I. Introduction

Ever since the Asian financial crisis, Asian policymakers have embarked in a number of initiatives to foster regional cooperation and financial integration. This drive has been motivated to a large extent by the desire to enhance resilience against the vagaries of global financial markets by developing a local-currency denominated bond market and beefing up regional reserves. The “Manila Framework” was developed in 1997 as a “new framework for enhanced Asian regional cooperation to promote financial stability”. Other important steps toward regional financial integration include liquidity support arrangements through the Chiang Mai Initiative Multilateralization, the Asian Bond Fund, the Asian Bond Market Initiative, and financial forums such as the Association of Southeast Asian Nations Plus Three and the Executives’ Meeting of East Asia–Pacific Central Banks. The Association of Southeast Asian Nations (ASEAN) has also outlined plans to foster capital market integration, including by building capital market infrastructure and harmonizing regulations.¹

In spite of these efforts, though, the empirical evidence indicates that regional financial integration lags behind trade integration (IMF, 2014), and that Asian economies maintain stronger financial links with the rest of the world than with other economies in the region (Borensztein and Loungani 2011; Eichengreen and Park 2004; Garcia-Herrero, Yang, and Wooldridge 2008; Pongsaparn and Unteroberdoerster 2011).

This paper takes a fresh look at the status of financial integration within Asia and at possible factors hindering progress, focusing on portfolio investment and banking claims. More specifically, it attempts to address the following questions: how financially integrated are Asian economies within the region? Has Asia’s regional financial integration increased? And how does it compare to other regions? What are the drivers of financial integration? And, hence, what are the implications for Asian policymakers pursuing deeper regional financial integration?

To answer these questions we first review recent trends in the share of cross-border holdings of portfolio investment assets and bank claims within Asia compared to outside the region. Next, we estimate the home bias—that is, the tendency to invest more in one’s home country than abroad—in Asia and other regions. Then, through a gravity model, we study the main drivers of financial integration—focusing in particular on the role of regulations—and use the results to draw implications for Asia.

The paper finds that the degree of financial integration within Asia has increased, but remains relatively low, especially when compared with Asia’s high degree of trade integration. Moreover, financial linkages within Asia are less strong than those within the euro area and the European Union, but tighter than those in Latin America. The home bias is found to be particularly strong in Asia, limiting cross-border financial transactions within the region.

The gravity model estimates indicate that cross-border portfolio investment assets and bank claims increase with the size and sophistication of financial systems and the extent of trade integration. In addition, restrictions on cross-border capital flows, informational asymmetries, barriers to foreign bank entry, and differences in regulatory and institutional quality create obstacles to financial integration.

Hence, initiatives to advance Asian policymakers’ agenda toward deeper regional integration could include steps to further promote financial market development and trade linkages, and reduce informational asymmetries through increased financial disclosure and reporting requirements. Lowering regulatory barriers to capital movements and foreign bank entry, as well as harmonizing regulation, especially for investor protection, contract enforcement, and bankruptcy procedures, appear particularly important.

II. Regional Financial Integration in Asia: Recent Trends

There is no single and universally accepted definition and measurement of financial integration. The term is sometimes used to indicate financial openness and free cross-border capital movements. In some studies financial integration is intended as equalization of prices among assets with similar risk and return profiles among a group of countries—the so called “law of one price” (e.g., Fukuda, 2011). In others, it is interpreted as reduction in the cost for trading financial assets (Martin, 2011).

This paper uses as indicator of regional financial integration the share of cross-border portfolio investment and bank claims that is intraregional.² We prefer to rely on quantity-based measures of integration, instead of price-based indicators—such as yields and returns co-movements—because the latter may be affected by global common factors that are unrelated to regional financial integration.

Unlike foreign direct investment (FDI), most of Asia’s portfolio investment is from or directed to outside the region (Figure 1 and 2). About 70 percent of direct investment is originated from within the region, and around 60 percent of Asian FDI is toward the region—with transactions between China and Hong Kong SAR accounting for nearly half of the intraregional total. On the other hand, most portfolio investment to Asia originates from the United States and advanced Europe, although the share of Asian origin increased from about 15 percent in 2001 to about 23 percent in 2013. The share of outward portfolio investment to the rest of the region grew from 10 percent to 24 percent over the same period, but North America and advanced Europe remained the main destinations. However, the shares of intraregional portfolio investment are higher when Japan—the largest portfolio investment source and destination country in Asia—is excluded, reaching 30 percent to 40 percent in 2013.³

View Full Size

Asia: Foreign Direct Investment

(Percent of total foreign direct investment to and from Asia)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: IMF, Coordinated Direct Investment Survey database; and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Asia: Foreign Portfolio Investment

(Percent of total foreign portfolio investment to and from Asia)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: IMF, Coordinated Portfolio Investment Survey database; and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

The share of regional inward portfolio investment is fairly homogeneous across Asian economies, with Japan and China being the main outliers (Figure 3). The high intraregional share in the latter reflects transactions between mainland and Hong Kong SAR. As expected, intraregional portfolio inward investment in Asia is low compared to the EU—only one third. On the other hand, intraregional portfolio inward investment in Asia is significantly higher than in Latin America. The share of Asia’s outward portfolio investment directed toward the region is rather heterogeneous across countries (Figure 4). Overall, though, it is smaller than in the EU, and higher than in Latin America.

View Full Size

Sources of Portfolio Inward Investment

(Percent; end-2013)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: IMF, Coordinated Portfolio Investment Survey database; and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Destinations of Portfolio Outward Investment

(Percent; end-2013)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: IMF, Coordinated Portfolio Investment Survey database; and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Hong Kong SAR and Singapore serve as two important financial centers, increasing financial transactions within Asia. Hong Kong SAR is often considered the “gateway” to China, while Singapore is the regional financial center for Southeast Asia (Le Leslé, at al., 2014). The share of Singapore’s foreign portfolio liabilities originating in Asia almost doubled from 13 percent in 2001 to 25 percent in 2013, with the share of portfolio assets in the rest of the region originating from Singapore increasing from 39 percent to 49 percent. For Hong Kong SAR, the rise in inward portfolio investment from Asia (excluding China) has been modest—from 15 percent to 18 percent—while portfolio assets from Hong Kong SAR to Asia (excluding China) have remained roughly stable at around 30 percent.

Asia’s cross-border banking linkages remain stronger between Asian economies and economies outside of Asia than among economies within the region, although intraregional foreign bank claims have increased. The share of foreign bank claims originating from within the region more than doubled, from 13 percent in 2001 to 30 percent in 2013, according to Bank for International Settlements (BIS) consolidated data (Figure 5).⁴ This surge reflects the expansion of Japanese and Australian banks in the region, especially since the global financial crisis, when European banks retrenched (IMF, 2015; Lam 2013). BIS locational data point to a similar degree of intraregional banking linkages.⁵ According to this metric, about 20 percent of foreign claims originated within the Asian region in 2013, and about 25 percent of Asia’s foreign bank claims were directed to the rest of that region (Figure 6).

View Full Size

Sources of Foreign Bank Claims on Asia

(Consolidated data; percent; end of period)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: Bank for International Settlements; and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Asia: Foreign Bank Claims

(Locational data; percent of total foreign bank claims to and from Asia)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: Bank for International Settlements; and IMF staff calculations.Note: Other includes remaining regions, unallocated locations, and offshore centers.

• Download Figure
• Download figure as PowerPoint slide

III. Home Bias in Asia

What accounts for the rather slow pace of regional financial integration in Asia, in spite of policymakers’ initiatives? One explanation is that most of Asia’s private financial investment remains within the domestic economy, rather than going abroad; in other words, home bias is strong in Asia. In fact, on average, Asian investors hold only 13 percent of their total equity portfolio in foreign markets (Figure 7). Conversely, the share of cross-border equity investment out of the total equity portfolio is much higher in other regions—31 percent in the EU and 22 percent in Latin America. When compared with the world portfolio allocation benchmarks, the gap between actual investment and the benchmark is lower for Asia’s intra-regional investments than for the inter-regional investment.⁶ This suggests that, once controlling for market size, Asian investors are not discriminating against their own region as a destination for investments. Nevertheless, the gap between actual intra-regional investment and the benchmark remains large for Asia, while it is very narrow for EU and Latin America.

View Full Size

Equity Holdings in Foreign Markets

(Percent of total equity investments; simple average)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: IMF Coordinated Portfolio Investment Survey database; and IMF staff calculations.1/ Exclude equity holdings in domestic market.2/ The sum of the blue bars represents the share of total equity portfolio invested in foreign markets by each source region.

