## Do Capital Controls Limit Inflow Surges? (Introduction and Data sections — WP/2023/050)

## Source details

**Canonical URL:** [Do Capital Controls Limit Inflow Surges? (Introduction and Data sections — WP/2023/050)](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023050-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023050-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023050-print-pdf.pdf.json)

---

### Context and motivation
- Capital flow surges and reversals are a persistent policy concern for emerging market and developing economies (EMDEs), often large relative to domestic financial systems.
- Periods of sustained inflows—partly driven by prolonged easy monetary policies in advanced economies (AEs)—were followed by sharp reversals (examples: the Taper Tantrum; COVID-19 pandemic).
- Typical policy responses:
  - AEs with open capital accounts and deep markets typically allow exchange rates to float.
  - Many EMDEs use capital account measures, macroprudential measures (MPMs), and foreign exchange intervention (FXI) alongside monetary and fiscal tools.

### Research questions (three main)
- Do countries respond to increases in capital inflows by introducing new or tightening previously existing capital controls?
  - Reported finding: only around 10 percent of measures that tightened controls were introduced during an inflow surge.
- Are tightened capital controls effective at limiting the probability of future inflow surges?
  - Reported finding: tightening of capital controls results in a lower probability of future surges at the aggregate flow level and more strongly when controls are mapped to the targeted asset.
- Do capital control actions produce multilateral spillovers to other countries?
  - Reported finding: strong multilateral spillovers—capital control actions result in flows being deflected to other countries in the region or with similar fundamentals.

### Contributions and approach
- Two surge definitions used for robustness: Forbes and Warnock (2012, 2021) and Ghosh and others (2014).
- New series for inflow tightening and easing measures constructed using capital control changes data based on Baba and others (forthcoming).
- Controls matched to the asset class they are applied to (portfolio debt, equity, FDI, other).
- Sample expanded from 21 to 40 countries; sample period 1995-2018.

### Sample and measures
- Sample: 40 countries over 1995-2018.
- Total initial measures identified: 184 measures tightening capital controls on nonresident transactions.
- Asset classes analyzed: portfolio debt, equity, foreign direct investment (FDI), and other (mainly bank and non-bank commercial and financial credits).
- Controls on derivatives excluded.

### Data sources and construction
- Policy-change data: changes section of the Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER), narrative converted into numerical measures.
- Nonresident capital inflow data: IMF’s Financial Flow Analytics (FFA) database.
  - Variable: difference between nonresident capital inflows and nonresident capital outflows as percent of Gross Domestic Product (GDP), excluding official flows.
  - A positive number indicates a positive net nonresident capital inflow.

### Classification dimensions for policy changes
- Direction of flows regulated: focus on controls on resident liabilities (nonresidents’ inflows and repatriation). Tightening on nonresident outflows (outside crises) represent only 5 percent of the measures that tighten controls.
- Direction of change: easing or tightening (analysis focuses on tightening; easing used in robustness checks).
- Asset class (IIP category): Equity, Debt, FDI, Other.
- Type of instrument: Quantitative (limits), price-based (taxes, unremunerated reserve requirements), administrative/monitoring (authorizations, notifications, registrations).

### Descriptive patterns (selected)
- Number of tightening actions increased progressively from the early 2000s and remained high in the post-GFC period.
- Bulk of tightening actions introduced by EMDEs, primarily from Asia and Latin America; some European countries (Iceland, Ukraine) also used controls.
- Distribution across asset classes: actions are evenly distributed across the four asset classes, with “Other” and Debt slightly dominating.
- Instrument types:
  - About one-third of tightened capital controls are price-based measures.
  - Administrative measures, limits, registration measures, and mandatory holdings each represent between 10 and 15 percent of measures.
  - Bans and notifications are rarely used.
- Timing patterns:
  - Some measures introduced after sustained and large outflows or after financial crises (examples listed in Annex Table I).
  - Rebounds in inflows after controls in post-crisis periods can reflect catching-up effects rather than policy ineffectiveness.

### Key descriptive statistic on timing relative to surges
- Controls tightened during a surge: 15 cases (or around 10 percent of the sample).
- Tightening actions within a year following a surge: 17.
- Tightening actions within a year preceding a surge: 9.
- Goodness-of-fit chi-square test p-value: 0.22 (implies cannot reject that timing of tightenings is not significantly different from random).

---

### Surges of capital inflows — definitions and measurement
- Three surge definitions used:
  - Forbes and Warnock (2012, 2021) (benchmark);
  - Ghosh and others (2014);
  - Simple threshold: gross total nonresident inflows above 5 percent of GDP over a given quarter.
- Note: when seven definitions are compared in literature, the number of episodes identified can vary by up to a factor of three.
- Comparison of data counts:
  - Using Ghosh methodology: of 3,075 observations considered (those with GDP data), 718 correspond to surges.
  - Using Forbes definition: 2,969 observations considered, 318 classified as surges.
  - Correlation between Forbes and Ghosh measures: 0.17.
- Simple threshold results (fraction of total observations out of 3,075):
  - 3 percent of GDP threshold: 1 Quarter = 42.0 percent, 2 Quarters = 34.4 percent.
  - 5 percent of GDP threshold: 1 Quarter = 26.9 percent, 2 Quarters = 19.5 percent.

---

### Econometric approach — overview
- Baseline: logit model; dependent variable = 1 if period t is an inflow surge (aggregate non-resident gross inflows), 0 otherwise.
- Key independent variable: Control = dummy = 1 if country introduces capital controls on non-resident inflows (and/or on non-resident outflows, outside crises) in the past four quarters.
  - A negative and significant coefficient on lagged Control indicates controls reduce the probability of future surges.
- Push-pull controls included (lagged):
  - Diff GDP growth = domestic real GDP growth minus U.S. real GDP growth.
  - Diff IR = domestic real interest rate minus U.S. real interest rate.
- Additional controls (lagged): foreign exchange reserves as % of GDP; composite risk index; Fernandez et al. (2018) index of capital restrictions on non-resident inflows.
- Fixed effects: country fixed effects (γ_i) and time fixed effects (θ_t).
- Endogeneity considerations: reverse causality could bias estimates; use of lags and descriptive timing evidence mitigate concerns.

### Main econometric findings — aggregate surge probability (selected coefficients from Table 1)
- "Whether tightening in the past 4 qtrs" coefficients (standard errors in parentheses):
  - -0.405 (0.270)
  - -0.456* (0.246)
  - -0.356** (0.160)
- Other coefficients (examples, standard errors in parentheses):
  - Diff GDPGrowth: 0.181*** (0.0612); 0.185*** (0.0611); 0.185*** (0.0613).
  - Diff IR: 0.0581*** (0.0219); 0.0570*** (0.0218); 0.0549** (0.0219).
  - Composite Risk Index: 0.123*** (0.0300); 0.160*** (0.0280); 0.113*** (0.0168).
  - Foreign Exchange reserves as % of GDP: -0.00294 (0.00298); 0.000394 (0.00246); -0.00260* (0.00157).
  - FD (Financial development index): 0.832 (1.600); 4.772*** (1.620); 3.462*** (1.092).
  - Inflow restrictions index (Fernandez et al.): 0.691* (0.369) in one specification.
- Robustness:
  - Alternative measures of tightening (net tightening; sum of tightenings/easings separately) produce consistent results.
  - Results robust across the three surge definitions and when using surge definition based on net inflows.
- Caveat:
  - Unweighted counts of measures assume equal importance across measures; controls are often used alongside other tools (MPMs, FXI) which could confound joint effects.

