I am currently working on back testing Joseph Piotroski's article Value investing: "The Use of Historical Financial Statement Information to Separate Winners from Losers". I am using Compustat data and CRSP data from Whartons research data services (wrds). The idea is to replicate the research article.
I am having trouble with the SAS code, finding and linking the returns between the two data sets. Returns, which have been defined by: " I measure firm-specific returns as one-year (two-year) buy-and-hold returns earned from the beginning of the fifth month after the firm’s fiscal year-end through the earliest subsequent date: one year (two years) after return compounding began or the last day of CRSP traded returns. If a firm delists, I assume the delisting return is zero. I chose the fifth month to ensure that the necessary annual financial information is available to investors at the time of portfolio formation. I define market-adjusted returns as the buy-and-hold return less the value-weighted market return over the corresponding time period. "
Issue: I am having trouble turning this concept into workable SAS code. CRSP has the returns information and the dates are displayed as ex: 19860131 (Jan 31 1986) and Cstat has the fiscal year end for each firm displayed as a number 1-12. As you will see below, adding 5 months will in some cases move the start of the returns into the next year. These two data sets link firms by a firm identifier code.
Variable descriptions: gvkey= company identifier,Cstat= (Compustat) financials dataset, CRSP = returns dataset, t= fiscal year end, RET = returns, (yymm = year*12 + mm)
This is the concept, compound returns from t+5 to t+16, where t is the fiscal year end. RET is the returns. This is the buy and hold return.
- i) So, for every month for every firm on CRSP, create 12 additional variables with returns from t+5 to t+16. Ex: if t is 10 (October) I need to find the returns from t+5, 3(March) through next 3 (march)
- ii) Then, get 1+RET each of the 12 months, multiply all 12, and subtract 1 to get one year buy and hold return (BHR) for firm. So, you will add one more variable, the 12 month BHR. Do same BHR for the market portfolio, so you can do market adjusted returns like the paper. So, you have 14 addl vars, 12 monthly rets from t+5 to t=16 and the 2BHRs.
iii) Now, suppose you are doing this for IBM. If IBM has Dec year end, then you would have used 2011 Dec financials from IBM to do the F Score calculation for 2011. This should be merged with the May 2012 record for IBM from CRSP, so you are aligning the Dec 2011 F-Score with the correct returns starting from t+5.
- iv) You can use the monthly returns to calculate the portfolio returns (VW and EW) and then do the Jensen and FF regressions to get alphas.
This being prefaced, the SAS code would look something like this?
If month >= 12 then month = month-12;
Confirm it is same firm (gvkey = lag1(gvkey);
Then, merge by gvkey match month;
Date from CSTAT matched with 5 months later on CRSP;
On CRSP match yymm = yymm; (yymm = year*12 + mm)
On CSTAT match yymm = fiscal end yymm + 5;
Ex from earlier,
For IBM 2011, FY end is Dec, yymm = (2011*12) + 12 = 24144
On CRSP, match yymm = year*12+mm,
On CStat, it will be 24149 to match on CRSP
Match May 2012 CRSP data with Cstat Dec 2011
This is a link to a PDF version of the research paper. https://www.chicagobooth.edu/~/media/FE874EE65F624AAEBD0166B1974FD74D.pdf