• Download Figure
• Download figure as PowerPoint slide

To further assess the size of home bias in Asia, also in comparison with other regions, a home bias index in equity markets is constructed for 50 countries over 2001-12.⁷ This measures the extent to which investors allocate a larger share of their portfolio in domestic equities, compared to the benchmark based on the size of the domestic market in the world stock market. The index ranges from 0 to 100, after normalization, with a higher number indicating greater home bias.

The average home bias in Asia—particularly in the ASEAN-5 economies (Indonesia, Malaysia, the Philippines, Singapore, Thailand)—according to the index is higher than that in the European Union and the United States, though it is lower than that in Latin America (Figure 8). Overall, there has been a clear downward trend in the home bias across all regions for most part of the 2000s, probably driven by increased financial globalization. However, this trend decline seems to have stalled in most regions after the global financial crisis (GFC), when international capital flows retrenched. Only in the European Union members the home bias continued to decline even after the GFC, as domestic investors moved out of their home stock market amidst market corrections and significant uncertainties over the region’s economic and financial outlook.

View Full Size

Home Bias Index across Regions

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: IMF Coordinated Portfolio Investment Survey database; and IMF staff calculations.Note: ASEAN-5 = Indonesia, Malaysia, the Philippines, Singapore and Thailand. The index range is from 0 to 100, with a higher number indicating greater home bias.

• Download Figure
• Download figure as PowerPoint slide

What explains the home bias in equity holdings? The large literature on determinants of financial investment destinations points to three main potentially explanatory factors, namely (i) the level of economic and financial development, (ii) policy restrictions, such as capital control measures, and (iii) implicit transaction costs arising from information frictions, real exchange rate risk, country risk, and corporate governance issues (Chan, Covrig and Ng, 2005; and Bekaert and Wang (2009).

Indeed, there is a negative correlation between the home bias and the level of economic development (Figure 9). A simple panel regression analysis confirms that GDP per capita, financial development (proxied by the share of domestic bank assets to GDP), stock market size, and the degree of capital account liberalization (measured by the Chinn-Ito index of financial openness) are significant determinants of home bias (Table 1).⁸ Interestingly, the estimated coefficient on the stock market size variable, which could potentially be a proxy for the level of financial development, has a positive sign. This is perhaps because a larger domestic stock market is more liquid and entails lower transaction costs, thus making domestic equity investment relatively more attractive, after controlling for the level of financial development.⁹

View Full Size

Home Bias and Economic Development

(GDP per capital in thousands of US dollars; average of 2001-2012)

Citation: IMF Working Papers 2015, 160; 10.5089/9781513554815.001.A001

Sources: IMF Coordinated Portfolio Investment Survey database; and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Table 1:

Home Bias Regressions

Source: IMF staff estimates. Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 1:

Home Bias Regressions

Dependent Variable:	(1)	(2)	(3)
Home Bias Index (0-100)
GDP per Capita	−0.353***	−0.317***	0.042
	(−4.615)	(−3.965)	(0.585)
Bank Asset Size / GDP	−0.122***	−0.140***	−0.050*
	(−5.976)	(−6.334)	(−1.773)
Stock Market Size / GDP	0.057***	0.054***	0.056***
	(6.014)	(4.978)	(3.544)
Capital Account Openness	−3.554***	−4.022***	−1.968**
	(−4.465)	(−3.975)	(−2.374)
Asia Dummy		10.213*** (3.925)
EU Dummy		6.169** (1.985)
Latin America Dummy		3.624 (1.406)
Constant	101.632***	95.944***	97.423***
	(36.599)	(27.105)	(38.038)
Time Effects	YES	YES	YES
Country Effects	NO	NO	YES
Observations	538	538	538
R-squared	0.407	0.429	0.937

Source: IMF staff estimates. Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

View Table

Table 1:

Home Bias Regressions

Dependent Variable:	(1)	(2)	(3)
Home Bias Index (0-100)
GDP per Capita	−0.353***	−0.317***	0.042
	(−4.615)	(−3.965)	(0.585)
Bank Asset Size / GDP	−0.122***	−0.140***	−0.050*
	(−5.976)	(−6.334)	(−1.773)
Stock Market Size / GDP	0.057***	0.054***	0.056***
	(6.014)	(4.978)	(3.544)
Capital Account Openness	−3.554***	−4.022***	−1.968**
	(−4.465)	(−3.975)	(−2.374)
Asia Dummy		10.213*** (3.925)
EU Dummy		6.169** (1.985)
Latin America Dummy		3.624 (1.406)
Constant	101.632***	95.944***	97.423***
	(36.599)	(27.105)	(38.038)
Time Effects	YES	YES	YES
Country Effects	NO	NO	YES
Observations	538	538	538
R-squared	0.407	0.429	0.937

Source: IMF staff estimates. Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

Another noteworthy result from the regressions is that, although the average home bias is lower in Asia than in Latin America (Figure 8), once the level of economic and financial development and capital account openness are controlled for, Asia seems to have much higher residual home bias than Latin America, as captured by the Asia dummy variable (Table 1, Column (2)). The fact that home bias has been particularly strong in Asian economies could be an important factor hindering intraregional financial integration in Asia as most financial investment remains within each country’s border instead of being directed toward other countries in the region.

IV. Drivers of Financial Integration

What are the main factors driving financial integration between countries? In other words, what are the determinants of cross-border bilateral financial investment? To answer these questions, we estimate a gravity model, based on the theoretical framework developed in Martin and Rey (2004) and Aviat and Coeurdacier (2007).¹⁰ More specifically, the basic estimating equation is as follows:

where Asset_ijt are the asset holdings of country i in country j. MktSize_i and MktSize_j are the market size of country i and country j, respectively. Z_ij are proxies for transaction costs on financial asset trading between the two countries. R_j is a set of variables affecting the expected return on asset holdings in the destination country.

We run two sets of regressions. In the first, the dependent variable is total portfolio assets (equities and bonds), obtained from the IMF’s Coordinated Portfolio Investment Survey (CPIS). In the second, the dependent variable is cross-border bank claims from the Bank of International Settlements.¹¹

When the dependent variable is total portfolio holdings, as a measure of market size MktSize_i and MktSize_j we use the sum of equity market capitalization and the value of the domestic bond market in each country. In regressions where the dependent variable is bilateral bank claims, nominal GDP is the proxy for market size.

Indicators for expected returns R_j include interest differentials between the source and destination country, past returns of stock indexes in the destination country, change in recipient country’s exchange rate vis-à-vis the source country’s currency, exchange rate volatility, as well as measures of political, macroeconomic, and financial risks in the destination country. To test whether portfolio diversification is a relevant factor in driving investor decisions, additional explanatory variables are the covariance between real GDP growth of the source and destination country, the covariance of their stock market returns, and the covariance between consumption growth in the source country and stock returns in the destination country, at various time horizons (Appendix III).

Transaction costs on financial asset trading are mainly driven by different types of frictions, which can be grouped into two broad categories, direct and indirect barriers.

Direct barriers are the restrictions imposed on foreign investors in acquiring assets in a particular country, and/or on domestic investors of that country in trading foreign assets. These are measured by the capital account openness indexes developed by Chinn and Ito (2006) and Quinn (1997).

Indirect barriers include informational asymmetries, poor financial market infrastructure, and differences in regulatory and institutional quality. As shown in Portes and Rey (2005), informational asymmetries can be well proxied by the distance between the two countries and the lack of a common language because these factors hinder the interaction among economic agents and, hence, the exchange of knowledge about market structures, corporate culture, and other information that may be important for investment decisions. Thus, we use the log of geographical distance between the two capital cities of country pairs as a measure of “informational distance”. A dummy for “common language” is also used to measure whether country pairs share the same language. Furthermore, the size of bilateral trade between two countries is included as an additional explanatory variable, as there can be information spillovers from goods trading into financial assets trading (Aviat and Coeurdacier, 2007; and Lane and Milesi-Ferretti, 2004).¹²

Limited financial market sophistication and infrastructure could also create indirect barriers to financial asset trading. Hence, per capita GDP is added to the explanatory variables set, as a proxy for financial markets sophistication and quality of transaction technology.

Or main hypothesis—and departure from the literature—is that differences in regulatory and institutional quality among countries can be important indirect barriers to financial integration. Indeed, investors may be reluctant to carry out financial transactions with countries whose regulations and institutions are very different from their own. Hence, we include several explanatory variables as proxies of regulatory and institutional quality differences, including indicators of the degree of investor protection, quality of insolvency law and contract enforcement (Appendix III). Also departing from the literature, we test whether a strong foreign bank presence in a county—or regulation favoring foreign bank penetration—support financial integration, by reducing informational asymmetry and transaction costs in cross-border financial transactions. The results are summarized in Table 2.

