European Stock Market Integration and EMU - CiteSeerX

6 downloads 265 Views 300KB Size Report
Shinasi, 1997). The recently announced cross-border stock market alliances and mergers, such as Euronext (Brussels, Amsterdam, and Paris), are important.
European Stock Market Integration and EMU Lieven Baele and Rudi Vander Vennet∗ Department of Financial Economics, Ghent University Paper in progress September 2001

Abstract

We examine the extent of capital market integration in a group of European countries following their intense efforts in the 1990s for closer economic and monetary integration. We estimate a conditional asset pricing model, which allows for a time-varying degree of integration. The degree of integration is determined by the importance of European risk factors relative to country-specific risks. The main innovation of the paper is that we allow the degree of stock market integration to vary with proxies for different aspects of EMU. More specifically, we test whether the evolution of exchange rate volatility, increased economic integration, and monetary policy coordination made increase European stock market integration. Preliminary results indicate that integration is strongly linked to the reduction in currency volatility, as well as to monetary policy convergence. Keywords: International CAPM, Stock Market Integration, EMU JEL Classification: C32, G12, G15

1

Introduction

The objective of this paper is to examine the extent and evolution of European stock market integration. The case of the European countries is particularly interesting because we can directly test whether the intense efforts aimed at strengthening the economic and monetary integration in the EU during the 1990s have also resulted in a higher degree of stock market integration. During the last two decades, several initiatives at the European level have been designed to promote securities market integration. By the end of the 1980s most capital controls were eliminated, allowing capital to move freely across ∗ Lieven Baele (contact person), Department of Financial Economics, Ghent University. [email protected]. Rudi Vander Vennet, Department of Financial Economics, Ghent University, E-mail: [email protected].

1

borders. Exchange rate stability and increased monetary policy coordination was sought with the creation of the European Monetary System (EMS) and its Exchange Rate Mechanism (ERM). In addition, the adoption of the Second Banking Directive (1989), the Capital Adequacy Directive (1993) and the Investment Services Directive (1993) significantly reduced remaining barriers between the European financial markets. The principles of a single passport and home country control allow unrestricted cross-border establishment and servicing and constitute an important step in the deregulation of securities market entry. The establishment of the Economic and Monetary Union (EMU) between 11 countries (Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, Netherlands, Portugal, and Spain; Greece joined in 2001) from January 1999 onwards is expected to eliminate most of the remaining obstacles to full financial integration. All financial contracts are denominated in euro and the European Central Bank (ECB) now conducts a common monetary policy for the whole Eurozone. In this paper, we focus on stock market integration. There are several reasons to expect that European stock market integration should increase. First, the introduction of the single currency eliminated all currency fluctuations within the Eurozone. As a result, the cost of hedging currency risk as a barrier to international investment has disappeared. Second, the introduction of the euro directly removed a number of existing barriers to cross-border investments. The EU matching rule, requiring that that liabilities in foreign currency be matched for a large percentage by assets in the same currency, can no longer prevent insurance companies, pension funds, and other financial institutions with liabilities denominated in Euro from investing in any country within EMU. The lower currency hedging costs and the elimination of currency risk should induce investors to increase their holdings of pan-European assets, leading to the elimination of any associated home bias. Third, after the elimination of currency risk, investors are likely to focus on other risk factors, such as credit, liquidity, settlement, and legal risks. As a result, issuers of securities as well as stock exchanges have strong incentives to reduce these risk components. They may do so by creating sufficient liquidity, by improving the financial and trading infrastructure, and by promoting transparent and cost-effective market practices (Prati and Shinasi, 1997). The recently announced cross-border stock market alliances and mergers, such as Euronext (Brussels, Amsterdam, and Paris), are important steps in that direction. Fourth, closer economic integration, increased monetary policy coordination and the rules for fiscal policy imposed by the Stability and Growth Pact caused a convergence of inflation rates. As a result, investors no longer need to hedge local inflation risk by e.g. investing proportionally more in local assets (see Adler and Dumas (1983), Stulz (1981b), Cooper and Kaplanis (1994)), resulting in a further breakdown of market segments in Europe. In addition, evidence in favor of business cycle synchronization within Europe (see e.g. Artis, Krolzig, and Toro, 1999) as well as more interdependence through trade should lead to a convergence of real cash flow expectations across the EMU countries, and hence to a more homegeneous valuation of equities. The issue of European stock market integration is of considerable impor2

tance for corporate managers, investors, consumers, and fiscal and monetary authorities. In integrated financial markets, corporate managers will face a lower cost of capital, since improved risk sharing possibilities will decrease risk exposures and the associated risk premia (see Errunza and Losq (1985), Bekaert and Harvey (1995), Stulz (1999), Bekaert and Harvey (2000), Henry (2000a), and Hardouvelis et. al. (2000)). The reduction of the cost of capital increases the number of positive net present value projects and should spur investment. Henry (2000b) reports that in a sample of 11 developing countries, stock market liberalizations were consistently followed by a temporary increase in the growth rate of real private investment. Bekaert, Harvey, and Lundblad (2001) find that equity market liberalizations lead on average to a one percent increase in annual real economic growth over a five-year period, even after correcting for other macro-economic effects. For investors, the gradual shift from local to global risk factors will affect their portfolio allocation strategies. First, in integrated markets, expected returns will no longer depend on the variance of local risk factors, but on the covariances with worldwide risk factors. Second, the correlation structure of stock markets is likely to be affected by the gradual process towards integrated markets. King and Whadhawani (1990), King, Sentana and Whadhawani (1994), Karolyi and Stulz (1996), and Bekaert and Harvey (1997) investigate time-varying linkages between international stock markets and find that correlations increase when global factors dominate domestic ones. In addition, several authors have documented that correlations are much higher when markets go simultaneously down, further reducing the insurance effect from international diversification (see Longin and Solnik (2001)). Consequently, increasing financial integration will require a substantial portfolio rebalancing, implied by the shift in relevant risk factors and their correlation structure. A third group of interest are consumers in general. In an integrated market, consumers will be able to share their consumption risks via cross-border ownership of productive assets. Sφrensen and Yosha (1998) find that the degree of risk sharing between a set of European countries is much lower than in the United States. They conclude that the lower degree of financial integration in Europe may be the reason for this apparent lack of risk sharing. Since consumption shocks in Europe are only to a small extent smoothed via other channels (labor mobility, fiscal policy), promoting further financial integration, and the associated possiblities for international risk sharing, is an important task. Finally, the degree of financial integration is also of interest for both monetary and fiscal authorities and for financial supervisors since the effects of their policy actions depend on the degree of financial integration. To determine the degree of European financial integration, we use the methodology proposed by Bekaert and Harvey (1995) and Hardouvelis et al. (2000). We estimate a conditional asset pricing model that contains European and local risk factors. Under full integration, rewards for risk should be the same in the different stock markets, and only EU-wide risk factors should be priced. In the opposite case of complete segmentation, only local risk factors should receive a premium. In the intermediate case of partial integration, expected returns will 3

be determined both by EU-wide and local factors. The degree of integration is said to increase if the time-varying weight of expected returns explained by EU-wide factors relative to local instruments increases. A contribution of this paper is the incorporation of a direct link between EMU-related economic factors and European stock market integration. Unlike Hardouvelis et al. (2000), who investigate whether (expected) participation to EMU increases stock market integration, we incorporate a set of economic factors affected by the EMU. More specifically, we test whether the reduction of exchange rate uncertainty, the increased monetary policy convergence, and the closer economic integration causes the degree of financial integration to increase. By including European countries that are not part of EMU, we can assess whether not participating in EMU has caused these stock markets to follow a different path towards integration compared to EMU stock markets. In addition, our paper also covers a period of two years following the introduction of the single currency in 1999. The rest of this paper is structured as follows. Section 2 discusses the impact of EMU on financial integration in Europe. Section 3 presents the model and the estimation procedures. Section 4 describes the data. Section 5 presents the main empirical results and a series of specification tests. Section 6 summarizes the principal findings and concludes.

2

Stock Market Integration and EMU

Asset pricing models have in common that they try to explain the cross-section of expected returns by differences in risk exposure. However, the risk factors investors will take into account depend on the perceived degree of market integration. If European capital markets were integrated, expected returns would only be determined by European-wide risk factors. In addition, the premiums associated with these risk factors would be identical in all countries. If, on the other hand, markets were segmented from the rest of the world, a rational investor would also have to add country-specific risk factors to the pricing kernel. Consequently, the identification of the relevant asset pricing model depends on how well markets are integrated and a different model may be needed when the degree of financial integration changes. Most asset pricing models assume either complete segmentation or complete integration. Examples of complete segmentation models are the domestic CAPM of Sharpe (1964), Lintner (1965) and Black (1972) and models linking expected returns to the variance of the local markets’ volatility (Merton (1980), French, Schwert, and Stambaugh (1987)). Asset pricing models in an international context usually assume markets to be perfectly integrated. Examples include a world CAPM (Harvey (1991), De Santis and Gerard (1997)), a world CAPM with currency risk (Solnik (1974), Sercu (1980), Stulz (1981b), Adler and Dumas (1983), Dumas and Solnik (1995), De Santis and Gerard (1998)), consumption-based asset pricing models (Wheatley (1988)), multiple risk factor models (Ferson and Harvey (1994, 1997), Sentana (2000)), and international la-

