A. Rules-Based Measures

IMF Measures

Every issue of the International Monetary Fund’s Annual Report on Exchange Arrangements and Exchange Restrictions published between 1967 (which refers to conditions in 1966) and 1996 (which refers to conditions in 1995) includes a summary table in which a single row directly addresses the presence of capital controls: line E.2, labeled “Restrictions on payments for capital transactions.”⁵ The information in this row has been the basis for generating an indicator of the existence of rules or restrictions that inhibit cross-border capital flows or discriminate on the basis of citizenship or residence of transacting agents (see Table 1, row 1).

Share Measure

A standard way to use the information from line E.2 of the International Monetary Fund’s Annual Report on Exchange Arrangements and Exchange Restrictions is to construct, for each country, a variable reflecting the proportion of years in which countries had liberalized capital accounts. We call this variable Share (Table 1, row 2).⁶For example, if the AREAER judged capital markets open for five years out of a 10-year period, then the openness measure Share would be 0.5. A potential problem with this approach is that a value of Share equal to 0.5 is consistent with a situation where a country had open capital markets for the first five years of a decade, for the last five years of a decade, for every other year of a decade, or for many other on-again, off-again patterns. In practice, however, as shown in Panel A of Table 2 (from Klein and Olivei, 2000), there are very few instances of on-again, off-again capital account controls, at least for the 10-year period that ends in 1995, the last year in which capital account liberalization reflects the entry in a single line of AREAER. Thus, over the period 1986 to 1995, a country with a value of Share equal to 0.1 had an open capital account in 1995 only, a country with a value of Share equal to 0.2 had an open capital account in 1994 and 1995, and so on. This pattern holds for all industrial countries and for 10 of the 12 developing countries that had some experience with open capital accounts during this period. Panel B of this table shows that this correspondence between the value of Share and the number of continuous years of open capital accounts continues to hold for industrial countries as the sample period is extended back to 1976. But, in this longer sample period, there are many more cases of on-again, off-again capital account liberalization among developing countries. In fact, among the developing countries that had any experience with open capital accounts over the period 1976–95, the only countries where capital accounts were not closed after having been opened were Indonesia and Malaysia, both of which had open capital accounts throughout the entire 1976–95 period.⁷

Table 2.

IMF Capital Account Restriction Measure

Source: IMF AREAER and Klein and Olivei 2001.

Table 2.

IMF Capital Account Restriction Measure

A. Value of Share and Years When Capital Markets Open, 1986 - 1995
(For countries that had open markets at some point in time, i.e. Share = 0)
	Years Open	Industrial Countries	Developing Countries
0.1	1995	Norway	Costa Rica, Niger
0.2	1994–95	Spain	Trinidad & Tobago
0.3	1993–95	Portugal, Sweden	Honduras, Peru
0.4	1992–95	Ireland
0.5	1991–95	Finland, Austria
0.6	1990–95	France, Italy
	1988–92, 1995		Ecuador
0.7	1989–95		Guatemala
	1986–92		Uruguay
0.8	1988–95	Denmark
1.0	1986–95	Australia, Belgium, Canada, Japan, Netherlands, New Zealand, United Kingdom, United States	Bolivia, Indonesia, Malaysia, Panama
B. The Evolution of Capital Account Restrictions, 1976-1985
(For countries that had open markets at some point in time during this Period)
Years Open	Industrial Countries	Years Open	Developing Countries
1984–1985	Australia, New Zealand	1980–1981	Costa Rica
1978–1985	Japan, United Kingdom	1976–1979	Guatemala, Honduras
		1976–1984	Ecuador
1976–1985	Belgium, Canada, Germany, Netherlands, United States	1976–1985	Indonesia, Malaysia

Source: IMF AREAER and Klein and Olivei 2001.

View Table

Table 2.

IMF Capital Account Restriction Measure

A. Value of Share and Years When Capital Markets Open, 1986 - 1995
(For countries that had open markets at some point in time, i.e. Share = 0)
	Years Open	Industrial Countries	Developing Countries
0.1	1995	Norway	Costa Rica, Niger
0.2	1994–95	Spain	Trinidad & Tobago
0.3	1993–95	Portugal, Sweden	Honduras, Peru
0.4	1992–95	Ireland
0.5	1991–95	Finland, Austria
0.6	1990–95	France, Italy
	1988–92, 1995		Ecuador
0.7	1989–95		Guatemala
	1986–92		Uruguay
0.8	1988–95	Denmark
1.0	1986–95	Australia, Belgium, Canada, Japan, Netherlands, New Zealand, United Kingdom, United States	Bolivia, Indonesia, Malaysia, Panama
B. The Evolution of Capital Account Restrictions, 1976-1985
(For countries that had open markets at some point in time during this Period)
Years Open	Industrial Countries	Years Open	Developing Countries
1984–1985	Australia, New Zealand	1980–1981	Costa Rica
1978–1985	Japan, United Kingdom	1976–1979	Guatemala, Honduras
		1976–1984	Ecuador
1976–1985	Belgium, Canada, Germany, Netherlands, United States	1976–1985	Indonesia, Malaysia

Source: IMF AREAER and Klein and Olivei 2001.

An Intensity Measure

The on/off indicator of capital controls presented in the summary table of the Annual Report on Exchange Arrangements and Exchange Restrictions described above does not distinguish between strongly administered capital controls and those that are somewhat more porous. Quinn (1997) attempts to capture the intensity of enforcement of controls on both the capital account and the current account through a careful reading of the narrative descriptions in the Annual Report on Exchange Arrangements and Exchange Restrictions (Table 1, row 3).⁸

We limit our discussion to Quinn’s measures of capital account liberalization.⁹ He scores separately the intensity of controls for capital account receipts and capital account payments. For each of these two categories the scoring method is as follows: a score of 0 indicates payments are forbidden, 0.5 indicates that there are quantitative or other regulatory restrictions, 1 indicates that transactions are subject to heavy taxes, 1.5 indicates that there are less severe taxes, and 2 indicates that transactions are free of restrictions or taxes. The sum of the values for the two categories is an indicator of overall capital account openness that ranges between 0 and 4. These indicators are available annually from 1950-1997 for 21 OECD countries, and for the years 1958,1973, 1982, and 1988 for 43 non-OECD countries.

A glance at Quinn’s dataset indicates that the overall trend towards liberalization is mostly driven by the OECD countries. Table 3 presents a tabulation of Quinn’s measure of capital account liberalization for 1973, 1982, and 1988, years for which these indicators are available for all countries. The top panel of this table presents the data for the full set of 63 countries while the lower panel presents the data for the 42 non-OECD countries only. The data in the first set of columns of these tables show that, in 1973, 37 of 63 countries, including 26 of the 42 developing countries, had capital account indicators equal to the mid-range (2) or lower. The overall trend towards greater capital account liberalization is reflected in the fact that, from 1973 to 1988, the total number of countries with capital account indicators equal to 2 or lower decreased to 33. But this overall trend hides significant heterogeneity: over the same period the number of developing countries with a score of 2 or lower actually increased to 32.

Table 3.

Quinn’s Indicators of Capital Account Liberalization

Source: Quinn (1997) and authors’ calculations.

Table 3.

Quinn’s Indicators of Capital Account Liberalization

Number of Overall Countries
Indicator Value	1973	1982	1988	With no change in value 1973–88	With increase in value 1973–1988	With decrease in value 1973–88
0 (restricted)	2	1	0	0	2	–
0.5	4	5	3	1	3	0
1	11	9	7	4	6	1
1.5	13	15	14	6	6	1
2	7	13	9	0	6	1
2.5	7	4	6	2	4	1
3	9	7	11	2	4	3
3.5	5	0	4	0	2	3
4 (liberalized)	5	9	9	3	–	2
Total	63	63	63	18	33	12
Number of Developing (non-OECD) Countries
Indicator Value	1973	1982	1988	With no change in value 1973–88	With increase in value 1973–88	With decrease in value 1973–88
0 (restricted)	2	1	0	0	2	_
0.5	4	5	3	1	6	0
1	10	8	7	4	5	1
1.5	9	14	13	6	2	1
2	1	8	9	0	0	1
2.5	5	1	3	2	2	1
3	4	0	2	0	1	3
3.5	3	0	2	0	0	3
4 (liberalized)	4	5	3	2	-	2
Total	42	42	42	15	15	12

Source: Quinn (1997) and authors’ calculations.

