Top minus Bottom Quintile Portfolio Returns with Newey West Adjusted Standard Erros

[Felix Moritz]

Dear all,

I have a panel of monthly stock market return (by firm and month) from the year 1953 to 2016.
I sorted the data into Quintiles according to the variable called ST each month. (high ST = 5 to low ST = 1)
Then, I calculated each month the equal weighted average of returns by Quintiles in order to have the monthly portfolio returns for each quintile each month.

bys quintiles month: egen portfolioST = mean(return)

Now, I calculated the time-series average of the quintile porfoltio returns for each Quintile:

tabstat portfolioST, by(quintiles) s(mean)

Finally, I want to calculate the difference between the top quintile portfolio (ST == 5) and the bottom quintile portfolio (ST == 1) each month and than take the time-series-average over all months:

I therefore use the following:

bys month: egen STq5 = mean(cond(quintiles == 5, ret, .) )

bys month: egen STq1 = mean(cond(quintiles == 1, ret, .))

gen STqdiff = STq5 - STq1

newey2 STqdiff, lag(12) force

Note: I use - newey2 - in order to estimate the mean with Newey West Adjusted Standard Errors.

Now my confusion(s):

1. When I calculate - mean(STq5) - I get a (slightly) different result than what I receive using - tabstat .
   I get the same results as with - tabstat - calculating for instance :
   
   

   	bys month : egen quintile1  = mean(retahead) if quintiles == 1
   	
   	mean(quintile1)

2. Is there any other way to calculate the top-bottom quintile difference with Newey West Adjusted Standard Errors?

Let me know if something is unclear of if you need any clarifications!

Thank you very much in advance for your help!

Best,

Felix

[Kate Lussy]

Hi Felix, I also have to calculate the difference of the top and bottom quintile portfolios with the Newey West adjustment. Did you by any chance figure out how to do so?

[Andres Salman]

Kate Lussy and Felix Moritz , by any change did you manage to find a way to get the t-statistics and adjust them accordingly?