International Capital Flows: Economic Impact and Policy Implications

Diploma Thesis, 2000

210 Pages, Grade: 1



1. Introduction and Overview

2. The Degree of International Financial Integration
2.1. Saving-Investment Correlations
2.1.1. Non-Fundamental Causes
Model-Specific Sources of Bias
Sampling and Measurement Bias
Evidence of Endogeneity
Evidence of a Sampling Bias
Evidence of the Large-Country Effect
Interim Assessment
2.1.2. Fundamental Causes
Capital Controls
International Risk Diversification
Ultrarational Households
2.1.3. Summary
2.2. Arbitrage Tests
2.2.1. Concepts of Interest Parity
Uncovered Interest Parity
Covered Interest Parity
Real Interest Parity
2.2.2. Measurement
2.2.3. Empirical Evidence
2.2.4. Summary
Additional Comments
2.3. Portfolio Tests
2.3.1. Mean-Variance Analysis
2.3.2. The International Capital Asset Pricing Model
2.3.3. Alternative Measures Based on Portfolio Theory
2.3.4. Empirical Evidence
2.3.5. The Home Bias Revisited
Barriers to International Investment
Econometric Sources of Bias
Data Constraints
Market Failure
2.3.6. Summary
Additional Comments
2.4. Degree of Monetary Autonomy
2.4.1. Basic Outline
2.4.2. Measurement
2.4.3. Empirical Evidence
Additional Comments
2.5. Summary and Conclusions

3. The Economic Impact of Capital Mobility
3.1. Benefits from Capital Market Integration
3.1.1. The Traditional View
3.1.2. Endogenous Growth
3.1.3. International Risk Sharing
3.1.4. Summary
3.2. Risks of Capital Market Integration
3.2.1. Market Failure
Public Commercial Interventionism
Capital Account Liberalisation
Wage Rigidities and Unemployment
Inequality of Private and Social Returns
Policy Implications
3.2.2. Monetary and Fiscal Policy
The Basic Mundell-Fleming Model
Monetary and Fiscal Policy in a Fixed Exchange Rate Regime
Monetary and Fiscal Policy with Flexible Exchange Rates
Interim Assessment
The Mundell-Fleming Model with Sticky Prices
Policy Implications
3.2.3. Balance of Payments Crises
Theories of Speculative Attacks
First-Generation Models
Second Generation Models
Third Generation Models
Policy Implications
3.3. Summary

4. Capital Controls: Theory and Evidence
4.1. Capital Controls in Modern Economic History
4.2. Types and Motivations of Capital Controls
4.3. The Effects of Capital Controls in Theory
4.3.1. The Traditional View
4.3.2. Changing the Volume and Composition of Capital Flows
4.3.3. Reducing Exchange Rate Volatility
Speculation versus Hedging Activities
4.3.4. Fending off a Speculative Attack
4.4. Measuring the Effectiveness of Capital Controls
4.5. Empirical Evidence
4.5.1. Inflow versus Outflow Controls: Some Stylised Facts
4.5.2. Controls on Capital Outflows
Restoring Policy Autonomy
Fending off Speculative Attacks
Recovering from a Speculative Attack
Interim Assessment
4.5.3. Controls on Capital Inflows
The Chilean Experience
Interest Rate Equivalence of the Chilean URR
The Effectiveness of Capital Controls in Chile
Methodological Issues
The Volume and Composition of Inflows
Domestic Interest Rates
Real Exchange Rates
Financial Stability
Selective Evidence from Other Countries
4.6. Conclusions

5. Summary and Policy Implications

6. Appendix
6.1. Purchasing Power Parity
6.1.1. The Law of One Price
6.1.2. Absolute Purchasing Power Parity
6.1.3. Relative Purchasing Power Parity
6.2. Concepts of the Real Exchange Rate
6.3. Exchange Rate Expectations
6.3.1. The Survey-Based Approach
6.3.2. The Rational Expectations Approach
6.4. Endogeneity of Regressors and Cointegration


1. Introduction and Overview

In the wake of the latest financial crisis in East Asia (1997-98) and other emerging markets, notably Mexico (1994), Russia (1998), and Brazil (1999), a debate has resurfaced that is reminiscent of the discussions in Bretton Woods more than 50 years ago. Again international economic policymakers are exchanging views about the future of international capital markets and how much of a political effort should be made towards reforming the international financial landscape.

To this end, a vast number of proposals have been published in recent years, many of which are contradictory and mutually exclusive. Some recommend that exchange rates should become more flexible to buffer international asymmetric shocks. Others prefer the re-establishment of a multilaterally binding fixed exchange rate mechanism, similar to that prevailing during the Bretton Woods era, in order to reduce the volatility of crucial macroeconomic variables. While some emphasise the need for more vigorous intervention in financial markets on behalf of the international community, (perhaps by giving greater legal authority to international institutions such as the IMF or the World Bank), opponents to this view argue that market forces should be allowed to take their “natural” course. Finally, while some insist that policymakers should increase their efforts towards greater liberalisation of their national capital accounts, contestants of this suggestion advocate the imposition of capital controls as a means of “throwing sands in the wheels of international finance”, which they believe will reduce the likelihood of systemic financial destabilisation. This paper focuses on the latter compound in this ongoing debate.

Arguably, observers offer such radically different recommendations because they have different preferential views on conflicting policy objectives and different opinions about the operating mechanisms of the international financial system.

With respect to the former, obviously, during the Bretton Woods period, economic stability was given priority over the efficiency of investment allocation,[1] while with regards to the latter, policymaker were deeply doubtful about the intrinsic self-stabilising powers of unrestricted financial markets. Accordingly, the break-down of the Bretton Woods arrangement reflected a general revaluation of this stability-efficiency trade-off. The initial post-Bretton Woods economic climate could therefore be described as one of wide-spread international market optimism leading to a broad dismantling round of barriers to international trade in financial assets.

Only recently have concerns re-emerged about the advantages of a fully deregulated international financial system. This re-visitation has occurred as policymakers and academics alike have been mesmerised by the volatility of international capital flows, the sheer size of daily transactions,[2] as well as their potentially destabilising impact on small open economies, as the above mentioned latest series of crises in emerging markets has dramatically demonstrated.

The objective of this paper is to have a closer look at some of the beliefs and assumptions in this renewed debate concerning the operation of the international financial system that have influenced – along with personal preferences – observers’ recommendations on how to reform the international financial architecture.

One of these common beliefs is that capital mobility has increased substantially during the last few decades and that capital markets are more integrated today compared with any previous period in history. The second chapter in this paper analyses this conjecture in some more detail.

A second belief is that liberalised financial markets have compelling benefits. They are deemed to encourage savings mobilisation and efficient investment allocation, while allowing more effective ways of portfolio diversification. This assertion, as well as alternative hypotheses are discussed in chapter 3.

Finally, the fourth chapter examines the orthodox conviction that restrictions on capital flows are never an optimal policy. In conclusion chapter 5 summarises the main findings in chapters 2-4 and draws some general conclusions concerning the conduct of economic policy.