Table 2:

Summary of the Results ^{1/ 2/}

Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/Color green indicates that the coefficient of the corresponding variable is statistically significant or highly significant using alternative estimation methods. Color yellow indicates that the coefficient of the corresponding variable is statistically significant or highly significant only with some estimation methods.

Table 2:

Summary of the Results ^{1/ 2/}

Dependent Variable:	Bilateral portfolio investment	Bilateral bank claims (consolidated)	Bilateral bank claims (locational)
Log (Market Size) - Source	+	+	+
Log (Market Size) - Destination	+	+	+
Log (Distance)	−	−	−
Common Language	+	+	+
GDP per Capita - S	+	+	+
GDP per Capita - D	+	+	+
Bilateral Trade	+	+	+
Capital Account Openness - S	+	+	+
Capital Account Openness - D	+	+	+
Foreign Bank Presence (Asset Share) - D	+
Foreign Bank Presence (Number Share) - D
Financial and Banking Regulation Index - S	+
Financial and Banking Regulation Index - D	Non significant
Financial and Banking Regulation Index - Difference	−
Regulation of Securities Exchanges - S	+
Regulation of Securities Exchanges - D	Non significant
Regulation of Securities Exchanges - Difference	−
Regulatory Quality - S	+
Regulatory Quality - D	+
Regulatory Quality - Difference	−
Rule of Law - S	+
Rule of Law - D	+
Rule of Law - Difference	−
Protecting Investors Index - S	+
Protecting Investors Index - D	+
Protecting Investors Index - Difference	−
Enforcing Contracts Index - S	Non significant
Enforcing Contracts Index - D	+
Enforcing Contracts Index - Difference	−
Covariance of Real GDP Growth (lagged)	+
Interest rate differential (lagged)	+
Stock returns in Destination (lagged)	Non significant
FX appreciation of D against S (lagged)	−
Political Risk - D (lagged)	−
Economic Risk - D (lagged)	Non significant
Financial Risk - D (lagged)	+
Bank Branches per Population - D (lagged)	+
Private Credit to GDP - D (lagged)	+
Stock Market Cap to GDP - D (lagged)	−
Mutual Fund Assets to GDP - D	+
Foreign Bank Entry Application Denied Ratio - D		−	−
Foreign Bank Entry Prohibition Index - D		−	−
Ability of Resolving Insolvency Index - Difference		−	Non significant
Depth of Credit Information Index - Difference		−	Non significant
Auditing Standard - Difference		−	Non significant
Stringency of Minimum Capital Requirements Index - Differerence		Non significant	−
Actual Capital Ratio - Difference		Non significant	−
Accounting Standard - Difference		Non significant	−
Auditing Standard - Difference		Non significant	−

Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/Color green indicates that the coefficient of the corresponding variable is statistically significant or highly significant using alternative estimation methods. Color yellow indicates that the coefficient of the corresponding variable is statistically significant or highly significant only with some estimation methods.

View Table

Table 2:

Summary of the Results ^{1/ 2/}

Dependent Variable:	Bilateral portfolio investment	Bilateral bank claims (consolidated)	Bilateral bank claims (locational)
Log (Market Size) - Source	+	+	+
Log (Market Size) - Destination	+	+	+
Log (Distance)	−	−	−
Common Language	+	+	+
GDP per Capita - S	+	+	+
GDP per Capita - D	+	+	+
Bilateral Trade	+	+	+
Capital Account Openness - S	+	+	+
Capital Account Openness - D	+	+	+
Foreign Bank Presence (Asset Share) - D	+
Foreign Bank Presence (Number Share) - D
Financial and Banking Regulation Index - S	+
Financial and Banking Regulation Index - D	Non significant
Financial and Banking Regulation Index - Difference	−
Regulation of Securities Exchanges - S	+
Regulation of Securities Exchanges - D	Non significant
Regulation of Securities Exchanges - Difference	−
Regulatory Quality - S	+
Regulatory Quality - D	+
Regulatory Quality - Difference	−
Rule of Law - S	+
Rule of Law - D	+
Rule of Law - Difference	−
Protecting Investors Index - S	+
Protecting Investors Index - D	+
Protecting Investors Index - Difference	−
Enforcing Contracts Index - S	Non significant
Enforcing Contracts Index - D	+
Enforcing Contracts Index - Difference	−
Covariance of Real GDP Growth (lagged)	+
Interest rate differential (lagged)	+
Stock returns in Destination (lagged)	Non significant
FX appreciation of D against S (lagged)	−
Political Risk - D (lagged)	−
Economic Risk - D (lagged)	Non significant
Financial Risk - D (lagged)	+
Bank Branches per Population - D (lagged)	+
Private Credit to GDP - D (lagged)	+
Stock Market Cap to GDP - D (lagged)	−
Mutual Fund Assets to GDP - D	+
Foreign Bank Entry Application Denied Ratio - D		−	−
Foreign Bank Entry Prohibition Index - D		−	−
Ability of Resolving Insolvency Index - Difference		−	Non significant
Depth of Credit Information Index - Difference		−	Non significant
Auditing Standard - Difference		−	Non significant
Stringency of Minimum Capital Requirements Index - Differerence		Non significant	−
Actual Capital Ratio - Difference		Non significant	−
Accounting Standard - Difference		Non significant	−
Auditing Standard - Difference		Non significant	−

Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/Color green indicates that the coefficient of the corresponding variable is statistically significant or highly significant using alternative estimation methods. Color yellow indicates that the coefficient of the corresponding variable is statistically significant or highly significant only with some estimation methods.

A. Results on the determinants of bilateral portfolio investment

Baseline regressions

The baseline model specification includes only the main explanatory variables typically featured in gravity-models, namely the market size of the source and destination country, geographic distance, and a common language dummy variable. The dependent variable is asset holdings by the source country in the destination country (Table 3), or the sum of assets and liabilities of the source country toward the destination country (Table 4). All equations include time dummies to control for aggregate shocks that are common across all country-pairs at each point in time. Standard errors are robust to heteroskedasticity, and clustered at the country-pair level. To check for robustness, different econometric estimation techniques are used: the pooled OLS, between effects, random effects, destination-country fixed effects, country-pair fixed effects and the Hausman-Taylor estimator.¹³

Table 3:

Financial Gravity Model - Portfolio Investment; Baseline Regressions

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

Table 3:

Financial Gravity Model - Portfolio Investment; Baseline Regressions

Dependent Variable:		(1)	(2)	(3)	(4)	(5)	(6)
Log (Portfolio Assets)		OLS	Between	Random	Destination	Pair-Fixed	Hausman-
			Effects	Effects	Country-FE	Effects	Taylor
Log (Stock and Bond Market Size) - Source		1.390***	1.151***	0.962***	1.383***	0.199***	0.224***
		(39.099)	(31.589)	(35.911)	(41.323)	(4.360)	(5.251)
Log (Stock and Bond Market Size) - Destination		1.640***	1.367***	1.319***	0.397***	0.476***	0.523***
		(70.859)	(48.602)	(60.691)	(5.856)	(11.611)	(13.749)
Log (Distance)		−2.834***	−2.500***	−2.781***	−3.062***		−2.722***
		(−35.219)	(−31.039)	(−34.895)	(−34.362)		(−4.381)
Common Language		1.534***	1.697***	1.755***	1.941***		26.327***
		(6.270)	(7.693)	(7.988)	(7.790)		(3.257)
Constant		20.473***	14.771***	21.640***	35.817***	5.451***	23.275***
		(28.914)	(3.953)	(32.498)	(31.417)	(22.251)	(4.719)
Observations		55,059	55,059	55,059	55,059	55,059	55,059
R-squared		0.465	0.579	0.590	0.519	0.054
Time Dummies		YES	YES	YES	YES	YES	YES
Number of Pairid		5,900	5,901	5,901	5,902	5,901	5,901
Hausman Test (p-value)				536 (0.00)			10.59 (0.47)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

View Table

Table 3:

Financial Gravity Model - Portfolio Investment; Baseline Regressions

Dependent Variable:		(1)	(2)	(3)	(4)	(5)	(6)
Log (Portfolio Assets)		OLS	Between	Random	Destination	Pair-Fixed	Hausman-
			Effects	Effects	Country-FE	Effects	Taylor
Log (Stock and Bond Market Size) - Source		1.390***	1.151***	0.962***	1.383***	0.199***	0.224***
		(39.099)	(31.589)	(35.911)	(41.323)	(4.360)	(5.251)
Log (Stock and Bond Market Size) - Destination		1.640***	1.367***	1.319***	0.397***	0.476***	0.523***
		(70.859)	(48.602)	(60.691)	(5.856)	(11.611)	(13.749)
Log (Distance)		−2.834***	−2.500***	−2.781***	−3.062***		−2.722***
		(−35.219)	(−31.039)	(−34.895)	(−34.362)		(−4.381)
Common Language		1.534***	1.697***	1.755***	1.941***		26.327***
		(6.270)	(7.693)	(7.988)	(7.790)		(3.257)
Constant		20.473***	14.771***	21.640***	35.817***	5.451***	23.275***
		(28.914)	(3.953)	(32.498)	(31.417)	(22.251)	(4.719)
Observations		55,059	55,059	55,059	55,059	55,059	55,059
R-squared		0.465	0.579	0.590	0.519	0.054
Time Dummies		YES	YES	YES	YES	YES	YES
Number of Pairid		5,900	5,901	5,901	5,902	5,901	5,901
Hausman Test (p-value)				536 (0.00)			10.59 (0.47)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

Table 4:

Financial Gravity Model - Portfolio Investment; Baseline Regressions

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

Table 4:

Financial Gravity Model - Portfolio Investment; Baseline Regressions

Dependent Variable:		(1)	(2)	(3)	(4)	(5)	(6)
Log (Portfolio Assets + Liabilities)		OLS	Between	Random	Destination	Pair-Fixed	Hausman-
			Effects	Effects	Country-FE	Effects	Taylor
Log (Stock and Bond Market Size) - Source		1.779***	1.583***	1.519***	−0.056	0.620***	0.661***
		(54.266)	(44.162)	(55.896)	(−0.592)	(11.682)	(13.299)
Log (Stock and Bond Market Size) - Destination		1.547***	1.348***	1.279***	0.641***	0.602***	0.635***
		(52.360)	(43.631)	(50.977)	(7.507)	(12.141)	(13.767)
Log (Distance)		−2.966***	−2.790***	−2.867***	−2.127***		−3.461***
		(−38.002)	(−34.858)	(−36.176)	(−20.926)		(−6.668)
Common Language		1.158***	1.119***	1.253***	2.551***		45.205***
		(4.107)	(4.579)	(5.138)	(10.409)		(5.359)
Constant		22.067***	21.767***	22.442***	5.459**	6.377***	28.092***
		(33.027)	(5.958)	(34.274)	(2.279)	(21.124)	(6.982)
Observations		42,465	42,465	42,465	42,465	42,465	42,465
R-squared		0.515	0.632	0.520	0.678	0.089
Time Dummies		YES	YES	YES	YES	YES	YES
Number of Pairid		4,731	4,732	4,732	4,733	4,732	4,732
Hausman Test (p-value)				787 (0.00)			11.53 (0.40)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

View Table

Table 4:

Financial Gravity Model - Portfolio Investment; Baseline Regressions

Dependent Variable:		(1)	(2)	(3)	(4)	(5)	(6)
Log (Portfolio Assets + Liabilities)		OLS	Between	Random	Destination	Pair-Fixed	Hausman-
			Effects	Effects	Country-FE	Effects	Taylor
Log (Stock and Bond Market Size) - Source		1.779***	1.583***	1.519***	−0.056	0.620***	0.661***
		(54.266)	(44.162)	(55.896)	(−0.592)	(11.682)	(13.299)
Log (Stock and Bond Market Size) - Destination		1.547***	1.348***	1.279***	0.641***	0.602***	0.635***
		(52.360)	(43.631)	(50.977)	(7.507)	(12.141)	(13.767)
Log (Distance)		−2.966***	−2.790***	−2.867***	−2.127***		−3.461***
		(−38.002)	(−34.858)	(−36.176)	(−20.926)		(−6.668)
Common Language		1.158***	1.119***	1.253***	2.551***		45.205***
		(4.107)	(4.579)	(5.138)	(10.409)		(5.359)
Constant		22.067***	21.767***	22.442***	5.459**	6.377***	28.092***
		(33.027)	(5.958)	(34.274)	(2.279)	(21.124)	(6.982)
Observations		42,465	42,465	42,465	42,465	42,465	42,465
R-squared		0.515	0.632	0.520	0.678	0.089
Time Dummies		YES	YES	YES	YES	YES	YES
Number of Pairid		4,731	4,732	4,732	4,733	4,732	4,732
Hausman Test (p-value)				787 (0.00)			11.53 (0.40)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

All the regressors have the expected signs and are highly significant, regardless of the econometric techniques, although the magnitude of the coefficients vary. This indicates that cross-border investment depends positively on market size of the source and destination country, and negatively on their physical distance, and is larger when the two countries share a common language, consistent with the results in Portes and Rey (2005).

The model including country-pair fixed effects can control for any time invariant omitted explanatory variable which is country-pair specific, but it is not suitable when some of the regressors are completely time invariant (e.g., common language or distance) or have limited variation over time, such as regulatory and institutional factors, which are the main focus of our analysis. The random effects estimator is not appropriate for our data since the null hypothesis of significant random effects is rejected by the Hausman test. In principle, the Hausman-Taylor estimation would be the best approach, since it allows both time-varying and pure cross-sectional regressors in the equation. However, most model specifications do not pass the Hausman’s specification test,¹⁴ and those that do tend to produce results that are quite sensitive. Therefore, we will rely mostly on the pooled OLS results for the rest of our empirical analysis, and perform robustness checks using fixed effects or Hausman-Taylor estimation when applicable.

Assessing intraregional financial integration

To investigate regional integration in Asia, and compare it to trends in other regions, intraregional dummy variables are added to the baseline specification. The Asia-intraregional dummy takes on the value of 1 if both source and destination countries are Asian, and 0 otherwise. The estimated coefficient on this variable measures the difference between the level of Asian economies among themselves relative to their level of integration with the rest of the world. Similar intraregional dummies are added for the EU, Latin America, and NAFTA. All intraregional dummies are significant when the market size of the source and destination countries are the only controlling variables (Table 5, column (1)). But when proxies for informational frictions are included, the coefficient of the dummies become smaller or insignificant, as proximity and common language may explain part of intraregional financial integration.

Table 5:

Financial Gravity Model - Portfolio Investment; ^1/ Regional Comparison

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/SGP=Singapore; HKG=Hong Kong SAR

Table 5:

Financial Gravity Model - Portfolio Investment; ^1/ Regional Comparison

Dependent Variable:		(1)	(2)	(3)	(4)	(5)
Log (Portfolio Assets)		OLS	OLS	OLS	OLS Excl.	Hausman-Taylor
					SGP&HKG
Log (Stock and Bond Market Size) - Source		1.333***	1.360***	1.380***	1.361***	0.226***
		(36.969)	(38.416)	(38.782)	(37.666)	(5.280)
Log (Stock and Bond Market Size) - Destination		1.554***	1.578***	1.611***	1.572***	0.528***
		(62.901)	(65.941)	(68.783)	(65.458)	(13.946)
Log (Distance)			−1.932***	−2.464**	−2.479***	−0.820
			(−18.136)	(−27.520)	(−27.244)	(−1.254)
Common Language			1.795***	1.615***	1.422***	29.556***
			(7.025)	(6.370)	(5.215)	(4.298)
Asia-Intra Dummy		2.261*** (5.449)	0.903** (2.392)			−1.681 (−0.572)
ASEAN-Intra Dummy				2.217*** (3.811)	1.287** (1.977)
Asia-Intra (excl. ASEAN) Dummy				−0.629 (−0.943)	−1.324* (−1.725)
European Union-Intra Dummy		7.613*** (38.959)	4.522*** (16.499)			11.055*** (5.676)
Euro area-Intra Dummy				4.355*** (16.578)	4.515*** (17.280)
European Union-Intra (excl. Euro area) Dummy				2.932***	2.947***
				(6.673)	(6.733)
Latin America-Intra Dummy		3.879***	1.294*	0.808	0.961	−17.399***
		(5.931)	(1.926)	(1.213)	(1.435)	(−2.924)
NAFTA-Intra Dummy		3.510***	0.655	−0.381	−0.037	4.540
		(2.836)	(0.472)	(−0.263)	(−0.026)	(0.317)
Constant		−3.356***	12.808***	17.360**	*17.690***	6.598
		(−17.209)	(14.022)	(22.322)	(22.411)	(1.213)
Observations		55,134	55,059	55,059	52,010	55,059
R-squared		0.453	0.480	0.472	0.470
Time Dummies		YES	YES	YES	YES	YES
Hausman Test						15.83
Hausman Test p-value						0.20

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/SGP=Singapore; HKG=Hong Kong SAR

View Table

Table 5:

Financial Gravity Model - Portfolio Investment; ^1/ Regional Comparison

Dependent Variable:		(1)	(2)	(3)	(4)	(5)
Log (Portfolio Assets)		OLS	OLS	OLS	OLS Excl.	Hausman-Taylor
					SGP&HKG
Log (Stock and Bond Market Size) - Source		1.333***	1.360***	1.380***	1.361***	0.226***
		(36.969)	(38.416)	(38.782)	(37.666)	(5.280)
Log (Stock and Bond Market Size) - Destination		1.554***	1.578***	1.611***	1.572***	0.528***
		(62.901)	(65.941)	(68.783)	(65.458)	(13.946)
Log (Distance)			−1.932***	−2.464**	−2.479***	−0.820
			(−18.136)	(−27.520)	(−27.244)	(−1.254)
Common Language			1.795***	1.615***	1.422***	29.556***
			(7.025)	(6.370)	(5.215)	(4.298)
Asia-Intra Dummy		2.261*** (5.449)	0.903** (2.392)			−1.681 (−0.572)
ASEAN-Intra Dummy				2.217*** (3.811)	1.287** (1.977)
Asia-Intra (excl. ASEAN) Dummy				−0.629 (−0.943)	−1.324* (−1.725)
European Union-Intra Dummy		7.613*** (38.959)	4.522*** (16.499)			11.055*** (5.676)
Euro area-Intra Dummy				4.355*** (16.578)	4.515*** (17.280)
European Union-Intra (excl. Euro area) Dummy				2.932***	2.947***
				(6.673)	(6.733)
Latin America-Intra Dummy		3.879***	1.294*	0.808	0.961	−17.399***
		(5.931)	(1.926)	(1.213)	(1.435)	(−2.924)
NAFTA-Intra Dummy		3.510***	0.655	−0.381	−0.037	4.540
		(2.836)	(0.472)	(−0.263)	(−0.026)	(0.317)
Constant		−3.356***	12.808***	17.360**	*17.690***	6.598
		(−17.209)	(14.022)	(22.322)	(22.411)	(1.213)
Observations		55,134	55,059	55,059	52,010	55,059
R-squared		0.453	0.480	0.472	0.470
Time Dummies		YES	YES	YES	YES	YES
Hausman Test						15.83
Hausman Test p-value						0.20

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/SGP=Singapore; HKG=Hong Kong SAR

The positive and significant coefficient of the Asia-intraregional dummy suggests that Asian economies are more integrated among themselves than with the rest of the world. The size of the coefficient indicates that integration is lower than in the EU, while comparable to the degree of integration in Latin America (Table 5, column (2)).¹⁵ However, the apparently higher intraregional integration in Asia is driven by ASEAN. In fact, when the Asia dummy is divided into an ASEAN-intraregional dummy (equal to 1 when both countries belong to ASEAN), and Non-ASEAN Asia intraregional (equal to 1 when both countries belong to Asia, but are outside of ASEAN), only the coefficient on the former is statistically significant (Table 5, column (3)). When Singapore and Hong Kong SAR—the two important financial centers in Asia—are removed from the sample, the coefficient on the ASEAN and Non-ASEAN Asia dummies became smaller, with the non-ASEAN Asia’s coefficient becoming negative and statistically significant (Table 5, Column (4)). These results suggest that most financial integration within Asia occurred among the ASEAN economies, with Singapore and Hong Kong SAR potentially playing an important role in facilitating cross-border financial asset holdings.¹⁶

When total bilateral portfolio investments are disaggregated by instrument, regression results indicate that ASEAN intraregional integration has been stronger in the equity and short-term debt securities markets (Table7). Conversely, Latin America’s financial integration seems more prominent in the long-term debt security market, while all portfolio investment markets are highly integrated in the Euro Area.

Assessing the determinants of bilateral portfolio investment: the role of regulation

We now expand the baseline model to include the additional variables discussed above.

The coefficients on GDP per capita of the source and destination countries—the proxy for market sophistication—are always positive and significant, and more so for the source than the destination country (Table 6).¹⁷ As expected, indicators of capital account openness are also found to have positive and significant coefficients, and openness in the source country seems to have a bigger impact on financial integration.¹⁸

Table 6:

Financial Gravity Model - Portfolio Investment; ^1/ Different Types of Portfolio Assets

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/Market size for:

Column (1) is measured as total size of domestic stock and bond markets in the country;

Column (2) is measured as stock market size only;

Column (3) and (4) is measured as bond market size only.

Table 6:

Financial Gravity Model - Portfolio Investment; ^1/ Different Types of Portfolio Assets

	(1)	(2)	(3)	(4)
Dependent Variable:	Total	Equity	ST	LT
	Portfolio		Debt	Debt
Log (Market Size) - Source	1.380***	1.251***	0.859***	1.526***
	(38.782)	(33.782)	(16.898)	(28.801)
Log (Market Size) - Destination	1.611***	1.687***	1.456***	1.513***
	(68.783)	(63.769)	(22.881)	(23.500)
Log (Distance)	−2.464***	−2.096***	−1.863***	−2.738***
	(−27.520)	(−23.104)	(−14.614)	(−23.192)
Common Language	1.615***	1.665***	3.250***	1.843***
	(6.370)	(6.365)	(7.422)	(4.381)
ASEAN-Intra Dummy	2.217***	1.944*	3.161**	1.024
	(3.811)	(1.912)	(2.249)	(0.972)
Asia-Intra (ex. ASEAN) Dummy	−0.629	0.539	−1.137	−2.005
	(−0.943)	(0.850)	(−0.803)	(−1.583)
Euro area-Intra Dummy	4.355***	5.126***	5.256***	4.332***
	(16.578)	(16.672)	(10.369)	(14.250)
European Union-Intra (excl. Euro area) Dummy	2.932***	2.213***	−0.250	0.875
	(6.673)	(4.879)	(−0.388)	(1.360)
Latin America-Intra Dummy	0.808	−1.851***	−1.927**	2.917***
	(1.213)	(−2.898)	(−2.561)	(2.767)
NAFTA-Intra Dummy	−0.381	1.727	2.111	−1.022
	(−0.263)	(0.885)	(0.928)	(−0.654)
Constant	17.360***	13.383***	8.781***	19.376***
	(22.322)	(17.264)	(7.791)	(18.250)
Observations	55,059	57,650	26,454	26,454
R-squared	0.472	0.428	0.296	0.370
Time Dummies	YES	YES	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/Market size for:

Column (1) is measured as total size of domestic stock and bond markets in the country;

Column (2) is measured as stock market size only;

Column (3) and (4) is measured as bond market size only.

View Table

Table 6:

Financial Gravity Model - Portfolio Investment; ^1/ Different Types of Portfolio Assets

	(1)	(2)	(3)	(4)
Dependent Variable:	Total	Equity	ST	LT
	Portfolio		Debt	Debt
Log (Market Size) - Source	1.380***	1.251***	0.859***	1.526***
	(38.782)	(33.782)	(16.898)	(28.801)
Log (Market Size) - Destination	1.611***	1.687***	1.456***	1.513***
	(68.783)	(63.769)	(22.881)	(23.500)
Log (Distance)	−2.464***	−2.096***	−1.863***	−2.738***
	(−27.520)	(−23.104)	(−14.614)	(−23.192)
Common Language	1.615***	1.665***	3.250***	1.843***
	(6.370)	(6.365)	(7.422)	(4.381)
ASEAN-Intra Dummy	2.217***	1.944*	3.161**	1.024
	(3.811)	(1.912)	(2.249)	(0.972)
Asia-Intra (ex. ASEAN) Dummy	−0.629	0.539	−1.137	−2.005
	(−0.943)	(0.850)	(−0.803)	(−1.583)
Euro area-Intra Dummy	4.355***	5.126***	5.256***	4.332***
	(16.578)	(16.672)	(10.369)	(14.250)
European Union-Intra (excl. Euro area) Dummy	2.932***	2.213***	−0.250	0.875
	(6.673)	(4.879)	(−0.388)	(1.360)
Latin America-Intra Dummy	0.808	−1.851***	−1.927**	2.917***
	(1.213)	(−2.898)	(−2.561)	(2.767)
NAFTA-Intra Dummy	−0.381	1.727	2.111	−1.022
	(−0.263)	(0.885)	(0.928)	(−0.654)
Constant	17.360***	13.383***	8.781***	19.376***
	(22.322)	(17.264)	(7.791)	(18.250)
Observations	55,059	57,650	26,454	26,454
R-squared	0.472	0.428	0.296	0.370
Time Dummies	YES	YES	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/Market size for:

Column (1) is measured as total size of domestic stock and bond markets in the country;

Column (2) is measured as stock market size only;

Column (3) and (4) is measured as bond market size only.

