The aggregate stock market index and the exchange rate are known to have a very low correlation with any other measurable macroeconomic variable except at very low frequencies (Frankel and Rose (1995), Rogoff (2001)). Financial economists interpret this very lack of predictability as evidence for efficiency, whereby only unpredictable news should move prices. But even gathering proxy variables for news ex-post does not seem to substantially increase the explanatory power of asset pricing models (Roll (1988)). This motivates us to examine a new financial market variable called order flow in its relationship to stock and exchange rate returns. Order flow is the net of buy minus sell initiated orders. In the foreign exchange market, daily exchange rate returns and daily order flow show a remarkably high correlation (Evans and Lyons (2002a, 2002b, 2002c) and Killeen, Lyons and Moore (2006)) and even permanent changes in the exchange rate appear to be explained by order flow. Unfortunately, most of the microstructure literature features order flow as an exogenous variable in a single market setting. Its very origin remains unexplained and this lack of economic structure constrains the analysis. In particular, issues of market interdependence between different international stock markets are generally ignored.
This paper contributes to the existing literature in four dimensions. First, we provide a market model in which order flow is the result of belief changes of various investor groups. This allows for a structural interpretation of order flow regressions. Second, we model a two country multi-market setting in which we can explore the relationship between equity, exchange rate and bond markets. In particular, we obtain testable restrictions which link equity returns to the various order flows. We explicitly model exchange rate determination unlike much of the international investment literature (see Albuquerque, Rui, Bauer and Schneider (2006)). Thirdly, we show that our empirical framework explains up to 60 percent of the daily return variations in the S&P 100 index. In accordance with the theory, both exchange rate returns and order flow into the overseas market have explanatory power for the domestic stock market returns. Fourth, our model can explain the asymmetry in the correlation structure of equity returns and exchange rates. U.S. equity market appreciations typically come with U.S. dollar appreciations, while European equity market returns correlate negatively with Euro appreciations.
The starting point of our analysis is a coherent interpretation of order flow itself. What motivates trades through market orders as opposed to limit orders? In most microstructure models of limit order markets those market participants with private asset valuations removed from the current midprice tend to pursue market order strategies. The intuition is straightforward. Execution uncertainty related to limit order submission is a multiplicative factor of the expected benefit of a trade. In the absence of risk aversion, the probability of non-execution reduces the expected trade benefit linearly as the difference between current midprice and the private value increases. The cost of market order submission by contrast is an additive cost related to the effective spread. It is unchanged by more extreme private asset valuations. A large change in the asset valuation by a segment of market participants will therefore tend to trigger predominantly market orders. This feature of modern limit order markets makes order flow a suitable proxy for (substantial) investor belief changes. Our simple market model captures this aspect, namely order flow is simply a linear function of belief changes. Hence, order flows can be used to identify heterogeneous belief changes within a segmented investor population. We do not deny that other trade motivations like (urgent) hedging or liquidation needs might also come with a preference for market over limit order implementation of the transaction. These trades are outside the model framework and feature as noise in the empirical analysis. We also highlight that we are agnostic about the source of the belief changes. These could be based on private information or have a behavioral explanation.
There is a growing literature that considers equity valuation in the context of dispersion of IBES (Institutional Brokers’ Estimate System) forecasts. For example, Basak (2000) studies the behavior of security prices in the presence of investors’ heterogeneous beliefs regarding the price of risk. In Basak (2005) the basic analysis is generalized to incorporate multiple sources of risk and disagreement about non-fundamentals. Anderson et al. (2005) provides a theoretical treatment of heterogeneous beliefs as well as empirical evidence showing that heterogeneous beliefs matter for asset pricing. A central feature of much of the theoretical literature is a reliance on the combination of behavioral constraints and heterogeneous beliefs. Miller (1977) argues that short-sale constraints could lead to an over-valuation effect because negative views are not acted upon to the same extent as positive ones. An example of the ‘investing with constraints’ literature is Boehme et al. (2007) where short-selling restrictions combined with dispersion in beliefs is shown to imply Miller’s overvaluation effects. This follows similar work such as Diether, Malloy and Scherbina (2002) who find that raw returns of stocks with higher dispersion of analysts’ earnings forecasts earn lower future returns than a control sample.
There are only a few contributions in the literature where belief changes are aggregated to the country level. Kothari et al. (2006) is of interest because they examine earnings announcements at an aggregate level. This is part of an empirical literature that tests whether stock prices move in response to cash-flow news or discount-rate news (Campbell and Shiller (1988)). In contrast with firm-level evidence, Kothari et al. find that aggregate returns and aggregate earnings growth are negatively correlated for the U.S. equity market. They also find that aggregate earnings growth is strongly correlated with discount rate proxies such as T-bill rates and that cash-flow news is largely idiosyncratic. In other words, positive market-wide earnings innovations are associated with increased discounting because of the macroeconomic policy reaction, such that a negative valuation reaction is found. But we also note that Bernard and Thomas (1990) find exceedingly slow and small price reaction to this kind of public news.
Download
International Order Flows: Explaining Equity and Exchange Rate Returns