4

tent factor models (Chang et al. (1991), Bekaert and Hodrick (1992), Campbell and Hamao (1992), and Harvey, Solnik, and Zhou (1994)). In reality, however, there may exist barriers that prevent investors to diversify their portfolios internationally. These barriers include various forms of transaction costs, withholding taxes, differential access to markets, deviating corporate governance practices, and language barriers or other cultural differences. In models of partial integration, these barriers are treated as additional costs faced by international investors. If these costs are higher than the diversification benefits, investors will overinvest in the local market and portfolios may be biased towards home assets. Examples of such models are Black (1974), Stehle (1977), Stulz (1981a), and Cooper and Kaplanis (1994, 1999), among others). While most direct barriers to intra-European investment have been the subject of deregulation initiatives (for a detailed discussion, see Licht (1993)), there may exist indirect barriers that prevent investors from exploiting all possible diversification benefits and cause markets to remain segmented. First, markets may segment during periods of significant exchange rate volatility. Increased uncertainty about a particular currency often reflects doubts about the countries’ economic situation and its ability to take appropriate action. If the (conditional) risk premium does not fully compensate for the the increased uncertainty, and if hedging costs are higher than the expected benefits from diversification, currency volatility will cause markets to segment. In the European case, countries that were hit by the ERM-crisis (France, Ireland, Italy, Portugal, Sweden1 , UK) and the Asian crisis (Sweden, Switzerland, Spain) may have segmented during these periods. Or, countries of which EMU participation was uncertain may have been confronted with some degree of segmentation. Hardouvelis et al. (2000) found that countries that were considered more likely to fulfill the EMU accession criteria were more integrated with the aggregate European markets. One of the explanations they offer is that institutions such as pension funds and insurance companies were obliged to hold a large part of their financial assets in their domestic currency. This obligation restricts the scope for diversification in foreign assets, and hence acts as a direct barrier to foreign investment2 . These barriers existed for all European countries until the introduction of the single curency in January 1999 and continue to exist for countries that are not (yet) part of the euro-zone3 . Second, if international investors face inflation risk and deviations from purchasing power parity (PPP), they will hold portfolios that differ by a component designed to hedge local inflation risk (see e.g. Stulz (1981), Adler and Dumas (1983), Cooper and Kaplanis (1994)). In the European context, the increased monetary policy coordination as well as the requirement for EMU members to 1 Sweden was not formally a member of the ERM, but its exchange rate was pegged to the ECU. 2 Errunza and Losq (1985) proposed a model that investigates a similar barrier to market segmentation. They distinguish between eligible and noneligible securities. In this setting, mild segmentation occurs because investors from one country cannot invest in in stocks from another country, while investors from the other county do not face such restrictions. 3 Hardouvelis et.al. (2000) report that pension funds in Germany and France need to keep respectively a 95 and 80 percent currency match between liabilities and assets.

5

fulfill an explicit inflation criterion resulted in a substantial convergence of inflation rates throughout the 1990s. This should have reduced the need for inflation hedging significantly. Hence, our hypothesis is that the degree of integration between the European stock markets has risen with the degree of convergence of inflation. Since this is also expected to cause a convergence of real risk-free rates across EMU member states, a more homogeneous valuation of equities should be the result. Given the central role of inflation as a target for central bank policy, we interpret inflation as an indicator of monetary policy. In this sense, the recent evidence of divergence of inflation within the eurozone can be interpreted as differences in the stance of monetary policy across countries. Since the ECB conducts its monetary policy for the whole Eurozone, it may be more restrictive or expansive for a specific country than the national central bank would have been in the absence of EMU, given the countries economic fundamentals. In addition, the monetary policy of the ECB may produce asymmetric effects, potentially leading to divergences in inflation rates. Third, if EMU results in more closely correlated business cylces across its member countries, European stock market integration should increase. If the economies of two countries are fully integrated, labor, capital, and information would be able to move freely between the two countries. Consequently, differences in production costs and the state of technology should disappear. Common shocks would have a symmetric impact on economic growth and on future expected corporate earnings in the two countries. As a result, business cycle convergence should result in a more homogeneous valuation of equities and in a higher degree of integration. However, Eichengreen (1992), Kenen (1969), and Krugman (1993) have argued that closer economic integration could result in countries becoming more specialized in the goods in which they have a competitive advantage. In this situation, countries (or regions) may be more sensitive to industry-specific shocks, resulting in diverging business cycles. While both cases are theoretically possible, empirical results by e.g. Artis, Krolzig, and Toro (1999), Artis and Zhang (1997, 1999), Frankel and Rose (1996) suggest that economic integration in Europe resulted in a convergence of business cycles across the EMU countries4 . Several authors have investigated the link between business cycle synchronization, country return correlations, and financial integration. Erb, Harvey, and Viskanta (1994) found some evidence that cross-equity correlations in the G-7 countries are affected by the business cycle5 . Ragunathan, Faff, Brooks (1999) found the same relationship between US and Australian markets. Bracker, Docking, and Koch (1999) found a statisticallly significant relationship between bilateral import dependence and the degree of 4 Barrios, Brülhar, and Elliott (2001) for instance report that over the 1966-1997 period UK macro movents were significantly less correlated with the euro zone than those of the other main EU economies. In addition, they found that the trend has been towards further cyclical divergence rather than convergence between the UK and the euro zone. 5 Notice that increased correlations between equity markets does not necesarily mean increased financial integration, as e.g. industry mixes within each country may be sufficiently different to induce low equity correlation.

6

stock market integration. Dumas, Harvey, and Ruiz (2000) take the opposite view, and calculate the theoretical degree of return correlations both under integration and segmentation, after controlling for the degree of commonality of country outputs. They find that the assumption of market integration leads to a better explanation of the level of observed correlations than the assumption of market segmentation. Finally, a number of institutional changes in European stock markets are also likely to increase the degree of integration. Increased competition between the stock markets to attract listings and trades creates an incentive for stock exchanges to provide an optimal trading environment, where transactions can be done at relatively low cost. Stock exchanges and listed companies try to increase the liquidity of the quoted shares. Moreover, competition between listed firms to attract buyers will push firms to provide international investors with timely and correct information based on unified accounting standards. This should significantly reduce the problem of asymmetric information in the stock valuation process. As a results of these economic and institutional determinants, we expect that the degree of stock market integration in Europe has increased over the 1990s. This should especially the case for the Eurozone members.

3

Empirical model

In integrated markets and in the absence of exchange rate risk, the conditional CAPM of Sharpe (1964) and Lintner (1965) is given by Et−1 [ri,t ] = λEU,t−1 covt−1 [ri,t , rEU,t ]

(1)

where Et−1 [ri,t ] is the conditionally expected excess return on the local stock market index, rEU,t is the excess return on a European benchmark index, covt−1 is the conditional covariance operator and λEU,t−1 is the conditional price of European risk for time t. In the international context however, because of deviations from Purchasing Power Parity (PPP), the standard CAPM has to be extended to include exchange rate risk premiums. If there are L currencies next to the measurement currency, we can write the international CAPM as follows:

Et−1 [ri,t ] = λEU,t−1 covt−1 [ri,t , rEU,t ] +

L X

λl,t−1 covt−1 [ri,t , rl,t ]

(2)

l=1

where rl,t is the return on nonmeasurement-currency deposit l and λl,t−1 , i = 1...L, are the time-varying prices of exchange rate risk. To keep the empirical analysis tractable, we use a single basket of currencies C and hence (2) simplifies to : Et−1 [ri,t ] = λEU,t−1 covt−1 [ri,t , rEU,t ] + λC,t−1 covt−1 [ri,t , rC,t ] 7

(3)

In the case of complete segmentation, each asset will be priced corresponding to its covariance with the local market index returns. At an aggregate level, we get : Et−1 [ri,t ] = λi,t−1 vart−1 [ri,t ]

(4)

This means that expected returns in a segmented market are determined by the variance of returns in that market times the price of variance. Merton (1980) showed that λi is a measure of the representative investor’s relative risk aversion. Hence, the price of variance will depend on the weighted relative risk aversions of the investors in country i. In the case of partial segmentation, we can write the conditional expected return for stock market i as: Et−1 [ri,t ] = φi,t−1 (λEU,t−1 covt−1 [ri,t , rEU,t ] + λC,t−1 covt−1 [ri,t , rC,t ]) ¡ ¢ + 1 − φi,t−1 (λi,t−1 vart−1 [ri,t ]) (5)

where φi,t−1 measures the conditional level of integration of market i based on information available at time t−1 (0 ≤ φi ≤ 1) . Equation (5) describes expected returns in a partially integrated market where both local and global risk is priced. The parameter φi,t−1 determines what proportion of expected returns is a reward for local risk and what proportion is due to global risk. In the case of perfect integration, φi,t−1 will be equal to 1, and all weight will be put on the asset’s covariance with global factors. In this case, the model reduces to the International CAPM. In the opposite situation of perfect segmentation, φi,t−1 will be 0, and only local risk will matter. In this case, the model collapses to the domestic CAPM. We model the time-varying degree of integration φi,t−1 as follows: φi,t−1 =

exp(ψ0 Xt−1 ) 1 + exp(ψ0 Xt−1 )

(6)

where ψ is a vector of parameters and Xt−1 a vector of predetermined information variables. This functional form guarantees φi,t−1 to lie between 0 and 1. In addition, its curvature is in accordance with economic intuition. If the domestic market is segmented, a lot of effort is needed to start the process of integration and the time path of φi,t−1 will be relatively flat. Once however the necessary environment has been created, integration with other capital markets can advance quickly and φi,t−1 will be steep. Inversely, adverse shocks to the domestic economy may cause the market to become segmented in a very short period of time.