The coefficient on the bilateral trade is always positive and significant, suggesting that trade integration buttresses financial integration, possibly because trade in goods and services can help alleviate informational asymmetries and, hence, transaction costs, as argued by Aviat and Coeurdacier (2007).¹⁹

Departing from the literature, measures of foreign bank presence (number and asset shares in the domestic banking system) are included as additional regressors (Table 7a, columns (1) and (2)). The positive and significant coefficients on these variable suggest that foreign bank participation in the domestic banking system of the destination country supports international financial integration, as foreign banks could be the bridge between foreign funds and domestic investment projects, or because they are likely to be equipped with expertise and technology that help facilitate cross-border financial investments.

Table 7a:

Financial Gravity Model - Portfolio Investment; ^1/ Regulatory and Institutional Quality (1)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

Table 7a:

Financial Gravity Model - Portfolio Investment; ^1/ Regulatory and Institutional Quality (1)

Dependent Variable:	(1)	(2)	(3)	(4)	(5)
Log (Portfolio Assets)
Log (Market Size) - Source	0.172***	0.235***	0.422***	0.184***	0.269***
	(3.205)	(4.954)	(5.452)	(3.346)	(5.745)
Log (Market Size) - Destination	1.036***	1.049***	0.784***	1.011***	0.941***
	(22.852)	(27.107)	(10.928)	(21.823)	(24.592)
Log (Distance)	−1.335***	−1.374***	−0.839***	−1.511***	−1.282***
	(−13.594)	(−15.693)	(−7.070)	(−15.252)	(−14.720)
Common Language	1.581***	1.341***	1.111***	1.293***	1.421***
	(6.394)	(6.001)	(3.946)	(5.140)	(6.382)
GDP per Capita - S	0.106***	0.103***	0.103***	0.087***	0.091***
	(19.962)	(20.270)	(17.187)	(15.535)	(16.933)
GDP per Capita - D	0.062***	0.066***	0.052***	0.053***	0.038***
	(10.921)	(12.151)	(8.074)	(8.977)	(6.831)
Bilateral Trade	0.798***	0.762***	1.256***	0.751***	0.691***
	(18.048)	(19.285)	(16.028)	(16.658)	(17.847)
Capital Account Openness - S	0.972***	1.006***	0.996***	1.006***	0.675***
	(13.294)	(15.431)	(11.264)	(13.715)	(9.186)
Capital Account Openness - D	−0.006	0.014	0.069	0.129**	−0.140**
	(−0.086)	(0.233)	(0.847)	(2.000)	(−2.172)
Foreign Bank Presence (Asset Share) - D	0.017*** (6.298)
Foreign Bank Presence (Number Share) - D		0.031***
		(10.379)
Financial and Banking Regulation Index - S			0.295***
			(4.669)
Financial and Banking Regulation Index - D			−0.078
			(−1.238)
Financial and Banking Regulation Index - Difference			−0.146**
			(−2.230)
Regulation of Securities Exchanges - S				0.729***
				(6.277)
Regulation of Securities Exchanges - D				0.003
				(0.026)
Regulation of Securities Exchanges - Difference				−0.675***
				(−6.333)
Regulatory Quality - S					1.059***
					(7.587)
Regulatory Quality - D					0.983***
					(7.026)
Regulatory Quality - Difference					−0.687***
					(−6.519)
Constant	6.444***	6.885***	0.808	6.369***	8.197***
	(7.743)	(9.378)	(0.714)	(6.550)	(11.341)
Observations	35,254	51,999	25,231	28,658	52,491
R-squared	0.551	0.563	0.575	0.557	0.566
Time Dummies	YES	YES	YES	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

View Table

Table 7a:

Financial Gravity Model - Portfolio Investment; ^1/ Regulatory and Institutional Quality (1)

Dependent Variable:	(1)	(2)	(3)	(4)	(5)
Log (Portfolio Assets)
Log (Market Size) - Source	0.172***	0.235***	0.422***	0.184***	0.269***
	(3.205)	(4.954)	(5.452)	(3.346)	(5.745)
Log (Market Size) - Destination	1.036***	1.049***	0.784***	1.011***	0.941***
	(22.852)	(27.107)	(10.928)	(21.823)	(24.592)
Log (Distance)	−1.335***	−1.374***	−0.839***	−1.511***	−1.282***
	(−13.594)	(−15.693)	(−7.070)	(−15.252)	(−14.720)
Common Language	1.581***	1.341***	1.111***	1.293***	1.421***
	(6.394)	(6.001)	(3.946)	(5.140)	(6.382)
GDP per Capita - S	0.106***	0.103***	0.103***	0.087***	0.091***
	(19.962)	(20.270)	(17.187)	(15.535)	(16.933)
GDP per Capita - D	0.062***	0.066***	0.052***	0.053***	0.038***
	(10.921)	(12.151)	(8.074)	(8.977)	(6.831)
Bilateral Trade	0.798***	0.762***	1.256***	0.751***	0.691***
	(18.048)	(19.285)	(16.028)	(16.658)	(17.847)
Capital Account Openness - S	0.972***	1.006***	0.996***	1.006***	0.675***
	(13.294)	(15.431)	(11.264)	(13.715)	(9.186)
Capital Account Openness - D	−0.006	0.014	0.069	0.129**	−0.140**
	(−0.086)	(0.233)	(0.847)	(2.000)	(−2.172)
Foreign Bank Presence (Asset Share) - D	0.017*** (6.298)
Foreign Bank Presence (Number Share) - D		0.031***
		(10.379)
Financial and Banking Regulation Index - S			0.295***
			(4.669)
Financial and Banking Regulation Index - D			−0.078
			(−1.238)
Financial and Banking Regulation Index - Difference			−0.146**
			(−2.230)
Regulation of Securities Exchanges - S				0.729***
				(6.277)
Regulation of Securities Exchanges - D				0.003
				(0.026)
Regulation of Securities Exchanges - Difference				−0.675***
				(−6.333)
Regulatory Quality - S					1.059***
					(7.587)
Regulatory Quality - D					0.983***
					(7.026)
Regulatory Quality - Difference					−0.687***
					(−6.519)
Constant	6.444***	6.885***	0.808	6.369***	8.197***
	(7.743)	(9.378)	(0.714)	(6.550)	(11.341)
Observations	35,254	51,999	25,231	28,658	52,491
R-squared	0.551	0.563	0.575	0.557	0.566
Time Dummies	YES	YES	YES	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

Our key departure from existing financial gravity literature is the investigation of the role of regulation and institutions, particularly differences in financial sector regulations between two countries, as implicit barriers to cross-border financial transactions. Hence, several measures of regulatory and institutional quality from various sources are used as additional explanatory variables (Appendix III). The coefficients of the regulation variables of the source and destination country are found to be positive and highly significant in most regressions (Table 7a, columns (3)-(5); Table 7b). Furthermore, differences between country pairs’ regulatory quality always have negative and significant coefficients. The estimates indicate that the more similar is the quality of financial and banking regulation, security exchange regulation, investor protection, and contract enforcement between two countries, the larger are their bilateral financial transactions. This is probably because similarities in regulatory frameworks lower information asymmetry and boost investor confidence.

Table 7b:

Financial Gravity Model - Portfolio Investment; ^1/ Regulatory and Institutional Quality (2)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

Table 7b:

Financial Gravity Model - Portfolio Investment; ^1/ Regulatory and Institutional Quality (2)

Dependent Variable:	(1)	(2)
Log (Portfolio Assets)
Log (Market Size) - Source	0.262***	0.277***
	(5.509)	(5.528)
Log (Market Size) - Destination	0.981***	1.019***
	(25.270)	(25.471)
Log (Distance)	−1.357***	−1.406***
	(−15.462)	(−15.169)
Common Language	1.243***	1.425***
	(5.582)	(6.160)
GDP per Capita - S	0.081***	0.099***
	(13.486)	(18.304)
GDP per Capita - D	0.046***	0.040***
	(7.094)	(7.153)
Bilateral Trade	0.699***	0.686***
	(17.832)	(17.189)
Capital Account Openness - S	0.658***	0.970***
	(8.979)	(14.312)
Capital Account Openness - D	0.076	0.148**
	(1.276)	(2.536)
Rule of Law - S	1.119***
	(7.912)
Rule of Law - D	0.376***
	(2.785)
Rule of Law - Difference	−0.389***
	(−4.058)
Protecting Investors Index - S		0.165***
		(3.513)
Protecting Investors Index - D		0.148***
		(3.144)
Protecting Investors Index - Difference		−0.135***
		(−2.579)
Enforcing Contracts Index - S		0.114
		(1.520)
Enforcing Contracts Index - D		0.500***
		(6.239)
Enforcing Contracts Index - Difference		−0.217***
		(−2.613)
Constant	6.981***	2.628**
	(9.495)	(2.381)
Observations	52,491	44,436
R-squared	0.561	0.564
Time Dummies	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