2. The Degree of International Financial Integration

Several dramatic incidents in recent years, where spill-over effects of originally localised financial tremors were felt even in very remote countries, have consolidated the conventional wisdom that financial markets are more integrated today compared to any previous period in history. The empirical evidence, however, is not as clear-cut.

The following chapter overviews ways that have been traditionally advanced to examine the hypothesis that international capital markets behave as one. While discussing some related empirical evidence it is further attempted to explain why researchers have drawn very different conclusions about the actual degree of international financial integration, despite looking at the same data sets.[3]

Prior to this, some definitional remarks are warranted to avoid the semantic confusion often caused by the difficulty of conceptually separating financial integration from notions of financial market openness, international financial markets efficiency, and international capital mobility. In this paper, complete international financial integration is defined by two conditions: (1) capital markets allocate efficiently; (2) capital is perfectly mobile, which is viewed to depend on the presence of explicit and implicit barriers to international capital flows (that is the degree of financial market openness).

2.1. Saving-Investment Correlations

If capital is immobile, investors cannot allocate funds outside their local economy and firms cannot borrow from non-residents. Consequently, domestic investment must be equal to domestic saving. Conversely, in a world of perfectly mobile capital, domestic saving would seek out the highest returns in the world capital market, irrespective of local investment demand. By the same implication, the world capital market should serve as a source of financing for domestic investment needs. Thus, if capital markets are integrated, the investment ratio should be independent of the savings ratio . This was also the basic idea of Feldstein and Horioka (1980), who argued that the correlation between investment and saving, in a cross-section of countries, might provide a test of international capital mobility. Commensurate with this conjecture they came up with the following econometric model:

Abbildung in dieser Leseprobe nicht enthalten

where b denotes the relevant regression coefficient and e t a random error term.

Under the null hypothesis of perfect financial integration among OECD countries, Feldstein and Horioka computed a theoretical (mean) value of 0.1 for a regression coefficient between savings and investment ratios to be consistent with the conjecture that domestic investment of OECD countries was not constrained by domestic saving during their sample period from 1960 to 1974. This calculation took into account that the coefficient should vary across countries in accordance with each country’s capital stock.

Against their own expectation, actual estimates of b were nowhere close to the hypothesised value and statistically indistinguishable from unity (see Table 1). Consequently, the authors concluded that most of any incremental saving remains in its country of origin, which contradicts the conventional wisdom of a high degree of capital mobility among OECD countries.[4]

Table 1. The Feldstein-Horioka Measure for OECD countries 1960-74

Abbildung in dieser Leseprobe nicht enthalten

Source: Feldstein and Horioka (1980), p. 321.

However, the empirical assertion of high savings-investment correlations may not be sufficient to prove a low degree of financial integration. This is because the possibility exists that other factors like, for instance, methodological and econometric deficiencies cause a spurious relationship between domestic saving and investment, without the implication being a rejection of perfect international capital mobility. The following passages discuss this conjecture and present some evidence to determine whether or not non-fundamental factors can resolve the Feldstein-Horioka-paradox.

2.1.1. Non-Fundamental Causes

Model-Specific Sources of Bias

The assumption that investors would shift funds in accordance with a country’s marginal product of capital, implies a very crude proxy of actual investors’ behaviour. It is likely that this assumption exaggerated the gap between the theoretical and the empirical b-value, as it does not take into account transaction costs, nor the fact that investors tend to adjust raw returns to capital by a discount premium that compensates them for the absorption of risk. There are several reasons why cross-border investment may impose a greater risk and higher transaction costs than a comparable domestic investment. With respect to the risk premium, the most obvious is, of course, the absence of exchange risk for domestic projects.[5] Political uncertainties and information asymmetries concerning foreign markets may also feed into higher risk premia on foreign assets, while additional transaction costs may be caused by unfamiliar contractual procedures abroad, for example. If risk premia and transaction costs for foreign investments are very large, investment-saving correlations would differ from zero for reasons other than a low degree of capital mobility.

By considering net capital flows only, the Feldstein-Horioka measure (FH-measure) excludes the possibility that a small volume of net capital flows may coincide with a large volume of gross in- and outflows which just happen to be of similar magnitude. Again the reason for the failure of the model is linked to the narrow concept of what constitutes an incentive for foreign investment. In assuming that there is only one representative national rate of return and that this return is the only relevant parameter for investment, the model implicitly postulates that financial transactions cannot take place in both directions at the same time. In reality, there are, of course, multiple reasons for international capital flows that are consistent with the simultaneity of capital in-and outflows.

Feldstein and Horioka have also been subject to severe criticism concerning their failure to account for the possible endogeneity of national saving and investment ratios, also known as the common-factor problem. According to this conjecture, the finding that relatively high saving ratios are associated with relatively high investment ratios could simply mean that factors that generate high saving ratios in a given country also generate high investment ratios.

Further, in their analysis the authors implicitly assumed that the world savings rate was exogenous, however, this may not hold for a very large country. The reason is that if a country is large enough – and arguably most OECD countries are - a fall in national saving might drive up interest rates and crowd out investment at home as well as elsewhere in the world.

In the event either of the latter two effects are of substantial economic relevance, it would be erroneous to conclude from a co-movement of domestic investment and domestic saving that capital mobility is low.

Sampling and Measurement Bias

Estimates have been shown to be extremely sensitive to the length of the sample period for which averages are computed. For instance, if calculated as decade averages, savings-investment correlations are prone to reflect the separate tendency of net saving (Sn) and net investment (In) to cancel out over time according to

Abbildung in dieser Leseprobe nicht enthalten

which is the same as the current account moving from surplus to deficit and vice versa in order to fulfil an intertemporal budget constraint (see Sinn, 1992). Therefore, in the Feldstein-Horioka study, the estimation period of subsamples may not have been sufficiently small to prove that savings-investment correlations are indeed high for every period.

In addition, a measurement bias follows from using aggregate savings and investment data from national account statistics, which tend to disregard intertemporal imbalances inbetween ends of periods. This was readily admitted by Feldstein and Horioka themselves, who argued that their test merely asserted that over a long enough period net savings and net investments cancel out

Abbildung in dieser Leseprobe nicht enthalten

further, this is inconsistent with the one-market-hypothesis under which savings and investment should be serially uncorrelated.

Finally, a source of downward measurement bias, although arguably most relevant for developing countries, is that official data on capital flows tend to be incomplete. One reason for this is that - per definition - they do not capture transactions in curb and illegal financial markets, both of which may actually absorb a substantial amount of a country’s overall financial activity depending on the state of evolution of the financial sector. Alternatively, due to innovations in financial engineering, data on the true volume and maturity of capital flows may be seriously distorted and thus figures on the total amount of net investment erroneous.

