive CFM total c × TT t: 0.005* (0.003).
  - Preemptive CFM inflows c × TT t: -0.02***, -0.06***, -0.02** (std. errors (0.004), (0.005), (0.01) across specs).
  - Preemptive CFM outflows c × TT t: 0.005*, 0.05***, 0.02** (std. errors (0.003), (0.004), (0.006)).
- Adjusted R2 values reported: 0.35 (most columns) and 0.74 (one column).
- Observations: 6816 in reported regressions.
- Fixed effects: Country FE (yes/no by spec), Month FE (yes), Country * Year FE (in some specs yes).
- Notes: Taper Tantrum (TT) dummy = 1 in months May-December 2013. Descriptions of MPM and CFM variable construction provided.

### Regression Evidence: COVID-19 (Table 4)
- Dependent variable: UIP c,t. Panel A (5-year window) and Panel B (longer window).
- Panel A key estimates (1996m1-2020m6, 5-year window):
  - Preemptive MPM domestic c × COVID t: -0.01** (0.005); -0.008 (0.005) in alternative spec.
  - Preemptive CFM/MPM c × COVID t: -0.01*** (0.003); -0.005* (0.003).
  - Preemptive CFM total c × COVID t: -0.004 (0.004).
  - Preemptive CFM inflows c × COVID t: -0.004 (0.004); -0.02** (0.009).
  - Preemptive CFM outflows c × COVID t: -0.0006 (0.005); 0.03** (0.01).
- Panel B key estimates (longer pre-period 2008Q1-2016Q4):
  - Preemptive MPM domestic c × COVID t: -0.02*** (0.005); -0.01** (0.005).
  - Preemptive CFM/MPM c × COVID t: -0.03*** (0.003); -0.03*** (0.004).
  - Preemptive CFM total c × COVID t: -0.006 (0.004).
  - Preemptive CFM inflows c × COVID t: -0.008* (0.004); -0.02*** (0.005).
  - Preemptive CFM outflows c × COVID t: -0.003 (0.005); 0.02*** (0.006).
- Adjusted R2 values: 0.35 across columns. Observations: 6816.

### Robustness and Granular Measures (Tables 5–7)
- Table 5 (Broader Measures of CFM/MPMs) reproduces Taper Tantrum interactions with:
  - Preemptive MPM domestic c × TT t: 0.02***, 0.03***, 0.02*** (std. errors (0.004), (0.004), (0.005)).
  - Preemptive CFM/MPM c × TT t: -0.03***, -0.03***, -0.002 (std. errors (0.003), (0.003), (0.004)).
  - Preemptive CFM total c × TT t: 0.005* (0.003).
  - Preemptive CFM inflows c × TT t: -0.02***, -0.06***, -0.02**.
  - Preemptive CFM outflows c × TT t: 0.005*, 0.04***, 0.01**.
- Table 6 (Granular Preemptive Measures) reports many specific interaction coefficients with TT:
  - MPM domestic c × TT t: 0.02*** to 0.04*** across many granular specs (std. errors typically (0.003) or (0.004)).
  - MPM/CFM c × TT t: -0.02*** to -0.04*** across granular specs (std. errors typically (0.003)).
  - Approval Requirement outflows c × TT t: 0.09*** (0.006).
  - Bans outflows c × TT t: 0.09*** (0.006).
  - Holding Period inflows c × TT t: -0.01 (0.009).
  - Limit inflows c × TT t: -0.002 (0.003).
  - Limit outflows c × TT t: 0.007*** (0.003).
  - Reserve Requirement inflows c × TT t: -0.01*** (0.002).
  - Surrender + Repatriation Requirement outflows c × TT t: 0.09*** (0.006).
  - Other inflows c × TT t: -0.04*** (0.008).
  - Other outflows c × TT t: 0.09*** (0.006).
- Table 7 (Role of Policies During TT) highlights:
  - Preemptive MPM domestic c × TT t: 0.02*** (std. errors around (0.003)-(0.004)).
  - Preemptive CFM/MPM c × TT t: -0.02*** to -0.12*** in different specs.
  - Preemptive CFM inflows c × TT t: -0.04***, -0.07***, 0.002, -0.06***, -0.06*** across specs.
  - Preemptive CFM outflows c × TT t: 0.04***, 0.05***, 0.06***, 0.05*** across specs.
  - Monetary Policy c × TT t: 0.61*** and 0.36*** (std. errors (0.06), (0.07)).
  - Loosening/Tightening CFMs c × TT t: 0.03*** and 0.05*** (0.005).
  - Actual FXI c × TT t: -1.68* (0.93) in one specification.
  - FX Reserves c × TT t: -0.001***, -0.001*** (0.0002).
- Adjusted R2 values range 0.34–0.36 in these robustness checks; observations vary by specification.

