Can risk-rebalancing explain the negative correlation between stock return differential and currency? Or, does source status drive it?

Highlights

• Hau and Rey׳s (2006) empirical evidence is not sufficient to support their risk-rebalancing hypothesis to explain negative stock return differential–currency correlation.

• Combination of home-wealth rebalancing and extrapolative expectations accounts for this correlation conditional on source status.

• We define source status as the relative activity of outbound and inbound international capital flows.

• Source status is a main predictor of the cross-section of stock return differential–currency correlation

• We also provide support for Filipe׳s (2012) relative volatility as a main predictor.

Abstract

We show that Hau and Rey׳s (2006) empirical evidence is not sufficient to support their risk-rebalancing theory as an explanation for the negative correlation between the stock market return differential and currency. A simple model combining home-wealth rebalancing and extrapolative expectations on the foreign stock predicts this negative correlation only when the host market is a source of international capital. Panel regressions indicate that the source status of the economy (i.e., whether it is a net receiver or source of international capital) is a main predictor of the stock return differential–currency correlation.

Introduction

Hau and Rey (2006, HR hereafter) present a model of joint determination of stock market returns, exchange rates, and equity portfolio flows, and provide empirical evidence that both suggest a negative correlation between the host currency and the return differential of the host stock market over the U.S. stock market. This negative correlation “contradicts the conventional wisdom that a strong equity market comes with a strong currency.”¹ The key mechanism that brings about such negative correlation in HR model is portfolio rebalancing by international equity investors to manage exchange rate risk. Accordingly, high returns in the host stock market should be associated with a depreciation of the host currency against the US dollar, since the consequent portfolio rebalancing by US investors to bring portfolio weights back to optimal levels will require selling the host currency. This risk-rebalancing behavior, combined with inelastic forex supply, generates the observed negative correlation, which has been termed “uncovered equity parity (UEP)” in recent literature (Capiello and De Santis, 2005, Kim, 2011, Curcuru et al., 2014).

We first re-examine the empirical work provided by HR, which consists of a battery of contemporaneous correlations. HR provide evidence of a negative correlation between return differentials of the host stock market over the U.S. market and the host currency (vis-à-vis U.S. dollar), and a positive correlation between net bilateral equity portfolio flows toward the host market and the host currency. However, they do not provide any evidence that U.S. equity investors negatively respond to return differentials of the host stock market, although risk-rebalancing is the key mechanism in their model. By examining this missing link, we assess whether the surprising negative correlation between the currency and the stock market return differential can really be explained by the price pressure resulting from rebalancing flows of unhedged equity portfolio investors. Using their data set and approach, we cannot find any evidence of rebalancing behavior by U.S. investors at the aggregate level on either HR׳s sample or on a more recent and larger sample. Instead, U.S. investor flows are usually positively associated with host market return differentials.

HR model predicts negative net bilateral flows in response to positive differentials of the host stock market return, thus it predicts the return differential–currency correlation to be always negative provided that equity portfolio investments are of significant size. However, the results from our extended sample suggest that this correlation is not uniformly negative: rather, it displays large variation across countries, and is significantly positive for most emerging markets, including those which attract large amounts of international equity flows. A simple model with home bias, home-wealth rebalancing, and extrapolative expectations on the foreign stock predicts a negative (positive) stock return differential–net bilateral equity flow correlation when the host market is a source (receiver) of international capital flows. Thus, it can account for the cross-section of the stock return differential–currency correlation, under HR׳s assumption of forex pressure of equity flows. In this model, home-wealth rebalancing induces a negative correlation, while extrapolative expectations of U.S. investors induce a positive correlation between the return differential and the host currency.

Within the context of this model, we propose the source status of an economy as a driver of the stock return differential–currency correlation. We define source status as an indicator of whether an economy has recently been more active as a source or receiver of international capital flows; it captures not only external balance but also host market characteristics, such as information asymmetry, relative riskiness, presence/absence of profitable investment opportunities, wealth, and market size. When a host country is a ‘receiver’, home-wealth rebalancing flows of host-country investors following a host market return shock will be small relative to inbound flows of U.S. investors positively (extrapolatively) responding to host country news, leading to a positive return differential–currency correlation. When the host country is a net supplier of international capital flows, host country investors׳ home-wealth rebalancing flows will dominate and lead to a negative correlation. This explanation is consistent with the observed variation in the stock market return differential–currency correlation: on a comprehensive panel, we document that our source status variable is a main driver of the variation in the return differential–currency correlation. HR׳s findings seem to be driven, at least partly, by their sample tilted towards source countries.

The impact of equity portfolio reallocation on exchange rate dynamics has been a recent focus in the international finance literature (e.g., Ding and Ma, 2013). Chang (2013) examines whether portfolio rebalancing can account for the forward premium puzzle. Two paths of this literature focus on HR׳s risk-rebalancing hypothesis. The first one attempts to refine the UEP: Chaban (2009) shows that HR׳s portfolio rebalancing story is not supported for commodity currencies. Kim (2011) attempts to explain violations of UEP with equity market risk. Filipe (2012) presents a model that can explain why portfolio rebalancing is not important for commodity currencies, or more generally, for currencies with higher fundamental volatility. The second path, which has the same motivation as the current study, addresses the missing link in HR׳s empirical evidence: In their more recent work, Hau and Rey (2009) provide evidence of rebalancing behavior using fund-level data at semi-annual frequency. Gyntelberg, Loretan, Subhanij, and Chan (2014) examine HR׳s hypotheses using data from Thailand. Ülkü and Karpova (2014) test the risk-rebalancing hypothesis using data from Greece. In a closely related work, Curcuru, Thomas, Warnock, and Wongswan (2014) employ a portfolio-based method and Treasury Capital International (TIC) bilateral transactions data corrected for financial center bias to search for evidence of rebalancing behavior.

The first contribution of the current article to this literature is an empirical re-examination of the role of alleged risk-rebalancing in driving the currency. For this purpose, we first stick to HR׳s data set and methodology, which follows from their model implications, and address, on an extended sample, the link left missing in Hau and Rey (2006). We find that the (supposedly positive) relation between U.S. investor flows and exchange rate changes is not robust across subperiods. More importantly, the contemporaneous response of U.S. investor flows to host return differentials is positive, instead of negative. Thus, risk-rebalancing is unlikely to explain the variation in the stock return differential–currency correlation in HR׳s data. Further analysis shows that U.S. investors display an extrapolative response to host market returns, in line with the models of Brennan and Cao (1997) and Griffin, Nardari, and Stulz (2004). Foreign-country investors׳ behavior appears to be characterized by rebalancing with respect to home wealth, in line with the model of Griffin, Nardari, and Stulz (2004).

Our second contribution is to introduce source status as a driver of portfolio flows׳ role in shaping the return differential–host currency correlation. Within the context of a simple model with the features of the above-mentioned models, but without risk-rebalancing, source status provides an intuitive explanation for the observed cross-sectional variation in the stock return differential–currency correlation. Panel regressions confirm source status as a significant driver of the stock return differential–currency correlation. They also provide strong support for Filipe׳s (2012) model variable, the fundamental volatility of the host market.

In Section 2, we reassess HR׳s empirical work and address the missing link. In Section 3, source status is introduced as a driver of the stock return differential–currency correlation. Then, a comprehensive panel is used to empirically investigate drivers of this correlation. Conclusions are summarized in Section 4.