View Table

Table 3.

Quinn’s Indicators of Capital Account Liberalization

Number of Overall Countries
Indicator Value	1973	1982	1988	With no change in value 1973–88	With increase in value 1973–1988	With decrease in value 1973–88
0 (restricted)	2	1	0	0	2	–
0.5	4	5	3	1	3	0
1	11	9	7	4	6	1
1.5	13	15	14	6	6	1
2	7	13	9	0	6	1
2.5	7	4	6	2	4	1
3	9	7	11	2	4	3
3.5	5	0	4	0	2	3
4 (liberalized)	5	9	9	3	–	2
Total	63	63	63	18	33	12
Number of Developing (non-OECD) Countries
Indicator Value	1973	1982	1988	With no change in value 1973–88	With increase in value 1973–88	With decrease in value 1973–88
0 (restricted)	2	1	0	0	2	_
0.5	4	5	3	1	6	0
1	10	8	7	4	5	1
1.5	9	14	13	6	2	1
2	1	8	9	0	0	1
2.5	5	1	3	2	2	1
3	4	0	2	0	1	3
3.5	3	0	2	0	0	3
4 (liberalized)	4	5	3	2	-	2
Total	42	42	42	15	15	12

Source: Quinn (1997) and authors’ calculations.

As discussed in more detail below, Quinn (1997) uses the change in the value of the indicator of capital account restrictions rather than its level in his growth regressions (Table 1, row 4). The last three columns of each of the two panels in Table 3 uncover interesting patterns in the churning of the capital account liberalization experience. While most OECD countries (18 out of 21) increased their degree of liberalization between 1973 and 1988, the developing countries were almost equally split between those that increased it (15 countries), reduced it (12), or kept it unchanged (15). Note also that by the end of the 1980s, developing countries had converged towards an intermediate level of liberalization, as those that increased it started in 1973 from low degrees of liberalization and those that decreased it started from high degrees of liberalization.

Quinn’s capital account restriction indicators are meant to be used as cardinal numbers and, therefore, there is the implication that a country in category 2 is literally twice as unfettered as one in category 1, or a change in the value of the indicator of 1 is exactly twice the amount of liberalization as a change in the indicator of 0.5. A less structured approach for using Quinn’s indicators in regressions would involve the creation of dummy variables representing each of the values of the indicators. Alternatively, one might decide to generate fewer dummy variables, say a single dummy variable that takes the value 1 when Quinn’s indicator takes the value 0.5, 1, or 1.5. In the extreme, one may wonder how to convert Quinn’s measure into a (0,1) dummy, i.e. how to identify which value of Quinn’s scale would correspond to a threshold for classifying countries as open or closed.

Table 4 suggests that the correspondence between Quinn’s multiple-measures and the 0/1 AREAER measures is approximately achieved if one chooses a threshold of 2 in Quinn’s scale. Each of the six panels in this Table offers a tabulation of the number of countries that have capital accounts classified as 0 (closed) or 1 (open) by the AREAER and are given values of capital account restrictions of 0–2 (more closed) or 3–4 (more open) by Quinn. There are three panels for the full sample and three for developing countries, with one pair for each of the more recent years for which Quinn has data for all countries, 1973, 1982, and 1988. A high correspondence between the AREAER and Quinn measures would be reflected in relatively large diagonal elements of the 2 × 2 matrix and relatively small off-diagonal elements. In fact, over 80 percent of the countries in the full sample and over 90 percent of the countries for the developing countries sample fall into one of the diagonal cells for each of the three years. For the full sample, there are few cases where the AREAER indicates open capital accounts while Quinn indicates closed capital accounts (that is, it is rare to find a relatively large number in the upper right cell of the 2 × 2 matrix). Most of the discrepancies between the two indicators are cases where the AREAER indicates closed capital accounts and Quinn gives a value of 2.5, 3, 3.5, or 4. A more detailed presentation would show that almost all the cases of non-zero entries in the lower-left cell represent a value of 2.5 or 3 assigned by Quinn. Thus, there seems to be quite a high correspondence between Quinn’s measures and those of the AREAER, when one chooses a threshold of 2 in Quinn’s scale in order to classify a country’s capital account as closed or open. The relevant question is whether the gradations presented by Quinn offer significant information, a point we return to below.

Table 4.

Comparison of Quinn and IMF Indicators

(Number of countries for each combination of indicators)

Source: Authors’ calculations.

Table 4.

Comparison of Quinn and IMF Indicators

(Number of countries for each combination of indicators)

Panel 1: Full Sample: Year 1973						Panel 4: Developing Countries Year 1973
		IMF Indicator		Total				IMF Indicator		Total
		0	1					0	1
Quinn	0–2	36	0	36		Quinn	0–2	26	0	26
Indicator	3–4	10	15	25		Indicator	3–1	4	12	16
Total		46	15	61		Total		30	12	42
Panel 2: Full Sample: Year 1982						Panel 5: Developing Countries Year 1982
		IMF Indicator		Total				IMF Indicator		Total
		0	1					0	1
Quinn	0–2	38	4	42		Quinn	0–2	32	4	36
Indicator	3–4	7	12	19		Indicator	3–4	0	6	6
Total		45	16	61		Total		32	10	42
Panel 3: Full Sample: Year 1988						Panel 6: Developing Countries Year 1988
		IMF Indicator		Total				IMF Indicator		Total
		0	1					0	1
Quinn	0–2	32	0	32		Quinn	0–2	32	0	32
Indicator	3–4	12	17	29		Indicator	3–4	2	8	10
Total		44	17	61		Total		34	8	42

Source: Authors’ calculations.

View Table

Table 4.

Comparison of Quinn and IMF Indicators

(Number of countries for each combination of indicators)

Panel 1: Full Sample: Year 1973						Panel 4: Developing Countries Year 1973
		IMF Indicator		Total				IMF Indicator		Total
		0	1					0	1
Quinn	0–2	36	0	36		Quinn	0–2	26	0	26
Indicator	3–4	10	15	25		Indicator	3–1	4	12	16
Total		46	15	61		Total		30	12	42
Panel 2: Full Sample: Year 1982						Panel 5: Developing Countries Year 1982
		IMF Indicator		Total				IMF Indicator		Total
		0	1					0	1
Quinn	0–2	38	4	42		Quinn	0–2	32	4	36
Indicator	3–4	7	12	19		Indicator	3–4	0	6	6
Total		45	16	61		Total		32	10	42
Panel 3: Full Sample: Year 1988						Panel 6: Developing Countries Year 1988
		IMF Indicator		Total				IMF Indicator		Total
		0	1					0	1
Quinn	0–2	32	0	32		Quinn	0–2	32	0	32
Indicator	3–4	12	17	29		Indicator	3–4	2	8	10
Total		44	17	61		Total		34	8	42

Source: Authors’ calculations.

Stock Market Liberalization Indicators

Several papers have focused on the liberalization of controls on the international sale or purchase of equities. Research in this area has typically focused on dating the opening of equity markets to foreign investors. Levine and Zervos (1998) and Henry (2000a, 2000b), who extended this work, have compiled dates from a variety of sources including The Wilson Directory of Emerging Market Funds, IFC Investable Indexes, various issues of The Economist Intelligence Unit and the IMF’s AREAER. For instance, for each of the 11 countries in his sample, Henry uses the dates reflecting official policy decrees as the first date in which a country fund is available, or a 10-percentage point jump in the IFC Investable Index (Table 1, row 7).¹⁴

Bekaert (1995) and Bekaert and Harvey (1995) also determine dates when equity markets in emerging economies opened to foreign investors based upon a variety of indicators including Official Liberalization Dates, the date of introduction of American Depository Receipts (ADR), the date of the introduction of country funds, and a date estimated through a regime-switching model based on the time series of net U.S. capital flows (see Table 1 in Bekaert and Harvey, 2000). The number of countries covered is extended in a paper by these two authors and Lundblad (2001) to 95 countries, 16 of which had liberalized stock markets over the full period 1980 to 1997 and 27 of which had some experience with stock market liberalizations over this period. (See their Appendix Table Al; these dates are referred to in our Table 1, row 8).

Table 5 presents the dates of “Official International Financial Liberalization” for the 30 countries classified as either emerging markets or frontier markets by the International Finance Corporation, as reported by Bekaert, Harvey, and Lundblad (2001). Note that, in comparing the entries in this table with the entries in Table 2, only Indonesia and Malaysia had both equity market liberalization and capital account liberalization (as reported by the IMF in the AREAER) and, equity market liberalization followed capital account liberalization for both of these countries.