Evidence of Endogeneity

If investment is determined by domestic saving only, a country’s investment rate must depend on the national rate of return but not on other variables that are correlated with domestic saving, as stated by the following econometric regression

Abbildung in dieser Leseprobe nicht enthalten

where a1 denotes a linear coefficient, and et is an error term. If investment is to be uncorrelated with domestic saving, it is crucial that this error term be purely random. In other words, if factors other than the cost of capital that determine investment, happen to be uncorrelated with national saving, then there will be no econometric problem. To demonstrate that such a lack of correlation is an absurdly strong condition, the most popular response is to assert that governments usually react systemically to offset current account imbalances. For instance, policymakers usually seek to attain a low target current account balance through appropriate fiscal or balance of payments policies. If cross-country targets are similar, a high correlation of national savings and investment across countries would follow automatically for reasons that have nothing to do with capital immobility or investors‘ myopia. Other variants of common factor problems arise from the fact that investment and saving are both closely associated with population and productivity growth (see Obstfeld, 1986, Vamvakidis and Wacziarg, 1998).[6],[7]

Since not controlling for common causes may have undermined Feldstein and Horioka’s original results, several authors have modified the basic study outline of Feldstein and Horioka through an instrumental variables approach.[8] The latter implies the replacement of distorting endogenous variables on the right hand side of equation [1] with so-called instrument variables that do not correlate with the error term e t.

Dooley, et al. (1987), for instance, used as an instrumental variable the ratio of military expenditure to GNP to estimate the saving-investment correlation for a sample of 14 industrial and 48 developing countries. Although some of the coefficients for developing countries lost their statistical significance, the coefficients for industrial even increased compared to an ordinary least square regression. His finding, therefore, only emphasises the puzzling discrepancy between regression estimates and the presumption of high financial integration among industrial countries.

In a similar, more recent study, Vamvakidis and Wacziarg (1998) employed instrumental variables that controlled for population growth and business cycle effects. Yet again, they found b -coefficients not to differ significantly from unity (0.896 on average). This reality forced them to reject the hypothesis of financial integration among high-income countries.

Overall, these results do not warrant much optimism about the conjecture that controlling for common causes alone could resolve the domestic saving-investment anomaly.

Evidence of a Sampling Bias

In a study of the US Economy between 1929 and 1987, Frankel (1993) revealed the sensitivity of b -coefficients to the dividing of observations into different sample periods. His results are recorded in Table 2.

Table 2. The Feldstein-Horioka Coefficient in the US 1955-87

Abbildung in dieser Leseprobe nicht enthalten

a Instrumental variables regression of US investment against national saving (as shares of GNP and cyclically adjusted)

b Constant term is automatic zero because cyclically adjusted rates are residuals from a 1955-87 regression against the GNP gap

Source: Frankel (1991), p. 32.

The highlighted figures in Table 2. show that, reducing the sample period from 30 to 15 years causes the b -coefficient to rise to a level almost 46 times greater than the 30 year coefficient. This cautions against an automatic interpretation of high decade- or five-year-period b -estimates as evidence in favour or against the financial integration hypothesis.

Likewise, Krol (1996) argued that the original approach to work with time averaged-data in cross-sectional regressions in order to eliminate business cycle effects would bias the results toward finding evidence for capital im mobility. he proposed to work with annual data in panel regressions and to control for business cycle effects by including a time dummy. Making these adjustments, Krol found in fact lower beta-coefficients in the order of magnitude of 0.2 for a panel of OECD countries, yet business cycle effects appear to be insignificant.

Gundlach and Sinn (1992) argued that using cross-section data clouds different institutional structures between countries. They therefore suggested to exploit the fact that the difference between saving and investment is the mirror-image of the current account balance. They proposed to test for the stationarity of the latter, since a non-stationary current account would imply that saving and investment move like independent random walks, which - following the original Feldstein and Horioka interpretation – should be taken as evidence for capital mobility. The authors found that Germany, Japan, and the US were integrated into the international capital market and that the degree of capital mobility increased during the post-Bretton-Woods-era.

Evidence of the Large-Country Effect

Amongst the Feldstein-Horioka critics, Murphy (1984) focused exclusively on the assumption of world interest rates to be exogenous. To this end he examined a sample of OECD vis-à-vis developing countries. He found that on average the 7 largest economies had a higher coefficient on saving (0.98) than the smaller countries (0.59). He interpreted these results to support the conjecture that the high saving-investment correlations could be attributed to country size rather than imperfect capital mobility.[9]

This view was confirmed by Vamvakidis and Wacziarg (1998) who also studied developing countries separate from the rest of their sample. Their result showed that b -coefficients for developing countries were lower on average than for OECD countries whose coefficients were again not significantly different from unity (Table 3.).

Table 3. The Feldstein-Horioka coefficient for OECD and Developing Countries 1975-93

Abbildung in dieser Leseprobe nicht enthalten

e empirical coefficient

The sample includes 103 countries in total of which 20 are classified as OECD countries and the rest as developing countries. Heteroscedasticity consistent standard errors are in parentheses

Source: Vamvakidis, A. and Wacziarg, R. (1998), p. 12.

On the contrary, Frankel (1985) demonstrated that high saving-investment correlations in the US economy could not be attributed to the large-country effect.

Because of the conflicting evidence, it is worthwhile to take a closer look at this phenomenon. One already mentioned reason why in most studies developing countries show up with substantially lower coefficients than industrial countries - which seems rather counter-intuitive – is that there is a possible measurement bias due to the low quality of data from less developed countries. In addition, as in fact Bhagwati (1978) famously argued, the saving rate in developing countries might be inversely related to net foreign inflows. This is a plausible assumption as governments in poorer countries tend to implement policies of substitution between foreign capital and domestic saving, to increase internal consumption and welfare in the short and medium run. Finally, low saving-investment correlations estimates could be spurious because a large portion of capital movements to developing countries is made up of official transactions, i.e. aid, etc., while private investors may indeed face tight capital controls. All three factors together could cause coefficients to be much lower than the true degree of capital mobility among developing countries and the rest of the world would suggest.[10] Thus, it remains somewhat unclear which portion of high saving-investment correlations may be attributed to a simple size effect.

Interim Assessment

Most empirical studies have demonstrated that it is difficult to ascribe the mysteries of high investment-savings correlations of industrial countries wholly to econometric and methodological problems, or simply insufficient information because of low quality data. It is therefore now warranted to turn to the alternative possibility according to which there are fundamental economic reasons that cause the Feldstein-Horioka-paradox.

2.1.2. Fundamental Causes

According to the definition given at the outset of this section, financial markets are considered to be integrated if two conditions hold: (1) markets show no signs of systematic allocation inefficiencies, and (2) capital is perfectly mobile which requires the absence of indirect and direct barriers to capital. If neither of these conditions apply, there is no reason for saving-investment correlations to show values anywhere near to zero, and the Feldstein-Horioka paradox may in fact yield an adequate diagnosis of low financial integration.

Capital Controls

With regards to the latter requirement, it has been conjectured that savings-investment correlations would proxy for governments’ balance of payments policies such as quantitative restrictions on private capital, capital taxation policies, or other selective regulations that limit trans-border investment (Bayoumi 1990). Since most of the sample period covered in the original Feldstein and Horioka study falls into the Bretton-Woods-era, where such policies were rather common among industrial countries, the latter might indeed be a plausible explanation for why their b-estimates assumed such high values.[11]

Contrary to this conjecture Dooley et al. (1987) rejected the hypothesis that the Feldstein-Horioka coefficient is associated with restrictive capital account policies. In their study b -coefficients show up as remarkably stable over time, including the period before and after the collapse of the Bretton-Woods-system in 1973 (Table 4.).

