Changes

Run the regressions on raw data (i.e., don,'t join to COMPUSTAT first). Specs:*Date Range: 1/1/78 -> 12/31/11*Company Codes: PERMNO*SEARCH ENTIRE DATABASE*ID Info: CUSIP, NAICS*Time Series: VOL (Share Volume, -99 is missing), RET (Holding Period Return, error codes -66.0 -77.0, -88.0, -99.0)*Share info: SHROUT (No. Shares Outstanding in K)*Mkt Info: VWRETD (Value-Weighted Return inc. Dists)*Output: tab-delimited txt*Date Format: YYMMDDn8 (corresponds to ISO 8601 and is Postgres compatible) RET errors:*E -44.0 No valid comparison for an excess return*D -55.0 No listing information*C -66.0 more than 10 periods between time t and the time of the preceding price t? *B -77.0 not trading on the current exchange at time t *A -88.0 no return, array index t not within range of BEGRET and ENDRET *'' -99.0 missing return due to missing price at time t  Get file:*e5da94cf0a760426.txt (3195.8 MB, 60127561 observations 8 variables) Plan: Pull into Postgres, Index, Cut into chunks (yearly? Want < 250mb?), Run regressions in STATA using a batch script with dispatch to Bear.

*No. of Shares Outstanding (SHROUTin K)