## _wp1580 - REFERENCES

## Source details

**Canonical URL:** [_wp1580 - REFERENCES](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1580.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1580.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1580.pdf.json)

---

### I. Introduction: motivation and purpose
- International capital flows affect monetary and exchange rate policy choices via the policy trilemma and allow domestic investment to diverge from domestic savings.
- Capital flows can promote consumption smoothing and risky investment but can also transmit shocks or trigger sudden stops.
- Historical context and shifts in attitudes toward capital controls summarized (pre-World War I favorable; 1933 Keynes; Bretton Woods pervasive controls; reductions from late 1970s through 1990s; renewed debate in late 1990s/2000s).
- Empirical literature is mixed on effectiveness and costs of controls.
- Purpose of the paper:
  - Introduce a new dataset (based on Schindler (2009) methodology) with more countries, more asset categories, and more years.
  - Dataset: annual presence/absence of capital controls for 100 countries over the period 1995 to 2013.
  - Distinguishing feature: disaggregation by inflows vs outflows and by 10 asset categories, enabling 32 transaction-category series and carefully targeted aggregate measures.

### II. Dataset overview and extensions
- Coverage and period:
  - 100 countries.
  - 1995 to 2013.
- Extensions relative to earlier datasets:
  - Asset categories expanded from six to 10 by adding derivatives, commercial credit, financial guarantees, and real estate.
  - Nine new countries added (bringing total to 100) — selected as the nine with the largest populations in 2012 not in Schindler’s original set and included in the AREAER.
  - Sample period extended to 1995–2013.
- Data availability note:
  - Dataset is publicly available for download at the National Bureau of Economic Research website or at request from the authors (as reported in the source).

### III. Asset and transaction categorization (granularity)
- Ten asset categories with two-letter abbreviations:
  - mm: Money market instruments (original maturity of one year or less; includes certificates of deposit, bills of exchange).
  - bo: Bonds or other debt securities with an original maturity of more than one year.
  - eq: Equity, shares or other securities of a participating nature, excluding foreign direct investment.
  - ci: Collective investment securities (mutual funds, investment trusts).
  - fc: Financial credit and credits other than commercial credits granted by residents to nonresidents or vice versa.
  - de: Derivatives (rights, warrants, financial options and futures, swaps, foreign exchange without other underlying transactions).
  - cc: Commercial credits directly linked to international trade transactions or rendering of international services.
  - gs: Guarantees, Sureties and Financial Back-Up Facilities (securities pledged for payment or performance, performance bonds, standby letters of credit, financial backup facilities).
  - re: Real Estate transactions not associated with direct investment (financial investments in real estate or acquisition for personal use).
  - di: Direct Investment (transactions to establish lasting economic relations).
- Transaction-level disaggregation:
  - For mm, bo, eq, ci, de: four transaction categories recorded:
    - _plbn: Purchase Locally By Non-Residents (inflow control).
    - _siar: Sale or Issue Abroad By Residents (inflow control).
    - _pabr: Purchase Abroad By Residents (outflow control).
    - _siln (or _slbn): Sale or Issue Locally By Non-Residents (outflow control).
  - Real Estate includes three series: re_pabr, re_slbn, re_plbn.
  - For gs, fc, cc: only inflow (i) or outflow (o) broad classifications: gsi/gso, fci/fco, cci/cco.
  - Direct Investment includes three series: dii, dio, ldi.
- Aggregate scope in most disaggregated format: 32 transaction categories.

### IV. Construction methodology and coding rules
- Primary source: IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER) — captures de jure legal restrictions (does not measure enforcement or de facto restrictions).
- Start of disaggregated reporting: AREAER volume covering 1995 conditions (1996 volume); series begin in 1995 and include data through 2013, except bonds series begins in 1997 due to limited 1995–1996 coverage.
- Coding approach:
  - Narrative descriptions in AREAER translated into binary indicators: 1 = presence of a restriction; 0 = no restriction.
  - If no narrative in column three, code from column two: assign 0 for NO and 1 for YES.
  - AREAER special entries: n.r., n.a., and d.n.e. appear; about 2.8 percent are n.r. or n.a., and d.n.e. appears 15 times (0.03 percent). The online dataset retains n.r., n.a., d.n.e.; for paper statistics these are set to missing.
  - Definition of a control:
    - Control if narrative explicitly requires “authorization,” “approval,” “permission,” or “clearance.”
    - “Reporting,” “registration,” or “notification” requirements are not counted as controls.
    - Quantity restrictions (e.g., “ceiling”) are coded as controls.
    - Explicit “prudential” restrictions are coded as controls.
    - Restrictions based on political or national security reasons preventing flows to/from specific countries are not considered capital controls.
    - Sector- or state-reserved exceptions: restrictions specific to one sector (except financial system or pension funds) or to state-reserved areas (defense, central banking) are not categorized as capital controls; if unspecified, then treated as a control; restrictions covering more than one sector where private entrepreneurship is common are counted as capital controls.
  - Harmonization and revision:
    - Discrepancies with Schindler (2009) for 2005 were revised backwards until no discrepancy remained where needed.
    - Total of 145 observations (less than one percent of the original dataset) were modified; these are listed in the master data file.
  - Technical Appendix contains more detailed rules and guiding principles.