Table 4. The Feldstein-Horioka Coefficient before and after Bretton Woods

Abbildung in dieser Leseprobe nicht enthalten

Instrumental variables: ratio of military expenditure to GNP, ratio of dependents to working population

b Significance at the 5 percent level

c Significance at the 1 percent level

Standard errors are given in parentheses below coefficients.

Source: Dooley, Frankel, Mathieson (1987), pp. 513-14.

The evidence reported in Table 4. seems particularly odd, given that in general, former Bretton-Woods-members substantially reduced capital control mechanisms once they had switched to flexible exchange rates. Also in the late 1970s and throughout the 1980s, considerable amounts of capital surpluses of the OECD countries were recycled through international lending, while the volume of Eurocurrency transactions increased.

Nonetheless, subsequent econometric research confirmed the basic finding of Feldstein and Horioka (Penati and Dooley, 1984; Frankel, 1985; Obstfeld, 1986). Consequently, unless capital controls had been wholly ineffective throughout the entire Bretton-Woods-period, assertions of international financial integration based on domestic saving-investment correlations alone would seem suspect.

In a study on the US economy alone, Frankel (1991) reported that during the 1980s saving-investment correlations, hitherto near unity, were effectively turned over (see Table 2.). This coincided with a dramatic and unprecedented change in policy from economic interventionism towards aggressive financial liberalisation under the Reagan administration.[12] Feldstein and Bachetta (1989) matched these results by finding a similar drop in the saving-investment coefficient in the 1980s for a cross-section of industrialised countries (though they do not use instrumental variables, and are thus liable to econometric criticism). Nonetheless, these studies provide some weak evidence that capital mobility increased during the 1980s, which confirms the conventional presumption of a higher degree of international financial integration in the post-Bretton-Woods-era.

International Risk Diversification

Several researchers have argued that high coefficients, derived from a cross-section analysis of a relatively homogenous group of countries, are only the natural outcome of what motivates foreign investment. Since more dissimilar countries provide better opportunities in terms of overall portfolio risk reduction, OECD investors should prefer foreign investment into non-OECD countries, who in addition tend to be in short supply of capital and thus should offer higher returns. Thus, only a small portion of overall net capital flows may be accounted for when industrial countries are analysed in isolation from the rest of the world. This is supported by various studies who recorded coefficients to drop substantially once a more heterogeneous sample of countries was analysed (see Vamvakidis and Wacziarg, 1998).

Ultrarational Households

Others have attempted to explain the Feldstein-Horioka paradox as the result of an ultrarational household sector (Miller, 1988).[13] Under the ultrarationality hypothesis, the household sector sees the business sector as an extension of itself, so that business saving decisions are incorporated directly into the household consumption-saving decisions, ex ante. The appropriate level of saving aggregation is therefore the entire private sector rather than only the household sector. The existence of an ultrarational household sector implies that changes in government expenditure are neutralised because debt financed (tax financed) public expenditure is substituted for domestic investment (consumption) in equal proportion. As a result, national saving and domestic investment rates are perfectly correlated independent of international capital mobility. However, the empirical evidence in favour of an ultrarational household sector is not overwhelmingly persuasive. Despite the fact that Demopoulos, et al. (1986) found some evidence in support of this conjecture in several OECD countries with integrated corporate tax systems, overall, it seems unlikely that this - even when accepting that households neutralise government behaviour - is also true for the multitude of other disturbances that could induce international capital flows (e.g. technological breakthroughs, etc.). Moreover, if the correct description of the world is one in which there are, ex ante, no reasons for net capital flows, then the whole issue of capital mobility would be of little interest as it would have no influence on economic welfare.

Since savings-investment correlations neither seem to capture the effects of explicit barriers to capital, nor what motivates capital flows in the first place, the answer to the question “what do savings-investment correlations tell us?” (Dooley et al., 1987, p. 503.) remains unresolved. Clearly more research is needed to fully understand the economic meaning of a lack of saving-investment imbalances among supposedly integrated countries.

2.1.3. Summary

Overall, the results of this section have shown that quite contrary to conventional wisdom, little evidence is found in the non-financial data to confirm the one-market hypothesis for international capital. Further, a reading of the empirical literature suggests that the original Feldstein-Horioka test is not informative, since conventional cross-sectional regressions are likely to produce high saving-investment correlations regardless of whether the degree of international capital mobility is high or low.

In addition to its methodological and econometric problems, the Feldstein-Horioka measure does not produce a benchmark that can indicate low or high integration. Hence, even if the Feldstein-Horioka criterion measured integration properly and the econometrics yielded a proper estimate, one would still be left without a yardstick that says what is “high” and what is “low”. That being said, potentially useful information may be obtained by analysing the changes over time in the correlation between saving and investment. Such an analysis has been done by Taylor (1996), who - by using a modified F-H measure - found a lower correlation among high-income countries. In addition, he uncovered a general decline in the correlation from 1980 onwards. Taylor concluded that in this modified framework, international markets exhibited a recent tendency towards increased integration. Although this conclusion seems plausible, it is subject to similar criticism as the one levelled against the original Feldstein and Horioka result in so far as Taylor fails to control for the importance of common shocks.

Recently, economists have taken a long historical view on non-financial data to analyse the question of whether capital mobility has increased. A surprising finding is that net capital flows tended to be of the same order of magnitude during the period of the international gold standard as compared to the present period. This has been confirmed by Sachs and Warner (1995) and Rodrik (1998), leading to the conclusion that today’s degree of capital mobility is nothing particularly notable compared to the situation a century ago. This reality is clearly counter to the conventional wisdom that exists in the popular press and in the large “globalisation” literature. More in line with conventional wisdom, Eichengreen (1999) has claimed, however, that the present degree of financial integration has increased relative to one hundred years ago, based on US data. Overall, nonetheless, there is a broad consensus that according to national current account statistics the degree of financial integration in the world has increased over the last few decades.

2.2. Arbitrage Tests

Tests of capital mobility on the basis of interest parity conditions make use of the fact that in frictionless financial markets assets must have the same price, irrespective of where they are traded. Generally, there are three concepts of interest parity. Each of these will be described and evaluated in the next passage alongside some empirical evidence.

2.2.1. Concepts of Interest Parity

Uncovered Interest Parity

Uncovered interest parity posits that differences in nominal rates of return on financial assets at home and abroad are equal to expected exchange rate changes in the spot rate, where denotes the expected exchange rate in period t. Thus, UIP can be written as

Abbildung in dieser Leseprobe nicht enthalten

where i (i*) denotes the domestic (foreign) interest rate. The exchange rate e is defined as the domestic currency price of one unit of the foreign currency. Deviations from uncovered interest parity consequently reflect additional exchange risk premia, that are unaccounted for in .

Covered Interest Parity

To include exchange risk, covered interest parity replaces expected spot rates with the forward discount, f, where , and denotes the contractual exchange rate that is applied when a transaction - agreed upon in period t – takes place in T. The forward discount could be interpreted as an implicit insurance against exchange rate variability that one obtains when purchasing foreign exchange today and simultaneously selling it forward at a guaranteed rate.

