The cost of capital in international financial markets: local or global?

https://doi.org/10.1016/S0261-5606(02)00028-1Get rights and content

Abstract

This paper analyzes to what extent international and domestic asset pricing models lead to a different estimate of the cost of capital for an individual firm under the maintained assumption of perfect international financial integration. We distinguish between (i) the multifactor Solnik–Sercu ICAPM including both the global market portfolio and exchange rate risk premia, and (ii) the single factor domestic CAPM. We use a sample of 3,293 stocks from nine countries in the period 1980–1999. The domestic CAPM yields a significantly different estimate of the cost of capital from the multifactor ICAPM for only five percent of the firms in our sample. We attribute the close correspondence between local and global pricing to strong country factors in individual stock returns, which are probably due to lack of real integration. Our results reinforce the home bias puzzle.

Introduction

Theory suggests the use of an international CAPM (ICAPM) for computing a firm’s cost of capital in a financially integrated world. In practice, however, a wide variety of asset pricing models that ignore the international dimension is used to compute the cost of capital.1 This may, among other things, be related to the fact that even though the ICAPM is theoretically preferable to the domestic CAPM, a firm’s beta calculated using the domestic CAPM does not necessarily provide an incorrect estimate of the cost of capital. The two asset pricing models could lead to the same cost of capital if the local stock market portfolio contains all the information that is relevant in order to price domestic assets internationally.2

The purpose of this paper is to empirically examine whether international and domestic asset pricing models really lead to a different estimate of the cost of capital. A partial answer is given by Stulz (1995b), who derives an expression for the difference in the estimation of a firm’s beta when computed with the domestic CAPM as compared to the single factor ICAPM of Grauer et al. (1976). Stulz refers to this difference as the pricing error, which is linearly related to the computed cost of capital differential. Stulz uses data on the Swiss multinational Nestlé and finds a substantial pricing error.

We generalize the analysis of Stulz (1995b) in three ways. First, we employ the multifactor ICAPM of Solnik (1983) and Sercu (1980) including both the global market portfolio and exchange rate risk premia.3 Second, we derive statistical tests for the significance of the pricing error. Third, we use data on 3,293 stocks from nine different countries to investigate the difference between each of these models empirically.4 We analyze the sample period 1980:02–1999:06.

We find that the pricing error in terms of the cost of capital computed with either the domestic CAPM or the multifactor ICAPM of Solnik–Sercu is marginal. Only for about 5 percent of all firms in our sample the domestic CAPM yields a significantly different cost of capital than the multifactor ICAPM at the 95% confidence level. We show that the absolute difference in the cost of capital amounts to about 50 basis points for the US, about 75 basis points for Germany and Japan, and similar amounts for the other countries in our sample. We argue that our findings may be attributed to strong country factors in the individual stock returns, consistent with the evidence of Heston and Rouwenhorst (1994) and Griffin and Karolyi (1998). A tentative explanation of this finding is a lack of real capital market integration, due to both cyclical and structural, and institutional country-specific factors. These closely tie together the fortunes of all firms operating in the same country. The observed differences between countries can and should be used by individual investors for the purpose of portfolio diversification. Diversification across industries within one country is insufficient to cope with a country’s systemic risk according to our results. Our evidence reinforces the home bias puzzle.

Testing for a pricing error turns out to be very similar to testing for foreign exchange exposure. We show how both methodologies are related and how pricing error tests can shed light on the well-known puzzle that firms from a variety of data sets show little exposure to exchange rate fluctuations.5

The paper is set up as follows. In Section 2, we review the international CAPM and the domestic CAPM and derive testable hypotheses. In Section 3, the data are described and summary statistics are discussed. Empirical results are presented in Section 4. Section 5 explores the results using a variance decomposition technique. We elaborate on the link between the pricing error tests and the foreign exchange exposure literature in Section 6. Summary and conclusions are presented in Section 7.

Section snippets

The international CAPM and the domestic CAPM

In this section, we develop tests to evaluate whether the domestic CAPM leads to a different cost of capital than the multifactor ICAPM. The starting point is the Solnik–Sercu version of the multifactor ICAPM. Assume a world with N+1 countries (currencies). The ICAPM has N+1 systematic risk factors: the global market portfolio and N exchange rates.6

Data

In the empirical analysis we use monthly data for nine industrialized countries: Australia, Canada, France, Germany, Japan, the Netherlands, Switzerland, United Kingdom, and the United States. Nominal exchange rates for all countries are taken from the International Financial Statistics (IFS) tape (line ae). In the empirical application we consider the period 1980:02–1999:06.

