Some data inconsistencies/errors

[MichaelChirico]

Per here:

http://www.lacledecountymissouri.org/clerk/electresults/files/naccum.pdf

The Dem/GOP totals are correct for 2008, but both exceed the total_2008 field.

The other_2008 field also appears to exclude write-ins? I don't know what the rule is in the overall data for this.

Appears this should be 16,379 (or 16,477 -- the former comes from totaling all counts for presidential votes).

2012 numbers are also out of sync, but the error margins appear minimal, so it cold just be a matter of data record timing...

http://www.lacledecountymissouri.org/clerk/electresults/files/totalsn12.pdf

Some other inconsistencies, all from 2008:

• Sauk County, WI. Democrat vote total should be 18,617 (transposed in this data to 18,167). Source
• Ottawa County, OH. Vote total should be 23,475. Democrat total should be 12,049. Other total should be 401. Source
• LaPorte County, IN. Total votes cast is 48,107. Democrats got 28,247. GOP got 17,911. Others had 842. Source
• Platte County, MO. Total should be 46,640. Dem is 21,459. Other is 721 (560 if excluding write-ins) Source

[tonmcg]

Those are good catches. I did not factor in error catches and should probably do so. Some counties publish incorrect data or miscalculate totals. I'm thinking of creating an output table in the notebook comparing total votes to the sum of its parts. Thoughts?

Thanks for researching those counties and providing links.

[MichaelChirico]

Sounds excellent! It would be a miracle if all the data came in perfectly clean. Hopefully you get some crowd-sourced help tracking down issues. Publishing that sort of notebook could provide useful visibility to the project & facilitate this.

[seesharp15]

Ugh - I wish I read through this prior to submitting the same -_-

Any plans to fix?

[tonmcg]

@seesharp15 Forgive my delayed response. It looks like you changed the files to factor in Shannon County's (Oglala Lakota) name and FIPS change. You also made updates to county-level totals for 08-16 for certain counties.

On the county name FIPS changes: this is a tough issue to address. The U.S. Census publishes changes in county names and FIPS codes every decennial that also include substantial changes in county boundaries as well. Whereas the 2010 Shannon County boundaries perfectly match the 2016 Oglala Lakota County boundaries, Bedford County boundaries changed significantly in 2013 from 2010.

This means county-level maps (Shapefile, GeoJSON, TopoJOSN, etc) for each election year would all be on a 2016 basis. If this were the case, then counties that existed in 2012 but don't exist in 2016 (Bedford city) would show up in a 2016 map but with no data. What are you thoughts on this?

On vote totals update: I'm fine with those updates, though we should be careful that we're not trying to manage and combine numerous data sources.