### V. Characteristics of the capital control indicators (values, aggregation)
- Construction of asset-direction indices:
  - Real Estate: one inflow category denoted rei and aggregate outflow reo formed by aggregating re_pabr and re_slbn.
  - For categories with two inflow/outflow subcomponents, aggregate inflow (or outflow) index = average of the two 0/½/1 indicators; thus mmi, mmo, boi, boo, eqi, eqo, cii, cio, dei, deo and reo take values 0, ½ or 1.
  - Footnote rule: when one subcategory missing, aggregate inflow/outflow entry scored with value of remaining subcategory.
  - Interpretation: 1 represents greater intensity of controls than ½.
- Full dataset contains 32 categories.
- Aggregation approach:
  - Construct indicators of inflow controls and outflow controls for the ten asset categories.
  - No aggregation needed for cc, fc, gs single-direction categories (cci/cco, fci/fco, gsi/gso).
  - Direct Investment categories (dii, dio, ldi) kept separate.

### VI. Prevalence, counts, and key statistics (Table 3 and aggregates)
- Prevalence summary (Figure 1 treatment counts ½ and 1 equally as "a control"):
  - 18 percent: liquidation of direct investment observations with controls.
  - 25 percent: inflow controls on Guarantees, Sureties and Financial Backup Facilities.
  - 50 percent or greater: inflow controls on Real Estate and outflow controls on Money Market Instruments, Bonds, Equities, Collective Investments, and Derivatives.
  - General pattern: except for Real Estate and Direct Investment, higher prevalence of controls on outflows than on inflows.
- Detailed counts for categories with values 0, 0.5, 1 (first block):
  - mmi: 0 -> 1,143; 0.5 -> 346; 1 -> 388; Total 1,877; Pr. Cntrl 0.39
  - mmo: 0 -> 917; 0.5 -> 367; 1 -> 589; Total 1,873; Pr. Cntrl 0.51
  - boi*: 0 -> 980; 0.5 -> 378; 1 -> 327; Total 1,685; Pr. Cntrl 0.42
  - boo*: 0 -> 807; 0.5 -> 356; 1 -> 517; Total 1,680; Pr. Cntrl 0.52
  - eqi: 0 -> 1,024; 0.5 -> 459; 1 -> 399; Total 1,882; Pr. Cntrl 0.46
  - eqo: 0 -> 914; 0.5 -> 388; 1 -> 584; Total 1,886; Pr. Cntrl 0.52
  - cii: 0 -> 1,152; 0.5 -> 360; 1 -> 335; Total 1,847; Pr. Cntrl 0.38
  - cio: 0 -> 892; 0.5 -> 398; 1 -> 577; Total 1,867; Pr. Cntrl 0.52
  - dei: 0 -> 1,073; 0.5 -> 219; 1 -> 452; Total 1,744; Pr. Cntrl 0.38
  - deo: 0 -> 890; 0.5 -> 310; 1 -> 585; Total 1,785; Pr. Cntrl 0.50
  - reo: 0 -> 1,084; 0.5 -> 395; 1 -> 388; Total 1,867; Pr. Cntrl 0.42
  - Note: *Data on Bonds available 1997-2013.
- Single-component categories (second block):
  - fci: 0 -> 1,205; 1 -> 685; Total 1,890; Pr. Cntrl 0.36
  - fco: 0 -> 1,119; 1 -> 767; Total 1,886; Pr. Cntrl 0.41
  - cci: 0 -> 1,337; 1 -> 546; Total 1,883; Pr. Cntrl 0.29
  - cco: 0 -> 1,225; 1 -> 644; Total 1,869; Pr. Cntrl 0.34
  - gsi: 0 -> 1,384; 1 -> 471; Total 1,855; Pr. Cntrl 0.25
  - gso: 0 -> 1,227; 1 -> 631; Total 1,858; Pr. Cntrl 0.34
  - dii: 0 -> 1,121; 1 -> 779; Total 1,900; Pr. Cntrl 0.41
  - dio: 0 -> 1,246; 1 -> 625; Total 1,871; Pr. Cntrl 0.33
  - ldi: 0 -> 1,546; 1 -> 334; Total 1,880; Pr. Cntrl 0.18
  - rei: 0 -> 828; 1 -> 1,034; Total 1,862; Pr. Cntrl 0.55
- Aggregate totals:
  - Total 0 entries: 23,469
  - Total 0.5 or 1 entries: 15,134
  - Grand total observations: 38,603
  - Overall proportion of observations with controls: 0.40
- Additional note: for the 11 asset/direction categories with 0/½/1 values, observations of 1 exceed observations of ½: 26 percent vs. 20 percent.

