Exchange Rate Determination and Forecasting: Can the Microstructure Approach Rescue Us from the Exchange Rate Disparity?

Table 6

Granger’s causality estimation of (6) (5-minute frequency).

	Day 1	Day 2	Day 3	Day 4	Day 5
Order flow
L1	−0.118071; (0.19506)	−0.137330; (0.16762)	0.218996; (0.16716)	−0.220818; (0.23302)	0.293388; (0.20688)
L2	−0.120591; (0.19228)	−0.088450; (0.16622)	−0.060422; (0.16300)	−0.013327; (0.23357)	−0.299061; (0.20852)
L3	0.108081; (0.19130)	−0.172152; (0.16703)	0.268007; (0.15693)	−0.447154; (0.23714)	−0.410111; (0.20512)
L4	−0.190258; (0.18857)	−0.311102; (0.16644)	−0.051263; (0.15932)	−0.014220; (0.23410)	0.235299; (0.21190)
L5	−0.022752; (0.18485)	0.270998; (0.15131)	−0.096563; (0.15590)	0.167554; (0.23372)	−0.269974; (0.20186)
Exchange rate return
L1	0.150455; (0.17953)	0.256772; (0.16881)	0.023109; (0.17056)	0.257228; (0.17044)	−0.135055; (0.17139)
L2	0.064654; (0.17867)	−0.043441; (0.16814)	−0.086769; (0.16403)	0.148246; (0.17029)	0.031216; (0.17199)
L3	−0.015274; (0.17925)	0.217579; (0.16979)	−0.239797; (0.16118)	0.330404; (0.17486)	0.255290; (0.16949)
L4	0.172611; (0.17717)	0.385293; (0.16870)	−0.093093; (0.16256)	0.106227; (0.17636)	−0.215438; (0.17468)
L5	−0.169169; (0.17404)	−0.245181; (0.15871)	−0.055984; (0.16353)	−0.232155; (0.17743)	0.121239; (0.17028)
Constant	2.816570; (4.10656)	−0.516888; (4.00396)	−5.881906; (3.26599)	−1.067474; (5.48496)	7.343517; (5.01448)
Adjusted 𝑅-sq	−0.013954	0.047594	0.020127	0.030615	0.038009
𝐹-statistic	0.844486	1.564692	1.232112	1.356878	1.446467
𝑃 value	0.5872	0.1277	0.2796	0.2110	0.1707

This table reports the results of the regressions of the exchange rate return at 5-minute frequency on five lags of order flow and five lags of exchange rate returns over the five days in the sample. Standard errors are reported in the parentheses below the coefficient estimates. The 𝐹-statistics and 𝑃 value are from the 𝐹-tests, which indicate lagged order flow coefficients are not jointly different from zero.