---

### Asset-level analysis (three-dimensional panel) — specification and findings
- Motivation: surges can be asset-specific; controls may target specific assets.
- Panel: country-time-asset. Dependent variable = 1 if asset-level surge; ControlAsset = 1 if controls introduced on nonresident inflows in that specific asset class in the past four quarters. Asset fixed effects μ_n included.
- Selected coefficients (Table 2 highlights):
  - "Whether tightening in the past 4 qtrs": -0.300* (0.182); -0.266* (0.139).
  - Sum of net tightening in past 4 qtrs: -0.107** (0.0443); -0.130*** (0.0365).
  - Sum of tightenings in past 4 qtrs: -0.198* (0.110); -0.201** (0.0900).
  - Sum of easings in past 4 qtrs: 0.0961** (0.0471); 0.121*** (0.0383).
  - Diff GDPGrowth and Diff IR: positive and often significant.
  - Composite Risk Index: 0.0943*** (0.0121) in example.
  - FD: 1.992*** (0.726); 3.402*** (0.590) in examples.
  - Inflow restrictions index at asset/flow level (Fernandez et al.): -0.573*** (0.186); -0.647*** (0.186); -0.625*** (0.187) in Forbes asset-level specifications.
- Interpretation:
  - Controls mapped directly to targeted assets yield stronger reductions in probability of future asset-specific surges.
  - Higher pre-existing restrictiveness at asset/flow level associated with lower future asset-specific surges in baseline specification.

---

### Multilateral spillovers — grouping, specification, and key results
- Spillover groupings:
  - Region based: seven regional groups (Asia, Sub-Saharan Africa, Middle East and North Africa, Western Europe, Eastern Europe and Central Asia, Latin America, North America).
  - Return based: four groups by average real GDP growth (bottom 25 percentile = low-return group).
  - Risk based: four groups by composite risk index (bottom/percentiles analogous).
  - BRICS based: BRICS (Brazil, Russia, India, China, South Africa) assumed to generate important spillovers; non-BRICS linked to regional BRICS.
- Spillover variable:
  - CIControls_s−i,t−1,t−4 = weighted average of controls introduced in past 4 quarters in other countries in the group; weights y_j are non-resident inflows (flow-weight).
  - BRICS-based contagion constructed as sum of policy changes in other BRICS (or regional BRICS for non-BRICS).

- Selected regression results (Table 3 highlights; standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1):
  - Own-country:
    - Sum of own tightenings in past 4 qtrs: -0.136* (0.0827); -0.140* (0.0823); -0.144* (0.0827); -0.152* (0.0811).
    - Sum of own easings in past 4 qtrs: 0.0358 (0.0338); 0.0379 (0.0336); 0.0421 (0.0336); 0.0425 (0.0340).
  - Lagged macro and risk controls (examples):
    - L. Diff GDPGrowth: 0.200*** (0.0628) ... values around 0.176–0.200 across specs.
    - L. Diff IR: 0.0526** (0.0219) ... values around 0.0528–0.0621 across specs.
    - Composite Risk Index: 0.111*** (0.0302) ... values around 0.111–0.124.
    - Foreign Exchange reserves as % of GDP: coefficients around -0.00215 to -0.00291 (standard errors ~0.00288–0.00302).
  - Multilateral spillover coefficients (selected):
    - Region based spillover — Tightenings: 0.0345 (0.0234); 0.0356 (0.0234).
    - Region based spillover — Easings: -0.0150* (0.00857); -0.0154* (0.00857).
    - Risk based spillover — Tightenings: 0.0931** (0.0381); 0.0948** (0.0377).
    - Risk based spillover — Easings: -0.0221 (0.0151); -0.0228 (0.0152).
    - Return based spillover — Tightenings: 0.108 (0.0791); 0.105 (0.0791).
    - BRICS based spillover — Tightenings: 0.0877* (0.0473); 0.0858* (0.0470).
    - BRICS based spillover — Easings: 0.0294 (0.0181); 0.0274 (0.0180).
  - Sample:
    - Observations: 2,382
    - Number of panels: 33

- Main spillover findings:
  - Own-country tightenings reduce domestic probability of future surges (negative and statistically significant coefficients).
  - Tightenings increase probability of surges in other countries with similar composite risk index (risk-based tightenings 0.0931** and 0.0948**) and when tightenings occur in BRICS (BRICS tightenings 0.0877* and 0.0858*).
  - Easings in a country can reduce probability of surges in other countries in the same region (region-based easings -0.0150* and -0.0154*).
  - Within-group spillovers are largest among countries with similar composite risk index and from controls introduced in BRICS countries.

---

### Policy implications and conclusions (as reported)
- Capital controls can be an important tool to limit inflow surges: tightening is associated with a reduced probability of future surges domestically.
- Capital controls generate significant adverse multilateral spillovers by deflecting flows to countries with similar fundamentals—particularly those with similar risk profiles as measured by the composite risk index.
- Presence of large multilateral spillovers, especially from BRICS and among countries with similar risk, suggests potential justification for international cooperation in the use of capital controls.
- An optimal policy response should weigh pros and cons of available tools depending on a country’s initial conditions and nature of the shock, consistent with the integrated policy framework and IMF guidance on liberalization and management of capital flows.

---

### Annex highlights — selected tightened measures and numeric thresholds (exact values preserved)
- Argentina 2005q2: minimum holding period raised to 365 days from 180 days; retention of 365 days on certain debts; 30 percent unremunerated deposit requirement of one year on foreign portfolio investment.
- Brazil IOF tax rates on inflows: introduced at 2 percent in 2009q4; increased to 4 percent and then to 6 percent in 2010q4.
- Iceland 2016q2: special 40 percent reserve requirement with a 12-month holding period at zero interest.
- Indonesia SBIs: one-month holding period in 2010q3; lengthened to six months in 2011q2.
- Colombia 2007q2: six-month URR of 40 percent for foreign portfolio investments.
- Ukraine FDI surrender requirement: 50 percent in 2013q3; increased to 100 percent in 2014q3.
- Canada 2013q4: review threshold intended to liberalize to Can$1 billion for private investors; foreign SOE review threshold remains Can$330 million.
- Thailand limits: B 50 million consolidated entity limit in 2006q4; limit decreased to B 10 million in 2008q1.
- Monetary and flow thresholds preserved verbatim in Annex Table I and Annex Table 2 summary statistics.

