The second hypothesis concerns cross-predictability of stock returns in a cross-country sample. Based on the objectives stated in subsection 1.1 this paper aims to test whether the limited- information models appear valid in a cross-country sample. In order to reach this objective, this paper examines cross-predictability effects using data on the Eurozone and EU27 countries. Even though the focus is on the limited-information model by Menzly and Ozbas (2010), it should be noted that performing an out-of-sample test of the cross-predictability effect provides insight on the validity of the limited-information models and information diffusion hypothesis in general.
As mentioned earlier, the limited-information model by MO (2010) posits that there are two types of investors, informed and uninformed. Informed investors specialize in one market for which they receive informative signals about the eventual cash flows on assets whereas uninformed investors do not receive any informative signals. In addition, at least some uninformed investors fail to process information due to either limited information-processing capabilities or costs to processing information. The interaction of these two investors generates return cross-predictability between economically linked assets based on publicly available information. In their article, MO provide empirical evidence on their model’s predictions by showing that lagged returns in supplier and customer industries cross-predict stock (industry) returns. Furthermore, they show that this cross- predictability effect is economically significant and lasts from one month up to 12 months. More empirical evidence on return cross-predictability is provided by Hong, Torous and Valkanov (2007) who show that several important industries predict the movements of stock markets in multiple countries. Consistent with their model, HTV also document that an industry’s predictive ability is
strongly correlated with its propensity to forecast indicators of economic activity.30
The above findings suggest that cross-predictability effects should exist, as predicted by the limited- information models, also in non-US samples. However, the use of European cross-country data in this paper may induce differences compared to the previous studies. As noted earlier, national borders are potentially a significant factor that may impact the correlation of industry fundamentals
30 For a detailed description of the limited-information models by Hong, Torous and Valkanov (2007) and Menzly and Ozbas (2010),
across countries. In addition, to the possible effect of national borders on industry correlations, borders may also have an impact on the way how price innovations are transmitted across industries in different countries. In other words, the cross-predictability effect in this study may differ from the one reported in earlier studies due to country-specific factors. Relevant country-specific factors that may affect stock returns are, for example, differences in transaction costs across stock exchanges and differences in information costs.
Academic literature on stock market comovements has documented that equity markets of many developed and economically integrated economies move to a larger extent independently of each
other in terms of returns and volatility.31 According to Rouwenhorst (1999b), the explanations for
the low correlations between country index returns can be divided into three groups: (i) home bias, (ii) country effects and (iii) industry effects. The first explanation attributes the low correlations to investors’ tendency to hold disproportionately large amounts of domestic shares in their portfolios which may cause country portfolios to reflect, at least partly, the different sentiments of domestic investors. The second explanation states that country-specific factors such as regional economic shocks, local monetary and fiscal policies and differences in institutional structures cause global economic shocks to have different effects on companies in different countries. In other words, there are country-specific factors that drive equity returns in different countries. The third explanation argues that the low country correlations are driven by differences in industrial structures between countries rather than country-specific differences. For example, Switzerland has a large banking sector and thus, the Swiss country index is imperfectly correlated with the Swedish stock market index which is heavy on the basic industries. (Heston and Rouwenhorst, 1994.)
In a sense, the stock market comovement literature is closely related to the business cycle literature discussed in the previous section. However, as the analysis in the stock market comovement literature is performed on the level of stock returns, reviewing it is important since there are several forces, other than the business cycle, that may affect stock returns across countries. Thus, while providing evidence on the correlation of industry fundamentals may satisfy the limited-information model assumption, the observed cross-predictability effect may still be different in a cross-country sample as compared to a single country sample. For example, if country effects dominate industry effects or if differences in investor sentiment drive a wedge between the returns of companies that are in the same industry but located in different countries, the cross-predictability effect may be weaker in a cross-country sample. This point is highlighted by Heston and Rouwenhorst (1994)
31 Grubel (1968), Levy and Sarnat (1970) and Solnik (1974) were among the first to document the low correlations between stock
who note that country-specific components of return variation may still dominate any industry effects even though the actual correlation between industries might be high.
A vast amount of empirical evidence from stock market comovement literature suggests that industry effects play a little role in explaining the low country correlations of stock returns (see e.g., Heston and Rouwenhorst, 1994; Beckers, et al., 1992). For example, Heston and Rouwenhorst (1994) study the European stock markets and find that country effects dominate industry effects even in European countries that are economically strongly integrated. Even though the industry portfolios in their sample are strongly positively correlated, the industrial structure explains only a small portion of the correlation of country index returns. Griffin and Karolyi (1998) later confirm the findings by Heston and Rouwenhorst by showing that less than 4% of the variation in country index returns can be attributed to their industrial composition. Furthermore, Griffin and Karolyi also document cross-sectional differences in the variances of industry effects for industry indexes. More specifically, they find that traded goods-industries tend to have higher industry effects than
nontraded-goods industries.32 In addition, Bekaert, Hodrick and Zhang (2009) study international
asset return comovements by using a linear factor model which they argue is a better model than the widely-applied Heston–Rouwenhorst (1994) dummy variable model. Also their results suggest that country factors dominate industry factors in Europe.
It is worth noting that the empirical evidence on country effects’ dominance on international stock return comovements is not unanimous. Roll (1992) was the first one to provide evidence that a significant part of the international stock market comovement can be explained by the industry compositions of the national stock market indices. Although his results have been questioned in later studies, there exists a number of more recent articles which claim that industry factors have become more dominant (see e.g,, Cavaglia, Brightman and Aked, 2000; and Baca, Garbe, and Weiss, 2000). For example, Ferreira and Ferreira (2006) study the Economic and Monetary Union countries over the period 1975 to 2001 and find evidence that industrial effects have become similar in magnitude to country effects in the post-euro period. They also document similar results for non- EMU countries. Also Brooks and Del Negro (2004) provide evidence that industry effects have become increasingly important in Europe. To summarize, based on the earlier research, it remains unclear whether national borders have an impact on the cross-predictability of stock returns.
32 Traded-goods industries are industries that produce goods which are traded internationally such as automobiles, computers, office
equipment, pharmaceuticals and semi-conductors. Nontraded-goods industries are industries that produce goods which are not traded internationally such as media, heavy construction, plantations, conglomerates and real estate. (Griffin and Karolyi, 1998.)
In addition to country and industry effects, researchers have also studied the exposure of the European equity markets to the US equity market. For example, Baele (2005) investigates the magnitude and the time-varying nature of volatility spillovers from aggregate European and US equity market indices to 13 local European equity markets. He documents that, while the relative importance of the regional European market has increased, the US equity market continues to be the dominating influence in European equity markets. He also finds some evidence of contagion effects from the US market to several local European markets in times of high equity market volatility. In another study, Fratzcher (2002) provides evidence that while the USA is the dominant market outside the Eurozone, it is no longer the only dominant market within the Eurozone. His results indicate that the euro area market has become the dominant market for individual Euro area countries since the mid-1990s. The importance of the US equity market to the European equity market may influence the return cross-predictability results obtained in this paper, particularly if the US market has a varying impact on equity markets in different European countries. However, it is extremely difficult to predict the possible influence of the US equity market on the cross- predictability effect.
Based on the above literature and empirical evidence on the cross-predictability effect, the second hypothesis of this study is divided into two parts:
H2.1: Lagged supplier industry returns cross-predict stock (industry) returns in a cross-country European sample.
H2.2: Lagged customer industry returns cross-predict stock (industry) returns in a cross-country European sample.
3.3 HYPOTHESIS 3: EFFECT OF INFORMED INVESTORS AND GEOGRAPHIC