Abbildung in dieser Leseprobe nicht enthalten

Equation [7] is a risk-free arbitrage condition that holds regardless of the preferences of investors. To the extent however, that investors are risk averse, the forward rate can differ from the expected future spot rate by a premium for the perceived risk of holding domestic versus foreign assets. Hence, deviations from covered interest parity usually reflect country-specific risk premia that compensate for contingent transaction costs – often inflicted upon investors by government interventions - and other country-related sources of capital markets imperfections, such as information asymmetries, etc. (Obstfeld and Rogoff, 1996).

Real Interest Parity

Real interest parity comprises covered and uncovered interest parity plus the additional condition that expectations of changes in the exchange rate are equal to inflation rate differentials according to

Abbildung in dieser Leseprobe nicht enthalten

where r-r* is the real interest differential, p e the expected domestic and p *e the expected foreign inflation rate. In order to disentangle further components of real interest parity, it can be rewritten as

Abbildung in dieser Leseprobe nicht enthalten

where the first term is the familiar covered interest parity condition. According to Frankel (1991), the second term could be interpreted as the real forward discount, or “currency premium” (p.36). A further decomposition of this term yields

Abbildung in dieser Leseprobe nicht enthalten

where the first term on the right-hand side represents the exchange risk premium while the second term reflects expectations of real depreciation. Given purchasing power parity (PPP) this latter term would equal zero, since in this case investors would form rational expectations of zero real exchange rate changes, known as the hypothesis of ex-ante relative PPP. However, since empirically there is little doubt about the failure of PPP on levels, there is no reason to assume this weaker version of PPP. Hence, unless investors believe that exchange rates follow a (driftless) random walk, the logical implication is that expectations of real exchange rate devaluation are different from zero.[14]

A summary of these different interest parity conditions, their inter-relationship, and their implicit assumptions about the behaviour of international investment is displayed in Table 5.

Table 5. Different Concepts of Interest Parity

Abbildung in dieser Leseprobe nicht enthalten

2.2.2. Measurement

Since real interest parity lumps together the effects of country-specific and exchange risk as well as non-zero real exchange rate change expectations, it is unlikely to be informative about the degree of international capital mobility. This is because expectations of real currency devaluations and the consequent failure of ex-ante relative PPP are purely caused by an imperfect integration of international goods markets. Similarly, a positive exchange risk premium, which would cause deviations of real and uncovered interest parity, neither reflects deficiencies in financial integration in a narrow sense, because it merely shows that risk-averse investors demand a compensation for holding currencies that are perceived to be risky or in oversupply. In fact, since less exchange risk is usually associated with fixed exchange regimes, the latter provides an explanation why uncovered and real interest parity conditions usually assign greater capital mobility to periods of fixed exchange rate regimes, despite high interventionism of central banks to influence capital flows:

“interest rates have been subjected to two conflicting forces: the removal of barriers leading to greater capital flows has resulted in the convergence of interest rates, while the shift to a flexible exchange rate system leading to greater exchange rate volatility, has resulted in greater divergence of interest rates.’ (Aburachis, 1993).

Overall, this would suggest that only the weakest condition, i.e. covered interest differentials, may sensibly be linked to capital mobility and financial integration. However, even tests of the covered interest parity condition pose substantial problems since - in a strict sense - they can only be applied to covered interest rates for identical assets denominated in the same currency in different financial centres. Hence, when comparing countries with heterogeneous assets, there is no reason to assume that the covered interest condition holds, since rational investors would incorporate these asset-specific risks into their forward premium (Buch, 1999).[15] Obstfeld (1995) therefore proposes to focus on off- and onshore deposit rates, for the same currency in testing for interest parity and capital mobility according to

Abbildung in dieser Leseprobe nicht enthalten

where iH (i*H) denotes the onshore iF (i*F) the offshore deposit rate respectively.

2.2.3. Empirical Evidence

Dooley and Isaard (1980) documented that the use of reserve requirements and other controls over capital inflows by Germany from 1970-74 generated substantial covered interest differentials between on- and offshore bank deposits over this period. This supported the conjecture that covered interest differentials were a robust measure of financial integration. This finding has also been affirmed by a cross-country comparison of presumably open versus closed economies conducted by Frankel (1991). His results are reported in Table 6. below.

The evidence reveals that individually as well as on average, substantially lower covered interest differentials pertain to open economies than to countries classified as closed developed and closed less developed countries.

However, these results are not repeated once the concept of real interest differentials is applied to the same data set. Although on average, lower mean real interest deviations are still smaller for open developed countries than for closed countries, former pairwise comparisons, e.g. of France and Switzerland or Spain and Germany, are effectively reversed. This finding underpins empirically, the former conceptual rejection of real interest differentials yielding a good proxy for financial integration. Besides, a further decomposition of real interest differentials in Table 6. would show that a large portion of this deviation can be ascribed to non-zero real exchange rate change expectations. This also warrants doubts about the supposed connection between real interest differentials and capital mobility (Frankel, 1991, p. 47).

Table 6. Mean Real Interest Differentials and Covered Interest Differentials 1982-1988

Abbildung in dieser Leseprobe nicht enthalten

a Denmark, Ireland, Italy were also included in the original table

Source: Frankel, J.A. (1991), p. 38 and 40.

Dooley and Chinn (1995a) arrived at a similar conclusion. There the authors compared covered interest differentials across various countries and reported evidence confirming conventional perceptions of presumably integrated versus segregated financial markets. Again, this finding was found to be less robust using the same data set and real interest differentials as a measure (Table 7.-8.).

Table 7. Covered Interest Differentials for a Number of Selected Countries 1982-1994

Abbildung in dieser Leseprobe nicht enthalten

Source: Dooley and Chinn (1995a), p. 20.

Table 8. Real Interest Differentials for a Number of Selected Countries 1982-1994

Abbildung in dieser Leseprobe nicht enthalten

Source: Dooley and Chinn (1995a), p. 21.

Instead of comparing absolute averages of differentials in a given period, other researchers have conducted econometric studies on the correlation between uncovered interest differentials and realised changes in the exchange rate according to a simple OLS regression analysis

Abbildung in dieser Leseprobe nicht enthalten

where the term on the left hand side equals the realised change in the exchange rate from t to t+k, and equals the difference between the k -period yield on the domestic and the foreign asset. On the assumption that the composite error term is orthogonal to the interest differential, the estimated slope parameter should be close to unity to accept the real interest parity hypothesis.[16]

Most studies, using values for k that range up to one year reject the null hypothesis (b =1). The study by Flood (1990), for instance, finds an average estimate for b of -0.88, which is similar in magnitude to the expected value of b under the null hypothesis, but of the opposite sign, thus suggesting an inverse relationship between exchange rate moves and uncovered interest differentials.[17] A more recent empirical analysis by Meredith and Chinn (1998) confirms this inverse relationship between uncovered interest parity and exchange rate changes over short horizons. Their estimates are displayed below (Table 9.).