Table 5.

Dates of Equity Market Liberalization of Emerging Market or Frontier Market Countries

Source: Bekaert, Harvey, and Lundblad, 2001.Note: Countries classified as emerging or frontier by the International Finance Corporation.NL refers to Not Liberalized.

Table 5.

Dates of Equity Market Liberalization of Emerging Market or Frontier Market Countries

Country	Liberalization Date	Country	Liberalization Date
Argentina	1989	Malaysia	1988
Bangladesh	NL	Mexico	1989
Brazil	1991	Morocco	1997
Chile	1992	Nigeria	1995
Colombia	1991	Pakistan	1991
Côte d’Ivoire	NL	Philippines	1991
Egypt	1997	Portugal	1986
Greece	1987	Sri Lanka	1992
India	1992	South Africa	1992
Indonesia	1989	Thailand	1987
Israel	1996	Trinidad & Tobago	NL
Jamaica	NL	Tunisia	NL
Jordan	1995	Turkey	1989
Kenya	NL	Venezuela	1990
Korea	1992	Zimbabwe	1993

Source: Bekaert, Harvey, and Lundblad, 2001.Note: Countries classified as emerging or frontier by the International Finance Corporation.NL refers to Not Liberalized.

View Table

Table 5.

Dates of Equity Market Liberalization of Emerging Market or Frontier Market Countries

Country	Liberalization Date	Country	Liberalization Date
Argentina	1989	Malaysia	1988
Bangladesh	NL	Mexico	1989
Brazil	1991	Morocco	1997
Chile	1992	Nigeria	1995
Colombia	1991	Pakistan	1991
Côte d’Ivoire	NL	Philippines	1991
Egypt	1997	Portugal	1986
Greece	1987	Sri Lanka	1992
India	1992	South Africa	1992
Indonesia	1989	Thailand	1987
Israel	1996	Trinidad & Tobago	NL
Jamaica	NL	Tunisia	NL
Jordan	1995	Turkey	1989
Kenya	NL	Venezuela	1990
Korea	1992	Zimbabwe	1993

Source: Bekaert, Harvey, and Lundblad, 2001.Note: Countries classified as emerging or frontier by the International Finance Corporation.NL refers to Not Liberalized.

A recent study by Edison and Warnock (2001) provides a new measure of restrictions on foreign ownership of domestic equities, which shows the intensity of controls at a point in time as well as their evolution over time. The measure, discussed in full details in their article and used in cross-sectional work of Ahearne, Griever, and Warnock (2000), builds on the work of the International Finance Corporation. In particular, for emerging market countries, the IFC publishes two indexes of equity prices, the Global (IFCG) and Investable (IFCI) indexes. Since the IFCI index is comprised of IFCG stocks minus the portion not available to foreigners, the ratio of the market capitalizations of the IFCI to IFCG indexes is a measure of the availability of a country’s stocks to foreigners, and one minus the ratio is a measure of foreign ownership restrictions (Table 1, row 9). This measure can be seen as an extension of the liberalization analysis of Bekaert and Harvey (2000) and Henry (2000a). Indeed, the initial relaxation of controls shown by the Edison-Warnock measure corresponds quite well with the Bekaert-Harvey liberalization date.¹⁵ This new measure provides additional information, giving an indication of the extent of the liberalization and its evolution over time. It shows that international financial liberalizations can be gradual—a point alsomade in Bekaert and Harvey (1995)—which would argue against the use of an event study approach or structural break analysis.

B. Quantitative Measures

An alternative to the practice discussed above, of constructing indicators from published regulations, is to derive quantitative measures of the limits placed on capital account transactions from the values of economic variables. There are three different sets of variables that researchers have considered in this vein: national saving rate paired with national investment rates, interest rate differentials, and international capital flows. While, at first blush, it may seem preferable to use actual performance rather than published regulations in order to gauge the restrictiveness of capital controls, we will see that, as with rules-based measures, there are conceptual and practical challenges associated with the use of these quantitative measures.

To the best of our knowledge, neither a comparison of national saving rates with national investment rates nor interest rate differentials has been used in studying the effects of capital account liberalization on long-run economic growth. There have been some recent efforts to use actual capital flows to estimate the effects of capital account restrictiveness on growth. Despite the limited use of these measures in analyzing the links between capital account liberalization and growth, we discuss each of these three sets of

Feldstein and Horioka (1980) published a study that was quite influential due, in part, to the fact that it was one of the first efforts to quantitatively measure the extent of capital mobility across a broad range of countries. They analyzed the behavior of savings and investment in a number of countries to measure the “true” degree of capital mobility, arguing that the degree of correlation between the two series was a good indicator of impediments to capital movements. In any particular year, savings matches investment in a country with stringent capital account restrictions while there need not be a link between savings and investment in a country with free capital mobility. Feldstein and Horioka found that, over the period 1960–74, as well as over the three 5-year subperiods making up this 15-year period, average saving rates and average investment rates were highly positively correlated. Based on this finding, they concluded that there were significant impediments to capital movements.¹⁶

But this conclusion has been criticized since the saving and investment rates of a country may be highly correlated, even if that country has no restrictions on international capital flows. For example, Obstfeld (1986) shows that, even with perfect capital mobility, saving and investment may be highly correlated just because of the types of shocks hitting an economy. Obstfeld also demonstrated that the saving-investment correlations were higher for large countries than for small countries, a result one would expect to find if there was free capital mobility and outcomes in large countries influenced worldwide economic conditions. Furthermore, Obstfeld documents a reduction in saving-investment correlations in the post-1973 period. A different criticism of Feldstein and Horioka’s conclusions concerning the extent of capital mobility based on saving-investment correlations comes from Bayoumi (1990) who shows that this correlation may just reflect efforts by governments to target the current account.

Another set of quantitative measures of capital mobility includes onshore-offshore interest rate differentials and deviations from covered interest rate parity. Unlike stock market returns or other quantitative measures, short-term interest rates can be analyzed without first transforming them in model-specific ways. However, data availability restricts this method to a limited number of countries.

Recently, efforts have been made to gauge the extent of capital mobility through the use of actual capital inflows and outflows, either as a percentage of GDP (Kraay 1998) or, as in Lane and Milesi-Ferretti (forthcoming), by using an annual measure of portfolio and direct investment assets and liabilities as a percent of GDP as a long-run indicator of financial openness (see IMF, 2001, Chanda, 2001, and O’Donnell, 2001). These measures are analogous to measures of trade openness, and can be thought of in a similar manner. For example, like the level of trade openness, which is typically calculated as the sum of imports and exports over GDP, the Lane and Milesi-Ferretti measure is a good indicator of openness at a point in time. But both the Lane and Milesi-Ferretti measure, and that of Kraay, may fluctuate from year to year, since capital flows are endogenous and there can be large valuation adjustments due to, say, a large swing in equity values (Eichengreen 2001). However, changes in these measures over longer periods are likely to be indicative of changes in openness. In the subsequent section, this stock measure will be referred to as the openness measure.

C. Comparisons of Rules-Based and Quantitative Measures of Capital Account Liberalization¹⁷

We would hope to find that a measure of capital account restrictiveness based on published regulations and rules is correlated with one that reflects the actual behavior of capital flows. Figure 1 offers some evidence that enables us to judge the correspondence between these two types of measures. This figure includes the time series of a rule-based measure, related to the restrictions on capital flows as reported to the IMF by national authorities, as well as a measure of capital account openness based on the estimated stocks of gross foreign assets and liabilities as a ratio of GDP.

Summary of Measures of Capital Account Openness¹

The two measures of liberalization show similar overall patterns, but the openness measure points progress in the 1990s.

Citation: IMF Working Papers 2002, 120; 10.5089/9781451854282.001.A001

Sources: IMF, Annual Report on Exchange Arrangements and Exchange Restrictions, various issues; International Financial Statistics; and IMF staff calculations.The restriction measure is calculated as the “average” value of the on/off measure for the country group. The openness measure is calculated as the average stock of accumulated capital flows (as percent of GDP) in a country group.For country coverage, see Table 4.2.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Summary of Measures of Capital Account Openness¹

The two measures of liberalization show similar overall patterns, but the openness measure points progress in the 1990s.