Key numeric summary statistics (Annex Table 3, exact values):
- Index of capital restrictions on NR inflows: mean 0.44, s.d. 0.32, p50 0.45, min 0.00, max 1.00, N 14,720
- GDP growth differential: mean 0.42, s.d. 1.76, p50 0.38, min -10.50, max 13.89, N 13,000
- Interest Rate differential: mean 6.89, s.d. 13.19, p50 4.25, min -6.50, max 208.60, N 12,976
- Foreign Reserves, percent of GDP: mean 62.68, s.d. 63.13, p50 45.84, min 0.82, max 448.32, N 14,508
- Financial Development Index: mean 0.44, s.d. 0.24, p50 0.38, min 0.05, max 1.00, N 14,352
- Composite Risk Index: mean 71.38, s.d. 8.53, p50 70.50, min 41.00, max 92.75, N 14,768
- BRICS dummy variable: mean 0.13, s.d. 0.33, p50 0.00, min 0.00, max 1.00, N 14,880
- Contagion (BRICS, Gross Tightening): mean 0.22, s.d. 0.74, p50 0.00, min 0.00, max 7.00, N 14,880

*IMF Working Papers — Do Capital Controls Limit Inflow Surges? (Introduction and Data sections, WP/2023/050)*

### 1. INTRODUCTION.........................................................................................................

### 1. INTRODUCTION

### Context and motivation
- Capital flow surges and reversals have been a persistent policy concern, especially for emerging market and developing economies (EMDEs), as flows can be large relative to domestic financial systems.
- Several periods of sustained inflows—partly driven by prolonged easy monetary policies in advanced economies (AEs)—were followed by sharp reversals (examples: the Taper Tantrum; COVID-19 pandemic), producing repeated surges and reversals.
- Policy responses vary: AEs with open capital accounts and deep markets typically allow exchange rates to float; many EMDEs use capital account measures, macroprudential measures (MPMs), and foreign exchange intervention (FXI) alongside monetary and fiscal tools.

### Research questions
The paper addresses three main questions:
- Do countries respond to increases in capital inflows by introducing new or tightening previously existing capital controls?
  - Finding reported: only around 10 percent of measures that tightened controls were introduced during an inflow surge.
- Are tightened capital controls effective at limiting the probability of future inflow surges?
  - Main finding reported: tightening of capital controls results in a lower probability of future surges at the aggregate flow level and more strongly when controls are mapped to the targeted asset.
- Do capital control actions produce multilateral spillovers to other countries?
  - Main finding reported: strong multilateral spillovers—capital control actions result in flows being deflected to other countries in the region or with similar fundamentals.

### Contributions and approach
- Uses two surge definitions to ensure robustness: Forbes and Warnock (2012, 2021) and Ghosh and others (2014).
- Constructs a new series for inflow tightening and easing measures using capital control changes data based on Baba and others (forthcoming).
- Matches controls to the asset class they are applied to (portfolio debt, equity, FDI, other).
- Expands sample coverage relative to prior work (sample expanded from 21 to 40 countries).

### 2. DATA AND DESCRIPTIVE STATISTICS

### Sample and measures
- Sample: 40 countries over 1995-2018.
- Total initial measures identified: 184 measures tightening capital controls on nonresident transactions.
- Asset classes analyzed: portfolio debt, equity, foreign direct investment (FDI), and other (mainly bank and non-bank commercial and financial credits).
- Controls on derivatives excluded due to lack of derivative flow information for a large proportion of the sample.

### Data sources and construction
- Policy-change data derived from the changes section of the Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER); narrative approach converted into numerical measures capturing direction of policy.
- Nonresident capital inflow data from IMF’s Financial Flow Analytics (FFA) database.
  - Variable defined as the difference between nonresident capital inflows and nonresident capital outflows as percent of Gross Domestic Product (GDP), excluding official flows.
  - A positive number indicates a positive net nonresident capital inflow.

### Classification dimensions for policy changes
Each policy change in the dataset has a unique classification along the following dimensions:
- Direction of flows regulated: focus on controls on resident liabilities (nonresidents’ inflows and repatriation). Controls on nonresident outflows during crisis times are excluded from main analysis.
  - Observation: tightening actions on nonresident outflows (outside of crises) represent only 5 percent of the measures that tighten controls.
- Direction of change: easing or tightening of restrictions (analysis focuses on tightening of inflow restrictions; easing considered in robustness checks).
- Asset class (IIP category): Equity (portfolio flows), Debt (portfolio flows), FDI, and Other flows.
- Type of instrument: Quantitative (limits), price-based (taxes, unremunerated reserve requirements), administrative/monitoring (authorizations, notifications, registrations). Some insignificant measures were dropped.

### Descriptive patterns and summary statistics (selected)
- Number of tightening actions increased progressively from the early 2000s and remained high in the post-GFC period.
- Bulk of tightening actions introduced by EMDEs, primarily from Asia and Latin America; some European countries (Iceland, Ukraine) also used controls.
- Distribution across asset classes: actions are evenly distributed across the four asset classes, with “Other” (commercial and financial credits) and Debt slightly dominating.
- Instrument types:
  - About one-third of tightened capital controls are price-based measures.
  - Administrative measures, limits, registration measures, and mandatory holdings each represent between 10 and 15 percent of measures.
  - Bans and notifications are rarely used.
- Timing patterns:
  - Some measures were introduced after sustained and large outflows or after financial crises (examples cited: India 2009Q1 on Other flows; Argentina 2005Q2 and 2005Q4 on Debt during sovereign restructuring; Brazil 2009Q4 on Debt after the GFC; Kazakhstan post-1999 currency crisis; Turkey post-1996 currency crisis).
  - Rebounds in inflows after the introduction of controls in post-crisis periods can reflect catching-up effects rather than policy ineffectiveness.

### 3. KEY EMPIRICAL FINDINGS (as reported in Introduction and Data sections)

- Only around 10 percent of tightened-control measures were introduced during an inflow surge—evidence of limited countercyclical use of capital controls.
- Tightening of capital controls is associated with a lower probability of future inflow surges:
  - Effect holds at aggregate flow level.
  - Effect is stronger when controls are directly mapped to the asset they target.
- Capital controls generate multilateral spillovers:
  - Capital control actions result in flows being deflected to other countries in the region or with similar fundamentals (risk, return, etc.).
  - Spillovers should be considered when assessing the multilateral implications of capital control policies.

### 4. Organization of the paper (sections cited)
- Section 2: data and descriptive statistics.
- Section 3: econometric approach and results (including effects on probability of surges, asset-specific surges, and international spillovers).
- Section 4: conclusion.
- Annex 1: methodology for identification of surges.

*IMF WORKING PAPERS — Do Capital Controls Limit Inflow Surges? (Introduction and Data sections)*

### 2.3 Surges of Capital Inflows

### 2.3 Surges of Capital Inflows

### Definitions and measurement of surges
- Three definitions of surges (using nonresident private inflows) are used:
  - Forbes and Warnock (2012, 2021) (benchmark);
  - Ghosh and others (2014);
  - Simple threshold: gross total nonresident inflows above 5 percent of GDP over a given quarter.
- Details of the Forbes/Warnock and Ghosh et al. methodologies are presented in Annex 1 (source document).
- Note from the literature: various definitions yield different frequency and length of surges; when seven definitions are compared, the number of episodes identified can vary by up to a factor of three.