Table 9. Short-term Horizon Regression Estimates of Uncovered Interest Parity 1980Q1-98Q1

Abbildung in dieser Leseprobe nicht enthalten

Standard errors in parentheses

* Different from null at 10 percent significance level

** Different from null at 5 percent significance level

*** Different from null at 1 percent significance level

Source: Meredith and Chinn (1998), p. 22-24.

Yet, in surprising stark contrast to short-term estimations, the authors find that the slope coefficient is closer to the hypothesised value of unity than zero when testing for long-term uncovered interest parity, as reported in Table 10.

Table 10. Long-term Horizon Regression Estimates of Uncovered Interest Parity 1983Q1-98Q1

Abbildung in dieser Leseprobe nicht enthalten

standard errors in parentheses

* Different from null at 10 percent significance level

** Different from null at 5 percent significance level

*** Different from null at 1 percent significance level

Source: Meredith and Chinn (1998), p. 22-24.

Meredith and Chinn (1998) explain the puzzling divergence between tests based on short-term versus long-term data by risk premium shocks. they contend that these shocks produce a perverse relationship between exchange rates and interest rates in the short run, the temporary effects of which, tend to fade over longer horizons where other dynamics consistent with uncovered interest parity dominate.

2.2.4. Summary

The covered interest differential has been shown to better capture factors related to the political jurisdiction in which the asset is issued, which is also why it may best proxy for international capital mobility as a function of explicit barriers to financial integration.

When applying the covered interest criterion to empirical data, most studies concede that over the past years, covered interest differentials have converged, thus, it may be concluded that capital mobility and financial integration have increased in many regions of the world since the late 1970s and early 1980s.

In contrast, the application of other concepts of interest parity, that do not separate exchange risk from other types of risk, may diagnose that international capital mobility has declined in recent years due to the increase in exchange rate volatility since the breakdown of the Bretton-Woods-system.

Additional Comments

As mentioned before, covered interest parity is commonly considered to yield a superior proxy of international capital mobility than alternative interest criteria because it overlooks the influence of real devaluation expectations. However, this assumption and the consequent interpretation of converging covered interest differentials to signal a trend towards higher financial integration may not be entirely innocent. Consider the extreme case where rational expectations of infinite real exchange rate appreciation effectively stop international capital flows. Should this be classified as perfect financial integration, as the covered interest parity condition would indeed suggest? Conversely, should a situation with country-risk premia so high that residents purely invest in domestic assets be interpreted to proxy for explicit barriers to international capital flows, given that these premia arise from systematic misperceptions of actual risk? Although both scenarios are extreme examples and thus very unlikely, they point to an important issue that so far has been elusive to the discussion: Tests based on interest differentials implicitly assume market efficiency to prevail everywhere else outside the model itself. Such model simplification can seriously distort the outcome of analysis, if it fails to hold in the real world. When testing for CIP, one likely source of such distortion - other than price distortions in the real sector - has been suggested by Frankel and Froot (1990) to arise from irrational aggregate market expectations, which they explained by the existence of heterogeneous traders in these markets. A related argument is that individual investors’ behaviour may not always be model-consistent. Yet, little research on the microstructure of investment allocation is available to confirm this hypothesis.

Another fundamental issue that questions the adequacy of interest differentials – regularly computed on the basis of interbank deposit rates – as a proxy for financial integration, is that these interest rates are obtained in very thin markets and therefore may not be representative of those that are actually relevant for most of a country’s financial activity. Non-financial firms borrow either through the commercial paper market or from the banking system. Since commercial paper rates usually follow interbank rates closely, the distortion introduced by tests based on highly tradable money market rates may in fact be small for countries with a developed commercial paper market. However, in most developing countries and emerging markets, firms typically obtain their financing means from the banking sector. Thus, the accurateness of tests of market integration based on money market rates in these countries strongly depends on the degree to which bank loan rates follow these interbank rates, or in how far bank loans are highly substitutable for commercial papers. Since the latter condition does not always hold, Brown and McNelis (1990) have suggested to replace covered interest differential by a measure of the correlation between changes in offshore and onshore domestic bank lending rates. However, in circumstances where bank loan rates are highly regulated market determined bank rates may not be observable.

In sum, even though interest differentials can provide valuable information about the mobility of international capital they should be interpreted with caution, since a failure to control for country-specific influences in the private financial sector is otherwise likely to cause erroneous conclusions.

2.3. Portfolio Tests

The previous section has assumed that international investment decisions are guided by simple arbitrage conditions only. Yet, according to portfolio theory, relative rates of return are just one subset of the parameters which influence the market opportunities faced by international investors. A more realistic picture of international investment decisions, however, should also include standard deviations of returns as a proxy for asset specific risk, and the correlation of rates of returns between individual assets as a proxy for the portfolio performance of a particular asset.

2.3.1. Mean-Variance Analysis

An uncomplicated way to evaluate international investment decisions is the simple mean-variance analysis originally developed by Markowitz (1959). The idea is to determine a minimal-risk portfolio and compare it with actual patterns of risk minimisation, in order to test for obstacles to international capital flows. The obvious advantage of this approach is that it offers a benchmark of international integration that captures the heterogeneity of assets across countries. The disadvantage is, of course, that a minimal variance of overall portfolio returns as the sole criterion to judge international investment imposes the unrealistically stringent condition that investors are driven by risk incentives only, or else the analysis must be confined to a comparison of assets with equal (mean) returns, if the parameters of investors’ risk aversion are unknown.

Here, for simplicity a two-country-two-asset-case with equal expected returns (x,y) will be considered. The overall return of a diversified portfolio r(x,y) can be written as

Abbildung in dieser Leseprobe nicht enthalten

where x and y denote the respective return on the domestic and the foreign asset, and a the portfolio weight of the domestic asset. It is further assumed that x and y are identically and independently distributed random variables, which according to the central limit theorem approximately renders a normal distribution of both . Thus, the variance of the return on the portfolio can be expressed as

Abbildung in dieser Leseprobe nicht enthalten

where r x,y is equal to the regression coefficient obtained from a regression of x onto y. A minimal overall variance of the portfolio can then be derived by setting the first partial derivative of the variance equal to zero.

Abbildung in dieser Leseprobe nicht enthalten

Obviously, if the covariance of x and y equals unity no additional risk diversification is available from foreign investment compared to purely domestic investment, and thus, a *=1. Conversely, if there was no co-movement of asset returns, a * would equal the variance of the foreign asset weighted by the sum of variances . Thus, if both variances were of the same magnitude, optimal portfolio weights would each amount to 0.5.

2.3.2. The International Capital Asset Pricing Model

Since pairwise mean-variance comparisons yield unequivocal dominance relations of alternative investments, only in the special case where either expected returns of different portfolios are equal, or if simultaneously both first two moments of the expected excess return of a particular portfolio are superior to all other alternatives, the scope of use of this analysis is quite limited. For a more general application, knowledge of the risk-aversion parameters of each investor would be necessary.