View Table

Table 7b:

Financial Gravity Model - Portfolio Investment; ^1/ Regulatory and Institutional Quality (2)

Dependent Variable:	(1)	(2)
Log (Portfolio Assets)
Log (Market Size) - Source	0.262***	0.277***
	(5.509)	(5.528)
Log (Market Size) - Destination	0.981***	1.019***
	(25.270)	(25.471)
Log (Distance)	−1.357***	−1.406***
	(−15.462)	(−15.169)
Common Language	1.243***	1.425***
	(5.582)	(6.160)
GDP per Capita - S	0.081***	0.099***
	(13.486)	(18.304)
GDP per Capita - D	0.046***	0.040***
	(7.094)	(7.153)
Bilateral Trade	0.699***	0.686***
	(17.832)	(17.189)
Capital Account Openness - S	0.658***	0.970***
	(8.979)	(14.312)
Capital Account Openness - D	0.076	0.148**
	(1.276)	(2.536)
Rule of Law - S	1.119***
	(7.912)
Rule of Law - D	0.376***
	(2.785)
Rule of Law - Difference	−0.389***
	(−4.058)
Protecting Investors Index - S		0.165***
		(3.513)
Protecting Investors Index - D		0.148***
		(3.144)
Protecting Investors Index - Difference		−0.135***
		(−2.579)
Enforcing Contracts Index - S		0.114
		(1.520)
Enforcing Contracts Index - D		0.500***
		(6.239)
Enforcing Contracts Index - Difference		−0.217***
		(−2.613)
Constant	6.981***	2.628**
	(9.495)	(2.381)
Observations	52,491	44,436
R-squared	0.561	0.564
Time Dummies	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

Additional regressions, where these regulatory differences are also interacted with intraregional dummies, suggest that lack of regulatory harmonization has a particularly large negative effects on Asian intraregional investment, suggesting that Asian investments may be more sensitive to these regulatory differences than the sample average.²⁰

Additional drivers of bilateral flows

Diversification does not seem to be a motive for bilateral portfolio investment. In fact, the coefficient on the variable measuring the (lagged) covariance between quarterly GDP growth of the country pair is found to be positive, indicating that countries are more likely to invest in economies with a synchronized business cycle (Table 8). This may be due to informational frictions discouraging transactions between countries located in different geographic regions, whose business cycle is typically less synchronized (Portes and Rey, 2005).

Table 8:

Financial Gravity Model - Portfolio Investment; ^{1/ 2/} Drivers of Cross-Border Investments (1)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/L. indicates that the variable is lagged.

Table 8:

Financial Gravity Model - Portfolio Investment; ^{1/ 2/} Drivers of Cross-Border Investments (1)

Dependent Variable:	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Log (Portfolio Assets)							FE
Log (Market Size) - Source	0.667***	0.780***	1.126***	0.851***	0.923***	1.375***	0.863***
	(16.449)	(19.798)	(21.643)	(15.395)	(18.744)	(19.033)	(7.911)
Log (Market Size) - Destination	1.435***	1.438***	1.449***	1.572***	1.539***	1.489***	0.854***
	(48.301)	(48.606)	(30.719)	(43.321)	(36.390)	(21.215)	(8.543)
Log (Distance)	−2.099***	−2.020***	−1.970***	−1.637***	−1.837***	−1.380***
	(−23.777)	(−23.134)	(−20.151)	(−15.025)	(−16.791)	(−11.861)
Common Language	1.482***	1.525***	2.315***	1.723***	2.082***	1.878***
	(6.213)	(6.496)	(8.012)	(6.344)	(7.467)	(6.311)
GDP per Capita - S	0.145***	0.099***	0.103***	0.090***	0.106***	0.091***	−0.029***
	(30.478)	(18.602)	(17.397)	(14.730)	(17.285)	(15.856)	(−3.507)
GDP per Capita - D	0.057***	0.051***	0.033***	0.049***	0.042***	0.027***	−0.005
	(11.513)	(9.130)	(5.273)	(7.322)	(6.739)	(3.869)	(−0.632)
Bilateral Trade	0.782***	0.779***	0.357***	0.599***	0.467***	0.355***	0.408**
	(4.562)	(4.650)	(3.252)	(4.550)	(2.981)	(2.935)	(2.015)
Capital Account Openess - S		0.950***	1.268***	1.147***	1.105***	1.335***	0.616***
		(14.323)	(13.370)	(12.439)	(12.908)	(11.728)	(6.852)
Capital Account Openess - D		0.120**	0.092	0.097	0.106	0.215**	0.283***
		(2.118)	(1.113)	(1.234)	(1.407)	(2.105)	(2.655)
L.(Covariance of Real GDP Growth)			0.039*** (4.670)			−0.004 (−0.312)	0.004 (0.589)
L.(Interest rate differential)				0.069*** (4.776)		0.012 (0.605)	−0.004 (−0.389)
L.(Stock returns in Destination)					0.001 (0.341)	−0.002 (−0.807)	−0.000 (−0.084)
L.(FX appreciation of D against S)					−0.007***	−0.009***	−0.000
					(−3.288)	(−3.136)	(−0.070)
Constant	15.484***	12.964***	11.676***	9.153***	8.579***	3.855***	4.807***
	(20.973)	(17.437)	(13.830)	(9.842)	(8.869)	(3.585)	(4.458)
Observations	53,532	53,115	26,742	32,397	32,304	17,728	17,728
R-squared	0.533	0.545	0.552	0.548	0.553	0.544	0.059
Time Dummies	YES	YES	YES	YES	YES	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/L. indicates that the variable is lagged.

View Table

Table 8:

Financial Gravity Model - Portfolio Investment; ^{1/ 2/} Drivers of Cross-Border Investments (1)

Dependent Variable:	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Log (Portfolio Assets)							FE
Log (Market Size) - Source	0.667***	0.780***	1.126***	0.851***	0.923***	1.375***	0.863***
	(16.449)	(19.798)	(21.643)	(15.395)	(18.744)	(19.033)	(7.911)
Log (Market Size) - Destination	1.435***	1.438***	1.449***	1.572***	1.539***	1.489***	0.854***
	(48.301)	(48.606)	(30.719)	(43.321)	(36.390)	(21.215)	(8.543)
Log (Distance)	−2.099***	−2.020***	−1.970***	−1.637***	−1.837***	−1.380***
	(−23.777)	(−23.134)	(−20.151)	(−15.025)	(−16.791)	(−11.861)
Common Language	1.482***	1.525***	2.315***	1.723***	2.082***	1.878***
	(6.213)	(6.496)	(8.012)	(6.344)	(7.467)	(6.311)
GDP per Capita - S	0.145***	0.099***	0.103***	0.090***	0.106***	0.091***	−0.029***
	(30.478)	(18.602)	(17.397)	(14.730)	(17.285)	(15.856)	(−3.507)
GDP per Capita - D	0.057***	0.051***	0.033***	0.049***	0.042***	0.027***	−0.005
	(11.513)	(9.130)	(5.273)	(7.322)	(6.739)	(3.869)	(−0.632)
Bilateral Trade	0.782***	0.779***	0.357***	0.599***	0.467***	0.355***	0.408**
	(4.562)	(4.650)	(3.252)	(4.550)	(2.981)	(2.935)	(2.015)
Capital Account Openess - S		0.950***	1.268***	1.147***	1.105***	1.335***	0.616***
		(14.323)	(13.370)	(12.439)	(12.908)	(11.728)	(6.852)
Capital Account Openess - D		0.120**	0.092	0.097	0.106	0.215**	0.283***
		(2.118)	(1.113)	(1.234)	(1.407)	(2.105)	(2.655)
L.(Covariance of Real GDP Growth)			0.039*** (4.670)			−0.004 (−0.312)	0.004 (0.589)
L.(Interest rate differential)				0.069*** (4.776)		0.012 (0.605)	−0.004 (−0.389)
L.(Stock returns in Destination)					0.001 (0.341)	−0.002 (−0.807)	−0.000 (−0.084)
L.(FX appreciation of D against S)					−0.007***	−0.009***	−0.000
					(−3.288)	(−3.136)	(−0.070)
Constant	15.484***	12.964***	11.676***	9.153***	8.579***	3.855***	4.807***
	(20.973)	(17.437)	(13.830)	(9.842)	(8.869)	(3.585)	(4.458)
Observations	53,532	53,115	26,742	32,397	32,304	17,728	17,728
R-squared	0.533	0.545	0.552	0.548	0.553	0.544	0.059
Time Dummies	YES	YES	YES	YES	YES	YES	YES

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/L. indicates that the variable is lagged.