### Descriptive evidence on timing of tightening relative to surges
- Aggregate patterns:
  - The number of countries experiencing a surge displays a broadly similar trend across the three definitions.
  - Asset-level surges do not always coincide with aggregate (total) surges: episodes can be “Surge both in asset and total”, “Surge in asset only”, or “Surge in total only” (illustrated for Direct Investment and Other Flows).
- Timing of tightening actions around surges (using Forbes gross inflow definition as baseline):
  - Controls were tightened in only 15 cases (or around 10 percent of the sample) during a surge.
  - 17 tightening actions were observed within a year following a surge.
  - 9 tightening actions happened within a year preceding a surge.
- Randomization and concentration tests:
  - A histogram comparison of actual tightening timing versus an artificial random distribution of quarters indicates actual tightenings are only very weakly more concentrated around surges than random.
  - Tightening actions are classified into timing-distance bins relative to closest surge (10+ quarters before; 9–6 before; 5–2 before; during or ±1 quarter; 2–5 after; 6–9 after; 10+ quarters away).
  - Goodness-of-fit chi-square test yields a p-value of 0.22.
    - Implication: do not reject the null that actual timing of tightenings is not significantly different from random timing. Conclusion: countries do not seem to enact controls specifically during or around surges.
  - Finding is robust to:
    - different surge definitions;
    - different random realizations;
    - restricting to episodes of net tightening;
    - specific types of controls (e.g., quantitative limits and prices).

### Econometric approach (overview)
- Baseline specification: logit model where dependent variable = 1 if period t is an inflow surge (aggregate non-resident gross inflows), 0 otherwise.
- Key independent variable: Control = dummy equal to 1 if country introduces capital controls on non-resident inflows (and/or on non-resident outflows, outside crises) in the past four quarters.
  - A negative and significant coefficient on the lagged Control variable indicates controls reduce the probability of future surges.
- Push-pull controls included (lagged):
  - Diff GDP growth = domestic real GDP growth minus U.S. real GDP growth (pull factor);
  - Diff IR = domestic real interest rate minus U.S. real interest rate (pull factor).
- Additional controls (lagged): foreign exchange reserves as % of GDP; composite risk index (political, financial, economic risk); Fernandez et al. (2018) index of capital restrictions on non-resident inflows.
- Fixed effects: country fixed effects (γ_i) and time fixed effects (θ_t).
- Endogeneity considerations:
  - Reverse causality (controls tightened when inflows increase) could bias estimates toward finding a positive effect of controls on inflows.
  - Using lags and the descriptive result that tightening timing is not clustered around surges helps mitigate endogeneity concerns.

### Main econometric findings — aggregate surge probability (Table 1 highlights)
- Coefficients for "Whether tightening in the past 4 qtrs" (standard errors in parentheses):
  - -0.405 (0.270)
  - -0.456* (0.246)
  - -0.356** (0.160)
- Other notable coefficient signs and significance (as reported in Table 1):
  - Diff GDPGrowth: 0.181*** (0.0612); 0.185*** (0.0611); 0.185*** (0.0613) and variants in other columns.
  - Diff IR: 0.0581*** (0.0219); 0.0570*** (0.0218); 0.0549** (0.0219) and variants.
  - Composite Risk Index: positive and significant across specifications (e.g., 0.123*** (0.0300) ... 0.160*** (0.0280) ... 0.113*** (0.0168)).
  - Foreign Exchange reserves as % of GDP: mixed signs and magnitudes (e.g., -0.00294 (0.00298), 0.000394 (0.00246), -0.00260* (0.00157)).
  - FD: Financial development index: positive coefficients, some significant (e.g., 0.832 (1.600); 4.772*** (1.620); 3.462*** (1.092) in various columns).
  - Inflow restrictions index (Fernandez et al.): mixed and sometimes significant (e.g., 0.691* (0.369) in one block).
- Robustness of main result:
  - Using alternative measures of tightening:
    - Net tightening in past four quarters (sum of tightenings minus sum of easings).
    - Sum of tightenings and easings separately.
  - Results are robust: tightening is effective in lowering probabilities of surges for all three surge definitions.
  - Results are also largely robust when using surge definition based on net inflows (gross nonresident inflows – gross resident outflows).
  - Caveat: unweighted counts of measures implicitly assume equal importance across measures; countries often use controls as part of a broader policy package (e.g., macroprudential tools, FX interventions), which are not explicitly controlled for and could confound joint effects.

### Asset-level analysis and findings (three-dimensional panel)
- Motivation:
  - Surges can be asset-specific; aggregate analysis may miss asset-level episodes and cannot match controls to the asset class where they are applied.
- Specification: three-dimensional panel (country-time-asset). Dependent variable = 1 if asset-level surge; ControlAsset = 1 if controls introduced on nonresident inflows in that specific asset class in the past four quarters. Asset fixed effects μ_n are included.
- Main results (Table 2 highlights):
  - "Whether tightening in the past 4 qtrs" coefficients (standard errors in parentheses):
    - -0.300* (0.182)
    - -0.266* (0.139)
  - Diff GDPGrowth positive and significant in several specifications (e.g., 0.0712*** (0.0257) ...).
  - Diff IR positive and significant in several specifications (e.g., 0.0407*** (0.00684) ...).
  - Composite Risk Index: positive and often significant (e.g., 0.0943*** (0.0121)).
  - FD: Financial development index: positive and significant (e.g., 1.992*** (0.726); 3.402*** (0.590)).
  - Inflow restrictions index at asset/flow level (Fernandez et al.): negative and significant in Forbes asset-level specifications (e.g., -0.573*** (0.186); -0.647*** (0.186); -0.625*** (0.187)).
  - Sum of net tightening in past 4 qtrs: -0.107** (0.0443); -0.130*** (0.0365) in asset-level specifications.
  - Sum of tightenings in past 4 qtrs: -0.198* (0.110); -0.201** (0.0900).
  - Sum of easings in past 4 qtrs: positive in some specifications (e.g., 0.0961** (0.0471); 0.121*** (0.0383)).
- Interpretation:
  - Controls mapped directly to assets to which they are applied yield generally stronger results: controls reduce the probability of future asset-specific surges.
  - Higher pre-existing restrictiveness (Fernandez index) at the asset/flow level is associated with lower future surges in the baseline (Forbes) asset-level specification.

