A Generalized Method of Moments estimation of the determinants of dollar/yen bid–ask spreads is undertaken. In particular, a long time-series of daily spot foreign exchange trading volumes is used for the first time. In line with standard spread models and volume theories, it can be shown that unpredictable foreign exchange turnover (a measure of the rate of information arrival) increases spreads, while predictable turnover decreases them. Both effects are strongly significant when employing spot turnover instead of proxies like forward turnover as in previous studies (Bessembinder, 1994). The results are also found to be robust when unpredictable Reuters quoting frequency is used as an instrument for unpredictable trading volumes to cope with their endogeneity. Spread estimations with plain (non-decomposed) volumes are rejected as misspecified. Finally, there is evidence for the conditional heteroscedasticity of unpredictable spot foreign exchange volumes.
