عوامل موثر بر بازده سهام در یک اقتصاد کوچک باز

This paper examines the determinants of stock returns in a small open economy using an APT framework. We focus on the Swiss stock market whose feature is to include a large proportion of firms that are exposed to foreign economic conditions. Both a statistical and a macroeconomic implementation of the model are performed on industrial sector indices for the period 1986–2002. It is found that statistical factors yield a better representation of the determinants of stock returns than macroeconomic variables. Stock returns are influenced by both global and local economic conditions, suggesting that the Swiss market is an internationally imperfectly integrated market.

Identifying the forces that drive stock returns is a major concern for practice and academic research. Financial theory provides several asset pricing models that relate expected returns to one or several variables representing various sources of risk. The identity of these variables depends on the assumptions on which the model is built. The most popular asset pricing models are the Capital Asset Pricing Model (CAPM, one source of risk) and the Arbitrage Pricing Theory (APT, several sources of risk). Such models are used, e.g., to assess the performance of managed funds or measure the cost of capital. Early versions of these models were developed under the assumption that investors have access to domestic securities only. This is a reasonable assumption if agents live in a closed economy or if a given country's financial market is totally segmented from other markets. These models have been tested extensively in the financial economics literature, but tests of the CAPM are at best unconvincing, and several anomalies have been reported. As factors are not explicitly specified by theory, two empirical versions of the APT have been implemented: factors are either extracted by means of statistical techniques or are prespecified. In their seminal paper, Chen, Roll, and Ross (1986) consider the influence of a set of six prespecified macroeconomic U.S. variables and find that three such factors are priced for the U.S. stock market. A number of authors have taken the same approach for various stock markets (e.g., Hamao, 1988, for Japan; Antoniou, Garrett, & Priestley, 1998, for UK). In another class of models, the assumption of investment being solely domestic is relaxed. In such a context, markets are assumed to be perfectly integrated because of the presence of arbitrageurs that trade stocks internationally. Such trading equalizes the price of stocks with the same payoffs across markets. This has led to the extension of domestic pricing models to international models, such as the various versions of the international CAPM or the international APT. Typically, the world market portfolio and the foreign exchange risk are considered as global sources of risk in the international CAPM (e.g., Korajczyk & Viallet, 1989), whereas various global factors are hypothesized to impact on stock prices in the international APT (e.g., Ferson & Harvey, 1994). Empirical evidence pertaining to these models is mixed (for surveys, see Heston et al., 1995 and Karolyi & Stulz, 2003). As such models rely on the joint hypothesis of the validity of the model and of the perfect integration of international stock markets, it is impossible, however, to attribute rejection to any one of the two assumptions. In an era of increasing globalization, it seems reasonable to assume that most developed markets would be integrated. The empirical international asset pricing literature suggests that this is unlikely to be the case, and that most markets are in fact imperfectly integrated. For instance, country effects have been shown to dominate industry effects (Heston & Rouwenhorst, 1994). Further, Griffin and Karolyi (1998) find that industries with internationally traded goods are more sensitive to global industry factors than firms that produce goods that are only domestically traded. Fedorov and Sarkissian (2000) and Griffin and Stulz (2001) also conclude that companies or industries with internationally traded goods are more integrated due to the cash flows of such firms being more sensitive to global factors. Finally, there is clear evidence that investors do not diversify their portfolios internationally as much as is suggested by portfolio theory. This phenomenon is known as the home bias and is a well-known anomaly in the international finance literature (for a review, see Lewis, 1999). This bias is the sign of the existence of market imperfections that prevent investors from diversifying their portfolio in an optimal way. Besides barriers to international investments, there are also additional costs related to such investments. These results constitute evidence against the assumption of perfect international market integration. In this case, pure international asset pricing models may not constitute a good representation of reality. Given this, theoretical asset pricing models assuming partial integration would appear to be better suited to explain stock returns. Such models provide a pricing equation both for securities that can and for securities that cannot be held by foreign investors Cooper & Kaplanis, 2000, Errunza & Losq, 1985 and Hietala, 1989. However, these models neither offer any indication on how to measure the level of integration of a specific market, nor do they provide a general equilibrium relationship that can be used for international asset pricing. Given the lack of theoretical guidance, some authors have used a more