*Source: IMF Working Paper excerpt — section 2.3 “Surges of Capital Inflows” (figures, tables, and text as provided).*

### 3.3 Multilateral Considerations of Capital Controls: International Spillovers

### 3.3 Multilateral Considerations of Capital Controls: International Spillovers

### Grouping and spillover metrics
- Region based spillover: seven regional groups—Asia, Sub-Saharan Africa, Middle East and North Africa, Western Europe, Eastern Europe and Central Asia, Latin America, and North America. Assess how controls affect inflows in other countries in the same group.
- Return based spillover: countries divided into four groups based on average real GDP growth rates across the sample period; the low-return group in period t consists of countries whose sample average growth rates are in the bottom 25 percentile in that period.
- Risk based spillover: same methodology as return based spillover but using the composite risk index as the measure.
- Large EMDEs (BRICS) based spillover: BRICS (Brazil, Russia, India, China, and South Africa) assumed to generate important spillovers to other BRICS; for non-BRICS, spillovers arise from the regional BRICS country (for example, Brazil for Latin America, India and China for Asia).

### Empirical specification to assess multilateral spillovers
- Regression estimated (binary dependent variable equal to one if a country experiences a surge based on aggregate non-resident inflow):
  - Includes own-country controls: sum of own tightenings/easings in past 4 quarters, lagged changes in GDP growth (L.Diff GDPGrowth) and interest rates (L.Diff IR), composite risk index, foreign exchange reserves as % of GDP, inflow restrictions index, usual controls X, country fixed effects γi, time fixed effects θt, and error term εi,t.
  - Adds a spillover term: CIControls_s−i,t−1,t−4, the weighted average of controls introduced in the past 4 quarters in other countries in the group.
  - For region/return/risk grouped spillovers:
    CIControls_s−i = (Σ_{j= s−i} y_j * CIControls_j) / (Σ_{j= s−i} y_j)
    where y_j is non-resident inflows in country j (flow-weight).
  - For BRICS-based spillovers:
    - For each BRICS, CIControls_s−i is the sum of the number of policy changes in any of the other BRICS in a given quarter.
    - For non-BRICS, the variable is the sum of the number of measures introduced by the regional BRICS country.

### Selected regression results (Table 3)
- Own-country policy actions (sum in past 4 quarters):
  - Sum of own tightenings in past 4 qtrs: -0.136* (0.0827), -0.140* (0.0823), -0.144* (0.0827), -0.152* (0.0811) across specifications.
  - Sum of own easings in past 4 qtrs: 0.0358 (0.0338), 0.0379 (0.0336), 0.0421 (0.0336), 0.0425 (0.0340) across specifications.
- Lagged macro and risk controls:
  - L. Diff GDPGrowth: 0.200*** (0.0628), 0.196*** (0.0621), 0.186*** (0.0612), 0.181*** (0.0606), 0.182*** (0.0612), 0.176*** (0.0607), 0.187*** (0.0620), 0.181*** (0.0612).
  - L. Diff IR: 0.0526** (0.0219), 0.0579*** (0.0219), 0.0563*** (0.0218), 0.0621*** (0.0218), 0.0553** (0.0220), 0.0613*** (0.0220), 0.0528** (0.0222), 0.0595*** (0.0222).
  - Composite Risk Index: 0.120*** (0.0302), 0.124*** (0.0300), 0.112*** (0.0301), 0.116*** (0.0299), 0.118*** (0.0301), 0.121*** (0.0299), 0.111*** (0.0302), 0.115*** (0.0299).
  - Foreign Exchange reserves as % of GDP: -0.00215 (0.00288), -0.00214 (0.00288), -0.00228 (0.00294), -0.00227 (0.00294), -0.00291 (0.00301), -0.00291 (0.00302), -0.00275 (0.00291), -0.00272 (0.00291).
  - Inflow restrictions index (Fernandez et.al.): -0.347 (0.626), -0.375 (0.589), -0.434 (0.628), -0.467 (0.591), -0.403 (0.623), -0.422 (0.588), -0.494 (0.630), -0.504 (0.592).
- Multilateral spillover coefficients (selected):
  - Region based spillover: Tightenings 0.0345 (0.0234), 0.0356 (0.0234).
  - Region based spillover: Easings -0.0150* (0.00857), -0.0154* (0.00857).
  - Risk based spillover: Tightenings 0.0931** (0.0381), 0.0948** (0.0377).
  - Risk based spillover: Easings -0.0221 (0.0151), -0.0228 (0.0152).
  - Return based spillover: Tightenings 0.108 (0.0791), 0.105 (0.0791).
  - Return based spillover: Easings -0.00996 (0.0100), -0.0105 (0.0102).
  - BRICS based spillover: Tightenings 0.0877* (0.0473), 0.0858* (0.0470).
  - BRICS based spillover: Easings 0.0294 (0.0181), 0.0274 (0.0180).
- Sample size and panels:
  - Observations: 2,382
  - Number of panels: 33
- Notes on significance: Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

### Main findings on spillovers
- Tightening of capital controls in one country:
  - Reduces the probability of future surges domestically (own-country negative and statistically significant coefficients for sum of own tightenings).
  - Increases the probability of surges in other countries with similar risk index (risk-based within-group spillover: tightenings 0.0931** and 0.0948**).
  - Increases the probability of surges in other countries when tightening occurs in BRICS (BRICS based spillover: tightenings 0.0877* and 0.0858*).
- Easing of capital controls in one country:
  - Can reduce the probability of surges in other countries in the same region (region-based easings -0.0150* and -0.0154*).
- Relative magnitudes:
  - Within-group spillovers are largest among countries with similar composite risk index and from controls introduced in BRICS countries.

### Policy implications and conclusions
- Capital controls can be an important tool to limit inflow surges: tightening is associated with a reduced probability of future surges domestically.
- Capital controls can generate significant adverse spillovers by deflecting flows to countries with similar fundamentals—particularly those with similar risk profiles as measured by the composite risk index.
- The presence of large multilateral spillovers, especially from BRICS and among countries with similar risk, suggests potential justification for international cooperation in the use of capital controls.
- An optimal policy response should weigh pros and cons of available tools depending on a country’s initial conditions and nature of the shock, as envisaged in the integrated policy framework and the IMF’s IV on the liberalization and management of capital flows.

*Source: 3.3 Multilateral Considerations of Capital Controls: International Spillovers — IMF Working Paper excerpt.*