The standard framework, therefore, in which international investment decisions are analysed, is the International Capital Asset Pricing Model (ICAPM), which compares returns from national portfolios against some neutral benchmark representing a “world” market portfolio. Traditional benchmark indices include independent estimations of national shares of world market capitalisation to derive a value weighted ‘world’ market portfolio. The ICAPM is further based on the assumption that there are no frictions in international financial markets, and that in an efficiently diversified portfolio, country-specific risk factors are diversified away. As a result, only systemic, i.e. global reaching risk factors represent relevant strategic parameters. Accordingly, the model predicts that the expected return on a financial asset j is given by the risk-free rate on this asset in terms of its home currency, plus the risk premium, pj, which compensates investors for the absorption of systemic risk.

Abbildung in dieser Leseprobe nicht enthalten

where E = expectations operators, rj = rate of return on asset j, =risk free rate of return on asset j, b j = the international systemic risk coefficient of asset j, rm = rate of return on the world market portfolio, = risk free rate of return on the world market portfolio. Thus, the greater the co-movement of the asset j with systemic (world) shocks the higher should be the risk premium.

Since , with correlation coefficient between the asset j and the world market portfolio, [16] can be transformed into

Abbildung in dieser Leseprobe nicht enthalten

From [18], it can be better inferred that a positive (negative) correlation between the expected return of asset j, captured by r j,m,and the world portfolio leads to a greater (smaller) risk premium. Also, the risk premium increases (decreases) with the standard deviation of the asset j (world market portfolio). Hence, while for a positive correlation, investors require a positive extra return to compensate for higher risk, they are willing to accept greater standard deviation of asset j even at lower rates of return, if the correlation of asset j ’s return with the world portfolio is sufficiently negative.

The main prediction of the ICAPM is that investors hold a combination of the risky world market portfolio and the riskless world market portfolio, the shares invested in each of these depend upon the investor’s degree of risk-aversion. The optimal composition of the world market portfolio, however, does not depend on individual preferences A test on the degree of capital mobility on the basis of ICAPM checks whether actual portfolio shares are allocated according to this prediction.

2.3.3. Alternative Measures Based on Portfolio Theory

All portfolio-theory based measures so far have reflected a static notion of financial integration. By comparing the variance and mean of different portfolios, irrespective of their connection to shocks on foreign markets, the Markowitz analysis, for example, implicitly assumes that only domestic factors are relevant risk parameters. In contrast, the ICAPM postulates that only the sensitivity of assets to systemic (global) changes determine their risk premium. To allow for more flexible notions of risk determination some scholars have attempted to construct models that allow for a continuum to exist between these extreme – so-called mild segmentation models (Errunza and Losq, 1992).

Bekaert and Harvey (1995), to give an example, developed a regime-switching model that resembles, in essence, an asset pricing model with time-varying risk determination.[18] It contains two components of equilibrium expected returns, one under complete isolation from world changes and one under complete dependence on world changes

Abbildung in dieser Leseprobe nicht enthalten

where the expected excess return on the domestic portfolio E(Rj) is equal to the price p per unit risk, which depends upon q, i.e. the probability that the overall portfolio is isolated from systemic shocks. Allowing expected returns to be determined by country-specific variability as well as the covariance with the world market, the authors argued, the model would yield a more realistic objective than either of the extreme variants.

Due to methodological pitfalls of mean-variance comparisons and ICAPM based measures, which will be discussed in more detail below, some scholars have advocated to compare prices of components of the conventional ICAPM objective.

However, since the observation of price levels of seemingly similar assets in a cross-country analysis may be biased by a failure of PPP and the existence of country or currency-related risk premia, most scholars have advocated instead to look at correlation figures, which tend to be more independent from the distortion on levels.

Eun and Sangdal (1993), for instance, reported evidence in favour of financial integration after an examination of correlations between changes of average excess returns on pairs of national stock exchanges. However, the interpretation of correlations between market indices arguably confuses causes with effects. In portfolio theory a high covariance of international stock exchanges cannot be interpreted as the result from increased international financial integration but must be considered the cause of the dynamics in international diversification patterns. In fact, in a heterogeneous world, a country could be perfectly integrated into the global financial system, but still have a low or negative correlation with other markets because its industry mix is very different from other countries. In addition, any approach based on correlation estimates is prone to criticism that originates from the influence of common causes. Accordingly, financial integration should not be the automatic conclusion from an observation of a simultaneous response of national financial markets to news that impinge on all national markets at the same time.

Others have interpreted the correlation coefficient between the world portfolio return and the return of a particular asset itself as an indicator of international capital mobility (Borio, Kennedy and Prowse, 1994). Yet, again, this measure is likely to be biased due to the influence of common causes.

Agmon (1992) argued that it is better to test the international integration hypothesis on the basis of individual share data. However, if markets conditions are dissimilar there is no reason to believe that even otherwise homgeneous assets are priced equally across different markets. Because of these confounding effects of heterogeneous assets it has been suggested to compare international prices per unit risk (Gultekin et al., 1989). Although there is an obvious advantage in this approach theoretically, in practice, this advantage is not easily forthcoming since the problem then is how to measure risk objectively.

A somewhat unusual approach has been taken by Claessen et al. (1993), who suggested a test for financial integration based on the hypothesis that predictable risk adjusted returns in excess of the world market return are inconsistent with national financial markets behaving as one. This is because in an efficient global market-place with no restrictions on foreign investment, instantaneous arbitrage must destroy any such predictable pattern of higher national margins. Thus, if international financial integration holds, successive asset price changes must be serially uncorrelated.

2.3.4. Empirical Evidence

Despite the largely accepted benefits of international portfolio diversification, in terms of both risk reduction and improved rates of return, an overwhelming majority of empirical studies suggests that there is a widespread tendency for individual and institutional investors to be active primarily in their home domestic markets. The next passage presents some evidence of this “home bias” in international portfolio allocation, and further reviews traditional causes advanced to explain this puzzling anomaly.

Tesar and Werner (1992) reported returns in excess of the risk free rate available from investment in five selected national markets during 1975-90. Their results are reported in Table 11. They are denominated in US dollars, which means that they are relevant for US investors who do not hedge their portfolios against exchange risk (their home investment options are highlighted).

Table 11. Moments of Monthly Excess Returns for Selected Countries 1975-90

Abbildung in dieser Leseprobe nicht enthalten

Source: Tesar and Werner (1992), p. 24.

The figures reveal that on the basis of mean excess returns and standard deviation alone, a value weighted world market portfolio of equity yielded a superior investment strategy than investment purely in the domestic (US) market portfolio. With respect to bonds, Canadian bonds offered the highest expected excess return and the lowest standard deviation in the sample. These remaining arbitrage incentives are not consistent with the one-market hypothesis, unless it can be wholly explained by exchange risk or substantial expectational errors

The empirical correlations between excess returns across the five markets further compound these puzzling results (Table 12.). Since correlations between equities and bonds are less than unity for all country pairs, theory predicts, ceteris paribus, that substantial benefits were available from international diversification. The data also indicate that excess returns for Canada, the UK and the US are more positively correlated than any of these countries are with Japan and Germany. Based on this observation, one would expect that investors in the US should prefer investing in German and Japanese markets to investing in British and Canadian equity.