Regression results provide some support to the hypothesis that bilateral investment is driven by the search for yield. Indeed, the coefficient on the (lagged) interest rate differential between the destination and the source country is positive and significant (Table 8, column (4)). However, another indicator of return differential (lagged stock market returns in destination country) is found not to be significant. There is also some indication that a stronger currency in the destination country vis-a-vis the source country deters bilateral flows (Table 8, column (5)). Overall, though, these results are generally not very robust to alternative econometric estimates.

High political risk in the destination country discourages bilateral financial investment (Table 9, column (1)), as indicated by the negative and significant coefficient of the corresponding variable. On the other hand, economic and financial risks do not seem to deter inward foreign portfolio investments. This could be because international investors may be able to hedge against some of such risks, e.g., exchange rate risk.

Table 9:

Financial Gravity Model - Portfolio Investment; ^{1/ 2/} Drivers of Cross-Border Investments (2)

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/L. indicates that the variable is lagged.

Table 9:

Financial Gravity Model - Portfolio Investment; ^{1/ 2/} Drivers of Cross-Border Investments (2)

Dependent Variable:	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Log (Portfolio Assets)							FE
Log (Market Size) - Source	0.234***	0.214***	0.282***	0.263***	0.156**	0.091	0.296***
	(4.746)	(3.897)	(5.796)	(5.499)	(2.455)	(1.246)	(2.802)
Log (Market Size) - Destination	1.028***	0.985***	0.942***	1.023***	0.846***	0.943***	0.527***
	(25.112)	(21.517)	(22.742)	(25.547)	(15.271)	(12.885)	(3.211)
Log (Distance)	−1.252***	−1.340***	−1.360***	−1.358***	−0.956***	−0.897***
	(−13.959)	(−13.573)	(−15.122)	(−15.396)	(−9.187)	(−7.502)
Common Language	1.418***	1.724***	1.448***	1.482***	1.344***	1.555***
	(6.372)	(6.641)	(6.387)	(6.514)	(5.157)	(5.066)
GDP per Capita - S	0.106***	0.106***	0.104***	0.104***	0.115***	0.118***	−0.017
	(20.303)	(19.106)	(19.864)	(20.213)	(19.689)	(18.738)	(−1.457)
GDP per Capita - D	0.043***	0.054***	0.051***	0.057***	0.045***	0.049***	0.028**
	(7.314)	(9.410)	(8.176)	(10.381)	(7.711)	(5.227)	(2.116)
Bilateral Trade	0.775***	0.748***	0.706***	0.722***	1.020***	1.044***	0.219***
	(18.562)	(16.363)	(17.723)	(18.328)	(16.411)	(14.460)	(2.941)
Capital Account Openess - S	1.057***	0.978***	0.995***	1.003***	1.178***	1.117***	0.352***
	(15.881)	(13.027)	(15.078)	(15.332)	(14.710)	(11.928)	(3.592)
Capital Account Openess - D	−0.016	0.088	0.113*	0.186***	0.225***	−0.087	−0.074
	(−0.266)	(1.258)	(1.955)	(3.333)	(3.129)	(−0.889)	(−0.747)
L. (Political Risk - D)	−0.059***					−0.052***	−0.017
	(−6.162)					(−3.364)	(−1.116)
L. (Economic Risk - D)	0.008					0.081***	0.014
	(0.459)					(−3.128)	(0.871)
L.(Financial Risk - D)	0.153***					0.111***	−0.010
	(10.485)					(5.125)	(−0.655)
L.(Bank Branches per Population - D)		0.014***				−0.003	−0.004
		(3.111)				(−0.518)	(−0.422)
L.(Private Credit to GDP - D)			0.009***			0.001	0.006
			(4.603)			(0.279)	(1.597)
L.(Stock Market Cap to GDP - D)				−0.003**		−0.003**	0.003
				(−2.402)		(−1.999)	(1.319)
L.(Mutual Fund Assets to GDP - D)					0.004***	0.005***	0.001
					(3.099)	(3.763)	(0.721)
Constant	9.569***	7.370***	8.541***	7.946***	4.823***	6.521***	5.166***
	(8.800)	(7.191)	(11.399)	(10.790)	(5.551)	(3.762)	(3.009)
Observations	49,751	35,795	50,336	52,491	33,328	19,895	19,895
R-squared	0.565	0.554	0.555	0.557	0.560	0.563	0.013
Time Dummies	YES	YES	YES	YES	YES	YES	YES
Number of Pairid							3,987

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/L. indicates that the variable is lagged.

View Table

Table 9:

Financial Gravity Model - Portfolio Investment; ^{1/ 2/} Drivers of Cross-Border Investments (2)

Dependent Variable:	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Log (Portfolio Assets)							FE
Log (Market Size) - Source	0.234***	0.214***	0.282***	0.263***	0.156**	0.091	0.296***
	(4.746)	(3.897)	(5.796)	(5.499)	(2.455)	(1.246)	(2.802)
Log (Market Size) - Destination	1.028***	0.985***	0.942***	1.023***	0.846***	0.943***	0.527***
	(25.112)	(21.517)	(22.742)	(25.547)	(15.271)	(12.885)	(3.211)
Log (Distance)	−1.252***	−1.340***	−1.360***	−1.358***	−0.956***	−0.897***
	(−13.959)	(−13.573)	(−15.122)	(−15.396)	(−9.187)	(−7.502)
Common Language	1.418***	1.724***	1.448***	1.482***	1.344***	1.555***
	(6.372)	(6.641)	(6.387)	(6.514)	(5.157)	(5.066)
GDP per Capita - S	0.106***	0.106***	0.104***	0.104***	0.115***	0.118***	−0.017
	(20.303)	(19.106)	(19.864)	(20.213)	(19.689)	(18.738)	(−1.457)
GDP per Capita - D	0.043***	0.054***	0.051***	0.057***	0.045***	0.049***	0.028**
	(7.314)	(9.410)	(8.176)	(10.381)	(7.711)	(5.227)	(2.116)
Bilateral Trade	0.775***	0.748***	0.706***	0.722***	1.020***	1.044***	0.219***
	(18.562)	(16.363)	(17.723)	(18.328)	(16.411)	(14.460)	(2.941)
Capital Account Openess - S	1.057***	0.978***	0.995***	1.003***	1.178***	1.117***	0.352***
	(15.881)	(13.027)	(15.078)	(15.332)	(14.710)	(11.928)	(3.592)
Capital Account Openess - D	−0.016	0.088	0.113*	0.186***	0.225***	−0.087	−0.074
	(−0.266)	(1.258)	(1.955)	(3.333)	(3.129)	(−0.889)	(−0.747)
L. (Political Risk - D)	−0.059***					−0.052***	−0.017
	(−6.162)					(−3.364)	(−1.116)
L. (Economic Risk - D)	0.008					0.081***	0.014
	(0.459)					(−3.128)	(0.871)
L.(Financial Risk - D)	0.153***					0.111***	−0.010
	(10.485)					(5.125)	(−0.655)
L.(Bank Branches per Population - D)		0.014***				−0.003	−0.004
		(3.111)				(−0.518)	(−0.422)
L.(Private Credit to GDP - D)			0.009***			0.001	0.006
			(4.603)			(0.279)	(1.597)
L.(Stock Market Cap to GDP - D)				−0.003**		−0.003**	0.003
				(−2.402)		(−1.999)	(1.319)
L.(Mutual Fund Assets to GDP - D)					0.004***	0.005***	0.001
					(3.099)	(3.763)	(0.721)
Constant	9.569***	7.370***	8.541***	7.946***	4.823***	6.521***	5.166***
	(8.800)	(7.191)	(11.399)	(10.790)	(5.551)	(3.762)	(3.009)
Observations	49,751	35,795	50,336	52,491	33,328	19,895	19,895
R-squared	0.565	0.554	0.555	0.557	0.560	0.563	0.013
Time Dummies	YES	YES	YES	YES	YES	YES	YES
Number of Pairid							3,987

Robust t-statistics in parentheses; Errors clustered at country-pair level *** p<0.01, ** p<0.05, * p<0.1 Source: IMF staff estimates.

^1/S=Source; D=Destination

^2/L. indicates that the variable is lagged.

Indicators of financial development (lagged), e.g., bank branch concentration, private credit to GDP and mutual fund assets to GDP in the destination country seem to have a significant impact on bilateral portfolio asset holdings (Table 9, columns (2)-(6)).