Citation: IMF Working Papers 2002, 120; 10.5089/9781451854282.001.A001

Sources: IMF, Annual Report on Exchange Arrangements and Exchange Restrictions, various issues; International Financial Statistics; and IMF staff calculations.The restriction measure is calculated as the “average” value of the on/off measure for the country group. The openness measure is calculated as the average stock of accumulated capital flows (as percent of GDP) in a country group.For country coverage, see Table 4.2.

• Download Figure
• Download figure as PowerPoint slide

Summary of Measures of Capital Account Openness¹

The two measures of liberalization show similar overall patterns, but the openness measure points progress in the 1990s.

Citation: IMF Working Papers 2002, 120; 10.5089/9781451854282.001.A001

Sources: IMF, Annual Report on Exchange Arrangements and Exchange Restrictions, various issues; International Financial Statistics; and IMF staff calculations.The restriction measure is calculated as the “average” value of the on/off measure for the country group. The openness measure is calculated as the average stock of accumulated capital flows (as percent of GDP) in a country group.For country coverage, see Table 4.2.

• Download Figure
• Download figure as PowerPoint slide

Figure 1 shows that both the rules-based and the quantitative measures indicate that capital accounts in industrial countries have become considerably less restricted over time. A particularly rapid decline in controls occurred during the 1980s, when the members of the European Community (now called the European Union) liberalized capital controls. Following this, there was a dramatic rise in cross-border capital flows. Among industrial countries, Canada, the Netherlands, Switzerland, and the United Kingdom appear to have the most unfettered international capital flows, based on the estimated stocks of gross foreign assets and liabilities as a percent of national income.

In developing countries, the story is more complex. In general, both measures suggest a less dramatic shift toward liberalization and openness than in industrial countries. For the developing countries as a whole, the rules-based restriction measure suggests that, after a period of liberalization in the 1970s, the trend toward openness reversed in the 1980s. Liberalization resumed in the early 1990s, but the pace has been relatively slow; the rules-based measure indicates that the current level of the indicator on average is only at the same level as it was in the late 1970s. By comparison, the capital account openness measure based on the estimated stocks of gross foreign assets and liabilities as a ratio of GDP shows a modest decline in openness to capital flows during the early 1970s, followed by a moderate increase in the 1980s, and a sharp acceleration in the early 1990s.

D. Efficacy of Capital Account Controls

Some of the rules-based indicators discussed above attempt to distinguish between the intensity with which capital account restrictions are imposed. But rules-based measures focus on de jure restrictions rather than the de facto efficacy of controls. Obviously, it is easier to determine whether laws controlling capital flows are on the books than whether these laws are enforced or, if they are enforced, whether they effectively stem the flow of capital that would otherwise be moving into or out of a country.

Various episodes suggest that capital controls have, at best, a temporary effect on stemming capital inflows or outflows. For example, Spain attempted to use capital controls to shield itself from the disruptions in the European Monetary System in September 1992, but the peseta was devalued that November anyway. Capital controls on inflows in Colombia (1993–98) and Brazil (1993–97) did not prevent the continued appreciation of their currencies. In Brazil, the presence of sophisticated financial institutions enabled people to get around controls. Likewise, the much-heralded controls on capital inflows to Chile, which took the form of unremunerated reserve requirements (URRs), lost their effectiveness as people found loopholes that enabled them to avoid these regulations. Capital controls in Thailand and Malaysia may have altered the maturity structure of capital inflows, though capital controls did not insulate the Thai baht from the consequences of speculative pressures in 1997.¹⁸

If, in fact, capital controls are generally full of holes and do not work, then we would not expect to find a strong link between capital account liberalization and any measures of economic performance, all else being equal. We next turn to a review of empirical analyses of whether or not capital account liberalization is significantly associated with economic growth.

A. Supporting Evidence of Capital Account Liberalization on Growth

Quinn (1997) is one of the first studies to identify a positive result between capital account liberalization and growth. Quinn augments the set of variables included in a standard growth regression with either a variable representing his indicator of the change in financial openness (described above) or the change in a broader measure of openness.²⁰The change in the financial openness variable and the change in the broader openness variable are calculated by subtracting the 1958 value of the index from its 1988 value. Quinn’s empirical estimates suggest that the change in capital account liberalization has a strongly significant effect on the growth in real GDP per capita in his cross section of 58 countries over the period 1960 to 1989. It is hard to disentangle the separate effects of financial openness and a broader measure of openness in Quinn’s results because he does not include a regression with both of these indicators and, therefore, to the extent that the change in financial openness is correlated with the change in the openness of trade in goods and services, the finding of a significant effect of the change in capital account liberalization on growth may reflect the correlation of changes in restrictions on the capital account and the current account.

Klein and Olivei (2000) find a positive effect of capital account liberalization on growth among industrial countries, but they do not find evidence that capital account liberalization promotes growth in non-industrial countries. This study follows a slightly different strategy from other research in this area by first focusing on the role of capital account liberalization on financial development and then considering the effect of financial development on growth. Klein and Olivei regress the capital account liberalization indicator Share described above (along with other standard regressors) on the change in financial depth over the period 1986 to 1995. They find a significant effect of capital account liberalization on the change in financial depth in the cross section of 82 developed and developing nations. This significant result seems to be due to the presence of the OECD countries in the sample. Klein and Olivei show that capital account liberalization significantly affects the change in financial depth in a sample consisting of the 20 OECD countries, but not in a sample of the non-OECD countries, nor in a narrower non-OECD sample of 18 Latin American countries, a group that had a relatively high incidence of capital account liberalizations. They also estimate a growth model that includes the change in financial depth as regressor and find that financial development is significant determinant of growth per capita. Klein and Olivei conclude that the beneficial effects of capital account liberalization, at least with respect to promoting financial depth, are achieved only in an environment in which there is a constellation of other institutions that can usefully support the changes brought about by the free flow of capital. Bailliu (2000) also finds that capital account liberalization spurs growth by promoting financial development.

This conjecture, that the growth effects of capital account liberalization depend upon the level of development of an economy, is supported by the results presented in Edwards (2001). Edwards finds that both the Quinn level and the ΔQuinn variables are significantly associated with growth in per capita income in the 1980s for a sample of about 60 countries, though his results are not robust when he instead uses Share. Edwards estimates regressions using weighted least squares, with the weights representing national income in 1985. He includes both the capital account liberalization variables and an interactive term representing the product of the variables and the logarithm of income per capita in 1980. His estimates show a negative coefficient on capital account openness and a positive coefficient on the openness-income interactive term. This suggests that capital account openness detracts from growth for countries at lower levels of income but promotes growth in industrial countries and in the richer emerging market countries.

Edwards’ methodology is scrutinized in Arteta, Eichengreen, and Wyplosz (2001). They question why Edwards obtains such strong results while other studies (cited below) fail to find any significant effect of capital account liberalization on growth. They point to a number of potential reasons for this discrepancy. They question whether Quinn’s measures, representing capital account openness in 1973 and 1988 only, is appropriate for the period Edwards studies. They also note that weighting observations by 1985 per capita GDP means that rich countries have much more influence in the regressions than do poor countries. There are also questions raised concerning the exogeneity of instruments used for capital account liberalization and the exclusion of other potentially relevant measures of openness, such as the openness of the current account, which may be correlated with capital account openness. Estimates by Arteta, Eichengreen, and Wyplosz suggest that Edwards’ results may be sensitive to a variety of factors and, therefore, they conclude that there is little evidence that capital account liberalization has more favorable effects in high-income and middle-income countries than in poorer developing countries.

But Arteta, Eichengreen, and Wyplosz do find some support for differences in the effect of capital account liberalization across countries, depending upon the degree of macroeconomic stability. They introduce two capital account interaction terms, multiplying the Quinn openness measure by both the Sachs-Warner (1995) openness measure and the black-market premium. They find that the interaction term representing the product of capital account openness and the black market premium is significant but the other interaction term, representing the product of the Sachs-Warner openness measure and capital account openness, is not significant. They interpret this to mean that countries that open their capital accounts grow faster, but only if they first eliminate the black market premium.