4 4.1

Estimation and Testing Estimation Issues

Equation (5) incorporates conditional EU market and currency returns. As a result, before we can estimate (5), we need to specify a model for both the EU 8

market and currency returns: rEU,t = λEU,t−1 vart−1 [rEU,t ] + λC,t−1 vart−1 [ri,t , rC,t ] + εEU,t

(7)

rC,t = λEU,t−1 covt−1 [rEU,t , rC,t ] + λC,t−1 vart−1 [rC,t ] + εC,t

(8)

The empirical version of (5) is given by ri,t

= φi,t−1 (λEU,t−1 covt−1 [ri,t , rEU,t ] + λC,t−1 covt−1 [ri,t , rC,t ]) ¡ ¢ + 1 − φi,t−1 (λi,t−1 vart−1 [ri,t ]) + εi,t

(9)

where all parameters are defined as in the previous section. As we made the variance and covariance factors time-varying, we also need to establish a model 0 of the second moments of all the returns. Let εt = [εEU,t , εC,t , εi,t ] | Ωt−1 ∼ N (0, Ht ) be the vector of unexpected excess returns, given the set of information Ωt−1 at time t − 1. Ht represents the conditional variance-covariance matrix of excess returns. Very often the standard GARCH(p, q) model is used, typically with a small p and q (see e.g. Hardouvelis et. al. (2000)). However, results by Pagan and Schwert (1990), Engle and Ng (1993), Bekaert and Wu (2000) suggest that GARCH models that account for asymmetric effects of return innovations generally outperform standard GARCH models. As in Bekaert and Wu (2000), we employ an asymmetric version of the BEKK model (Baba et.al. (1989), Engle and Kroner (1995), Kroner and Ng (1998)): Ht = C0 C + A0 εt−1 ε0t−1 A + B0 Ht−1 B + D0 ηt−1 η 0t−1 D

(10)

where for N assets C is a N + 2 by N + 2 lower triangular matrix of coefficients, and A, B,and D are N + 2 by N + 2 matrices of coefficients. We define η t−1 as follows:   ηEU,t−1  ηC,t−1  ½   −εi,t if εi,t < 0  η1,t−1  ∀i, EU, C η t−1 =   , η i,t−1 = 0 otherwise   ..   . ηN,t−1

In this specification, the conditional variance and covariance of each excess return is dependent upon the past conditional variances and covariances, past squared residuals and cross-residuals, and past squared asymmetric shocks and cross-asymmetric shocks. The advantage of the BEKK specification is that it enforces positive semi-definiteness on the covariance matrix Ht .On the other hand, one drawback of this model is the large number of parameters that must be estimated. To keep the size of the parameter space manageable, we assume the matrices A, B, and C to be diagonal. This implies that the conditional variances 9

only depend upon their past conditional variance, their past squared residuals, and their past squared asymmetric shocks, while covariances are determined only by their past covariance, the cross-product of lagged errors and asymmetric shocks. Finally, we have to specify a process for the evolution of the prices of risk: ¢ ¡ (11) λEU,t−1 = exp ξ 0EU XEU t−1 λC,t−1 = ξ 0C XEU t−1

(12)

¢ ¡ λi,t−1 = exp ψ0i XL t−1

(13)

¢ ¡ EU λi,t−1 = exp ψ0i XL represent European information variables, t−1 where X L X represents local information variables specific to country i, and ξ 0EU , ξ 0C , and ψ0i are vectors of coefficients. Merton (1980) and Adler and Dumas (1983) have shown that under risk aversion λEU,t−1 and λi,t−1 must be positive. To make sure that these restrictions are satisfied, we assume that λEU,t−1 and λi,t−1 are exponential functions of their instruments. However, there is no theoretical restriction on the sign of the price of currency risk. Investors may actually attach a negative price of currency risk if the expected devaluation of that currency vis-a-vis the base currency is larger than the interest rate differential. Therefore, the price of currency risk λC,t−1 is a linear function of its instruments. The parameters were estimated by Maximum likelihood (ML) assuming normally distributed errors. The loglikelihood function is given by T

T

1X 1X T (N + 2) ln 2π − ln |Ht (Θ| − εt (Θ)0 Ht (Θ)−1 εt (Θ) ln L(Θ) = − 2 2 t=1 2 t=1

(14)

where Θ is a vector of parameters to be estimated, T is the total number of observations, and N the number of series. Parameter estimates for Θ are obtained using the BFGS algorithm. In the estimation process we followed the two-step procedure proposed by Hardouvelis et.al.(1999). First, a bivariate model for the market and currency returns is estimated. This corresponds to equations (7), (8), (11), and (12). This produces an estimate of the conditional variances and the covariance of the market excess return and the excess currency return, as well as the price of both currency and market risk. Second, we impose these estimates to N equations (5), one for each country. Imposing that prices of risk are equal to our model is conform with our model, since in integrated markets price of risk must be equal across markets.This two-step procedure necessarily leads to some loss of efficiency. However, a simultaneous estimation of the full model is not practically feasible. In order to avoid problems due to non-normality in excess returns, we provide Quasi-ML estimates (QML), as proposed by Bollerslev and Woolridge (1992). However, we do not correct for the sampling error of the world market parameters in the first-stage estimation. Consequently, this approach yields consistent but not necessarily efficient estimates. 10

4.2

Specification Tests

To check whether our model is correctly specified, we follow a procedure similar to Nelson (1991), Bekaert and Harvey (1997), q and Bekaert and Wu (2000). ˆ i,i,t , for i = 1, .., N, EU, C, We calculate standardized residuals, zˆi,t = ˆεi,t / H which should follow a standard normal distribution conditional on time t − 1 information if the model is correctly specified. We then have the following orthogonality conditions to test: (a) E[ zˆi,t ] = 0 (b) E[ zˆi,t , zˆi,t−j ] = 0, 2 − 1] = 0 (c) E[ zˆi,t 2 2 (d) E[( zˆi,t − 1)( zˆi,t−j − 1)] = 0, 3 (e) E[ zˆi,t − ski ] = 0 4 (f) E[ zˆi,t − kurti ] = 0 zi,t |] = 0 (g) E[ˆ zi,t . |ˆ zl,t−j ] = 0 (h) E[ˆ zi,t .ˆ

j = 1..k j = 1..k

j = 1..k,

l = 1..N, j 6= i

All moment restrictions are tested using the generalized method of moments (Hansen (1982)). A test on correct specification on the conditional mean of the standardized residuals is implicit in (b), which provides us for k = 4 with a χ2 -statistic with four degrees of freedom. A similar test is conducted on the conditional variance, using moment condition (d). The distributional assumptions of the model are tested by examining conditions (a), (c), (e), and (f). This results in a χ2 -statistic with four degrees of freedom. Finally, we will jointly test all restrictions, which implies for k = 4 a test with 12 degrees of freedom.

5 5.1

Data and preliminary results Returns

We collect data from ten EMU countries (Austria, Belgium, Finland, France, Germany, Ireland, Italy, Netherlands, Portual, and Spain). The stock markets of these countries are expected to follow a path of steadily increasing integration as a result of the gradual intensification of the economic and monetary integration. We also include five European non-EMU countries : Denmark, Norway, Sweden, Switzerland and UK. The data are weekly, Deutschmarkdenominated, dividend-adjusted continuously compounded stock returns based on Friday closing prices for the period January 1990 - December 2000 downloaded from Datastream. The market index is a European benchmark index, calculated as the capitalization-weighted price index of the 15 country indices. Since we are interested in the total equity market (not only high capitalization, but also medium and small capitalization stocks), we prefer to use the total market indices compiled by Datastream International, which capture more than

11

75 percent of the market, as opposed to the the widely-used Morgan Stanley Capital International (MSCI) indices, which capure only approximately 60 percent of the total market. Larger companies, which are more involved in international trade, are likely to be more affected by European-wide risk factors than small companies. As a result, conducting the analysis using indices with a high proportion of large stocks may bias the results towards finding integrated markets. The weekly excess currency return is calculated as the ex-post deviation from uncovered interest rate parity vis-à-vis the Deutschmark: · ¸ 1 1 ln (1 + Ri,t−1 ) − ln (1 + RGE,t−1 ) rc,i,t ≡ ln(ei,t ) − ln(ei,t−1 ) + 52 52 where ei,t is the exchange rate (Deutschmark per unit of currency i), and Ri , RGE are the annualized one-month eurocurrency interest rates of currency i and the Deutschmark, respectively. In other words, excess currency return (positive or negative) occurs when investing currency i at t − 1 at Ri,t−1 and exchanging it in Deutschmarks at time t (for ei,t ) does not yield the same return as exchanging the local currency for Deutschmarks at time t − 1 (at ei,t ) and investing at for one period at the rate RGE,t−1 . Eurocurrency interest rates are London Friday closing rates, Deutchmark exchange rates are taken from Datastream. Since it is empirically not feasible to break down currency risk into a seperate component for each currency, we approximate currency risk by an aggegrate variable, calculated as the trade-weighted excess returns of the Belgian Frank, the Dutch Guilder, the French Frank, the Italian Lira, the Spanish Peseta, and the British Pound vis-a-vis the Deutchmark. Table 2 provides summary information for the DEM-denominated excess returns of the individual countries, the EU-15 value-weighted index, and the trade-weighted currency returns. Weekly excess returns are calculated by subtracting the one-month euro-DEM interest rate (adjusted as to reflect weekly returns) from the return of the local stock market over that week. In panel A, we report the mean excess return, the standard deviation, skewness, excess kurtosis, and the Jarque-Bera statistic for normality. One can observe considerable cross-sectional variation both in mean excess returns and in standard deviation. For example, Finland not only has the highest weekly mean return, 0.377%, but also the highest standard deviation, 3.968%. Belgium, on the other hand, performs worst with a mean of −0.086% per week and a standard deviation of 2.611%. Importantly, the EU-15 return/risk outcome domintates all local market indices (Austria has a slightly lower standard deviation, but also a considerably lower mean excess return), suggesting that investors can gain from international diversification. Excess currency returns are approximately zero and the standard deviation is very small compared to the one for excess stock returns. This suggests that European currency risk has been very small during the 1990s. Skewness and excess kurtosis are significantly higher than they should be under normality, and hence the Jarque-Bera test rejects normality for all excess return indices. Panel B reports first and second order autocorrelation coefficients of both excess returns and squared excess returns, as well as 12

contemporaneous correlation of local indices with the EU-15 and currency returns. We find that several autocorrelation coefficients are significantly different from zero for both the excess returns and squared excess returns. The former suggest that past returns can be useful predictors of future returns. The latter indicates that the use of a conditional heteroscedasticity model is advisable. All contemporaneous correlations between local excess returns and EU-15 returns are significant at the 1% level, and lie between 0.568 (Belgium) and 0.877 (UK). Contemporaneous correlations with aggregate currency returns are significantly positive, and range from 0.071 for Belgium to 0.437 for Italy.