### Annex Table I – Tightened Capital Control

### Annex Table I – Tightened Capital Control

### Debt measures (measures on non-resident inflows in debt instruments)
- Argentina 2005q2: (1) The minimum period that nonresident investors must hold incoming capital (excluding foreign trade operations and direct investment) in the country was raised to 365 days from 180 days; (2) All new local market debts denominated in foreign currency and any new debts or rollover of existing debts by residents with external creditors were subject to a retention of 365 days; (3) A 30 percent unremunerated deposit requirement of one year was imposed on foreign portfolio investment, excluding initial public offerings, foreign direct investment, and trade finance flows.
- Argentina 2009q2: BCRA approval requirement was introduced to access the MULC for the purpose of obtaining foreign exchange to repatriate the proceeds of portfolio investments by nonresidents if the foreign beneficiary is an individual or legal entity resident, incorporated, or domiciled in associated dominions, jurisdictions, territories, or states included on the list in Decree No. 1344/98 (Communication A 4940).
- Brazil 2009q4: A 2 percent tax (IOF) was introduced on foreign equity and fixed income inflows (but not FDI), with no discrimination between long- and short-term flows.
- Brazil 2010q4: (1) The IOF tax rate for foreign inflows in fixed-income instruments was increased from 2 percent to 4 percent; (2) The IOF tax rate for foreign inflows in fixed-income instruments was increased from 4 percent to 6 percent.
- Iceland 2016q2: A special 40 percent reserve requirement (with a 12-month holding period, at a zero interest rate) on specific debt inflows was introduced.
- Indonesia 2010q3: BI imposed a one-month minimum holding period for all investors (both domestic and foreign) for purchases of SBIs in both the primary and secondary markets.
- Indonesia 2011q2: The minimum holding period for all investors (both domestic and foreign) for purchases of Bank Indonesia certificates (SBIs) in both the primary and secondary markets was lengthened from one month to six months.
- Thailand 2006q4: (1) Thai financial institutions were not allowed to issue or sell bills of exchange in baht of any maturity to nonresidents. (2) Investments of more than B 50 million a consolidated entity in short-term debt and related products (not exceeding six months) issued by local financial institutions in the primary market without underlying transactions are not allowed. (3) Financial institutions were not allowed to sell and buy foreign currencies to and from nonresidents or credit or debit their baht account resulting from investments in government or BOT debt securities if the duration of such investments was less than or equal to three months. (4) The limit of B 50 million was applied to debt securities with maturities not exceeding six months (previously, three months) issued by domestic financial institutions and sold to nonresidents without underlying transaction. (5) Financial institutions were not allowed to sell or buy any types of debt securities to and from nonresidents through sell-and-buy-back transactions for any maturity.
- Ukraine 2010q1: Hryvnia funds from foreign investments in Ukraine must be deposited for five days in an analytical account of an authorized bank before conversion into foreign currency and transfer abroad.

### Equity measures (measures on non-resident inflows in equity/portfolio instruments)
- Argentina 2005q2: (1) The minimum period that nonresident investors must hold incoming capital (excluding foreign trade operations and direct investment) in the country was raised to 365 days from 180 days; (2) All new local market debts denominated in foreign currency and any new debts or rollover of existing debts by residents with external creditors were subject to a retention of 365 days; (3) A 30 percent unremunerated deposit requirement of one year was imposed on foreign portfolio investment, excluding initial public offerings, foreign direct investment, and trade finance flows.
- Argentina 2005q4: The exemption from the 30 percent deposit requirement for the primary issuance of trust securities was removed. All receipts on the local market of foreign exchange to be used to underwrite primary issues of securities, bonds, or equity certificates issued by the trustee of a trust, which may or may not be publicly offered and listed on self-regulated markets, when the stated requirements apply to any of the assets of the trust are also covered.
- Brazil 1999q3: A 15 percent tax was levied on foreign investment profits from Brazilian fixed-income funds.
- Brazil 2009q4: (1) A 2 percent tax (IOF) was introduced on foreign equity and fixed income inflows (but not FDI), with no discrimination between long- and short-term flows; (2) A 1.5 percent tax was introduced on certain trades involving ADRs issued by Brazilian companies. The tax is charged when foreign investors convert ADRs for Brazilian companies into receipts for shares issued locally.
- Brazil 2010q4: (1) The IOF rate was increased from 2 percent to 4 percent on inflows, including through simultaneous foreign exchange operations by foreign investors to purchase shares of participation investment funds, emerging companies investment funds, or investment funds in shares of equity funds regulated by the CVM; (2) The IOF rate was increased from 4 percent to 6 percent on inflows, including through simultaneous foreign exchange operations by foreign investors to purchase shares of participation investment funds, emerging companies investment funds, or investment funds in shares of equity funds regulated by the CVM.
- Colombia 2007q2: Foreign portfolio investments were made subject to a six-month URR of 40 percent to be held in the BR.
- Iceland 2008q4: The purchase of securities for foreign currency was prohibited except for reinvestment of investments made prior to that date. The proceeds must be reinvested in the same type of instrument within two weeks. The issuance and sale of securities denominated in foreign currency were prohibited if the settlement takes place in Icelandic krónur. If the issuance is denominated in Icelandic krónur, the proceeds from the sale must be deposited to a króna-denominated account, in the issuer’s name, with an AD in Iceland. Króna-denominated financial instruments may not be settled in foreign currency, and the proceeds must be deposited to the nonresident’s account with an Icelandic AD. However, residents may issue foreign-currency-denominated securities abroad.
- Malaysia 1998q3: Nonresident sellers of Malaysian securities were required to hold on to their ringgit proceeds for at least one year and to carry out all purchases and sales of ringgit securities through authorized depository institutions.
- Pakistan 2010q3: A 10 percent tax on capital gains on stocks/shares/securities held for six months or less applies. A 7.5 percent capital gains tax applies to holdings of stocks/shares/securities exceeding six months. No tax applies to capital gains arising from holdings exceeding 12 months.

### FDI measures (measures on non-resident inflows classified as foreign direct investment)
- Argentina 2003q4: Repatriation of inward direct investments by nonresidents was subject to a monthly limit of US$5,000 or the equivalent on sales of foreign exchange to nonresidents.
- Argentina 2004q4: Nonresidents’ purchases of foreign exchange were limited to US$2 million a month.
- Argentina 2011q4: The CB introduced new regulations setting additional requirements for the repatriation of direct investments made in Argentina by foreign investors (Communication A 5237). Since that date, access to the foreign exchange market for the repatriation of all forms of direct investment in the country without prior BCRA approval require evidence that funds for new investment involving new contributions and purchases of shares in domestic enterprises and real estate were brought in through the MULC and were disbursed in foreign exchange by the foreign investor on or after October 28, 2011. Transactions to repatriate direct investment subject to but not in compliance with the above requirements requires CB approval.
- Australia 2010q2: Residential real estate purchased by temporary residents directly or through Australian companies or trusts, where the real estate is an established dwelling intended to be the acquirer’s place of residence, constitutes single blocks of vacant land, or is a new dwelling, is subject to notification and approval requirements under the Foreign Acquisitions and Takeovers Act 1975. Previously, this was allowed under exemptions from the notification and approval requirements.
- Canada 2013q4: New investment rules for state-owned enterprises (SOEs), announced in December 2012 and supplementing 2007 SOE guidelines, stipulate that future SOE bids for control of a Canadian oil-sands business will be approved on an “exceptional basis only.” The definition of a SOE was also altered to include entities that are influenced directly or indirectly by a foreign government. The government intends to liberalize the review threshold under the Investment Canada Act to Can$1 billion in enterprise value over four years only for private sector investors. The review threshold for foreign SOEs will remain Can$330 million in asset value.
- Singapore 2011q4: A new additional buyer’s stamp duty (ABSD) was imposed on purchases of certain categories of residential property: (1) Foreigners and nonindividuals (corporate entities) that buy residential property must pay a 10 percent ABSD. (2) Permanent residents who own one residence and buy second and subsequent residential property must pay a 3 percent ABSD. (3) Singapore citizens (Singaporeans) who own two and buy third and subsequent residential property must pay a 3 percent ABSD. The ABSD is applied on top of the existing BSD rates, which have not been modified.
- Singapore 2013q1: Additional Buyer’s Stamp Duty (ABSD) is imposed on certain categories of residential property purchases as follows: (1) Foreigners and corporate entities buying residential property must pay ABSD of 15 percent (previously 10 percent). (2) Permanent residents (a) buying their first residential property must pay ABSD of 5 percent (previously zero); and (b) owning one and buying second and subsequent residential property must pay ABSD of 10 percent (previously 3 percent). (3) Singaporeans (a) owning one and buying a second residential property must pay ABSD of 7 percent (previously zero); and (b) owning two and buying a third or subsequent residential property must pay ABSD of 10 percent (previously 3 percent). The ABSD is applied on top of the existing Buyer’s Stamp Duty rates, which have not been modified.
- Vietnam 2014q3 FDI: Circular 19/2014 TT-NHNN of August 11, 2014, included further supplemental regulations on inward direct investment in Vietnam.