Bekaert, Harvey, and Lundblad (2001) (henceforth BHL) examine the impact of stock market liberalization on economic growth. As previous researchers have done, they begin their analysis by augmenting the standard set of growth model variables with their variable indicating stock market liberalization. To maximize the time-series content in their regression, they use a moving average panel data method. This means that they create overlapping data and therefore have to deal with the resulting moving average error component by adjusting their standard errors. In general, BHL find that financial liberalization leads to a 1 percent increase in annual per capital GDP growth over a five- year period and that this effect is statistically significant. BHL investigate the robustness of this result with respect to alternative sets of liberalization dates, different country groupings, and different time horizons for measuring economic growth. These results, along with those of Quinn, appear to be the strongest evidence supporting the hypothesis that capital account liberalization promotes growth among developing countries.

Following a slightly different tact, O’Donnell (2001) examines the impact of capital account liberalization using both IMF rules-based measure and a quantitative based measure of financial openness. Using a rather standard setup, he finds that the rules-based measure tends to be too coarse an indicator of the degree of capital account liberalization as it does not take account the nature of different types of controls. However, using the quantitative measure, he finds that capital account liberalization does seem to speed up economic growth. However, like other researchers he finds that the benefits to all countries are not equal. This difference in impact was also echoed in Chanda (2001). He suggests that the impact may vary with the level of ethnic and linguistic heterogeneity in the society, a proxy for the number of interest groups. In particular, he finds that capital controls lead to greater inefficiencies and lower growth among countries with a high degree of ethnic and linguistic heterogeneity.

B. Studies Not Supporting the Hypothesis that Liberalization Promotes Growth

Several studies have found no correlation between openness and growth. In fact, one of the first efforts to determine whether capital account liberalization promotes growth using a cross-section of countries was Grilli and Milesi-Ferretti (1995), although this was not the main focus of their paper. This study considers average growth of per capita income for five nonoverlapping five-year periods between 1966 and 1989. Their sample includes 61 countries although, with 181 observations in one set of regressions and 238 in another, not every country appears in each of the five sub-periods. They regress five-year growth rates on Share and comparable measures that capture the presence of current account controls (CurrAcct) and a multiple exchange rate system (MultEx) from Exchange Arrangements and Exchange Restrictions. In addition they include variables such as initial income, political variables and the level of schooling. The three variables Share, CurrAcct, and MultEx enter as predicted values from instrumental variable regressions using lagged values as the instruments. Their results do not support the hypothesis that capital account liberalization promotes growth. In some cases, capital account controls enter with a positive sign (that is, Share enters negatively) while the indicator of current account controls sometime enters with a negative sign (CurrAcct is positive).²¹

Rodrik (1998), in a widely cited paper, also casts doubt on the effect of capital account liberalization on growth. In a sample that includes almost 100 countries, developing as well as developed, he finds no significant effect of capital account liberalization, as measured by Share, on the percentage change in real income per capita over the period 1975 to 1989 in growth regressions that also include initial per capita income, initial secondary-school enrollment rate, an index of the quality of governmental institutions and regional dummy variables. Likewise, he finds no relationship between capital account liberalization and investment-to-income, nor between capital account liberalization and inflation. Eichengreen (2001) offers several possible reasons for the differences in the results between the Quinn and Rodrik papers, including the fact that Quinn’s sample includes fewer developing countries than Rodrik, the smaller proportion of years in Quinn’s study representing the “lost decade” of the 1980s and differences in the capital account indicator. The regressions presented in the next section allow us to explore the role of these potential sources of the differences in the two studies.

Kraay (1998) also finds no significant relationship, using a variety of measures of capital account openness, including Share, Quinn’s capital account openness indicator (in levels, with values from 0 to 4), and a measure based on actual net capital flows. Each of these measures is associated with a different sample size. His regressions take the form of cross sections, with one observation per country, where the dependent variable is the growth in output between 1985 and 1997. He uses both OLS and an approach in which the capital account liberalization variables are instrumented by their own past values. He fails to find a significant effect of Share or the Quinn indicator on growth, but, when these indicators are interacted with the average balance of the financial account (from the balance of payments statistics), he does find some significant effects.

A. The Basic Growth Model and the Impact of Capital Account Liberalization

The regressand in all of the estimates presented in this section is the growth in real per capita income over the period 1976 to 1995 (ΔlnY_76–95). The control variables used in the growth model include the logarithm of real per capita income in 1976 (In (Y₁₉₇₆)), the logarithm of secondary school enrollment rate in 1976 (ln(Educ.)), the average investment to GDP ratio for the years 1974 to 1978 (Invest._74–78), the population growth rate from 1976 to 1995 (ΔPop._76–95) and a dummy variable for countries in sub- Saharan Africa (Africa). Column 1 of Table 7 shows the results of a growth regression using these regressors for a sample of 89 countries. Each of the coefficients in the regression are significant at the 95 percent level of confidence or better, but for the coefficient on the average investment to GDP ratio.

Table 7.

Growth and Capital Account Liberalization

Regressions include a constant (not reported) Robust standard errors.Definition of Variables:ΔlnY_{76 – 95} = growth in In real per capita income, 1976 to 1995; In Y₁₉₇₆ = In real per capita income in 1976; ln(Educ.) = ln(secondary school enrollment rate); Invest. _74–78 = Average Investment, 1974 to 1978; ΔPop._76–95 = Population growth, 1976 to 1995; Africa = Dummy variable for African countries; Share_76–95 = Proportion of years with open capital accounts, 1976 to 1995, from IMF’s AREAER-, Quinn₈₂ = 0 to 4 measure of capital account openness in 1982 and ΔQuinn_{73 – 88} = Difference in 0 to 4 measure of capital account openness between 1973 and 1988, both from Quinn (1997); BHL_80–95 = Proportion of years between 1980 and 1995 with liberalized stock market, using dates of stock market liberalization from Bekaert, Harvey and Lundblad (2001).Instruments for Capital Account Liberalization Measures:Government consumption as proportion of GDP in 1976; Imports as proportion of GDP in 1976; Dummy variables for Latin American countries and East Asian countries; Quinn’s 0 to 4 measure of capital account openness in 1973 (for Quinn₈₂) or IMF’s AREAER 0/1 value in 1973 (for Shares _76–95 and for BHL_80–95).Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

Table 7.

Growth and Capital Account Liberalization

Dep. Var. ΔlnY76 – 95	OLS					IV
	1	2	3	4	5	6	7	8
In Y1976 (s.e.)	-0.24 (0.04)	-0.31 (0.05)	-0.48 (0.08)	-0.41 (0.09)	-0.33 (0.05)	-0.35 (0.06)	-0.49 (0.08)	-0.42 (0.06)
In (Educ.) (s.e.)	0.12 (0.05)	0.12 (0.05)	0.16 (0.08)	0.22 (0.10)	0.15 (0.05)	0.13 (0.05)	0.15 (0.09)	0.14 (0.05)
Invest.74-78 (s.e.)	0.007 (0.006)	0.01 (0.005)	0.02 (0.007)	0.02 (0.01)	0.007 (0.005)	0.01 (0.006)	0.02 (0.007)	0.008 (0.004)
ΔPop.75 – 95 (s.e.)	-0.80 (0.17)	-0.90 (0.17)	-1.09 (0.31)	-1.11 (0.35)	-0.55 (0.27)	-0.97 (0.18)	-1.06 (0.33)	-0.02 (0.33)
Africa (s.e.)	-0.37 (0.11)	-0.33 (0.10)	-0.41 (0.17)	-0.45 (0.17)	-0.33 (0.11)	-0.32 (0.11)	-0.41 (0.17)	-0.35 (0.13)
Share76–95 (s.e.)		0.34 (0.12)				0.56 (0.17)
Quinng82 (s.e.)			0.19 (0.06)				0.22 (0.06)
ΔQuinn73–88 (s.e.)				0.05 (0.08)
BHL80–95 (s.e.)					0.43 (0.15)			1.07 (0.28)
DM p-value						0.13	0.52	0.01
R2	0.47	0.52	0.59	0,44	0.53	0.50	0.58	0.34
no. of obs.	89	89	52	51	82	87	52	81

Regressions include a constant (not reported) Robust standard errors.Definition of Variables:ΔlnY_{76 – 95} = growth in In real per capita income, 1976 to 1995; In Y₁₉₇₆ = In real per capita income in 1976; ln(Educ.) = ln(secondary school enrollment rate); Invest. _74–78 = Average Investment, 1974 to 1978; ΔPop._76–95 = Population growth, 1976 to 1995; Africa = Dummy variable for African countries; Share_76–95 = Proportion of years with open capital accounts, 1976 to 1995, from IMF’s AREAER-, Quinn₈₂ = 0 to 4 measure of capital account openness in 1982 and ΔQuinn_{73 – 88} = Difference in 0 to 4 measure of capital account openness between 1973 and 1988, both from Quinn (1997); BHL_80–95 = Proportion of years between 1980 and 1995 with liberalized stock market, using dates of stock market liberalization from Bekaert, Harvey and Lundblad (2001).Instruments for Capital Account Liberalization Measures:Government consumption as proportion of GDP in 1976; Imports as proportion of GDP in 1976; Dummy variables for Latin American countries and East Asian countries; Quinn’s 0 to 4 measure of capital account openness in 1973 (for Quinn₈₂) or IMF’s AREAER 0/1 value in 1973 (for Shares _76–95 and for BHL_80–95).Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

View Table

Table 7.

