Ken Nyholm, "Inferring the Private Information Content of Trades: A Regime-Switching Approach", Journal of Applied Econometrics, Vol. 18, No. 4, 2003, pp. 457-470. All data are in the zipped file 'tq.zip'. There are data for 20 NYSE traded stocks. The data for each individual stock are stored separately in DOS ASCII format. The data are a subset of the August 1995 NYSE trade and quote data cd-rom. Individual stock files include one observation per line and one column per variable. The following variables are included: Date Time Price (times 10,000) Mid-quote (times 10,000) Volume (number of stocks traded) Trade indicator (1=ask, -1=bid) Spread (times 10,000) The tq.zip file contains 20 files, for each of the following stocks (number of observations in brackets): AXP [4,351] CHV [4,230] DD [5,064] DOW [3,962] EK [4,366] GE [10,779] GM [5,493] IBM [12,806] IP [3,839] JNJ [6,487] KO [7,605] MMM [4,454] MO [6,792] MOB [4,125] MRK [9,164] PG [4,798] S [4,418] T [13,416] X [2,814] XON [6,035] The number of observations given above is larger (by 23) than the observation numbers stated in the article (Table1). This is because mid-quote changes calculated from closing of day t and opening of day t+1 are excluded in the estimations (as mentioned in the article).