OpenSourceAP / CrossSection

Code to accompany our paper Chen and Zimmermann (2020), "Open source cross-sectional asset pricing"
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3604626
GNU General Public License v2.0
716 stars 215 forks source link

Low coverage ForecastDispersion in January/December #145

Closed assetpricer closed 1 month ago

assetpricer commented 1 month ago

Hi CZ,

I was going through the ForecastDispersion file that replicates Analyst Forecast Dispersion from Diether, Malloy and Scherbina. I noticed many missing dates for Decembers, Januarys and February (but bit lesser). Below I count the amount of rows for each month in the file:

image

In Diether's paper (table 1), it is mentioned that approx 3200 stocks are eligible in their sample in December 2000 already only

Could you please help me understand why the coverage in these months are so low?

tomz23 commented 1 month ago

Good catch @assetpricer ! The low coverage is due to the following line in the code: keep if fpedats != . & fpedats > statpers + 30 This line effectively drops all earnings forecast in the one month before and the two months after a fiscal period end date. Since most firms have fiscal period ends at the end of December this affects mostly December, January and February.

The OP does not seem to have this filter, so I remove it from our replication as well. Coverage in terms of number of firms is then similar to OP:

grafik

tomz23 commented 1 month ago

Fixed in https://github.com/OpenSourceAP/CrossSection/commit/b57ecb22252db46ab40ef0b0819082165e51e41b

assetpricer commented 1 month ago

Dear Tom, thank you so much for your response and the fix. Will the ForecastDispersion file in Google Drive be updated for the next update 😃?

tomz23 commented 1 month ago

Yes, the signal will be updated with the next release (probably by the end of the month).