Growth and Capital Account Liberalization

Dep. Var. ΔlnY76 – 95	OLS					IV
	1	2	3	4	5	6	7	8
In Y1976 (s.e.)	-0.24 (0.04)	-0.31 (0.05)	-0.48 (0.08)	-0.41 (0.09)	-0.33 (0.05)	-0.35 (0.06)	-0.49 (0.08)	-0.42 (0.06)
In (Educ.) (s.e.)	0.12 (0.05)	0.12 (0.05)	0.16 (0.08)	0.22 (0.10)	0.15 (0.05)	0.13 (0.05)	0.15 (0.09)	0.14 (0.05)
Invest.74-78 (s.e.)	0.007 (0.006)	0.01 (0.005)	0.02 (0.007)	0.02 (0.01)	0.007 (0.005)	0.01 (0.006)	0.02 (0.007)	0.008 (0.004)
ΔPop.75 – 95 (s.e.)	-0.80 (0.17)	-0.90 (0.17)	-1.09 (0.31)	-1.11 (0.35)	-0.55 (0.27)	-0.97 (0.18)	-1.06 (0.33)	-0.02 (0.33)
Africa (s.e.)	-0.37 (0.11)	-0.33 (0.10)	-0.41 (0.17)	-0.45 (0.17)	-0.33 (0.11)	-0.32 (0.11)	-0.41 (0.17)	-0.35 (0.13)
Share76–95 (s.e.)		0.34 (0.12)				0.56 (0.17)
Quinng82 (s.e.)			0.19 (0.06)				0.22 (0.06)
ΔQuinn73–88 (s.e.)				0.05 (0.08)
BHL80–95 (s.e.)					0.43 (0.15)			1.07 (0.28)
DM p-value						0.13	0.52	0.01
R2	0.47	0.52	0.59	0,44	0.53	0.50	0.58	0.34
no. of obs.	89	89	52	51	82	87	52	81

Regressions include a constant (not reported) Robust standard errors.Definition of Variables:ΔlnY_{76 – 95} = growth in In real per capita income, 1976 to 1995; In Y₁₉₇₆ = In real per capita income in 1976; ln(Educ.) = ln(secondary school enrollment rate); Invest. _74–78 = Average Investment, 1974 to 1978; ΔPop._76–95 = Population growth, 1976 to 1995; Africa = Dummy variable for African countries; Share_76–95 = Proportion of years with open capital accounts, 1976 to 1995, from IMF’s AREAER-, Quinn₈₂ = 0 to 4 measure of capital account openness in 1982 and ΔQuinn_{73 – 88} = Difference in 0 to 4 measure of capital account openness between 1973 and 1988, both from Quinn (1997); BHL_80–95 = Proportion of years between 1980 and 1995 with liberalized stock market, using dates of stock market liberalization from Bekaert, Harvey and Lundblad (2001).Instruments for Capital Account Liberalization Measures:Government consumption as proportion of GDP in 1976; Imports as proportion of GDP in 1976; Dummy variables for Latin American countries and East Asian countries; Quinn’s 0 to 4 measure of capital account openness in 1973 (for Quinn₈₂) or IMF’s AREAER 0/1 value in 1973 (for Shares _76–95 and for BHL_80–95).Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

We use three different indicators of capital account liberalization. First, as discussed in Section II, Share_76-95 represents the proportion of years in the period 1976 to 1995 that a country had open capital accounts, based on information from line E.2 of the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions. Second, we use the 0 to 4 measure of capital account openness in 1982 from Quinn (1997), which we denote as Quinn_82. As discussed above, Quinn used the change in the value of his indicator and, therefore, we also present a regression in which we use the difference in the value of the 0 to 4 measure of capital account openness between 1973 and 1988, ΔQuinn_73-88.²⁴ Third, we use the dates of stock market liberalization from Bekaert, Harvey and Lundblad (2001) to calculate the proportion of years between 1980 (when the data are first available) and 1995 that a country had a liberalized stock market. This variable, BHL₈o-95, is therefore analogous in its construction to Share._76-95.

Columns 2 through 5 of Table 7 present growth regressions in which the standard growth model presented in Column 1 is augmented with, in turn, each of these indicators of capital account liberalization. The results in this column generally support the hypothesis that liberalization of the capital account or of the equity market promotes growth, ceteris paribus. The coefficients on Share_76–95, Quinn₈₂ and BHL_8o–95 each have a p-value of less than 0.01. The coefficient on ΔQuinn_73–88 is positive but insignificant (the p-value is 0.50), a result at odds with those presented in Quinn (1997) in which the estimated coefficient on the change in the capital account openness measure is positive and significant.²⁵

A question that arises when considering the results in Columns 2 through 5 of Table 7 is whether there is an upward bias on the coefficients on liberalization because of reverse causality; perhaps the countries that are most likely to have liberalized their capital accounts are the ones that grew most quickly during the relevant period. To address this question, we present, in Columns 6 through 8 of Table 1, two-stage regressions that instrument for the values of the liberalization measures. The instruments that we use include government consumption as proportion of GDP in 1976, imports as proportion of GDP in 1976, a dummy variable for Latin American countries and another dummy variable for East Asian countries. In addition, for Share_76-95 and for BHL _{80- 95} we use, as an instrument, the 0 or 1 value of the capital account openness measure for 1973 from the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions while for the Quinn 0 to 4 measure of capital account openness in 1982 we use, as an instrument, the 1973 value of Quinn’s openness measure.

The results presented in Columns 6 through 8 of Table 7 indicate that, in all three cases, there is not an upward bias on the liberalization measures coefficients when we use OLS estimation since, in each case, the coefficient is larger using instrumental variables than when using OLS. Each of the liberalization coefficients presented in Columns 6 through 8 is significant at the 99 percent level of confidence or better. The adjusted R² for the first stage regression (which is not reported in the table) is 0.61 for Share₇₆ - ₉₅, 0.69 for Quinn₈₂, and 0.59 for BHL_80-95.

The results of a formal test concerning the possible endogeneity of the capital account openness variables are presented in Table 7 in the row labeled “DM p-value.” The statistics in this row are obtained using the artificial regression technique discussed in Davidson and MacKinnon (1989).²⁶ As shown in this row, the p-values on the coefficients on the instrumented regressors are 0.13 in the regression using Share_{76 – 95}, 0.52 in the regression using Quinn₈₂, and 0.01 in the regression using BHL_80-95 Therefore, estimates using OLS are not plagued by a significant (at the 10 percent level) problem with respect to consistency because of the endogeneity of the capital account openness variables in the regressions using either Share_{76 - 95}, or Quinn_82. There is evidence, however, that supports the use of instrumental variables in the regression using BHL_80-95. In the remainder of this section, we present results of the Davidson and MacKinnon test for all regressions and, where appropriate as indicated by this test, we estimate the regressions using instrumental variables rather than OLS.²⁷

B. The Robustness of the Results

An issue that arises in the literature surveyed above is whether the effect of capital account liberalization differs between industrial counties and developing countries. We investigate this in Columns 1 through 3 of Table 8 by reporting separate slope coefficients for industrial and developing countries (countries are classified as industrial if they were members of the OECD in 1986).²⁸ These results show that the estimated effect on economic growth of capital account openness or stock market liberalization is larger in developing countries than in industrial countries.²⁹ For developing countries, coefficients on both measures of capital account openness, Share76-95, and Quinn₈₂, are significant at better than the 99 percent level. In addition, the coefficient on Quinn₈₂ for industrial countries is also significant at better than the 99 percent level. The results for stock market liberalization, estimated using instrumental variables, also indicate strongly significant effects for both developing countries and industrial countries.

