A clear conclusion is that risk rebalancing is unlikely to unconditionally explain the variation in the stock return gap – the currency correlation in HR data. The conclusions of Curcuru et al. (2014) therefore pose a new challenge for the HR model, even if the response of US investors to the return shocks in the domestic markets is negative. First, the negative correlation between stock returns and the host currency in the HR model is concurrent, while the negative response of portfolio shifts documented by Curcuru et al.
US net purchases of foreign stocks, the key mechanism in HR's risk-rebalancing hypothesis, are not significantly correlated with exchange rate changes. In contrast to HR's sample period results, the correlations between bilateral net flows and changes in the host currency value are close to zero this time. HR's model relies on a chain of two links to create the observed negative correlation between stock market return differentials and foreign exchange returns: risk-rebalancing behavior and price pressure of equity-related flows in the foreign exchange market.
HR's Table 5 (our Table 1) supports only the second of these two necessary links (and still not fully, as our breakdown and out-of-sample results above show). Since HR's model implications involve contemporaneous relationships and their empirical work focuses on contemporaneous correlations, in this section we look at the contemporaneous relationships between stock market return differentials and net stock flows. Net bilateral flows' response to foreign stock return differentials is insignificant as in HR's sample period, as the two components offset each other.
The second column shows that foreign investors' net purchase of US stocks is positively but insignificantly correlated with the currency value (a positive relationship is against HR's model).
A new explanation
Filipe's (2012) model characterizes this correlation as a function of the fundamental volatility of the host market relative to the home market. Based on a more standard underlying model, we suggest that the source status of the host country (ie, whether it is a net recipient or a net supplier of international capital flows) drives the variation in stock return differentials—the currency correlation. Next, we use our combined sample for a comprehensive empirical investigation of the drivers of the correlation between stock return spreads and the host currency.
We show that the resource status of the host country can account for a significant part of the variation in the stock return differential – currency correlation. We derive forex order flow implications from this model and show how it can account for the cross section of the stock return differential - currency correlation. We argue and show that the resource status of the host economy determines which effect will dominate.
We operationalize the concept of source status (SS) as a measure of the activity of residents investing abroad relative to non-residents investing in the host country. In this section, we show that a combination of the features incorporated in more standard models of stock portfolio flows, rather than HR's risk-rebalancing mechanism, can account for the cross-sectional stock return differential–currency correlation and predict a negative correlation when the host economy is one. The second feature of the model is extrapolative expectations from US (host) investors for the host's (US) stock returns.
After an unexpectedly high return on the host stock, US investors revise their expected returns on the host stock. The difference in the number of host market stocks owned by the US investor from the world portfolio will be a function of DE. Taking the first derivative of Eq. 6) describes the flows of American investors to the host market.
The sign of the second term is determined by relative volatilities of host and US stocks: 𝜕𝐷𝐸 𝜕𝑃. For example, a large WH relative to PHNH will lead to greater variance of 𝑑𝑁𝑈𝑆𝐻, increasing the host country's SS. We estimate the explanatory power of SS after controlling for other potential drivers of the stock market–currency context.
More broadly, the inclusion of this variable serves to capture the potential regime dependence of the stock return differential - currency correlation (see Christiansen et al. Volatility of host stock market returns relative to the US stock market (rcv). The dependent variable (iDif ,T ,dE) is the correlation between the monthly difference in the return of the host stock market over the US stock market (Difi,t = Ri−RUS) and changes in the value of the base currency (−dEi,t) in the subperiod T .
Conclusions
We show that more support for Filipe's model comes from a sample that includes emerging markets, which is intuitive since emerging markets have higher fundamental volatility. Importantly, however, columns 7 and 8 show that SS, especially SS-Equity, remains significant when rcv is included in the model. Thus, the SS variable we introduce in the current paper captures additional factors used in our model that drive the stock return differential - currency correlation.
Second, countries with more volatile stock markets tend to have a more positive stock return differential - currency correlation. We have proposed a model where a combination of two opposing effects, the rebalancing of home wealth by host market investors and the reaction of international investors to local information, drives the stock return differential–currency correlation. Using a comprehensive panel, we have investigated the drivers of variation in the stock return differential – currency correlation.
The results show that the source status of the host country can significantly explain the variance of the stock return difference and the currency correlation and predict a negative correlation when the host country is a net source of international capital flows. Our empirical analysis also provides strong support for Filipe's (2012) relative fundamental volatility as a driver of the sign of stock return divergence and currency correlation. For US stocks, demand from US (host) investors equals 𝑁𝑈𝑆𝑈𝑆= 𝜇𝑈𝑆. 𝛾𝜎𝑈𝑆2 or with the same risk aversion coefficients 𝜇𝜎𝐻. 𝜎𝑈𝑆2 , so that US and host investors would invest in proportion to their assets.
If the US investor were the only one to decide how much less stock to hold in the host market due to international barriers, he would reduce the "barrier-free" holdings in the host country by 𝛾𝜎𝛿. However, this reduction must be equal to the amount of host shares by which the host investor will exceed his optimal "no hurdle" holdings. This sum is proportional to 𝑊𝑊𝐻, which we define as 1 − Ω, where Ω =𝑊𝑊𝑈𝑆. Excluding short positions leads to the market clearing condition 𝑁𝐻𝑈𝑆+ 𝑁𝐻𝐻 = 𝑁𝐻. On the other hand, if the host investor were the only one to decide how many more shares to hold in the host market, he would exceed his "unhindered" holdings in the host country by 𝛾𝜎𝛿. However, this increase must be equal to the amount of host shares by which the American investor will reduce it.
Now, with this system of (A.1) and (A.2) we check whether the market clearing condition is satisfied. Column dN indicates that the sign of dN, thus currency's correlation with the host stock return, changes under different numerical assumptions of model parameters and their interaction.