### Exchange Rate Volatility and Policies (Table 8)
- Dependent variables: Standard Dev. of Exchange Rate Index and ∆ Log Exchange Rate Index.
- Selected interaction coefficients with TT:
  - MPM domestic c × TT t: -0.27 (0.38), -0.77*** (0.23), -0.002 (0.003), -0.002 (0.003) across columns.
  - CFM/MPM c × TT t: -0.5*** (0.16), -0.43* (0.23), -0.002 (0.002), 0.0001 (0.002).
  - CFM inflows c × TT t: -2.22*** (0.31), -2.24*** (0.34), -0.001 (0.006), -0.003 (0.006).
  - CFM outflows c × TT t: 2.39*** (0.41), 2.47*** (0.56), 0.01** (0.005), 0.01** (0.005).
  - FX Reserves c × TT t: -0.01 (0.02), -0.0004*** (0.0001).
  - Monetary Policy c × TT t: 8.83* (5.13), -0.08 (0.05).
  - Loosening/Tightening CFMs c × TT t: 2.6*** (0.33), -0.001 (0.004).
- Adjusted R2: 0.33, 0.32, 0.01, 0.01. Observations vary across columns.

### Placebo and FX Debt Robustness (Tables 9–10)
- Table 9 (Placebo Shocks):
  - Preemptive MPM domestic c × hypotheticalshock t: -0.02* (0.01); other specs: -0.004 (0.008), 0.005 (0.009), 0.004 (0.008).
  - Preemptive CFM/MPM c × hypotheticalshock t: 0.01 (0.009); 0.0004 (0.009); -0.004 (0.007); -0.006 (0.008).
  - Preemptive CFM inflows c × hypotheticalshock t: 0.007 (0.005); -0.03*** (0.009); 0.002 (0.008); 0.03*** (0.01).
  - Preemptive CFM outflows c × hypotheticalshock t: -0.004 (0.003); 0.02** (0.009); 0.02** (0.007); -0.02 (0.01).
  - Adjusted R2: 0.72–0.74. Observations vary.
  - Hypothetical shock dates and preemptive periods listed for each regression.
- Table 10 (FX-denominated Debt and UIP Shocks):
  - Dependent variable: UIP c,t. Sample: Emerging Market Economies, 1996m1-2020m6.
  - Preemptive MPM domestic c × COVID t × FXdebt: -0.67*** (0.09) in one specification.
  - Preemptive MPM domestic c × COVID t: 0.11*** (0.02) in another spec.
  - COVID t × FXdebt: 0.32*** (0.05) and 0.02 (0.05) across columns.
  - Preemptive CFM/MPM c × COVID t × FXdebt: -1.19*** (0.29).
  - Preemptive CFM/MPM c × COVID t: 0.12*** (0.04).
  - Adjusted R2: 0.435 and 0.431. Observations: 1954.

### Data Availability and Definitions (Tables 11–14)
- Table 11: Country-level data availability ranges provided for EMDEs (e.g., Albania 2007-2020; China 1996-2020; India 1999-2020**; Malaysia 1996-2020; many entries include notes: * Some data are missing with this date range; ** Only missing data are 3 missing observations in 2002; *** Only missing data are 5 missing observations in 2015).
- Table 12: Description of Domestic MPM Measures. Examples:
  - CCB: Requirements for banks to maintain countercyclical capital buffers.
  - Conservation: Requirements for banks to maintain capital conservation buffers.
  - LTV: Limits to loan-to-value ratios.
  - SIFI: Measures to mitigate risks from global and domestic systemically important financial institutions (SIFIs).
  - Notes: Source Alam et al. (2019).
- Table 13: Description of MPM/CFM Measures. Examples:
  - RR: Reserve requirements for macroprudential purposes (domestic or foreign currency).
  - LFCL: Limits on foreign currency lending.
  - LFX: Limits on foreign exchange exposures and funding, currency mismatch regulations, and limits on net or gross open FX positions.
  - Notes: Source Alam et al. (2019).
- Table 14: Description of CFM Measures. Examples:
  - Approval Requirement: Administrative control on cross-border transactions requiring approval.
  - Limit: Regulations to limit commercial banks’ positions in foreign currencies, investment abroad, foreign exchange transfers abroad, etc.
  - Ban: Prohibition of transfer, purchase or sale of certain types of financial assets by residents and/or non-residents.
  - Fee: Additional charge/cost for transactions that require remittance or settlement in foreign currency.
  - Holding period requirement, Reserve Requirement (differential treatment), Repatriation Requirement, Surrender Requirement, Surrender / Repatriation Requirement, Tax, Other.
  - Source: Binici, Das and Pugacheva (forthcoming).

*Italic: Content extracted from the References and Tables sections of wpiea2022003-print-pdf.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022003-print-pdf.pdf_