Table 8.

Differential Effects by OECD Membership or Region

All estimates using OLS or IV with robust standard errors. Row labeled “p-value of sum” of capital account liberalization and capital account liberalization times OECD dummy variable coefficients. Regressions include a constant (not reported). All variables as listed in notes to Table 7. Gov. Rep. = Knack and Keefer (1995) measure of degree to which governments do not repudiate contracts, range is 1–10 and larger valued indicate government less likely to repudiate contracts.Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

Table 8.

Differential Effects by OECD Membership or Region

Dep. Var. ΔlnY76–95	1	2	3	4	5	6
	Share76–95	Quinn82	BHL80–95	Share76–95	Quinn82	BHL80–95
Ln Y1976 (s.e.)	-0.29 (0.05)	-0.44 (0.07)	-0.35 (0.06)	-0.29 (0.07)	-0.37 (0.09)	-0,24 (0.06)
In (Educ.) (s.e.)	0.13 (0.05)	0.19 (0.08)	0.14 (0.05)	0.12 (0.05)	0.16 (0.09)	0.14 (0.05)
Invest.74–78 (s.e.)	0.01 (0.005)	0.02 (0.007)	0.004 (0.004)	0.01 (0.006)	0.009 (0.006)	0.002 (0.004)
Δpop.76–95 (s.e.)	-1.00 (0.20)	-1.33 (0.42)	-0.17 (0.33)	-1.03 (0.31)	-1.16 (0.37)	-0.69 (0.26)
Africa (s.e.)	-0.30 (0.11)	-0.35 (0.14)	-0.22 (0.14)	-0.33 (0.11)	-0.43 (0.10)	-0.22 (0.11)
Openness Measures by Region
OECD (s.e.)	0.20 (0.11)	0.13 (0.05)	0.92 (0.28)	0.14 (0.11)	0.08 (0.05)	0.22 (0.13)
non–OECD (s.e.)	0.43 (0.16)	0.21 (0.07)	1.67 (0.49)
East Asia (s.e.)				0.69 (0.12)	0.26 (0.04)	1.15 (0.17)
Latin America (s.e.)				-0.25 (0.11)	0.05 (0.08)	0.20 (0.27)
Other non–OECD (s.e.)				0.52 (0.38)	0.14 (0.12)	0.82 (0.33)
DM p-value	0.38	0.82	0.02	0.23	0.89	0.10
(OLS or IV)	(OLS)	(OLS)	(IV)	(OLS)	(OLS)	(OLS)
R2	0.53	0.61	0.37	0.61	0.72	0.62
no. of obs.	89	52	81	89	52	82

All estimates using OLS or IV with robust standard errors. Row labeled “p-value of sum” of capital account liberalization and capital account liberalization times OECD dummy variable coefficients. Regressions include a constant (not reported). All variables as listed in notes to Table 7. Gov. Rep. = Knack and Keefer (1995) measure of degree to which governments do not repudiate contracts, range is 1–10 and larger valued indicate government less likely to repudiate contracts.Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

View Table

Table 8.

Differential Effects by OECD Membership or Region

Dep. Var. ΔlnY76–95	1	2	3	4	5	6
	Share76–95	Quinn82	BHL80–95	Share76–95	Quinn82	BHL80–95
Ln Y1976 (s.e.)	-0.29 (0.05)	-0.44 (0.07)	-0.35 (0.06)	-0.29 (0.07)	-0.37 (0.09)	-0,24 (0.06)
In (Educ.) (s.e.)	0.13 (0.05)	0.19 (0.08)	0.14 (0.05)	0.12 (0.05)	0.16 (0.09)	0.14 (0.05)
Invest.74–78 (s.e.)	0.01 (0.005)	0.02 (0.007)	0.004 (0.004)	0.01 (0.006)	0.009 (0.006)	0.002 (0.004)
Δpop.76–95 (s.e.)	-1.00 (0.20)	-1.33 (0.42)	-0.17 (0.33)	-1.03 (0.31)	-1.16 (0.37)	-0.69 (0.26)
Africa (s.e.)	-0.30 (0.11)	-0.35 (0.14)	-0.22 (0.14)	-0.33 (0.11)	-0.43 (0.10)	-0.22 (0.11)
Openness Measures by Region
OECD (s.e.)	0.20 (0.11)	0.13 (0.05)	0.92 (0.28)	0.14 (0.11)	0.08 (0.05)	0.22 (0.13)
non–OECD (s.e.)	0.43 (0.16)	0.21 (0.07)	1.67 (0.49)
East Asia (s.e.)				0.69 (0.12)	0.26 (0.04)	1.15 (0.17)
Latin America (s.e.)				-0.25 (0.11)	0.05 (0.08)	0.20 (0.27)
Other non–OECD (s.e.)				0.52 (0.38)	0.14 (0.12)	0.82 (0.33)
DM p-value	0.38	0.82	0.02	0.23	0.89	0.10
(OLS or IV)	(OLS)	(OLS)	(IV)	(OLS)	(OLS)	(OLS)
R2	0.53	0.61	0.37	0.61	0.72	0.62
no. of obs.	89	52	81	89	52	82

All estimates using OLS or IV with robust standard errors. Row labeled “p-value of sum” of capital account liberalization and capital account liberalization times OECD dummy variable coefficients. Regressions include a constant (not reported). All variables as listed in notes to Table 7. Gov. Rep. = Knack and Keefer (1995) measure of degree to which governments do not repudiate contracts, range is 1–10 and larger valued indicate government less likely to repudiate contracts.Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

All three regressions reported in Columns 1 through 3 of Table 8 show a significant effect of capital account liberalization on growth among non-OECD countries but differ in the reported effect among OECD countries. One possible source of this difference is the samples used in each regression. We ran regressions using Share_76-95, allowing for different effects across industrial and developing countries, using the subsamples employed when using either Quinn₈₂ or BHL_80-95 as a regressor. The Quinn₈₂ sample of 52 countries yields estimates of the effect of capital account liberalization on growth of 0.17 for industrial countries and 0.34 for developing countries, with associated p-values of 0.20 and 0.06, respectively. These point estimates are both a bit smaller and somewhat less significant than those obtained with the full sample of 89 countries. With the sample of 82 countries used in the estimation of the effects of BHL_{80- 95} on growth, the estimated effect of Share_76-95 on growth for industrial countries is 0.20 (with a p- value of 0.07) and for developing countries it is 0.32 (with a p-value of 0.07). This suggests that differences across indicators due to differences in samples necessitated by data availability are not pronounced.

The discussion in Section II shows that the majority of cases of capital account liberalization and stock market liberalization outside of the OECD occurred among countries in Latin America and East Asia. Columns 4 through 6 of Table 8 continue our investigation of differential effects of liberalization across sets of countries by allowing for separate slopes across countries in Latin America, countries in East Asia, OECD member countries, and all other countries. These estimates suggest that the largest and most significant effect of capital account liberalization on growth occurred among East Asian countries. For each of the three variables, the effect of liberalization on growth in East Asia is significant at greater than the 99 percent level of confidence. In other regions, there is less evidence of a significant effect of capital account liberalization on growth. This effect is significant for the OECD countries in this set of regressions only when using the Quinn₈₂ indicator of capital account openness. Capital account liberalization is estimated to have a negative effect on growth among Latin American countries when using the Share_76-95 indicator.

Differences in the effect of capital account liberalization on growth across the set of industrial and developing countries are present in some of the research discussed in the literature review above. But, in those papers, (such as Klein and Olivei (2000) and Edwards (2001)), capital account liberalization is usually found to have a more significant effect in promoting growth in industrial countries. The results of this paper, however, suggest that open capital accounts and liberalized stock markets have a significant effect on growth among East Asian countries, but there is less consistent evidence of these effects elsewhere, even among industrial countries.

One concern with the results presented in Tables 7 and 8 is that it is possible that capital account liberalization is really just a proxy for some other characteristic of the countries in the sample. Rodrik (1998) argues that capital account liberalization measures serve as a proxy for the quality of government. We investigate this possibility in Table 9. This specifications used in the regressions reported in this table differ from those used in Table 8 only by the inclusion of a measure of government reputation. This measure, from Knack and Keefer (1995), draws on information from various volumes of the International Country Risk Guide. The variable potentially ranges from 1 to 10, with larger values indicating that a government is less likely to repudiate contracts. For most countries, the variable reflects reputation in 1982.