empirical approach in that they include both local and global variables in the pricing equation; the relative importance of such factors being weighted by the degree of integration of the market Bekaert & Harvey, 1995 and Hardouvelis et al., 2002. Yet another approach that has been used is to add a single international variable (exports, a world index or the foreign exchange rate) to a domestic APT model with prespecified macroeconomic variables. Examples include Martikainen, Yli-Olli, and Gunasekaran (1991) for Finland, Kryzanowski and Zhang (1992) for Canada, Clare and Thomas (1994) for the UK, Kaneko and Lee (1995) for Japan, Groenewold and Fraser (1997) for Australia, and Clare and Priestley (1998) and Bilson, Brailsford, and Hooper (2001) for emerging markets. We are aware of one study which has used several prespecified local and global factors, but for emerging markets only (Rendu de Lint, 2002). Following up on this literature, this paper assumes explicitly that the determinants of stock returns can be either local, global, or a combination of both. As risk is likely to be multidimensional, the aim of this paper is to identify empirically the determinants of stock returns by using techniques that have been devised to implement and test the APT. This paper focuses on the Swiss stock market for the following important reason. As can be seen from Table 1, Switzerland belongs to a group of countries with developed stock markets that have a very large fraction of their quoted companies doing business abroad. Among these, only the Netherlands, Hong Kong, Finland, and Sweden have a larger exposition to international trade. Switzerland is the ninth largest stock market in the world and more than 80% of its companies have overseas sales. More than 50% of sales are foreign sales on average. This is in sharp contrast with results for the United States, for instance, where these figures amount to 31% and 10%, respectively. Additionally, there are no limitations to foreign ownership of stocks of Swiss companies. This would clearly speak for the integration of the Swiss market. Despite the international dimension of the Swiss stock market, it is one of the few developed markets that have been found not to be integrated (Heston et al., 1995). This is puzzling and the Swiss market is therefore an ideal candidate for using an approach with both local and global factors. Moreover, the Swiss market has been the focus of limited research only. The extant literature for Switzerland suffers from drawbacks as it has considered domestic factors only Beckers et al., 1993, Cuenot & Reyes, 1992 and Vessereau, 2000, or used a limited sample of stocks (Broillet, 1991), or a very short time period (Gallati, 1993). Table 1. Measures of market openness Country Market cap. (X+M)/Y Average foreign sales (%) % Firms # Firms 1. US 16635 0.29 10.19 31.43 6068 2. Japan 4547 0.20 15.15 59.69 707 3. UK 2933 0.74 24.45 56.15 1131 4. France 1475 0.56 36.96 77.67 430 5. Germany 1432 0.65 35.00 76.19 420 6. Canada 801 0.86 32.41 58.15 227 7. Italy 728 0.56 32.22 65.69 137 8. Netherlands 695 1.31 58.31 95.90 122 9. Switzerland 693 0.98 50.76 81.76 148 10. Hong Kong 609 2.89 52.88 91.59 321 11. Australia 478 0.45 19.68 52.87 174 12. Spain 432 0.63 24.00 65.67 67 13. Taiwan 376 1.06 17.12 47.42 213 14. Sweden 373 0.93 53.95 96.39 83 15. Finland 349 0.77 56.36 94.03 67 Market capitalizations are from Dimson, Marsh, and Staunton (2002), the macroeconomic data from Datastream Thomson Financial and the quoted company data from Thompson Analytics. The data are for year 2000. “Market cap.” gives the total market capitalizations in US$ billion. “(X+M)/Y” is the ratio of the sum of exports and imports of a country divided by its GDP. “Average foreign sales” represents the average ratio of overseas sales divided by total sales of a company for firms in a given country. “% firms” represents the percentage of firms that have sales abroad among the total number of companies in a country. “# firms” indicates the number of firms for which data on foreign sales are available in each country. Table options This paper makes the following contributions. First, it distinguishes itself from the extant literature by considering the influence on stock returns of both global and local variables for a developed market where most of the firms are internationally oriented. Second, the determinants are obtained from a broad set of variables representing macroeconomic influences on financial markets. These variables represent the evolution of economic conditions in Switzerland and in countries of the G7, the major trade partners of Switzerland. They can be classified in the following broad groups: business cycle, inflation, interest rates, and financial markets. These variables are all related to expected cash flows and/or discount rates, and therefore to stock prices. The selection of variables is made endogenously, starting from a large set of potential determinants, and then finding the best set of variables by means of cluster analysis. From a methodological point of view, the paper also provides significant improvements. It extends the methodology of Pettengill, Sundaram, and Mathur (1995) to a multifactor setting to assess the significance of the risk premia both for positive and negative occurrences of the factors. As our paper is set in an APT framework, variables should be represented by innovations. Among the methods used to determine such innovations, the most popular ones are the first differences and the ARIMA-type adjustments. We use an alternative approach in that we