5.2

Instruments

There is considerable evidence that expected returns are time-varying and related to the business cycle. Expected returns are found to be lower when economic conditions are strong and higher when conditions are weak. A number of variables related to business cycle movements have been identified as useful predictors of expected returns6 . The European-wide instruments include a constant, the first lag of the EU-15 index dividend yield in excess of the onemonth euro-DEM deposit rate (DY 1M IR), the first lag of the change in the term structure (∆T S), the first lag of the change in the one-month EU-15 deposit rate (∆DR), as well as a first and second lag in returns (lag1/2 ). The local instruments are a constant, the first lag of the local market index dividend yield in excess of the local market one-month deposit rate, the first lag in the local short-term interest rate, the first lag of the change in local market term structure, and the first and second lag of the local market excess return. All data are taken from Datastream. In table 3, we regress excess returns of the 15 local indices, the EU-15 excess return, and the aggregate currency return on a constant, the four local information variables and the four European instruments. We report Wald tests for the exclusion of the EU-15 instruments, the exclusion of the local instruments, and the exclusion of both. In many markets, it is not possible to exclude both sets of instruments. In addition, most local information variables seem to offer additional explanatory power in explaining excess returns in their respective markets. Adjusted R2 range from 0.07% to 3.78%..

5.3

Integration instruments

The main aim of this paper is to link European stock market integration to the Economic and Monetary Union. In section 2, we argued that there are three important evolutions related to EMU that should influence the degree of European stock market integration: the reduction of intra-European currency risk, business cycle convergence and increased monetary policy coordination. We examine the importance of each of these factors. 6 See e.g. Keim and Stambauch (1986), Campbell (1987), Fama and French (1988, 1989), Ferson and Harvey (1994,1997) , De Santis and Gerard (1997), Hardouvelis et.al (1999).

13

5.3.1

Local Exchange Rate Risk

We first test whether exchange rate volatility plays an important role in the financial integration process in Europe. There are several reasons why high exchange rate volatility may cause markets to segment. First, hedging costs will be higher for currencies going through a period of high volatility. Because of this extra cost, if not fully compensated by diversification benefits or a currency risk premium, investors may avoid a specific market. This argument may be especially important for the countries hit by the EMS crisis and the Asian crisis. Second, countries that were likely to participate to the Euro typically had lower exchange rate volatility at the end of the sample period. In addition, volatility decreased as the uncertainty about the participation to EMU gradually vanished. By proxying for the likelihood of participation in EMU, currency volatility gave useful information to pension funds, insurance companies, and other institutional investors since the switch to the Euro would abolish the restrictions on the currency composition of their portfolios. Since these investors are assumed to fully exploit all diversification benefits within the Eurozone, they should include more European assets in their portfolios, which should lead to a more homegeneous valuation of assets across the Eurozone. We measure the local currency volatility as a 52-week moving average of the standard deviation of the local currency returns. The exchange rates are denominated as Deutchmarks per unit of local currency. Table 4 reports some interesting summary statistics for volatility of the different currencies through time. We can observe that the path towards the euro in terms of reduction in currency volatility was quite uneven for the different member countries. While e.g. Austria, Belgium, and the Netherlands had already very small currency volatility from 1995 onwards, Finland, Ireland, Italy, and Spain experienced a substantial decrease of their currency volatility only in the last years before the introduction of the euro. This can be explained by the longer uncertainty about their participation in the euro. The currency volatility of the non-EMU currencies - the Danish Krone, the Norwegian Krone, the Swedish Krona, the Swiss Franc, and the UK pound - has been higher than that of the currencies that are now within the Eurozone in all subperiods. In addition, volatility has not decreased since 1995, contrary to the trend observed for the pre-euro currencies. 5.3.2

Monetary Policy Convergence

Given the importance for stock markets of monetary policy decisions, we hypothese that the convergence of monetary policies among European, and in particular the former Eurozone, countries had a positive influence on the degree of stock market integration. We assume that monetary policy convergence can be proxied by the convergence of inflation rates across the sample countries. Lower inflation reduces the need for inflation hedging significantly. In addition, the convergence of inflation rates lead to a convergence of real risk-free rates, resulting in a more homogeneous valuation of equities across Europe.

14

We construct several instruments to proxy for monetary policy convergence. The first measure is based on real interest rates. Therefore, we calculate the local real interest rates by substracting local inflation from the one month eurodeposit rate from the local currency. Next we calculate a benchmark real inflation rate for the EU-15, which is a GDP-weighted average of the individual countries’ real inflation rates. The weights used are calculated as previous year’s percentage of a country’s GDP in the total EU-15 GDP. Our first instrument is then the 12-month moving average of of differences between local real interest rates and the benchmark EU-15 real interest rate. Our second and third model are based on inflation rates. First of all we calculate a EU-15 benchmark inflation rate in the same way as in the real interest rate case. Next we calculated differences of local inflation with the EU-15 average inflation rate. As a second measure, we use a 12-month moving average of these differences. This measure focuses on the size of inflation relative to the EU-15 average. If we want to focus on inflation hedging costs as an impediment for international investment, we are more interested in the volatility of local inflation relative to the average EU-15 inflation volatility. Hence, a third measure is the difference between the 12-month standard deviation of local inflation rates and the inflation volatility of the EU-15 inflation. Summary statistics for the three intruments can be found in table 4, 5, and 6 respectively. The second column of table 5 reports the mean of differences between the local real interest rates and the EU-15 real interest rates over de full sample (January 1990 - December 2000), while column 3-7 report averages over different subperiods. Several countries had lower real interest rates than the EU-15 average over most of the sample. These countries include Austria, Belgium, France, Germany, the Netherlands, and Switzerland. For other countries, the convergence of real interest rates did only occur in the late 1990s (Italy, Spain and Portugal), or did not happen at all (U.K.). Table 6 reports summary statistics for the 12-month moving average of differences in local inflation rates with the aggregate EU-15 inflation rates. Several observations are worth noticing. First, Austria, Belgium, and France systematically had lower inflation rates than the EU-15 average. Also Germany and Switzerland had less than average inflation over the whole sample, with the exception of the period 1993-1994, during which inflation increased significantly. This is the main reason why we do not take Germany as a benchmark, as is often done in other studies. Ireland, Denmark, Norway, and to some extent also the Netherlands had fairly low levels of inflation during most of the 1990s, but were confronted with rising inflation in the period 1999-2000. Other countries, such as Italy, Spain, and Portugal had fairly high levels of inflation until the middle of 1990s, but were able to control their inflation in the second half of the decade. In table 7, the 12-month standard deviation of local inflation rates is compared to that of the EU-15 average inflation. Countries with periods of high inflation variability include Portugal, Sweden, Switzerland, and the UK. On the other hand, Austria, France, the Netherlands, and Denmark had, on average, a relatively low variability in inflation. Inflation rates in Germany were relatively volatile during the period 1990-1994. Not surprisingly, Ireland is confronted with an upsurge 15

in inflation variability at the end of the sample. 5.3.3

Business Cycle Convergence

The degree of stock market integration is not only allowed to vary with exchange rate variability and monetary policy convergence, but also with business cylce convergence. If economies are becoming more integrated, the correlation of corporate cash flows across countries should increase, leading to a more homogeneous valuation of European equities. We construct two proxies for business cycle convergence, based on monthly observations for seasonally adjusted industrial production (IP) for all countries and the EU-15 (all taken from Datastream). Results could be biased however if the country under investigation has a big weight in the aggregate EU-15 IP growth rate. Therefore, we constructed a benchmark IP growth rate by calculating a GDP-weighted average of the IP growth rates of all countries in sample, but the country under investigation. The weights are the same ones as used for the aggregate real interest and inflation rate in the previous section. The first integration instrument we will use is the 12-month moving standard deviation of the difference between the local IP growth rates and the growth rate of the aggregate European IP growth rate exclusive the country under investigation (STDIP). Second, we calculated 12-month moving correlations between the 15 countries’ IP growth rates and those of the aggregate EU-15 (again, excluding the country under investigation) (CORIP). All data is sampled monthly (January 1989-December 2000) and taken from Datastream. Monthly data were not available for Switzerland and for Portugal over the period January 1988 - December 1989. Summary statistics of both measures can be found in table 7 and 8 respectively. From table 7, we see that seven of the ten EMU members had their lowest STDIP value in the period 1997-2000, suggesting business cycle convergence. For Ireland however, STDIP merely suggests business cycle divergence. As for the non-EMU countries, convergence has happened for Norway, but not for Sweden and Denmark. The U.K. has low values for STDIP over the whole sample. However, we cannot observe any up- or downward pattern. The results in table 8 strongly suggest business cycle convergence. With the exception of Ireland, all EMU countries show a higher correlation with the aggregate IP growth rate in the last two years relative to their average. For most contries, correlations in the last period are in general quite high, lying between 40.76% and 51.22%. On the other hand, CORIP is low for Finland, Ireland, and to some extent also for the Netherlands. Correlations are lower van the non-EMU countries, and lie between 30.16% and -9.75%. In addition, there does not seem to be any increase, suggesting that there is no business cycle convergence with the EU average for these countries.