Table 9.

Testing Rodrik: Including Government Reputation as a Regressor

All estimates using OLS with robust standard errors. Row labeled “p-value of sum” is p-value of sum of capital account liberalization and capital account liberalization times OECD dummy variable coefficients. Regressions include a constant (not reported). All variables as listed in notes to Table 7. Gov. Rep. = Knack and Keefer (1995) measure of degree to which governments do not repudiate contracts, range is 1-10 and larger values indicate government less likely to repudiate contract.Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

Table 9.

Testing Rodrik: Including Government Reputation as a Regressor

Dep. Var. ΔlnY76–95	1	2	3	4	5	6
	Share76-95	Quinn82	BHL80–95	Share76-95	Quinn82	BHL80-95
Ln Y1976 (s.e.)	-0.41 (0.06)	-0.42 (0.06)	-0.40 (0.05)	-0.36 (0.06)	-0.33 (0.05)	-0.37 (0.05)
ln (Educ.) (s.e.)	0.16 (0.04)	0.12 (0.05)	0.13 (0.04)	0.15 (0.04)	0.10 (0.06)	0.13 (0.04)
Invest.74–78 (s.e.)	0.004 (0.004)	0.01 (0.004)	-0.0002 (0.003)	-0.001 (0.004)	0.001 (0.003)	-0.002 (0.003)
ΔPop.76-95 (s.e.)	-0.09 (0.25)	-0.46 (0.07)	-0.02 (0.21)	-0.10 (0.24)	-0.43 (0.15)	-0.01 (0.21)
Africa (s.e.)	-0.46 (0.08)	-0.46 (0.33)	-0.48 (0.05)	-0.50 (0.07)	-0.53 (0.06)	-0.49 (0.05)
Gov’t. Reputation (s.e.)	0.17 (0.03)	0.18 (0.04)	0.17 (0.03)	0.14 (0.04)	0.14 (0.03)	0.15 (0.03)
Openness Measures by Region
OECD (s.e.)	-0.09 (0.09)	-0.06 (0.05)	0.05 (0.09)	-0.09 (0.08)	-0.07 (0.04)	0.05 (0.09)
Non-OECD (s.e.)	0.22 (0.12)	0.06 (0.06)	0.54 (0.12)
East Asia (s.e.)				0.43 (0.11)	0.13 (0.02)	0.72 (0.11)
Latin America (s.e.)				-0.23 (0.09)	-0.05 (0.04)	0.18 (0.28)
Other non–OECD (s.e.)				0.76 (0.39)	0.03 (0.06)	0.28 (0.30)
DM p-value	0.23	0.59	0.15	0.41	0.53	0.27
R2	0.74	0.78	0.82	0.79	0.86	0.84
no. of obs.	71	50	68	71	50	68

All estimates using OLS with robust standard errors. Row labeled “p-value of sum” is p-value of sum of capital account liberalization and capital account liberalization times OECD dummy variable coefficients. Regressions include a constant (not reported). All variables as listed in notes to Table 7. Gov. Rep. = Knack and Keefer (1995) measure of degree to which governments do not repudiate contracts, range is 1-10 and larger values indicate government less likely to repudiate contract.Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

View Table

Table 9.

Testing Rodrik: Including Government Reputation as a Regressor

Dep. Var. ΔlnY76–95	1	2	3	4	5	6
	Share76-95	Quinn82	BHL80–95	Share76-95	Quinn82	BHL80-95
Ln Y1976 (s.e.)	-0.41 (0.06)	-0.42 (0.06)	-0.40 (0.05)	-0.36 (0.06)	-0.33 (0.05)	-0.37 (0.05)
ln (Educ.) (s.e.)	0.16 (0.04)	0.12 (0.05)	0.13 (0.04)	0.15 (0.04)	0.10 (0.06)	0.13 (0.04)
Invest.74–78 (s.e.)	0.004 (0.004)	0.01 (0.004)	-0.0002 (0.003)	-0.001 (0.004)	0.001 (0.003)	-0.002 (0.003)
ΔPop.76-95 (s.e.)	-0.09 (0.25)	-0.46 (0.07)	-0.02 (0.21)	-0.10 (0.24)	-0.43 (0.15)	-0.01 (0.21)
Africa (s.e.)	-0.46 (0.08)	-0.46 (0.33)	-0.48 (0.05)	-0.50 (0.07)	-0.53 (0.06)	-0.49 (0.05)
Gov’t. Reputation (s.e.)	0.17 (0.03)	0.18 (0.04)	0.17 (0.03)	0.14 (0.04)	0.14 (0.03)	0.15 (0.03)
Openness Measures by Region
OECD (s.e.)	-0.09 (0.09)	-0.06 (0.05)	0.05 (0.09)	-0.09 (0.08)	-0.07 (0.04)	0.05 (0.09)
Non-OECD (s.e.)	0.22 (0.12)	0.06 (0.06)	0.54 (0.12)
East Asia (s.e.)				0.43 (0.11)	0.13 (0.02)	0.72 (0.11)
Latin America (s.e.)				-0.23 (0.09)	-0.05 (0.04)	0.18 (0.28)
Other non–OECD (s.e.)				0.76 (0.39)	0.03 (0.06)	0.28 (0.30)
DM p-value	0.23	0.59	0.15	0.41	0.53	0.27
R2	0.74	0.78	0.82	0.79	0.86	0.84
no. of obs.	71	50	68	71	50	68

All estimates using OLS with robust standard errors. Row labeled “p-value of sum” is p-value of sum of capital account liberalization and capital account liberalization times OECD dummy variable coefficients. Regressions include a constant (not reported). All variables as listed in notes to Table 7. Gov. Rep. = Knack and Keefer (1995) measure of degree to which governments do not repudiate contracts, range is 1-10 and larger values indicate government less likely to repudiate contract.Bold denotes p-value of coefficient 0.05 or less.Italic denotes p-value of coefficient 0.05 to 0.10.

The results in Table 9 support the importance of government reputation as a determinant of growth but do not support the claim that capital account liberalization merely serves as a proxy for government reputation. In each of the six regressions, the coefficient on government reputation is significant at better than the 99 percent level of confidence. While it is true that the neither the coefficient on Quinn₈₂ for industrial nor for developing countries is significant in Column 2 (in contrast to the results presented in Table 8), and that the coefficient on Share_{76 – 95} for developing countries has a p-value of 0.07 as compared to 0.01 when government reputation is not included as a regressor (Column 1 in each table), the coefficient on BHL_80-95 for developing countries reported in Column 3 as well as each of the coefficients on the capital account openness or stock market liberalization variables for East Asia in Columns 4 through 6 continue to be significant at better than the 99 percent level of confidence.

A careful comparison of Tables 8 and 9 shows that the introduction of the government reputation variable reduces the number of observations in the regressions using Share_{86 – 95} by about one-fifth, from 89 to 71. This is part of the reason for the differences in the results concerning the coefficients on capital account liberalization in Columns 1 and 4 across these two tables. A regression with a specification like that in Column 1 of Table 8 that uses the sample of 71 countries that have non-missing values of the government repudiation variable (i.e. the sample used in Columns 1 and 4 of Table 9) has estimated coefficients of 0.17 (with a p-value of 0.09) for industrial countries and 0.34 (with a p-value of 0.06) for developing countries. These point estimates are lower than those reported in Column 1 of Table 8 and these coefficients are less significant in the more restricted sample. When allowing for different estimates across regions of non- OECD countries, the main difference in the value and pattern of significance between the samples of 71 and 89 countries is that the former yields a significant effect of capital account liberalization on growth for non-OECD countries outside of Latin America and East Asia; the estimated coefficient on Share_56-95 for this group of countries is 1.13, with a standard error of 0.47 (and therefore a p-value of 0.02). The pattern of significance of the effects of stock market liberalization on growth presented in Column 6 of Table 8 do not change very much if we restrict the sample to the 68 countries that have non-missing observations for both the government reputation variable and BHL_80-95. Therefore, the loss of significance of the coefficient on BHL_{80 – 95} for non-OECD countries outside of East Asia and Latin America is likely due to the correlation of government reputation with BHL_80-95 for these countries. But note that the coefficient on the BHL _{80- 95} for East Asian countries retains its significance in the results presented in Column 6 of Table 9.