### VII. Correlations across asset/direction categories and implications for aggregation
- Correlation computation: correlations across all observations x(t), y(t) where x and y are asset/direction categories and t is time; correlations missing if variance of an indicator is zero.
- Selected exact correlations (Table 4, All 100 countries, 1995-2013):
  - Diagonal inflow vs outflow (same asset):
    - mmi vs mmo: 0.78
    - eqi vs eqo: 0.72
    - dei vs deo: 0.86
    - dii vs dio: 0.37
    - rei vs reo: 0.30
  - Inflow vs inflow example: eqi and cii = 0.70
  - Outflow vs outflow example: gso and cco = 0.74
- Broad findings:
  - Correlation between inflow and outflow controls for a given asset tends to be high; highest for Derivatives (0.86), lowest for Direct Investment (0.37) and Real Estate (0.30).
  - Correlations highest among Money Market Instruments, Bonds, Equities, Collective Investments, and Derivatives for both inflow and outflow controls.
  - Lowest inflow correlations involve Real Estate relative to other asset categories.
  - Correlations generally higher among outflow controls than among inflow controls.
- Open/Gate/Wall subgroup patterns:
  - Gate countries (48): lower correlations; only one correlation > 80 percent and 40 less than 40 percent.
  - Open and Wall combined (52): very high correlations: all outflow correlations > 80 percent; majority of inflow correlations (except those involving Real Estate) > 60 percent; one fifth of inflow correlations > 80 percent.
  - Implication: correlation structure informs which asset categories to include in aggregate indices, and Gate-country heterogeneity implies potential sensitivity of empirical results to aggregation choice.

### VIII. Aggregate indicators, nested indexes, and correlations (KC series)
- Definitions:
  - KC10: Average of Inflows and Outflows for mm, bo, eq, ci, de, re, fc, cc, gs, di.
  - KC9: Average of Inflows and Outflows for mm, bo, eq, ci, de, re, fc, cc, gs (all but di).
  - KC5: Average of Inflows and Outflows for mm, bo, eq, ci, de.
  - KC2: Average of Inflows and Outflows for mm, bo.
- Table 6 — correlations between a 9-Asset aggregate (excluding one asset) and the excluded asset:
  - Excluded asset correlations: mm 0.87, bo 0.83, eq 0.87, Fc 0.83, ci 0.88, de 0.87, re 0.61, cc 0.71, gs 0.79, di 0.77.
  - Interpretation: Real Estate, Commercial Credits, Direct Investment, and Guarantees are least correlated with the nine-asset aggregate; Money Market Instruments, Collective Investments, Derivatives and Equities are most highly correlated.
- Table 7 — correlations among aggregate measures (all 100 countries upper-triangular; 48 Gate countries lower-triangular):
  - All 100 countries:
    - KC10 with KC9: 0.995
    - KC10 with KC5: 0.954
    - KC10 with KC2: 0.924
    - KC9 with KC5: 0.958
    - KC9 with KC2: 0.928
    - KC5 with KC2: 0.971
  - Gate countries (48):
    - KC9 with KC10: 0.992
    - KC5 with KC10: 0.901
    - KC2 with KC10: 0.873
    - KC5 with KC9: 0.910
    - KC2 with KC9: 0.877
    - KC2 with KC5: 0.953
- Key implications:
  - For full sample, correlations among aggregates are very high (range 0.924 to 0.995).
  - For Gate countries, correlations are lower with greater range; e.g., KC2 vs KC10 = 0.873.
  - Thus, aggregate choice may matter more for Gate-country analyses.

### IX. Comparison with Quinn and Chinn-Ito indices — regression evidence
- Quinn (1997) index and Chinn-Ito index converted to [0,1] where larger values = more restrictions.
- Regressions of country-average Quinn and Chinn-Ito indices on country-average KC10:
  - Quinn regression reported as: "݊݊݅ݑܳ௜ൌ 0.004 (ଶ.଴ଵଽ) ൅0.71 (ଶ.଴ସଵ) 10ܥܭ ௜ ܴଶ ൌ0.77; ݊ൌ 90" (preserved as in source).
  - Chinn-Ito regression reported as: "݋ݐܫ݄݊݊݅ܥ௜ൌ 0.049 (ଶ.଴ଶହ) ൅0.91 (ଶ.଴ହଵ) 10ܥܭ ௜ ܴଶ ൌ0.77; ݊ൌ 99." (preserved as in source).
- Statistical interpretation:
  - Coefficient on KC10_i is significantly different from zero at very high levels of confidence in both regressions.
  - Test whether coefficient equals 1:
    - t-statistic for Chinn-Ito regression = 1.71.
    - t-statistic for Quinn regression = 7.21.
    - Null that coefficients equal 1 can be rejected at the 95 percent level in both cases, but not at the 90 percent level for the Chinn-Ito indicator.
- Diagnostic plotting:
  - Regression plots and scatter plots shown in Figure 4; large regression errors flagged:
    - Quinn: absolute regression error > 0.25 identified.
    - Chinn-Ito: absolute regression error > 0.20 identified.

