This model can be summarized in the following way:
and ^fand XT are decay coefficients to be estimated. £fand Sr represent are the unexpected inflow and shocks to returns; a and b are respective persistence and trend following parameters for order flow, e describes the price impact of unexpected order flow on return, and с represents the extent to which price pressure offsets the information content of inflows. Our structural model can be thought of as a restricted (and over identified) version of the following reduced form model add comment:
where the distributed lag is the same as in the structural model. Parameters 71,t and K2l show the incremental predictability of lagged flows for future flows and returns, respectively. Similarly, TC,2 and 7I22 show the incremental predictability of lagged returns for future flows and returns.
The structural model (9) can be recovered from the reduced form model (11) by using the following restrictions:
We estimate equation (11) using nonlinear least squares equation-by-equation (i.e., without taking account of any correlation between the residuals, uf and ur. The results from equation (11) are presented in Table 10. Most of the n coefficients are positive and, for most regions, statistically so. For all countries combined, a 1 basis point increase in today’s cross-border holdings is associated with a 0.36 basis point increase in tomorrow’s cross-border holdings, and with a 3.50 basis point increase in tomorrow’s return.
For emerging markets, this latter number is more than four times as large. Similarly, for the world as a whole, a 1 percent increase in today’s returns is associated with a 0.02 basis point inflow (relative to market capitalization) and a 4 basis point increase in future returns. For emerging markets, many of these coefficients are statistically greater than zero. The only exception is K2b the incremental impact of flows on future returns. The same basic pattern applies across most regions.