16

6

Empirical Results

The estimation of the expected returns, risk, and level of integration is done in two steps. First, we estimate a bivariate model for the EU-15 and aggregate expected returns and risk. This is the subject of the first paragraph. In the second paragraph, we impose the estimates for the EU-15 and the currency prices of risk, as well as the innovations in expected market and currency returns on the full specification. Third, we investigate whether the model is robust to alternative specifications. Fourth, we decompose expected returns in premia for local and European-wide risk. Fifth, we calculate the change in cost of capital due to a changing degree of European financial integration.

6.1

Price of EU-15 Market and Currency Risk

To quantify the expected EU-15 excess market return, the expected aggregate currency returns as well as their prices, we estimated a bivariate model corresponding to equations (7), (8), (10),(11), and (12). The estimation results are reported in table 9. Panel A reports estimation results for the conditional variance and covariance specification, panel B for the conditional prices of market and currency risk. Panel C offers some residual statistics and parameter tests. To avoid problems due to non-normality in excess returns, we provide quasi-maximum likelihood standard errors in all tables. All diagonal elements of C, A,and B are significant at the 1% level, suggesting time-variation in quantities of risk. The off-diagonal elements of A and B however are not significantly different from zero. In Panel C, under the heading DIAG, a robust wald test indicates that the hypotheses that the diagonal elements of A and B are equal to zero cannot be rejected. This gives additional evidence that using the diagonal V ech model does not lead to an important loss of information. We also test whether the parameters adding asymmetry, matrix D, are significantly different from zero. The Wald test (ASYM) clearly rejects the hypothesis that all parameters are equal to zero. Panel B reports estimation results for the conditional mean equations. The coefficients for the market price of risk are significantly different from zero at least at the 5 percent level, suggesting time-variation in prices of market risk. However, the parameters linking the instruments to the price of aggregate currency risk are not significantly different from zero. This result continues to hold when we use different proxies for aggregate currency returns and for the instruments.. The proxies we used for aggregate currency returns include a tradeand a GDP-weighted average of deviations from uncovered interest rate parity for all 15 countries. Other instruments used include the change in the 1 month DM deposit rate, and various proxies for a European default spread. In panel C, we report robust wald tests of the null hypothesis that the price of market and currency risk are zero. The null hypothesis is easily rejected for the price of market risk, but not for the price of currency risk. This result suggests that currency risk - from the point of view of a German investor - has been very small 17

throughout the 1990s, and that it was not priced in the market. As the estimation results suggest that the price of currency risk is not significantly different from zero, we are not going to take into account currency effects in the second step. As a result, our second step reduces to imposing the price of EU-15 market risk and the innovations from the market equation on the individual country’s equations7 . Finally, panel C reports some test statistics on the residuals. The MEAN test investigates whether the first four autocorrelations of the standardized residuals are equal to zero. The VARIANCE checks whether there is any autocorrelation left (until the fourth order) in the squared residuals. Finally, the DISTRIB test investigates whether the standardized residuals are normally distributed or not. All test statistics are obtained through a GMM test procedure (see section 4.2.) and distributed as a χ2 distribution with 4 degrees of freedom. None of the test statistics can be rejected at the 5 percent level. However, the MEAN and DISTRIB test statistic for the standardized residuals from the EU-15 market returns are significant at the 10 percent level, suggesting that there might me some autocorrelation left in the standardized residuals, and that those may not be normally distributed. The latter confirms the need to use quasi-maximum likelihood standard errors. As the former is concerned, we will try to improve on that during the following weeks.

6.2

Estimates of Time-Varying Integration

Estimation results of the time-varying integration model for the individual countries can be found in table 10. These are obtained by imposing the price of EU-15 market risk and the conditional variance of the EU-15 excess market returns on equation (5). The currency part has been omitted for reasons discussed in the previous paragraph. We report preliminary results for France, Italy, Spain, and the U.K. Panel A of table 10 reports estimates for the local price of risk specification. Many local instruments are significant, suggesting time-variation in the local prices of risk and at least imperfect integration (local risk should not be priced in perfectly integrated markets). Panel B offers estimation results for the parameters linking the integration instruments to the time-varying degree of integration. The proxy for currency volatility is strongly significant for all countries, suggesting that the reduction in currency volatity has been very important in making European stock markets more integrated. The proxy for monetary policy convergence - a 12 month moving average of the difference between local inflation and aggregate EU-15 inflation is is significant for Spain (at the 1% level), and for France (at the 5 % level). The result for Spain is not surprising, as its convergence in inflation rates was particularly important. The proxy for business cycle convergence - 12 month correlation between local industrial production growth rates and the aggregate EU growth rate (excluding local country) - is not significant in any of the countries. 7 We will investigate further on this. More specifically, we will try other ways of constructing the aggregate currency returns, or to account for a structural break.

18

Panel C reports various checks on the residuals, as well as some parameter restriction tests (to be done). The results are still very incomplete, as well as preliminary. Further results are currently being processed for other countries, and for different proxies for monetary and economic integration.

6.3

Integration and the Cost of Capital

• Decomposition of expected returns in premia for local and EU-wide risk • what are the consequences of time-varying integration for the cost of capital? • what pricing errors to we make by (wrongly) assuming markets are or completely integrated, or completely segmented?

7

Conclusion

This paper relates the degree of European stock market integration to the economic and monetary integration in the EU during the 1990s. We hypothese that three EMU factors have contributed to a higher degree of European stock market integration: a reduction in currency volatility, increased economic integration, and monetary policy coordination (and unification). First, currency volatility may constitute a barrier to international investment if hedging costs are particularly high, which is likely to have been the case for the countries involved in the EMS crisis, for the countries of which participation to EMU was doubtfull, and for the countries who decided not to participate to the euro. In addition, the introduction of the euro was also the end of currency matching rules faced by many pension funds and insurance companies (at least within the euro zone), hereby removing a direct barrier to international investment. Second, monetary policy convergence (and unification) resulted in an inflation convergence across countries. The resulting convergence in real risk-free rates should lead to a more homogeneous valuation of equities across EMU countries. Third, if increased economic integration resulted in business cycle convergence, companies’ cash flows will be more and more determined by common shocks, leading to an increased cross-country correlation in cash flows and their volatilities. In this paper, we investigate whether these three effects resulted in a convergence of stochastic discount factors across countries. We estimate a conditional asset pricing model, which accounts for a time-varying degree of integration. The degree of integration is determined by the importance of European-wide risk factors relative to country-specific risk factors. Hardouvelis et.al. (2000)

19

- using the same methodology - investigated whether the probability of participating to EMU increased stock market integration. We go one step further, and try to find out what exactly of EMU did influence stock market integration. Therefore, we allow the degree of integration to vary with proxies for reduced currency volatility, monetary policy convergence, and business cycle convergence. The model accounts for intra-European currency risk, as well as time-varying prices and quantities of risk. The quantities of risk are obtainted using a multivariate GARCH-in-mean model, augmented with an asymmetric component. The prices of risk are allowed to vary with both local and EU-wide instruments. Preliminary results suggest that the main driving factor for increased European stock market integration is the reduction in currency volatility. In addition, monetary integration seems to be especially important for countries that were confronted with a particularly strong convergence in inflation rates. Business cycle convergence does not seem to be important in the countries investigated yet. In due course, we will investigate whether the results confirmed for other countries, and for different proxies for the integration instruments.

References [1] Adler, Michael, and Bernard Dumas (1983), International portfolio selection and corporate finance: a synthesis, Journal of Finance, 46, 925-984. [2] Artis, Michael J., Hans-Martin Krolzig, and Juan Toro (1999), The European Business Cycle, CEPR Discussion Paper No. 2242, September. [3] Bekaert, G. and C.R. Harvey (2000), Foreign Speculators and Emerging Equity Markets, Journal of Finance 55, p. 565-614. [4] Bekaert, G, C.R. Harvey, and C.Lundblad (2001), Does Financial Liberalization Spur Growth?, Working Paper. [5] Bekaert, Geert and Campbell Harvey (1997), Emerging equity market volatility, Journal of Financial Economics, 43, pp. 29-77. [6] Bekaert, Geert and Campbell Harvey (1995), Time-varying world market integration, Journal of Finance, 50, 403-444. [7] Bekaert, Geert, and Robert Hodrick (1992), Characterizing predictable components in excess returns on equity and foreign exchange markets, Journal of Finance 47, 467-509. [8] Beyer Andreas, Jurgen A. Doornik, and David F. Hendry (2001), Constructing Historical Euro-Zone Data, Economic Journal, 111, February, 102-121. [9] Black, Fisher (1972), Capital Market Equilibrium with Restricted Borrowing, Journal of Business, 45, pp. 444-455. 20

[10] Black, Fisher (1974), International Capital Market Equilibrium with Investment Barriers, Journal of Financial Economics, p.337-352. [11] Bodard, Vincent, and Paul Reding (1999), Exchange rate regime, volatility and international correlations on bond and stock markets, Journal of International Money and Finance 18, pp.133-151. [12] Bollerslev, Tim, Robert F. Engle and Jeffrey M. Wooldridge (1988), A Capital Asset Pricing Model with Time Varying Covariances, Journal of Political Economy, 96, pp.116-131. [13] Bollerslev, Tim, Ray Y. Chou, and Kenneth F. Kroner (1992), ARCH modeling in finance: a review of the theory and empirical evidence, Journal of Econometrics, 52, pp. 5-59. [14] Bollerslev, Tim, and Jeffrey M. Woolridge (1992), Quasi-Maximum likelihood estimation and inference in dynamic models with time-varying covariances, Econometric Reviews 11, 143-172. [15] Bracker, Kevin, Diane Scott Docking, and Paul D. Koch, Economic Determinants of evolution in international stock market integration, Journal of Empirical Finance 6, p. 1-27. [16] Bracker, Kevin, Diane Scott Docking, Paul D. Koch (1999), Economic Determinants of Evolution in International stock market integration, Journal of Empirical Finance 6, 1-27. [17] Campbell, John Y. (1987), Stock Returns and the Term Structure, Journal of Financial Economics, 18, 373-399. [18] Campbell, John Y., and Yasushi Hamao (1992), Predictable stock returns in the United States and Japan: A study of long-term capital market integration, Journal of Finance, 47, 43-69. [19] Cooper, I.A. and E. Kaplanis (1999), Partially segmented international capital markets and international capital budgeting, Journal of International Money and Finance 19, pp. 309-329. [20] Cooper, I.A. and E. Kaplanis (1994), Home Bias in Equity Portfolios, Inflation Hedging, and International Capital Market Equilibrium, Review of Financial Studies, Vol. 7, nr. 1, pp.45-60. [21] Chang, Eric C., J. Michael Pinegar and R. Ravichandran (1991), Latent Variables Tests of the Integration of European Equity Markets [22] De Santis, Giorgio and Bruno Gerard (1997), International Asset Pricing and Portfolio Diversification with Time-Varying Risk, Journal of Finance, 52:1881-912. [23] De Santis, Giorgio and Bruno Gerard (1998), How big is the premium for currency risk?, Journal of Financial Economics, 49, 375-412. 21