### X. Patterns over time, income groups, and country-level relationships
- Dataset coverage by World Bank income group:
  - 42 high income countries, 32 upper middle income countries, 18 lower middle income countries, and eight low income countries.
- Average capital control indices over time and by income group:
  - On average, capital control index inversely related to income.
  - Left axis midpoint for High Income: about 0.15 (inflows) and about 0.17 (outflows).
  - Right axes midpoints for Upper Middle, Lower Middle, Low: about 0.53 (inflows) and 0.60 (outflows).
- Temporal patterns (sample-wide averages):
  - Average controls fell from "0.20 for inflows and 0.22 for outflows in the first years of the sample period" to:
    - "less than 0.10 in 2008 for inflows"
    - "0.12 in 2004 for outflows"
    - followed by a rise in subsequent years.
- Income-group specifics:
  - Low Income: large decline in early sample years, then increase, especially in outflow controls.
  - Middle Income groups: averages and ranges lower than Low Income but more than twice High Income averages.
  - High Income: lowest averages and ranges among groups.
- Country-level inflow vs outflow relationships:
  - For each country, KCINFLOW_i and KCOUTFLOW_i are defined as average controls on inflows and outflows over full sample.
  - Scatterplots show somewhat higher prevalence of outflow than inflow controls; difference more pronounced for Medium and Lower Income countries than for High Income countries.
  - Country-by-country correlation of inflow and outflow controls: "about 0.8" for both sets of countries (Open and Walls subsets noted).

### XI. Conclusions and dataset contribution
- The role of capital controls in macroeconomic toolkits remains contested; careful theoretical and empirical research needed.
- Contribution:
  - New granular dataset enabling detailed empirical investigation of capital controls and their effects.
  - Demonstrated construction, properties of granular series, and behavior of multiple aggregate indicators.
- Authors' expressed hope: "this dataset proves useful in moving forward our understanding of this important topic."

*Source: _wp1580 - REFERENCES*

### REFERENCES

### _wp1580 - REFERENCES

### I. Introduction: motivation and context
- International capital flows affect monetary and exchange rate policy choices via the policy trilemma and allow domestic investment to diverge from domestic savings.
- Capital flows can promote consumption smoothing and risky investment but can also transmit shocks or trigger sudden stops.
- Historical shifts in attitudes toward capital controls:
  - Pre-World War I: favorable view (Keynes quoted).
  - 1933: Keynes advocated making finance national.
  - Bretton Woods era: pervasive capital controls; these were reduced or eliminated beginning in the late 1970s and increasingly through the 1980s and 1990s.
  - Late 1990s and 2000s: renewed debate after crises; some IMF work accepts capital controls as part of a “policy toolkit” under certain circumstances.
- Empirical literature is mixed: some theoretical work finds controls can contribute to financial stability; other empirical studies emphasize ineffectiveness and potential costs.
- Research challenge: availability and granularity of indicators of capital controls for cross-country panel analysis.
- Purpose of this paper:
  - Introduce a new dataset (based on Schindler (2009) methodology) with more countries, more asset categories, and more years.
  - The new dataset: annual presence/absence of capital controls for 100 countries over the period 1995 to 2013.
  - Distinguishing feature: disaggregation by inflows vs outflows and by 10 asset categories, enabling 32 transaction-category series and carefully targeted aggregate measures.

### II. Dataset overview and extensions
- Coverage and period:
  - 100 countries.
  - 1995 to 2013.
- Relation to previous datasets:
  - Schindler (2009): covers 91 countries, 1995 to 2005, six asset categories, inflow and outflow restrictions.
  - Klein (2012): extended Schindler to 2006–2010, limited to 44 countries and inflow restrictions.
  - Fernández, Rebucci and Uribe (2014): extended to 2011 for the original 91 countries and considered inflows and outflows.
- Extensions in this dataset:
  - Asset categories: expanded from Schindler’s six to 10 by adding derivatives, commercial credit, financial guarantees, and real estate.
  - Countries: nine new countries added (bringing total to 100) — selected as the nine with the largest populations in 2012 that were not in Schindler’s original set and were included in the AREAER.
  - Sample period: extended to 1995–2013.
- Data availability note:
  - Dataset is publicly available for download at the National Bureau of Economic Research website or at request from the authors (as reported in the source).

### III. Asset and transaction categorization (granularity)
- Ten asset categories with two-letter abbreviations:
  - mm: Money market instruments (original maturity of one year or less; includes certificates of deposit, bills of exchange).
  - bo: Bonds or other debt securities with an original maturity of more than one year.
  - eq: Equity, shares or other securities of a participating nature, excluding foreign direct investment.
  - ci: Collective investment securities (mutual funds, investment trusts).
  - fc: Financial credit and credits other than commercial credits granted by residents to nonresidents or vice versa.
  - de: Derivatives (rights, warrants, financial options and futures, swaps, foreign exchange without other underlying transactions).
  - cc: Commercial credits directly linked to international trade transactions or rendering of international services.
  - gs: Guarantees, Sureties and Financial Back-Up Facilities (securities pledged for payment or performance, performance bonds, standby letters of credit, financial backup facilities).
  - re: Real Estate transactions not associated with direct investment (financial investments in real estate or acquisition for personal use).
  - di: Direct Investment (transactions to establish lasting economic relations).
- Transaction-level disaggregation:
  - For five asset categories (mm, bo, eq, ci, de), four transaction categories are recorded:
    - _plbn: Purchase Locally By Non-Residents (inflow control).
    - _siar: Sale or Issue Abroad By Residents (inflow control).
    - _pabr: Purchase Abroad By Residents (outflow control).
    - _siln (or _slbn for some mentions): Sale or Issue Locally By Non-Residents (outflow control).
  - Real Estate includes three series:
    - re_pabr, re_slbn, re_plbn.
  - For three asset categories (gs, fc, cc), only inflow (i) or outflow (o) broad classifications are available: gsi/gso, fci/fco, cci/cco.
  - Direct Investment includes three series: dii (inflows), dio (outflows), ldi (liquidation of direct investment).
- Aggregate scope in most disaggregated format:
  - Provides information on 32 transaction categories (across assets and transaction directions).