The market weighted local equity indices and the market weighted global market index are from Morgan Stanley Capital

Empirical results

In this section, we discuss the main results we have obtained by applying the testing methodology introduced in Section 2 to the sample of 3,293 stocks. Throughout, we assume that the MSCI world and local indexes are good proxies for the global and local market portfolios, respectively. We apply the Pricing Error and Global Beta tests as discussed in Section 2 to each individual firm in order to assess the magnitude and significance of the pricing error made by the domestic CAPM as compared to

Local, global and currency factors: a variance decomposition

This section further explores our finding that the pricing error is rarely statistically significant in our sample of almost 3,300 international stocks. We investigate how much of the risk that is specific from a local country perspective is systematic from a global perspective. For this we use a variance decomposition metric that allows for an assessment of the respective contributions of the local market, the global market and the vector of exchange rate changes to an individual asset i’s

Foreign exchange exposure

In Section 2 we showed that in general testing for a pricing error can be implemented by examining the statistical significance of a set of instrumental variables in a time series regression of the stock return of an individual firm on an intercept and the domestic market return (see eq. (8)). These pricing error tests are very similar to the well-known tests for exchange rate exposure. In this section, we perform several exchange rate exposure tests and show that the results of Section 5 can

Conclusions

While theory suggests the use of an international CAPM in integrated capital markets, the domestic CAPM does not necessarily imply an incorrect estimate of the cost of capital. In this paper, we examine to what extent international and domestic asset pricing models imply a different estimate of the cost of capital for a sample of monthly data for 3,293 firms from nine major industrialized countries from 1980 to 1999. We distinguish between: (i) the multifactor ICAPM of Solnik–Sercu including

Acknowledgements

We thank Geert Bekaert, Frans de Roon, Casper de Vries, Bernard Dumas, Alex Lammertsma, Jim Lothian (the editor), Kate Phylaktis (the referee), Piet Sercu, René Stulz, Jean-Pierre Urbain, Tom Vinaimont, and seminar participants at Indiana University (Bloomington), Washington University (St. Louis), Catholic University of Louvain, University of Amsterdam, the Federal Reserve Bank of St. Louis, the 1999 Econometric Society European Meeting in Santiago de Compostela, the 1999 European Finance

References (32)

  • M. Adler et al.

    International portfolio choice and corporation finance: a synthesis

    Journal of Finance

    (1983)
  • M. Adler et al.

    Exposure to currency risk: definition and measurement

    Financial Management

    (1984)
  • Y. Amihud

    Evidence on exchange rates and valuation of equity shares

  • E. Bartov et al.

    Firm valuation, earnings expectations, and the exchange-rate exposure effect

    Journal of Finance

    (1994)
  • G. De Ménil

    Real capital market integration in the EU: how far has it gone? What will the effect of the Euro be? (with discussion)

    Economic Policy

    (1999)
  • B. Dumas et al.

    The world price of exchange rate risk

    Journal of Finance

    (1995)
  • Cited by (40)

    • From dotcom to Covid-19: A convergence analysis of Islamic investments

      2021, Journal of International Financial Markets, Institutions and Money
      Citation Excerpt :

      We assume that the Covid-19 financial crisis between February 2020 and July 2020. Several techniques may be used to infer a time-variant measure of interrelationship across markets, such as asymmetric dynamic conditional correlation (ADCC)-GARCH models (Gjika and Horváth, 2013), wavelet techniques (Aloui et al., 2015; Bodart & Candelon, 2009), BEKK models (Koedijk et al., 2002), and dynamic copulae (Bhatti & Nguyen, 2012; Jayech, 2016; Nguyen & Bhatti, 2012; Okimoto, 2008; Ye et al., 2012).13 The asymmetric dynamic conditional correlation (ADCC)-GARCH model (Cappiello et al., 2006) provides a robust analysis of time-varying linkages by allowing conditional asymmetries in both volatilities and correlations, and investigates the second-order moment dynamics of financial time-series, while overcoming the heteroscedasticity concern of Forbes and Rigobon (2002).

    • Country and industry factors in tests of Capital Asset Pricing Models for partially integrated emerging markets

      2020, Economic Modelling
      Citation Excerpt :

      Then in the second sub-sample from 1996, there is some degree of integration among ESMs which is better accounted for by GCAPM. The results lend support to the existing literature (e.g. Bruner et al., 2008; Koedijk et al., 2002). Furthermore, it is noticeable that the increase of relative importance of industry effects is much stronger for beta than for returns.

    • Which market integration measure?

      2017, Journal of Banking and Finance
    View all citing articles on Scopus
    View full text