compute innovations using Kalman filters. Priestley (1996) shows that such an approach leads to more reliable inferences regarding tests and applications of the APT. Finally, the best set of macroeconomic variables is compared to statistical factors. The latter are extracted using a technique recently proposed by Xu (2003) that assumes heteroskedasticity both in time series and in cross section. The empirical investigation is conducted on a set of portfolios representing the industrial sectors of the Swiss stock market to avoid the noise associated with individual stocks and to capture the major macroeconomic influences on the stock market. The results of our paper confirm our hypothesis that the Swiss stock market is influenced by both local and global factors. Four macroeconomic variables emerge from the analysis, two of which being related to global economic conditions and two reflecting domestic influences. A pure statistical approach yields five factors with a high explanatory power. These factors appear to be related to both global and local macroeconomic variables, providing further support for the partial integration of the Swiss stock market. This result has important practical implications in that both types of factors should be considered, e.g., for hedging purposes, performance measurement and cost of capital computations. The remainder of the paper is organized as follows. Section 2 presents the method used, while Section 3 introduces our data. Section 4 presents the results of our empirical analysis. Finally, Section 5 contains some concluding remarks.

This paper provides an analysis of the determinants of stock returns in a small open economy in an APT framework. The empirical investigation is conducted for the Swiss stock market which has the particularity of including a large proportion of firms that are exposed to economic conditions prevailing outside the country as they sell and purchase their products and services overseas. As it is a developed market with no barriers to international investments, it could be considered as a market that is integrated with the rest of the world. However, this market has been found not to be integrated in previous literature. For this reason, we include both local and global variables in the set of potential macroeconomic explanatory variables. The global variables are aggregates constructed from countries of the G7, the main trade partners of Switzerland. Two types of implementations of the model are investigated and compared: a statistical one and a macroeconomic one. We use monthly returns on 19 industrial sector portfolios over the period 1986–2002. The statistical implementation of the model yields five factors. The best macroeconomic version includes four variables, two of which are clearly linked to global economic conditions (innovations to G7 industrial production and changes in expected inflation) while two are linked to local factors (Swiss market return and innovations to the term structure). Interestingly, all belong to one of our four broad categories of variables: the general level of economic activity, price levels, credit conditions, and the stock market environment. These results confirm the identity of the relevant factors chosen by Chen et al. (1986) for the U.S. market, but emphasize the importance of international influences on the Swiss market. The two-pass standard FM tests show that neither the statistical nor the macroeconomic version displays significant relations between risk and return. However, when positive and negative realizations of the factors are taken into account, the risk–return relationship becomes highly significant for the statistical model, but only weakly for the macroeconomic model. This result clearly shows that the statistically determined factors yield a better representation of the determinants of stock returns than the macroeconomic variables. This is confirmed by formal comparisons of the explanatory power of both types of models. Finally, an analysis of the links existing between risk premia generated by the statistical model and those of the macroeconomic model shows that both types of premia are significantly related. However, the macroeconomic risk premia explain at best 65% of the variance of statistical risk premia, which suggests that other forces are at work. Their precise identification is left for further research. This paper has important implications for both researchers and practitioners. From a research point of view, this paper stresses the urgent need to develop a theoretical model of asset pricing of partial integration that explicitly defines the level of integration of a market and determines the relevant set of variables to represent the sources of risk. From a practical point of view, too, it has several implications. First, it shows that when managing a portfolio that includes Swiss equities, fund managers should not only take into account the sensitivity of stocks to the Swiss market, but also the sensitivities to innovations in the Swiss term structure, G7 industrial production and to changes in expected inflation in the G7. These factors should be borne in mind when devising hedging strategies. Second, when assessing the performance of a portfolio of Swiss stocks, the investors should take into account these additional risk factors. Finally, when selecting upon alternative investment projects, the calculation of the cost of capital for Swiss companies should consider these factors. Omitting these variables could yield seriously biased results and lead to erroneous investment decisions.