### IV. Construction methodology and coding rules
- Primary source:
  - IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER) — de jure legal restrictions (does not measure enforcement or de facto restrictions).
- Start of detailed reporting:
  - Fundamental change in reporting began with the 1996 volume of the AREAER (covering conditions in 1995), which included disaggregation by asset and by inflow/outflow distinctions; hence series begin in 1995 and include data through 2013.
  - Exception: very limited coverage for 1995 and 1996 for one category (bonds with maturity greater than one year), so that asset series begins in 1997.
- Coding approach:
  - Narrative descriptions in the AREAER are translated into binary indicators: 1 = presence of a restriction; 0 = no restriction.
  - Coding rules build on Schindler (2009) and are clarified and expanded; explicit criteria are provided (with further detail in a technical appendix available from the authors).
  - Harmonization and revision:
    - When discrepancies arose between Schindler’s dataset and this dataset for 2005 (the last year of Schindler’s dataset), revisions were made for that category in that year and backwards until no discrepancy remained.
    - If no discrepancy existed in 2005, no backward revision was done for that country/asset subcategory.
    - In total, 145 observations (less than one percent of the original dataset) were modified; these observations are listed in the master data file.

### V. Uses and analytical advantages
- The disaggregated data enable:
  - Analysis of co-movements of controls across asset types and between inflows and outflows.
  - Construction of aggregate measures targeted to the specific nature of research questions.
  - Time variation of aggregate measures as indicators of intensity of restrictions on international capital movements.
- The dataset’s wider country coverage and extended time span make it potentially valuable for empirical research and policy analysis on capital account regulations.

*Source: _wp1580 - REFERENCES*

### 1. The annual information from the AREAER reports comes with three columns;

### _wp1580 - 1. The annual information from the AREAER reports comes with three columns;

### Coding rules for AREAER entries
- The AREAER report structure:
  - Column 1: asset subcategory.
  - Column 2: contains a YES, a NO, or no entry.
  - Column 3: narrative information.
- Primary coding procedure:
  - If there is no narrative information in column three, code based on column two: assign 0 for NO and 1 for YES.
  - If there is information in column three, code based on the narrative information in that column.
- Special narrative entries and treatment:
  - The AREAER narrative is limited to either n.r. or n.a. in about 2.8 percent of the cases in the data.
  - The entry n.a. is used by the IMF “when it is unclear whether a particular category or measure exists – because pertinent information is not available at the time of publication.”
  - The entry n.r. is used when “members have provided the IMF staff with information that a category or an item is not regulated.”
  - The dataset includes the category d.n.e. representing “does not exist”; this appears only 15 times in the entire dataset (0.03 percent of the dataset).
  - The dataset available online retains the n.r., n.a., and d.n.e. entries, but for the statistics presented in the paper entries with any of these three classifications are set to missing.
- Additional documentation:
  - A more detailed description of rules and guiding principles is contained in the Technical Appendix.

### Definitions of what constitutes a control
- A control is deemed to be in place when the narrative alludes to a transaction explicitly requiring “authorization,” “approval,” “permission,” or “clearance” from a public institution.
- A requirement of “reporting,” “registration,” or “notification” is not counted as constituting a control.
- A quantity restriction on any investment (e.g., in the form of “ceiling”) is coded as a control.
- An explicit allusion to a restriction for “prudential” considerations is deemed to be a control.
- Restrictions based on political or national security reasons that prevent capital flows to and from specific countries are not considered capital controls.
- Sector- or state-reserved exceptions:
  - Restrictions that apply specifically to transactions for only one sector (except the financial system or for pension funds) and/or to areas reserved for state control (such as defense, security, central banking, etc.) are not categorized as capital controls.
  - If a restriction does not specify which areas other than defense are reserved for state control, then the restriction is categorized as a control.
  - Restrictions are counted as a capital control if they cover more than one sector in which private entrepreneurship is common; such restrictions are deemed to have a macroeconomic impact.

### Aggregation and indicator construction
- The full dataset contains 32 categories (the full set of 32 categories presented in Table 1).
- Basic aggregation approach:
  - Construct indicators of inflow controls and outflow controls for the ten asset categories.
  - No aggregation is required for the asset categories of Commercial Credits, Financial Credits or Guaranties, Sureties and Financial Backup Facilities because the dataset only includes their inflow (cci, fci and gsi) and outflow (cco, fco and gso) categories; the value of each of these indicators will be either 0 or 1.
- Treatment of Direct Investment categories:
  - Direct Investment is not aggregated in this paper; the three categories are kept separate and denoted as dii, dio, and ldi, all of which will have values of either 0 or 1.

