Timeline for Fama/French momentum replication: risk-free rate missing on one of the legs?
Current License: CC BY-SA 3.0
16 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 18, 2017 at 17:04 | comment | added | João | I seem to be having the same problem as you. Winners working fine, losers seem to be overestimated. Did you eventually find what was wrong with your replication? | |
| Jun 10, 2017 at 2:38 | comment | added | Matthew Gunn | Another issue: If the marketcap cutoff for the Russell 1000 is above the median market cap of the NYSE at some points in time, the Russell 1000 may not have the biggest losers of the Fama-French big portfolio because the biggest losers may leave the index. | |
| Jun 10, 2017 at 0:36 | answer | added | Matthew Gunn | timeline score: 5 | |
| Jun 10, 2017 at 0:16 | comment | added | Juan Q | Alex - thanks for the suggestion. For the time being I am still treating the low prior as a long, not doing anything funny to it, so I don't think so but I will double check. There must be a fairly obvious bug along those lines staring back at me! | |
| Jun 10, 2017 at 0:10 | comment | added | Juan Q | Matt - thanks a lot for your preliminary numbers. They're close enough to prof. French's, so I'm clear the issue is on my side. Not that I expected otherwise! | |
| Jun 9, 2017 at 23:50 | comment | added | Alex C | A random thought: the LoPrior are the ones you are supposed to short, I wonder if that has something to do with how you calculate their return? Somehow you give up the riskfree rate because you are short ?? | |
| Jun 9, 2017 at 23:27 | comment | added | Matthew Gunn | Out of curiosity, I took a rough stab at trying to calculate those portfolio returns using CRSP data. I know I'm not matching (yet) whatever Fama-French did precisely because the number of firms in each of my portfolios don't match them. I would think it's easier to match the value weight portfolios because big firms are going to be less sensitive to exact sample selection issues... (good price data on dates t-13, t-1, t-3 etc...) mattgunn.com/share/PRELIMINARY_big_lowpriorret.csv | |
| Jun 9, 2017 at 22:20 | history | edited | Juan Q | CC BY-SA 3.0 |
My comment on the UMD equation wasn't expressed correctly - what I meant is that I hadn't checked all the components so if the error crept up for instance on Small HiPrior as well, then it could wash out with the one showing up in Big LoPrior.
|
| Jun 9, 2017 at 22:08 | comment | added | Juan Q | In particular, for the "equal weighted" variety, which is where I started from as it is easier to replicate. Having said that, I am in the process of replicating the "value weight" tables and I think the issue appears as well. | |
| Jun 9, 2017 at 22:06 | comment | added | Juan Q | Yes, that is exactly what I meant. | |
| Jun 9, 2017 at 20:34 | comment | added | Matthew Gunn | So on French's website, you think that of the "6 Portfolios Formed on Size and Momentum (2 x 3)," you think the big, low prior returns portfolio may be missing the risk free rate? (Or that your code is bugged?) Did I understand your post properly? | |
| Jun 9, 2017 at 12:49 | comment | added | Juan Q | Hi - in building my data set I took into account all stocks that ever belonged to the index, not just the ones that exist today. Plus if it was the case, surely I should see an effect both on the HiPrior and LoPrior portfolios, not just on the latter? | |
| Jun 9, 2017 at 11:55 | history | edited | LocalVolatility | CC BY-SA 3.0 |
re-uploaded images - they weren't properly displayed before
|
| Jun 9, 2017 at 11:25 | comment | added | Richard | Have you considered the survivorship bias? | |
| Jun 9, 2017 at 10:26 | review | First posts | |||
| Jun 9, 2017 at 11:55 | |||||
| Jun 9, 2017 at 10:18 | history | asked | Juan Q | CC BY-SA 3.0 |