### Other measures (broader restrictions, reserve requirements, maturities, taxes, reporting)
- Argentina 2003q2: The minimum maturity requirement for new external borrowing from nonresidents to residents was increased to 180 days from 90 days.
- Brazil 1998q1: The minimum average maturities for external loans were increased to two years for new loans and to one year for renewed loans.
- Brazil 2011q1: The IOF rate for inflows related to external loans with a maximum maturity of 360 days was increased from zero to 6 percent.
- China 2007q1: The SAFE reduced the 2007 short-term external debt quotas of Chinese-funded banks to 30 percent and those of nonbank financial institutions and foreign-funded banks to 60 percent of their 2006 equivalents. In addition, Chinese-funded banks were required to reduce their outstanding short-term external debts to 30 percent or less and nonbank financial institutions and foreign-funded banks to 60 percent or less of their 2006 quotas by March 31, 2008.
- Colombia 1996q1: Foreign loans with maturities ranging from 1 day to 3 years were subject to a nonremunerated deposit requirement of 50 percent of the loan. Loans of over 6 months for imports are subject to a nonremunerated deposit requirement; these deposits are held for 18 months.
- Ukraine 2013q3: FDI coming into Ukraine is subject to a 50 percent surrender requirement.
- Ukraine 2014q3: The surrender requirement on inward FDI was increased from 50 percent to 100 percent.
- Uruguay 2012q4: The Central Bank of Uruguay imposes a 40 percent reserve requirement on the daily average (weekends and holidays included) of securities holdings in pesos (for the penultimate calendar month) in excess of the August 16, 2012, balance.
- Uruguay 2013q3: The reserve requirement of 50 percent was introduced on the average daily securities holdings (for the penultimate calendar month) of locally issued government securities in excess of the balance of June 7, 2013 held for and on behalf of nonresidents. The reserve requirement applies on government securities denominated in local currency or inflation-indexed units, must be met in monthly fixed-term deposits in pesos and does not earn interest.
- Vietnam 2014q1 Other: For short-term credit, borrowing firms must observe SBV conditions. For medium- and long-term loans, firms must register the borrowing and repayment schedules with the SBV before disbursement, within 30 working days of signing the loan contract. For extended loans, firms must register loans within 30 working days of signing an extension contract for short-term loans whose total period, including the extension, exceeds one year. Firms’ external borrowing must be within the government-approved annual limit and meet the SBV’s requirements. The borrowing and repayment schedules must be reported periodically to the SBV. Resident individuals were required to borrow and repay overseas loans based on the principle of self-borrowing and are self-liable in accordance with government regulations (Ordinance No. 06/2013/UBTVQH13 of March 18, 2013, amending and supplementing the Ordinance on Foreign Exchange No. 28/2005/PL-UBTVQH of December 13, 2005).

### Notable numeric thresholds and requirements (exact values preserved)
- 365 days (Argentina minimum holding period raised from 180 days to 365 days; retention of 365 days on new local market foreign-currency debt).
- 30 percent (Argentina unremunerated deposit requirement of 30 percent for foreign portfolio investment).
- 2 percent, 4 percent, 6 percent (Brazil IOF tax rates on inflows: introduced at 2 percent in 2009q4; increased to 4 percent and then to 6 percent in 2010q4).
- 15 percent (Brazil 1999q3 tax on foreign investment profits from fixed-income funds).
- US$5,000 (Argentina 2003q4 monthly limit on repatriation of inward direct investments by nonresidents).
- US$2 million (Argentina 2004q4 nonresidents’ purchases of foreign exchange limited to US$2 million a month).
- 40 percent (Colombia URR of 40 percent for foreign portfolio investments in 2007q2; Uruguay 2012q4 reserve requirement of 40 percent on excess securities holdings).
- 50 percent (Uruguay 2013q3 reserve requirement of 50 percent on average daily holdings in excess of benchmark; Ukraine 2013q3 FDI surrender requirement of 50 percent before 2014q3).
- 100 percent (Ukraine 2014q3 surrender requirement increased from 50 percent to 100 percent).
- One-month, six-month, 12-month holding periods (Indonesia SBIs holding period changes: one month in 2010q3; lengthened to six months in 2011q2. Iceland 2016q2 reserve with 12-month holding period).
- B 50 million, B 10 million (Thailand limits: B 50 million consolidated entity limit in 2006q4; limit decreased to B 10 million in 2008q1).
- Can$1 billion, Can$330 million (Canada 2013q4 Investment Canada Act review thresholds: intended liberalization to Can$1 billion for private investors; foreign SOE threshold remains Can$330 million in asset value).

*Annex Table I – Tightened Capital Control (measures on non-resident inflows) — extracted entries as provided in the source content.*

### Annex Table 2 –   Variable Description and

### Annex Table 2 –   Variable Description and Sources

### Variable descriptions and sources
- Capital Controls  
  - Changes in measures restricting non-resident gross inflows, excluding measures on derivatives.  
  - Source: IMF AREAER database.

- Non-resident Gross Inflows  
  - Non-resident capital inflows minus outflows (Net Flows and Net Flows as percent of GDP).  
  - Source: IMF’s Financial Flow Analytics (FFA) database.

- GDP growth Differential  
  - Difference between domestic and U.S. real GDP growths.  
  - Source: IMF WEO database, Haver.

- Interest Rate Differential  
  - Difference between domestic and U.S. interest rates.  
  - Source: IMF WEO database, Haver.

- Foreign Exchange Reserves  
  - Foreign Exchange Reserves, as percentage of Nominal GDP.  
  - Source: IMF IFS.

- Composite Risk Index  
  - Composite Political, Financial, Economic Risk Rating for a country (CPFER). Ranging from Very High Risk (0 - 49.5) to Very Low Risk (80 - 100).  
  - Source: Political Risk Service (PRS).

- FD  
  - Financial Development Index.  
  - Source: Sahay and others (2015).

- Inflow Restrictions Index  
  - index of capital restrictions on non-resident inflows (All asset categories).  
  - Source: Fernandez and others (2018).

