search

feomike / slope

makes a bunch of data objects for time series line charts
1 stars 0 forks source link

create a list of example FIs #6

Closed feomike closed 8 years ago

feomike commented 8 years ago

this list is developed purely for academic reasons. the reasons for selection to this are as much random as anything else.

feomike commented 8 years ago

the idea here is consistent w/ the 80:20 rule. in this case we know one struggle for doing something like this is the significant parent/top holder corporate structure that exists for many banks we would prefer to avoid these complications in the beginning. moreover, as a general rule we want to establish as long as possible line charts (hopefully from 1990 - 2015). so financial institutions which a) have no parent or top holder status in a given year AND b) have at least 1000 rows in the years 2010 - 2014 are likely to be large enough to be around for a while. these become good candidate FIs to test the protocol with. Moreover, it seems worth while to test the example with at least one financial institution in each of the regulators purviews (e.g. one from OCC, FRS, FDIC, NCUA, CFPB and HUD).

feomike commented 8 years ago

the unique ID for a given financial institution in HMDA data is agency code + RID. i get confused all the time about the current status identifiers in HMDA, it is something like RID (respondent id) is the ID field in the LAR. the RSSD is the id field in the NIC (national information center) data at the FRS. for FRS filers the RSSD is the RID, but for others it is not. in this task, i am seeing 1 financial institution per regulator, so here goes;

feomike commented 8 years ago

it is possible that the agency code for these institutions has changed over the time period for the test (e.g. 1990 - 2015). i will run tests on the panel/ts to establish the appropriate year/agencycode/rid for consistency sake in this code.