*Source: _wp1580 - 1. The annual information from the AREAER reports comes with three columns;*

### 1. In the case of Real Estate, there is only one inflow category (which we denote rei), but there

### III. CHARACTERISTICS OF THE CAPITAL CONTROL INDICATORS

### Construction of asset-direction indices
- Real Estate has one inflow category denoted rei and an aggregate outflow category reo formed by aggregating re_pabr and re_slbn.
- For categories with two inflow/outflow subcomponents, aggregate inflow (or outflow) index = average of the two 0/½/1 indicators.
- As a result the values of mmi, mmo, boi, boo, eqi, eqo, cii, cio, dei, deo and reo take values 0, ½ or 1.
- Interpretation note: an entry of 1 can be interpreted as representing greater intensity of controls than an entry of ½.
- Footnote rule: When there is a missing value in one of the two inflow or outflow subcategories, score the aggregate inflow or outflow entry with the value taken by the remaining subcategory.

### Dataset coverage and country classification
- Dataset covers 100 countries over the period 1995 to 2013.
- Country counts by World Bank Income Group: 42 high income countries, 32 upper middle income countries, 18 lower middle income countries, and eight low income countries.
- Klein (2012) Open/Gate/Wall classification:
  - Open: virtually no capital controls over the sample period.
  - Wall: pervasive controls across all, or almost all, categories of assets.
  - Gate: uses capital controls episodically.
- Definition thresholds (note text):
  - Open countries: capital controls on less than 10 percent of transactions subcategories on average over the sample period and do not have any years in which controls are on more than 20 percent of subcategories. (Note in Table 2 a slightly different phrasing: less than 15 percent and no year >25 percent is used for the 36 Open countries in practice.)
  - Walls: capital controls on more than 70 percent of subcategories on average and no year with controls on less than 60 percent.
  - Gate: neither Walls nor Open.

### Prevalence of controls by asset and direction (Figure 1 summary)
- Figure 1 treats ½ and 1 equally as "a control" for prevalence calculations.
- Prevalence ranges reported in text:
  - 18 percent of observations: liquidation of direct investment.
  - 25 percent: inflow controls on Guarantees, Sureties and Financial Backup Facilities.
  - 50 percent or greater: inflow controls on Real Estate and outflow controls on Money Market Instruments, Bonds, Equities, Collective Investments, and Derivatives.
- General pattern: except for Real Estate and Direct Investment, higher prevalence of controls on outflows than on inflows.

### Detailed prevalence by asset sub-categories (Table 3 exact figures)
- For categories that take values 0, 0.5 or 1 (first block), counts by value and total observations and proportion with controls:
  - mmi: 0 -> 1,143; 0.5 -> 346; 1 -> 388; Total 1,877; Pr. Cntrl 0.39
  - mmo: 0 -> 917; 0.5 -> 367; 1 -> 589; Total 1,873; Pr. Cntrl 0.51
  - boi*: 0 -> 980; 0.5 -> 378; 1 -> 327; Total 1,685; Pr. Cntrl 0.42
  - boo*: 0 -> 807; 0.5 -> 356; 1 -> 517; Total 1,680; Pr. Cntrl 0.52
  - eqi: 0 -> 1,024; 0.5 -> 459; 1 -> 399; Total 1,882; Pr. Cntrl 0.46
  - eqo: 0 -> 914; 0.5 -> 388; 1 -> 584; Total 1,886; Pr. Cntrl 0.52
  - cii: 0 -> 1,152; 0.5 -> 360; 1 -> 335; Total 1,847; Pr. Cntrl 0.38
  - cio: 0 -> 892; 0.5 -> 398; 1 -> 577; Total 1,867; Pr. Cntrl 0.52
  - dei: 0 -> 1,073; 0.5 -> 219; 1 -> 452; Total 1,744; Pr. Cntrl 0.38
  - deo: 0 -> 890; 0.5 -> 310; 1 -> 585; Total 1,785; Pr. Cntrl 0.50
  - reo: 0 -> 1,084; 0.5 -> 395; 1 -> 388; Total 1,867; Pr. Cntrl 0.42
  - Note: *Data on Bonds available 1997-2013.
- For single-component categories (second block), counts of 0 and 1, total and Pr. Cntrl:
  - fci: 0 -> 1,205; 1 -> 685; Total 1,890; Pr. Cntrl 0.36
  - fco: 0 -> 1,119; 1 -> 767; Total 1,886; Pr. Cntrl 0.41
  - cci: 0 -> 1,337; 1 -> 546; Total 1,883; Pr. Cntrl 0.29
  - cco: 0 -> 1,225; 1 -> 644; Total 1,869; Pr. Cntrl 0.34
  - gsi: 0 -> 1,384; 1 -> 471; Total 1,855; Pr. Cntrl 0.25
  - gso: 0 -> 1,227; 1 -> 631; Total 1,858; Pr. Cntrl 0.34
  - dii: 0 -> 1,121; 1 -> 779; Total 1,900; Pr. Cntrl 0.41
  - dio: 0 -> 1,246; 1 -> 625; Total 1,871; Pr. Cntrl 0.33
  - ldi: 0 -> 1,546; 1 -> 334; Total 1,880; Pr. Cntrl 0.18
  - rei: 0 -> 828; 1 -> 1,034; Total 1,862; Pr. Cntrl 0.55
- Aggregate totals:
  - Total 0 entries: 23,469
  - Total 0.5 or 1 entries: 15,134
  - Grand total observations: 38,603
  - Overall proportion of observations with controls: 0.40
- Additional note from text: for the 11 asset/direction categories with 0/½/1 values, there are more observations of 1 than of ½: 26 percent vs. 20 percent.