[24] Dumas, Bernard, Campbell R. Harvey, and Pierre Ruiz (2000), Are Correlations of Stock Returns Justified by Subsequent Changes in National Output, Working Paper. [25] Dumas, Bernard and Bruno Solnik (1995), The World Price of foreign exchange risk, Journal of Finance, 50, 445-479. [26] Engle, Robert F. and K. F. Kroner (1995), Multivariate Simulaneous Generalized ARCH, Econometric Theory 11, 122-150. [27] Erb, Claude B., Campbell Harvey, and Tadas Viskanta (1994), Forecasting International Equity Correlations, Financial Analysts Journal, November/december, pp. 32-45. [28] Errunza, Vihang, and Etienne Losq (1985), International Asset Pricing under Mild Segmentation: Theory and Tests, Journal of Finance XL, no. 1, pp. 105-124. [29] Fama, Eugene F. (1981), Stock Returns, real activity, inflation, and money, American Economic Review 71, 545-565. [30] Fama, Eugene F. and Kenneth R. French (1988), Dividend Yields and Expected Stock Returns, Journal of Financial Economics, 22, 3-25. [31] Fama, Eugene F. and Kenneth R. French (1989), Business Conditions and Expected Returns on Stocks and Bonds, Journal of Financial Economics, 25, 23-49. [32] Fama, Eugene F. and Kenneth R. French (1992), The Cross-Section of Expected Stock Returns, Journal of Financial Economics, 33, February, 3-56. [33] Ferson, Wayne E., and Campbell R. Harvey (1994), Sources of risk and expected returns in global equity markets, Journal of banking and finance, 18, 775-803. [34] Ferson, Wayne E., and Campbell R. Harvey (1997), The fundamental determinants of national equity market returns: A perspective of country risk in asset pricing, Journal of banking and finance, 18, 775-803. [35] Fratzscher, Marcel (2000), Financial Market Integration in Europe: On the Effects of EMU, working paper. [36] French, Kenneth R. and James M. Poterba (1991), International Diversification and International Equity Markets, American Economic Review, 81:2, pp. 222-226. [37] French, Kenneth, G. William Schwert, and Robert F. Stambaugh, 1987, Expected stock returns and volatility, Journal of Financial Economics, 19, 3-29. 22

[38] Geske, Robert and Richard Roll (1983), The monetary and fiscal linkage between stock returns and inflationm Journal of Finance 38, 1-33. [39] Gehrig, Thomas P. (1993), An information based explanation of the domestic bias in international equity investment, The Scandinavian Journal of Economics, 21, 7-109. [40] Glen, Jack, and Philippe Jorion (1993), Currency Hedging for International portfolios, Journal of Finance, Volume 48, Issue 5 (December), p. 18651886). [41] Hansen, Lars. P. (1982), Large Sample Properties of Generalized Method of Moments Estimators, Econometrica, 50, 1029-1054.s [42] Hardouvelis, Gikas A., Dimitrios Malliaropulos, and Richard Priestley (1999), EMU and European Stock Market Integration, CEPR Discussion Paper No. 2124. [43] Harvey, Campbell (1991), The World Price of Covariance Risk, Journal of Finance 46 (1991): 111-157. [44] Harvey, Campbell, Bruno Solnik and Guofu Zhou (1994), What Determines Expected International Asset Returns?, NBER working paper 4660. [45] Henry, P. (2000a), Stock Market Liberalization, Economic Reform, and Emerging Market Equity Prices, Journal of Finance 55, pp.529-564. [46] Henry, P. (2000b), Do Stock Market Liberalizations Cause Investment Booms, Journal of Financial Economics 58, pp. 301-334. [47] James, C., S. Koreisha, and M. Partch (1985), A VARMA Analysis of the Causal Relations Among Stock Returns, Real Output, and Real Interest Rates, Journal of Finance 40, 1375-1384. [48] Karolyi, G. and R. Stulz (1996), Why do markets move together? An investigation of U.S.-Japan stock return comovements using ADR’s, Journal of Finance 51, pp.951-986. [49] Kaul, Gautam (1987), Stock returns and inflation: the role of the monetary sector, Journal of Financial Economics 18, 253-276. [50] Keim, Donald B. and Robert F. Stambauch (1986), Predicting Returns in the Stock and Bond Markets, 17, 357-390. [51] King, M. and S. Whadhawani (1990), Transmission of Volatility between Stock Markets, Review of Financial Studies 3, pp. 3-33. [52] King, M., Sentana, E. and Wadhawani (1994), Volatility and the links between national stock markets, Econometrica 62, pp. 901-934.

23

[53] Lee Daniel Y., Purchasing power parity and dynamic error correction, International Review Of Economics And Finance (8)2 (1999) pp. 199-212 [54] Lintner, John (1965), The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets, Review of Economics and Statistics, 47, 13-37. [55] Licht, Amir N. (1998), Stock market integration in Europe, CAER II Discussion Paper No. 15, Harvard Institute for International Development, Cambridge, MA. [56] Longin, F. and B. Solnik (2001), Extreme Correlation of International Equity Markets, Journal of Finance 56, pp. 649-676. [57] Merton, Robert (1980), On Estimating the Expected Return on the Market: An Exploratory Investigation,” Journal of Financial Economics, 8, December 1980. [58] Nelson, Daniel B. (1991), Conditional Heteroskedasticity in Asset Returns: A New Approach, Econometrica, Volume 59, pp. 347-370. [59] Prati, Alessandro and Garry J. Schinasi (1997), European Monetary Union and International Capital Markets: Structural Implications and Risks, IMF Working Paper (WP/97/62). [60] Ragunathan, Vanitha, Robert W. Faff, and Robert D. Brooks (1999), Correlations,business cycles and integration, Journal of International Financial Markets, Institutions, and Money 9, 75-95. [61] Ram, Rati and David E. Spencer (1983), Stock returns, real activity, inflation and money: Comment, American Economic Review 73, 463-470. [62] Sentana, Enrique (2000), Did the EMS reduce the Cost of Capital?, CEMFI Working Paper 9514, [63] Sercu, Piet (1980), A Generalization of the International Asset Pricing Modelm Journal de l’Association Francaise de Finance 1(1). [64] Sharpe, William (1964), Capital asset prices: A theory of market equilibrium under conditions of risk, Journal of Finance 19, pp. 425-442. [65] Solnik, Bruno (1974), An Equilibrium model of the international capital market, Journal of Economic Theory 8, 500-524. [66] Sφrensen, Bent E., and Oved Yosha (1998), International Risk Sharing and European Monetary Unification, Journal of International Economics, vol. 45, August 1998, 211-38. [67] Stehle, Richard (1977), An empirical test of the alternative hypotheses of national and international pricing of risky assets, Journal of Finance, 32, 493-502. 24

[68] Stulz, René M. (1981a), On the effects of barriers to international investment, Journal of Finance, 36, 923-934. [69] Stulz, René M. (1981b), A model of international asset pricing, Journal of Financial Economics 9, 383-406. [70] Stulz, René M. (1986), Asset pricing and expected inflation, Journal of Finance 41, 209-223. [71] Stulz, René M. (1999), Globalization of equity markets and the cost of capital, Working Paper, February 1999, Charles A. Dice Center for Research in Financial Economics. [72] Wheatley, Simon (1988), Some Tests of International Equity Integration, Journal of Financial Economics, 21, p. 177-212.

