Understanding Order Flow

Motivation(s)

Standard microstructure models assume variables relevant to exchange rates are common knowledge, and thus trades have no causal effect on price. These models operate on aggregated total flow and ignore order flow heterogeneity, the transactions initiated by agents of different types (e.g. corporations, hedge funds, mutual funds). Recent empirical findings from Evans and Lyons (2004) on order flow heterogeneity indicate a new model is needed:

  • The price impact of order flow is different across end-user segments.

  • Order flow can forecast future flow, which in turn affects exchange rates.

While existing models can add more variables to their regression to model different customer segments, the overall analysis would be wrong:

  • The order flow data from different segments are significantly correlated and even exhibit autocorrelation.

  • The regression focuses on a single bank and ignores market-wide flows from all segments that move the exchange rate.

Proposed Solution(s)

The authors adapt the Evans and Lyons (2004) model to focus on how differences across customer types affects the information contained in customer order flow, which in turn affects the joint dynamics of order flows and exchange rates. In contrast to the original model that focuses on the information aggregation process of FX trading, the adapted model distinguishes between liquidity-motivated traders, short-term investors, and long-term investors.

Evaluation(s)

The USD/EUR customer orders from Citibank between 1993 and 1999 were used in the simulations and empirical analysis. Note that this data is only reflective of approximately 15% of the entire market.

The simulations demonstrate that:

  • customer flows provide more precise information about fundamentals when the mix of customers is tilted toward longer-horizon participants;

  • flows from customer segments can produce negative coefficients in contemporaneous return regressions, even when positively correlated with fundamentals; and

  • customer flows forecast returns because they are correlated with the future market-wide information flow that dealers use to revise their FX prices.

The authors’ empirical analysis suggest that:

  • both the aggregate and disaggregated customer flows received by Citibank are positively auto-correlated;

  • contemporaneous correlations across flow segments are low at the daily frequency, but high at the monthly frequency;

  • some customer segments do produce negative coefficients in contemporaneous return regressions;

  • the proportion of excess return variation that segment flows explain rises with the horizon; and

  • about one-third of the order flow’s power to forecast exchange rates one month ahead comes from its ability to forecast future flow, with the remaining two-thirds applying to price components unrelated to future flow.

Although the authors’ sequencing of quotes and trades oversimplifies actual trading in the market, this simplification enables them to focus on how information conveyed by customer orders is transmitted among dealers and embedded in spot rates.

Future Direction(s)

  • Are public order flows still effective from a trader’s perspective?

  • Does there exist style factors for forex?

Question(s)

  • The way the technical details were written was not smooth at all. Isn’t the main point simply customer order flows is another good indicator?

Analysis

Order flows can serve as a powerful indicator in forex. Unfortunately, this information is gated by market makers. The authors should have examined whether traders can still benefit from order flows under adversarial market dealers. Another issue is the convoluted presentation of the analysis: simple results are presented in an incomprehensible manner. The arguments are too centered on statistical significance. Nonetheless, the notion of an equilibrium depreciation process does seem useful in modeling forex.

References

EL06

Martin DD Evans and Richard K Lyons. Understanding order flow. International Journal of Finance & Economics, 11(1):3–23, 2006.