### Correlations across asset/direction categories (Tables 4, 5A, 5B)
- Correlations are computed across all observations x(t), y(t) where x and y are asset/direction categories and t is time; correlations missing if variance of an indicator is zero.
- Selected exact correlations and summaries from Table 4 (All 100 countries, 1995-2013):
  - Diagonal (inflow vs outflow for same asset):
    - mmi vs mmo: 0.78
    - eqi vs eqo: 0.72
    - dei vs deo: 0.86 (text: Derivatives highest correlation 86 percent)
    - dii vs dio: 0.37 (Direct Investment lowest 37 percent)
    - rei vs reo: 0.30 (Real Estate 30 percent)
  - Inflow vs inflow example: eqi and cii = 0.70
  - Outflow vs outflow example: gso and cco = 0.74
- Broad findings:
  - Correlation between inflow and outflow controls for a given asset tends to be high; highest for Derivatives (86 percent), lowest for Direct Investment (37 percent) and Real Estate (30 percent).
  - Correlations are highest among Money Market Instruments, Bonds, Equities, Collective Investments, and Derivatives for both inflow and outflow controls.
  - Lowest inflow correlations involve Real Estate relative to other asset categories.
  - Correlations are generally higher among outflow controls than among inflow controls.
- Open/Gate/Wall subgroup correlations:
  - Gate countries (48) have lower correlations; only one correlation > 80 percent and 40 less than 40 percent.
  - Open and Wall combined (52) have very high correlations: all outflow correlations > 80 percent; majority of inflow correlations (except those involving Real Estate) > 60 percent; one fifth of inflow correlations > 80 percent.
  - For Gate countries, highest correlations are between outflow controls on Money Market Instruments, Bonds, Equities, Collective Investments and Derivatives; lowest correlations are for inflow controls with Commercial Credits and Real Estate.
- Implication: correlations inform choices of which asset categories to include in aggregate indices.

### Aggregate indicators and cross-country patterns
- Aggregate inflow indicator for country i in year t = average value of controls on inflows for the 10 asset categories (formula provided in text).
- Aggregate outflow indicator for country i in year t = average value of controls on outflows for the 10 asset categories (formula provided in text).
- For visualization, averages of these aggregates are taken for each of the four income groups: High, Upper Middle, Lower Middle, Low.
- Textual summaries of Figures 2a and 2b:
  - On average, the capital control index is inversely related to income.
  - Left axis midpoint (High Income): about 0.15 in Figure 2a (inflows) and about 0.17 in Figure 2b (outflows).
  - Right axes midpoints (Upper Middle, Lower Middle, Low): about 0.53 (inflows) and 0.60 (outflows).
  - Interpretation: difference consistent with larger proportion of High Income countries classified as Open and more Gate/Wall countries in other income groups.
  - Reference: Fernández, Uribe and Rebucci (2014) also find an inverse relation between capital controls and income levels (their sample more limited).

*Source: _wp1580 - 1. In the case of Real Estate, there is only one inflow category (which we denote rei), but there*

### 0.20 for inflows and 0.22 for outflows in the first years of the sample period to less than 0.10 in

### _wp1580 - 0.20 for inflows and 0.22 for outflows in the first years of the sample period to less than 0.10 in

### Patterns of capital controls by income group and over time
- Average controls fell from "0.20 for inflows and 0.22 for outflows in the first years of the sample period" to:
  - "less than 0.10 in 2008 for inflows"
  - "0.12 in 2004 for outflows"
  - followed by a rise in subsequent years.
- Low Income countries as a group:
  - "see a large decline in their average inflow and outflow controls in the first years of the sample period, and then an increase, especially in average controls on outflows."
- Middle Income groups (Upper Middle and Lower Middle):
  - "The range of the averages across time for both inflow controls and outflow controls for the two Middle Income groups is lower than the other groups, and the averages themselves are lower than the Low Income group but more than twice as high as those for the High Income group."
- High Income group:
  - averages and ranges are the lowest among the income groups described.

### Inflow controls versus outflow controls — country-level relationships
- For each country, average controls on inflows and outflows over the full sample period are defined as KCINFLOW_i and KCOUTFLOW_i.
- Scatterplots (Figure 3) show:
  - A somewhat higher prevalence of outflow controls than inflow controls (consistent with Table 3 and Figure 1).
  - The difference in prevalence is more pronounced for Medium and Lower Income countries than for High Income countries.
  - A relatively high correlation of inflow and outflow controls on a country-by-country basis: "for both sets of countries, the correlation is about 0.8."
  - For subsets of countries:
    - 36 Open countries show this relationship necessarily (high correlation).
    - 16 Wall countries show it to a somewhat lesser extent.