25

Table 1 Summary Statistics on Excess Returns Panel A of table 1 presents summary statistics of the excess returns of the 15 country indices used in this study, as well as for the value-weighted index of the EU-15 market and the trade-weighted aggregate currency return. All data are weekly returns denominated in Deutschmarks minus the one month euro-DM deposit rate (transformed as to reflect a weekly riskfree return). The aggregate currency return is calculated as a trade-weighted return of the Franch Franc, the British Pound, the Italian Lira, the Dutch Guilder, the Spanish Peseta, and the Belgian Franc. Returns are calculated as deviations from uncovered interest rate parity. The sample period is 03/01/1990-27/12/2000, which is 574 observations. Column 2-5 report report respectively the average weekly return in percentage points, the standard deviation, skewness and kurtosis. Column 6 en 7 report the Jarque-Bera test statistic for normality in excess returns and its p-value (distributed as a χ2 with two degrees of freedom). Panel B reports autocorrelation coefficients of the first and second lag of excess returns. Panel A: Summary Statistics

Countries

Mean %

Std. Dev. %

Skewness

Kurtosis

Jarque-Bera

p-value

Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden Switzerland UK EU-15 CUR

-0.032 0.105 0.401 0.198 0.127 0.201 0.119 0.234 0.152 0.08 0.153 0.145 0.227 0.247 0.197 0.179 0.0190

2.630 1.967 4.406 2.489 2.392 2.591 3.469 2.060 2.878 2.372 2.156 3.202 3.539 2.077 2.281 2.032 0.3013

0.028 -0.243 0.011 -0.331 -0.529 -0.162 -0.010 -0.409 -0.437 0.251 -0.025 0.392 -0.146 -0.387 0.192 -0.444 -1.2129

6.155 0.898 4.129 0.560 1.173 2.187 0.543 2.387 0.884 2.616 0.613 2.700 0.698 1.298 1.812 0.801 8.7553

246.89 115.42 31.62 158.48 110.49 19.00 149.65 25.89 129.99 9.89 141.33 133.44 133.44 86.65 38.63 139.41 931.3

0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00

26

Panel B : Autocorrelations

Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark

Norway Sweden Switzerland UK EU-15 CUR

Autocorrelation of ri,t

2 Autocorrelation of ri,t

Lag 1 0.0059 0.0536 0.0324 -0.0601 -0.0734* -0.0166 0.0045 -0.0732* -0.0263

Lag 1 0.3792*** 0.1998*** 0.0762* 0.1763*** 0.1752*** 0.0579 0.0220 0.3189*** 0.1225*** 0.2213*** 0.0754* 0.0259 0.0898** 0.1383*** 0.0326 0.1839***

0.0824∗ 0.0508 0.0622 -0.0259 0.0047 -0.0705* -0.0411

Lag 2 0.1044** 0.1199*** 0.0937** 0.0697* 0.0759* 0.0674 0.0872** 0.1600*** 0.1637*** 0.1173*** 0.0724* 0.0770* 0.0785* 0.1067*** 0.0864** 0.1333***

27

Lag 2 0.2291*** 0.1772*** 0.2690*** 0.1507*** 0.2238*** 0.0646 0.1059** 0.1802*** 0.0971** 0.1535*** 0.0320 0.2271*** 0.0877** 0.2126*** 0.0384 0.2108***

Table 2 Predictability of Excess Returns

Countries

Exclude EU variables 2

Austria Belgium Finland France Germany Ireland Italy Netherlands Portugal Spain Denmark Switzerland Norway UK EU-15 CUR

Exclude local variables 2

Exclude Both 2

χ

p-value

χ

p-value

χ

p-value

5.885 5.175 12.340 13.475 8.440 10.578 9.476 6.719 11.210 12.787 13.691 1.727 10.187 10.438 10.922 5.183

0.328 0.395 0.030 0.019 0.134 0.060 0.092 0.242 0.047 0.025 0.017 0.886 0.070 0.064 0.027 0.371

8.779 1.860 18.717 14.039 11.884 12.544 5.305 22.069 12.989 10.477 14.289 6.362 15.618 13.163

0.118 0.868 0.002 0.015 0.036 0.028 0.379 0.000 0.023 0.063 0.014 0.273 0.008 0.022

10.486 6.114 22.682 17.248 16.534 15.891 10.373 25.149 14.819 13.937 18.151 9.635 21.039 14.544

0.313 0.728 0.007 0.045 0.057 0.069 0.320 0.002 0.096 0.125 0.033 0.381 0.012 0.104

28

¯2 R 1.765 1.041 3.258 2.573 2.716 2.476 1.892 3.349 2.228 2.274 2.967 0.070 3.778 2.011 3.325 1.993

Table 3 Summary Statistics for Local Currency Volatility This table reports summary statistics for the average 52-week standard deviation in local exchange rates for different subperiods. The weekly exchange rates are taken from Datastream, and are expressed as Deutschmarks per unit of local currency. Column 2 reports the average 52-week standard deviation of the local currency returns over the whole sample (03/01/1990-27/12/2000), while columns 3 to 6 report average 52-week standard deviations of the local currency over different subperiods.

Countries Austria Belgium Finland France Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden Switzerland UK

Average 52-week Standard Deviation of Weekly Currency Return (in %) Full Sample 0.082 0.181 0.720 0.231 0.582 0.615 0.087 0.545 0.429 0.316 0.640 0.967 0.682 0.986

1990-1992 0.109 0.146 0.940 0.194 0.309 0.420 0.094 0.543 0.540 0.272 0.418 0.651 0.733 0.871

1993-1994 0.017 0.388 1.477 0.375 1.007 1.343 0.074 1.030 0.748 0.529 0.607 1.608 0.741 1.090

29

1995-1996 0.013 0.088 0.785 0.444 0.793 1.252 0.053 0.801 0.360 0.367 0.444 1.350 0.797 0.961

1997-1998 0.009 0.028 0.458 0.125 0.712 0.440 0.038 0.219 0.163 0.212 0.886 0.925 0.721 1.124

1999-2000 0.002 0.004 0.015 0.004 0.047 0.019 0.009 0.011 0.014 0.239 0.948 0.928 0.644 1.091

Table 4 Summary Statistics for differences in real interest rates with the EU-15 real interest rate This table reports summary statistics for a monetary policy convergence instrument based on real interest rates. Real interest rates are calculated by substracting local inflation from the one-month euro-deposit rate for the local currency. All data are taken from Datastream. A benchmark EU-15 real interest rate is calculated as a GDP-weighted average of the individual countries’ real interest rates. The weights used are previous year’s percentage of a country’s GDP in the total EU-15 GDP. The measure reported here is a 12-month moving average of differences between local real interest rates and the benchmark EU-15 real interest rate. Column 2 reports an average over the full sample (January 1990 - December 2000), while columns 3 till 7 report averages over different subperiods.

Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden Switzerland UK

real interest differentials with aggegate EU-15 real interest rate Full Sample -1.162 -0.784 0.222 -0.404 -1.117 0.618 1.845 -1.197 1.891 2.466 0.237 0.625 1.214 -2.494 1.099

1990-1992 -1.566 -1.096 2.728 -0.592 -1.648 0.830 1.853 -1.585 2.798 3.213 0.264 0.716 3.099 -2.273 1.492

1993-1994 -1.291 -0.066 -0.616 0.137 -0.752 0.623 1.980 -1.033 3.053 5.360 1.747 -0.799 0.831 -2.632 -1.503

30

1995-1996 -1.690 -1.540 -0.976 -0.357 -1.683 0.162 3.967 -1.912 2.840 2.881 -0.571 -0.420 1.708 -3.066 0.710

1997-1998 -0.806 -0.783 -1.020 -0.830 -0.956 1.612 1.700 -0.983 0.593 0.791 -0.395 0.610 -0.151 -2.786 2.713

1999-2000 -0.257 -0.278 -0.271 -0.278 -0.278 -0.257 -0.278 -0.278 -0.278 -0.278 0.131 2.974 -0.363 -1.822 1.867

Table 5 Summary Statistics for differences in local inflation rates with the aggegate EU-15 inflation rate This table reports summary statistics for a monetary policy convergence instrument based on an inflation differential with the aggregate EU-15 inflation rate. All inflation rates are taken from Datastream. A benchmark EU-15 inflation rate is calculated as a GDP-weighted average of the individual countries’ inflation rates. The weights used are previous year’s percentage of a country’s GDP in the total EU-15 GDP. The measure reported here is a 12-month moving average of differences between local inflation rates and the benchmark EU-15 inflation rate. Column 2 reports an average over the full sample (January 1990 December 2000), while columns 3 till 7 report averages over different subperiods.

Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden Switzerland UK

MA of difference of local CPI-rates w. EU-15 CPI (in %) Full Sample -0.150 -0.338 0.158 -0.479 -0.228 0.021 1.735 -0.278 1.898 4.043 -0.052 0.293 1.022 -0.175 1.556

1990-1992 -0.703 -0.450 1.956 -0.476 -1.496 -0.423 2.540 -1.571 2.734 8.964 -0.760 0.193 5.546 1.362 4.467

1993-1994 0.437 -0.680 -0.536 -0.863 0.958 -0.614 1.874 -0.284 2.178 5.398 -1.429 -0.785 1.083 0.937 0.078

31

1995-1996 0.115 -0.366 -1.170 -0.323 -0.300 0.114 2.545 -0.044 2.210 1.799 -0.092 -0.275 -0.112 -0.887 0.799

1997-1998 -0.275 -0.057 -0.456 -0.398 -0.067 0.092 0.709 0.479 0.663 0.889 0.493 0.586 -1.309 -1.198 1.435

1999-2000 -0.240 0.125 0.418 -0.444 -0.324 1.468 0.663 0.891 1.152 1.261 1.196 1.236 -0.931 -0.428 1.009

Table 6 Summary Statistics for differences in standard deviations in local inflation and the EU-15 inflation rate. This table reports summary statistics for a monetary policy convergence instrument based on an inflation differential with the aggregate EU-15 inflation rate. All inflation rates are taken from Datastream. A benchmark EU-15 inflation rate is calculated as a GDP-weighted average of the individual countries’ inflation rates. The weights used are previous year’s percentage of a country’s GDP in the total EU-15 GDP. The measure reported here is a difference between the 12-month standard deviation of local inflation rates and that of the benchmark EU-15 inflation rate. Column 2 reports an average over the full sample (January 1990 - December 2000), while columns 3 till 7 report averages over different subperiods.

Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden Switzerland UK

diff. of standard deviation between local inflation and EU-15 inflation Full Sample 0.071 0.120 0.241 0.024 0.311 0.253 0.116 0.058 0.141 0.414 0.103 0.138 0.711 0.371 0.412

1990-1992 -0.021 0.063 0.240 -0.059 0.610 0.079 -0.049 0.050 0.045 0.558 0.172 0.038 1.326 0.776 0.727

1993-1994 -0.058 -0.022 0.223 -0.042 0.404 0.258 0.0005 0.016 0.153 0.517 0.069 0.031 0.900 0.338 0.336

32

1995-1996 0.185 0.232 0.379 0.087 0.139 0.174 0.509 0.200 0.274 0.469 0.128 0.424 0.350 0.424 0.311

1997-1998 0.175 0.210 0.123 0.136 0.156 0.202 0.206 0.063 0.218 0.336 0.081 0.256 0.484 0.003 0.160

1999-2000 0.123 0.144 0.240 0.041 0.098 0.639 -0.004 -0.035 0.063 0.117 0.030 -0.009 0.185 0.112 0.368

Table 7 Summary statistics for a 12 month moving standard deviation of the difference between local IP growth and EU-15 IP growth (excl. country under investigation) (%) This table reports summary statistics for a business cycle convergence indicator based on growth rates in industrial production. The variable reported here is a 12 month moving standard deviation of the difference between local IP growth rates and the aggregate IP growth rate, excluding the country under investigation. The aggregate IP growth rate is calculated as a GDP-weighted average of the IP growth rates of the individual countries, but excluding the country under investigation. The weights used are previous year’s percentage of a country’s GDP in the aggreage GDP. Column 2 reports an average over the full sample (January 1990 - December 2000), while columns 3 till 7 report averages over different subperiods.

Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden UK

Stdev of differences between local IP growth and EU-15 IP growth (%) Full Sample 2.466 3.908 2.220 1.147 1.561 3.902 2.122 2.597 1.320 2.646 2.553 3.541 2.596 0.940

1990-1992 2.061 3.562 2.527 1.155 1.449 3.350 2.127 3.794 1.408 2.764 2.206 4.836 3.140 1.052

1993-1994 3.551 4.072 2.205 0.895 1.714 3.225 1.916 2.352 1.110 3.024 2.504 3.543 2.270 0.890

33

1995-1996 1.886 4.655 1.396 1.275 1.681 3.944 2.321 2.328 1.118 3.231 2.483 2.360 2.550 0.812

1997-1998 1.387 3.914 1.812 0.890 1.460 4.268 1.457 1.807 1.503 1.756 2.928 2.324 3.039 0.930

1999-2000 2.155 3.295 2.739 0.859 1.161 5.146 2.447 1.979 1.443 1.496 3.222 2.314 2.585 1.010

Table 8 Summary statistics for moving correlations between local IP growth and EU-15 IP growth ( excl. country under investigation) (%) This table reports summary statistics for a business cycle convergence indicator based on growth rates in industrial production. The variable reported here is a 12 month moving correlation between local IP growth rates and the aggregate IP growth rate, excluding the country under investigation. The aggregate IP growth rate is calculated as a GDP-weighted average of the IP growth rates of the individual countries, but excluding the country under investigation. The weights used are previous year’s percentage of a country’s GDP in the aggreage GDP. Column 2 reports an average over the full sample (January 1990 December 2000), while columns 3 till 7 report averages over different subperiods. Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden UK

correlation betw local IP growth and EU-15 IP growth excl local country(%) Full Sample 21.35 12.11 0.66 24.59 15.30 13.95 23.17 15.73 20.38 18.73 17.75 23.53 -9.75 30.16

1990-1992 13.77 14.04 -20.49 18.76 29.38 23.82 13.75 4.26 13.23 NA 1.17 24.61 -8.05 28.03

1993-1994 18.46 -21.96 -27.87 35.09 -40.02 49.82 11.00 39.20 19.64 13.27 16.75 -31.75 -0.44 6.67

34

1995-1996 -9.85 -18.84 -6.79 -3.13 7.82 13.94 11.95 37.20 43.22 -9.88 23.76 33.98 7.28 59.89

1997-1998 48.86 38.08 62.93 40.69 30.12 -19.97 27.44 -14.44 19.66 40.99 27.08 39.47 -13.34 23.56

1999-2000 43.55 48.53 7.44 51.22 48.03 1.33 50.43 28.78 43.20 40.76 33.33 32.66 -31.14 24.34

Table 9 Estimation results for EU-15 prices of market and currency risk

Panel A & B report results from estimating the quantities and prices of EU-15 market risk and currency risk. The estimated model corresponds to equations (7), (8), (10),(11), and (12) described in the paper. Panel C reports some tests on the parameters and residuals. The EU15 market index is calculated as a capitalization-weighted average of the total returns of the 15 individual country indices. The aggregate currency return is calculated as a trade-weighted average of deviations from uncovered interest parity for the Belgian Franc, the Dutch Guilder, the French Frank, the Italian Liara, the Spanish Peseta, and the British Pound vis-a-vis the Deutschmark. The instruments related to the price of market risk are a constant, the first lag of the EU-15 index dividend yield in excess of the one-month euro-DM deposit rate, the first lag of the change in the term structure, as well as the first and second lag in the EU-15 market returns. For the price of currency risk, the same instruments are used as to quantify the price of market risk, with the exception of the lags wich have been replaced by the first and second lag of the aggregate currency returns. The p-values are calculated using quasi-maximum likelihood standard errors. Panel C reports some diagnostics on ¯ 2 is the adjusted R2 , the error terms, as well as some parameter restriction tests. The R PRICE tests whether the prices of market and currency risk are significantly different from zero, and is distributed as a χ2 distribution with 5 degrees of freedom. MEAN investigates whether the first four autocorrelations of the standardized residuals are equal to zero. VARIANCE checks whether there is any autocorrelation left (until the fourth order) in the squared residuals. Finally, DISTRIB tests whether the standardized residuals are normally distributed or not. All test statistics are obtained through a GMM test procedure (see section 4.2.) and distributed as a χ2 distribution with 4 degrees of freedom. Finally, DIAG tests whether the diagonal elements of the matrices A and B (results not reported here) are significantly different from zero, and ASYM investigates whether the elements adding asymmetry to the variance equation are significantly different from zero (see equation (10)). Both test statistics are distributed as a χ2 distribution with 4 degrees of freedom.

35

Table 9 (continued) Panel A: Estimation Results for Conditional Variance and Covariance

Estim. p-val

c11

c21

c22

α11

α22

b11

b22

d11

d12

d21

d22

0.003 [0.00]

1e-04 [0.00]

0.0043 [0.00]

0.309 [0.00]

0.221 [0.00]

0.908 [0.00]

0.927 [0.00]

0.222 [0.04]

0.014 [0.30]

0.833 [0.00]

0.213 [0.04]

Panel B: Estimates for conditional mean model

Estim. p-val

λEU,0

λEU,1

λEU,2

λEU,3

λEU,4

λC,0

λC,1

λC,2

λC,3

λC,4

0.973 [0.00]

0.271 [0.00]

-6.35 [0.05]

-16.84 [0.00]

17.26 [0.04]

29.93 [0.21]

8.718 [0.13]

-48.67 [0.35]

-33.56 [0.37]

-54.19 [0.39]

Panel C: Residual Analysis

EU-15 C

¯ 2 (%) R

PRICE

MEAN

VARIANCE

DISTRIB

3.72 4.95

[0.00] [0.57]

[0.09] [0.19]

[0.49] [0.58]

[0.07] [0.23]

36

W p-val

DIAG

ASYM

2.405 [0.66]

53.86 [0.00]

Table 10 Time-Varying Integration

This table reports results for the time-varying integration model as defined in equation (5). Table A reports model estimates for the parameters linking local instruments to the local price of market risk; The instruments used are a constant, the first lag of the local market index dividend yield in excess of the local market one-month deposit rate, the first lag in the local short-term interest rate, the first lag of the change in local market term structure, and the first and second lag of the local market excess return. All data are taken from Datastream. Panel A reports estimates for the integration instruments. The integration instruments used here are a constant, the 12 moving average of differences between local inflation and weighted EU-15 inflation (MP1), the 12 month moving correlation between local industrial production growth rates and the aggregate EU-15 IP growth rate (CorIP), and a 52 week moving standard deviation of the local currency returns against the Deutschmark (CurVol). Panel C reports some diagnostics on the error terms, as well ¯ 2 is the adjusted R2 , PRICE tests as some parameter restriction tests. The R whether the prices of local market risk is significantly different from zero, and is distributed as a χ2 distribution with 5 degrees of freedom. MEAN investigates whether the first four autocorrelations of the standardized residuals are equal to zero. VARIANCE checks whether there is any autocorrelation left (until the fourth order) in the squared residuals. Finally, DISTRIB tests whether the standardized residuals are normally distributed or not. All test statistics are obtained through a GMM test procedure (see section 4.2.) and distributed as a χ2 distribution with 4 degrees of freedom. Finally, DIAG tests whether the diagonal elements of the matrices A and B (results not reported here) are significantly different from zero, and ASYM investigates whether the elements adding asymmetry to the variance equation are significantly different from zero (see equation (10)). Both test statistics are distributed as a χ2 distribution with 4 degrees of freedom.

37

Table 10 (continued) Panel A Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden UK

Local Instruments term 1mir

constant

dy1mir

lag1

lag2

9.65∗∗∗

73.65∗∗

20.42∗∗

-18.26

-38.44∗

19.68

2.361∗

51.361∗∗

-1.012

-8.324∗

-3.257∗

1.627∗∗

11.915∗∗

41.319∗

3.151

-4.395

1.601∗∗

2.019

5.598∗∗

35.391∗∗∗

-2.567∗∗

-12.982

-24.361∗

15.367

Panel B

Countries constant Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden UK

Integration Instruments MP1 corIP CurVol

4.856∗∗

-5.585∗∗

0.414

-1098.22∗∗∗

-1.934∗∗

-16.737∗

30.386

-3902.90∗∗∗

-0.0167

-14.537∗∗∗

12.001

-2502.31∗∗∗

0.715∗∗

-6.504

6.520

-4455.90∗∗∗

38

Table 10 (continued) Panel C Countries Austria Belgium Finland France Germany Ireland Italy Netherlands Spain Portugal Denmark Norway Sweden UK

¯2 R

MEAN

VARIANCE

39

DISTRIB

PRICE

ASYM

DIAG