Table 12. Correlations of Excess Returns Across Markets 1975-1990

Abbildung in dieser Leseprobe nicht enthalten

Source: Tesar and Werner (1992), p. 24.

In combination, the data reported, generally suggest that US investors let substantial benefits available from international portfolio diversification pass. However, this preliminary diagnosis could be seriously biased if investors considering diversification across different countries are not only concerned with portfolio specific risks but also exchange risk. Tesar and Werner therefore computed Sharpe ratios - defined as the mean excess return divided by the standard deviation – for the national portfolios (denominated in own currencies), the world market portfolio (denominated in each of the five currencies), and the world market portfolio hedged against exchange risk. Their results are reported in Table 13., which should be read as follows: For example, a British investor holding a portfolio of UK equities earned a Sharpe ratio of 0.019 between 1980 and 1990. By diversifying across different national markets, he could have earned an almost seven fold Sharpe ratio of 0.129. If he would have hedged this world market portfolio against exchange risk, the ratio would have further increased to 0.132.

Table 13. Sharpe Ratios on Hedged and Unhedged Portfolios 1980-90

Abbildung in dieser Leseprobe nicht enthalten

Source: Tesas and Werner (1992), p. 25

Hedging for risk does not always seem beneficial. Comparing equity portfolios across markets, it improves the Sharpe ratio only in two of five cases. All Sharpe ratios of the bond portfolios, however, improve with hedging, which would be consistent with the fact that bonds tend to be more sensitive to inflation risk.

As an alternative to the comparison of national portfolio returns to a value weighted world portfolio, Tesar and Werner (1992) also examined the portfolio allocation that would be implied by a simple mean-variance model, based on ICAPM. Again using historical moments of return, they determined optimal national portfolio shares assuming that investors choose a set of risky securities (i.e. equity-indices and bond-indices respectively of each of the national markets), given a degree of absolute risk aversion equal to l =10. The model yielded the following objective

Abbildung in dieser Leseprobe nicht enthalten

where denotes a vector of portfolio weights, which sum to one, and Rj denotes a vector of expected excess returns of assets in all available markets. Optimal country shares can be derived from the partial derivative of [20] :

Abbildung in dieser Leseprobe nicht enthalten

Table 14. contrasts these optimal portfolios in own currencies as well as hedged against exchange rate risk with actually observed portfolios (two hypothetical portfolios Canb and USb are added which adjust for underestimation of international investment positions due to a lack of accurate data from the US and Canada.).


[1] Article VI. Section 3. of the IMF Articles of Agreement signed at the Bretton Woods conference in 1944 states that „members may exercise such (capital) controls as are necessary to regulate international capital movements, [...]“.

[2] The latest BIS estimates are that these could amount to $1.5 trillion.

[3] The debate on the degree of international financial integration is closely linked to the question of what causes it. Obviously each measure of financial integration reflects a somewhat different interpretation of these causes by making assumptions about investors‘ behaviour. The latter is taken as a sufficient enough concept to subsume indirect effects of the role of progress in information and telecommunication technologies, financial engineering, as well as political reforms aimed at facilitating international financial trading.

[4] The regression was also run for saving and investment net of depreciation, but it was thought that errors of measurement in the depreciation estimates could cause a spurious bias of the correlation coefficient. This is so because inflation discourages real saving and at the same time can lead to a high discount rate for depreciation, which causes the investment ratio to fall, without there being any causality of savings on investment.

[5] Adjusting returns for exchange rate risk can reflect two things (in reality both apply simultaneously): (1) Investors do not have perfect foresight. Therefore, with the exception where investors form zero expectations of currency devaluation (the latter applies for a random walk model of exchange rate changes), a positive risk premium will be deducted from the raw rate of return before comparing it to available domestic inmvestment; (2) Purchasing power parity does not hold (see Annex I).

[6] Not all of these factors necessarily bias the estimated correlation upwards. For example, if a government decides to grant subsidies to certain industries, investment should rise but the budget surplus, and therefore national saving should lower.

[7] One other obvious version of the endogeneity problem arises in time-series studies from the strongly pro-cyclical nature of saving and investment (even when expressed as shares of GNP). For this reason Feldstein and Horioka restricted their analysis to cross-section data. Another possibility is to adjust saving and investment data cyclically. In addition, many other factors could be thought of, such as energy shocks, real wages, strikes, or the presence of non-traded goods or of immobile production factors, etc. that will influence domestic saving and investment in the same direction, without the implication being a rejection of financial integration

[8] Feldstein and Horioka (1980) themselves used instrumental variables, including the ratio of retirees over the age of 65 to the population aged over 20-65, the ratio of younger dependents to the working- age population, the labour-force participation rate of older men, and the benefit-earnings replacement ratio under social security.

[9] Other supporters of this finding include Tobin (1983), Obstfeld (1986), and Dooley, et al. (1987).

[10] Another reason to expect the Feldstein-Horioka results to drop when developing countries are included in the sample is the following: according to either factor endowment theory or portfolio theory international capital flows are driven by cross-country heterogeneity. Thus, we should expect relatively little capital flows among „similar“ countries. As the size of the sample is increased to include more diverse countries, the correlation of savings and investment should drop (see Vamvakidis and Wacziarg, 1998). However, this does not explain low coefficients among subsamples where only developing countries are included (also see the extra section 2.3. on financial integration in portfolio theory).

[11] As mentioned in the first chapter, sacrificing capital mobility was a common policy for those countries committed to the maintenance of fixed exchange rates, the objective being to achieve at least some monetary autonomy.

[12] At the same time, fiscal and monetary policy were tightended substantially compared with former periods, which led to a sharp drop in private saving but left investment basically unaffected.

[13] The idea of an ultrarational household sector evolves from the empirical regularity referred to as Denison’s law, that is, the stability of the gross private saving rate.

[14] A random walk process of the real exchange rate would imply that , where the expected value of the error term is equal to zero. In this special case, expected exchange rate changes would also equal zero: .

[15] Such additional risk measures (e.g. foreign exchange risk) may include differences in the default assessment of foreign relative to domestic assets, or political risk related to the likelihood that assets in a particular country could be confiscated.

[16] A violation of the orthogonality assumption would mean that some elements of the regressor vector correlate with the error term. This would introduce serious problems of endogeneity into the equation of the same sort discussed in the previous section.

[17] This was also confirmed by MacDonald and Taylor (1992).

[18] Bekaert and Harvey actually considered the evidence of high covariation between national and foreign markets to be equivalent to accepting the financial integration hypothesis. For reasons stated previously, this paper would generally reject this view, unless the world was made of a homogenous group of countries.

Excerpt out of 210 pages


International Capital Flows: Economic Impact and Policy Implications
Christian-Albrechts-University of Kiel
Catalog Number
ISBN (eBook)
ISBN (Book)
File size
1448 KB
international, capital, flows, economic, impact, policy, implications
Quote paper
Nina Gillmann (Author), 2000, International Capital Flows: Economic Impact and Policy Implications, Munich, GRIN Verlag,


  • No comments yet.
Look inside the ebook
Title: International Capital Flows: Economic Impact and Policy Implications

Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free