### Asset-level correlations and construction of aggregate indicators
- Correlation between inflow and outflow series varies by asset:
  - Derivatives: correlation "0.86"
  - Real Estate: correlation "0.30"
- Pairwise correlations among certain asset outflow controls:
  - Money Market Instruments, Bonds, Equities, and Collective Investments: "each of the six pairwise correlations is greater than 80 percent."
  - Commercial Credits with those four: correlations "range from 55 percent to 64 percent."
- Table 6 — correlations between a 9-Asset aggregate (excluding one asset) and the excluded asset:
  - Excluded asset correlations: mm 0.87, bo 0.83, eq 0.87, Fc 0.83, ci 0.88, de 0.87, re 0.61, cc 0.71, gs 0.79, di 0.77.
  - Interpretation: "controls on Real Estate, Commercial Credits, Direct Investment, and Guarantees, Sureties, and Financial Backup Facilities are least correlated with the aggregate of the respective nine remaining categories while Money Market Instruments, Collective Investments, Derivatives and Equities are most highly correlated."

### Nested aggregate indicators (KC10, KC9, KC5, KC2) and their correlations
- Definitions:
  - KC10: "Average of Inflows and Outflows for mm, bo, eq, ci, de, re, fc, cc, gs, di."
  - KC9: "Average of Inflows and Outflows for mm, bo, eq, ci, de, re, fc, cc, gs (all but di)."
  - KC5: "Average of Inflows and Outflows for mm, bo, eq, ci, de."
  - KC2: "Average of Inflows and Outflows for mm, bo."
- Table 7 — correlations among aggregate measures for all 100 countries (upper triangular elements) and 48 Gate countries (lower triangular elements):
  - Upper-triangular (all 100 countries):
    - Correlation KC10 with KC9: 0.995
    - Correlation KC10 with KC5: 0.954
    - Correlation KC10 with KC2: 0.924
    - Correlation KC9 with KC5: 0.958
    - Correlation KC9 with KC2: 0.928
    - Correlation KC5 with KC2: 0.971
  - Lower-triangular (48 Gate countries):
    - Correlation KC9 with KC10: 0.992
    - Correlation KC5 with KC10: 0.901
    - Correlation KC2 with KC10: 0.873
    - Correlation KC5 with KC9: 0.910
    - Correlation KC2 with KC9: 0.877
    - Correlation KC2 with KC5: 0.953
- Key implications:
  - For the full set of countries, correlations are "very high" with a range from "0.924 (for the correlation between KC10 and KC2) to 0.995 (for the correlation between KC9 and KC10)."
  - For Gate countries correlations are lower and have a greater range; example: "the correlation between the two-asset and 10-asset indicators is 0.873."
  - "Thus, there could be differences in the estimated effect of capital controls in an analysis in which the identification depends upon the pattern of controls for Gates countries."

### Comparison with Quinn and Chinn-Ito indices — regression evidence
- Two widely used indices are converted to the range [0,1] so that larger values represent more restrictions:
  - Quinn (1997) index: "uses a five-point scale" derived from AREAER narratives, does not distinguish inflows vs outflows.
  - Chinn-Ito index: "takes the first principal component of the AREAER summary binary codings" and is converted to [0,1] for comparison.
- Regressions of country-average Quinn and Chinn-Ito indices on country-average KC10:
  - For Quinn indicator regression:
    - Coefficients and statistics reported as: "݊݊݅ݑܳ௜ൌ 0.004 (ଶ.଴ଵଽ) ൅0.71 (ଶ.଴ସଵ) 10ܥܭ ௜ ܴଶ ൌ0.77; ݊ൌ 90"
    - (As presented in the source; preserves original numeric and formatting.)
  - For Chinn-Ito indicator regression:
    - Coefficients and statistics reported as: "݋ݐܫ݄݊݊݅ܥ௜ൌ 0.049 (ଶ.଴ଶହ) ൅0.91 (ଶ.଴ହଵ) 10ܥܭ ௜ ܴଶ ൌ0.77; ݊ൌ 99."
    - (As presented in the source; preserves original numeric and formatting.)
- Statistical interpretation:
  - "In both of these regressions, the coefficient on KC10_i is significantly different from zero at very high levels of confidence."
  - Test whether coefficients equal 1:
    - "The t-statistic for this test in the regression with the Chinn-Ito indicator is 1.71 and the t-statistic for the Quinn regression is 7.21."
    - "Thus, the null hypothesis that the coefficients equal 1 can be rejected at the 95 percent level of confidence in both cases, but not at the 90 percent level of confidence in the case of the Chinn-Ito indicator."
- Diagnostic plotting:
  - "Plots of the regression lines, and the scatter plots of the points, are presented in the two panels of Figure 4."
  - Countries with large regression errors were identified:
    - For Quinn regression: points with absolute regression error > 0.25 are identified.
    - For Chinn-Ito regression: points with absolute regression error > 0.20 are identified.

### Conclusions and dataset contribution
- The role of capital controls in macroeconomic toolkits remains contested. The paper emphasizes:
  - Need for "careful, high-quality theoretical and empirical research."
  - The authors "contribute to this debate by making available a new dataset" enabling more detailed empirical investigation of capital controls and their effects.
- The paper illustrates data construction and presents properties of granular series and aggregates.
- The authors express the hope that "this dataset proves useful in moving forward our understanding of this important topic."

*Italic: Extracted content from the provided PDF content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1580.pdf_