Dynamic linkages among equity markets: local versus basket currencies

Abstract

This article compares the stability of interrelationships among nine national equity markets by employing basket currencies versus individual base currencies to measure rates of return. We show that the average correlations of equity indexes based on basket currencies are much lower than the average correlations of equity indexes based on individual home base currencies. Additionally, empirical directed graph results from using basket currencies are more consistent with each other than the results from using different individual currencies. We conclude that basket currencies provide more robust results for studying equity market comovements than individual currencies. By implication, studies on equity market interrelationships, including studies on potential financial contagion between equity markets, should use basket currencies to avoid currency dependent results.

Notes

¹ In this regard, studies by Jorion (1990, 1991), Amihud (1993), Bodnar and Gentry (1993), Heston and Rouwenhorst (1994), and others find little or no relationship between equity returns and exchange rate risk. However, Roll (1992), Dumas and Solnik (1995), Bartov et al. (1996), Chow et al. (1997), De Santis and Gérard (1998), He and Ng (1998), Patro et al. (2002) and Bodnar and Wong (2003) find evidence that equity returns are influenced by exchange rates (see also Eun and Resnick (1988) and Kaplanis and Schaefer (1991) concerning exchange risk and diversification effects).

² To conserve space we report only six directed graphs based on selected base currencies. Results for other currencies are available upon request from the authors.

³ Due to the time overlap between the European and North American markets, we also test the pattern of causal flow by imposing the assumptions that innovations in both of the European and North American markets can cause one another, but they cannot cause the innovations in Asia-Pacific markets. To save space, we do not report the directed graph results based on these restrictions. In terms of the patterns of causal flow, the results are similar to those in Fig. 3. For instance, compared to Fig. 3(a), this newly restricted directed graph based on SAC shows that Switzerland and Germany now cause UK, but Hong Kong no longer causes UK. These results are available from authors if requested.

⁴ In Fig. 3(a), the PC algorithm was not able to sort out the causal flow between UK−Germany, UK−Switzerland, Germany−Switzerland or Japan−Hong Kong. Similarly, in Fig. 3(b) and (c), the algorithm could not direct the edges between Germany−Switzerland, Germany−Italy, Switzerland−Italy or Japan−Hong Kong. A direct scoring is used to replace these undirected edges with the directed ones. The directed edges are scored using a modified Schwarz-loss metric (Haigh and Bessler, 2004). Based on the directed graphs scored using the Schwarz-loss metric and a Bernanke ordering, decompositions of forecast error variance of each country are computed.