How to compute difference in years between daily dates?

[Ariel Karlinsky]

I have daily stock data, inputted and formatted properly for stata. I wish to calculate the return on that stock from one period to 1,2,3...10 years prior.

My current code is:

gen date = date(date_str,"YMD")
format %td date
tsset date

forvalues t=1(1)10 {
    local days = `t' * 365
    gen d`t'y_close = close - L`days'.close
}

forvalues t=1(1)10 {
    di "`t' years return"
    di "number of days in which return was negative or zero"
    count if d`t'y_close <= 0
    di "total number of non missing days"
    count if d`t'y_close 
}

I have two questions regarding this:
1. Currently, I manually calculate the number of days (as the data is daily) as products of 365 days a year and number of years I'm interested in. Is there a more elegant way to do so? a command that decreases X years from a given date, and handles leap years etc. appropriately?

2. Sometimes, "5 years prior" to a specific date falls on a day which doesn't exist in the data as there's no trade on this day (sunday, holidays, etc.). Is there a way to create an expression (i thought using cond but I think it's gonna amount to lots of nesting conds), in which the logic is "if X years prior to given date is missing, go the smallest amount of days backward and use that date instead"?

Thank you

[David Benson]

There are probably more elegant solutions, but three options you might consider:

1) Since you have all trading days, you might use 250 trading days as your measure of a year (most years have 250, 251 or 252 trading days) rather than 365 or 366 calendar days.

2) I assume you pulled your data from CRSP or WRDS. If you have access to EVENTUS through WRDS, EVENTUS will take care of this for you.

3) There are several user-contributed commands that you could use: ascol (SSC) converts daily stock prices to annual returns. You might also look at asrol and asreg. rangerun and rangestat would also work. (All are on SSC)