- Contagion (BRICS)  
  - For each of the BRICS, is the sum of the number of policy changes in any of the other BRICS in each quarter.  
  - For the non-BRICS (non-Advanced), the variable is the sum of the number of measures introduced by the regional BRICS country (i.e. Brazil for Latin America, China and India for Asia, Russia for emerging Europe and South Africa for Africa) in a given quarter.  
  - Advanced countries: the variable is the sum of the number of measures introduced by the Advanced countries in each quarter.  
  - Source: Pasricha and others (2018), authors’ calculations.

- Contagion (Risk, Return)  
  - Computed as the flow weighted fraction of controls for countries in the group that tightened inflow controls. Groups are defined based on Real GDP growth (return) and the PRS composite risk index (risk).  
  - Source: Giordani and others (2017), authors’ calculations.

- Contagion (Regional)  
  - Computed as the flow weighted fraction of controls for countries in the region that tightened inflow controls. We have 7 regional groups: Asia, Eastern Europe & Central Asia, Latin America & Caribbean, Middle East & North Africa, North America, Sub-Saharan Africa and Western Europe.  
  - Source: Giordani and others (2017), authors’ calculations.

### Summary statistics (Annex Table 3)
- Index of capital restrictions on NR inflows: mean 0.44, s.d. 0.32, p50 0.45, min 0.00, max 1.00, N 14,720
- GDP growth differential: mean 0.42, s.d. 1.76, p50 0.38, min -10.50, max 13.89, N 13,000
- Interest Rate differential: mean 6.89, s.d. 13.19, p50 4.25, min -6.50, max 208.60, N 12,976
- Foreign Reserves, percent of GDP: mean 62.68, s.d. 63.13, p50 45.84, min 0.82, max 448.32, N 14,508
- Financial Development Index: mean 0.44, s.d. 0.24, p50 0.38, min 0.05, max 1.00, N 14,352
- BRICS dummy variable: mean 0.13, s.d. 0.33, p50 0.00, min 0.00, max 1.00, N 14,880
- Contagion (BRICS, Gross Tightening): mean 0.22, s.d. 0.74, p50 0.00, min 0.00, max 7.00, N 14,880
- Contagion (BRICS, Gross Easing): mean 1.18, s.d. 2.28, p50 0.00, min 0.00, max 20.00, N 14,880
- Composite Risk Index: mean 71.38, s.d. 8.53, p50 70.50, min 41.00, max 92.75, N 14,768
- Risk Contagion (Flow weighted*) Time Variant (Gross Tightening): mean 0.12, s.d. 1.23, p50 0.00, min -8.53, max 22.12, N 14,880
- Risk Contagion (Flow weighted*) Time Variant (Gross Easing): mean 0.55, s.d. 3.81, p50 0.03, min -95.18, max 91.70, N 14,880
- Return Contagion (Flow weighted*) Time Variant (Gross Tightening): mean 0.13, s.d. 1.00, p50 0.00, min -0.74, max 19.60, N 14,880
- Return Contagion (Flow weighted*) Time Variant (Gross Easing): mean 0.79, s.d. 11.47, p50 0.00, min -9.99, max 246.47, N 14,880
- Regional Contagion (Flow weighted*) Gross Tightening: mean -1.04, s.d. 20.09, p50 0.00, min -353.49, max 32.80, N 14,880
- Regional Contagion (Flow weighted*) Gross Easing: mean -2.94, s.d. 55.78, p50 0.00, min -987.90, max 25.43, N 14,880
- * Excludes controls on Derivatives

### Methodology for the identification of surges — key points and comparisons
- Definitions and focus  
  - Use of terminology: gross inflows = net non-resident flows (transactions creating liabilities for residents); gross outflows = net resident flows (assets on residents’ balance sheet); net capital inflow = gross inflows minus gross outflows.  
  - Focus on private or nonofficial flows; flows resulting from FXI or from international financial institutions such as the IMF are not considered.  
  - The analysis uses gross definitions at aggregate and asset level (debt, equity, FDI, and other) to construct the surge variable.

- Issues with detrending and HP filter  
  - HP-filter choice of λ is not trivial and affects deviations from trend; HP filter depends on entire sample (past and future) and is sensitive to endpoints.  
  - Detrending with HP filter can obscure absolute magnitude of flows.

- Forbes and Warnock (2012, 2021) definition (gross flows, year-on-year changes in four-quarter rolling sum)  
  - Quarterly data adjusted by taking a year-on-year difference of the four-quarter rolling sum of capital flows.  
  - Uses a 5-year rolling window to calculate rolling means and standard deviations.  
  - Surge criteria (three conditions):  
    1. current year-over-year changes in four-quarter gross capital inflows is more than two standard deviations above the historical average during at least one quarter of the episode;  
    2. the episode lasts for all consecutive quarters for which the year-over-year change in annual gross capital flows is more than one standard deviation above the historical average; and  
    3. the length of the episode is greater than one quarter.  
  - Captures both intensity (at least one quarter ≥ two standard deviations above mean) and persistence (episode > one quarter).

- Ghosh and others (2014) definition (net flows, percent of GDP)  
  - Defines a surge as a net capital flow observation that lies in the top thirtieth percentile of both the country-specific and the full sample’s distribution of net capital flows, expressed in percent of GDP.  
  - Measures consider both absolute and relative dimensions via country-specific and full-sample distributions.  
  - Sensitive to short-lived large increases at quarterly frequency and dependent on sample used for distributions and percentiles.

- Data sources and counts  
  - Main database: IMF’s Financial Flow Analytics (FFA), drawing from annual and quarterly BoP and GDP WEO data over the period 1995 to 2018..  
  - Due to data gaps, the sample contains 3,349 observations of net capital flows.  
  - Using Ghosh methodology: of the 3,075 observations considered (those with GDP data), 718 correspond to surges.  
  - Using Forbes definition: 2,969 observations considered, 318 classified as surges.

- Comparison of definitions and outcomes  
  - Low correlation between the two measures: 0.17.  
  - Both measures capture similar time-series trends in fraction of countries experiencing surges (Figure A.1 in source).  
  - Cross-country fraction of quarters classified as surges varies by country and methodology; under Ghosh many countries have fraction of observations close to 30 percent.

- Simpler constant-threshold approach (net inflows thresholds as percent of GDP)  
  - Definition: country experiences a surge if net capital inflows are above a set threshold and optionally if surge lasts at least two quarters.  
  - Fraction of total observations (out of 3,075 datapoints) classified as surge:  
    - 3 percent of GDP threshold: 1 Quarter = 42.0 percent, 2 Quarters = 34.4 percent  
    - 5 percent of GDP threshold: 1 Quarter = 26.9 percent, 2 Quarters = 19.5 percent  
  - Even with 5 percent of GDP for at least two consecutive quarters, the fraction of periods classified as surges is relatively high.

*Do Capital Controls Limit Inflow Surges? Working Paper No. WP/2023/050*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023050-print-pdf.pdf_