2016 general precinct discrepancies

[dwillis]

Compared to county-level totals.

• Dutchess - Clinton has 9 more DEM votes than county totals
• Essex - has both town and precinct rows, need to remove the former.

[jamesdunham]

Here's what I've found so far:

• For St. Lawrence, drop rows where precinct is TOTAL
• In Essex, drop the total row for each town's precincts, indicated e.g. by precinct of Ticonderoga, after Ticonderoga 1, Ticonderoga 2
• Similarly, in Niagara, we have Wilson/000/1, Wilson/000/2, ... Wilson/000
• In Columbia, mode votes = AbsenteeAffidavit + ElectionDayVotes
• Same in Herkimer, but votes = absentee_affidavit + polling_place
• In Lewis and Seneca, absentee + election_day = votes
• Finally, drop where candidate or vote mode matches total; and where precinct matches cumulative|total
• In NYC counties, drop where candidate is Public Counter
• St. Lawrence data are wide, with columns votes, machine_votes, absentee, absentee_hc, and affidavit. The first of these is the sum of the others.

[jamesdunham]

After the steps in my previous comment, I get this:

   jurisdiction            party votes.precincts votes.counties vote_diff
1:     Dutchess       democratic           58919          58910         9
2:      Ontario       democratic           21055          21062        -7
3:      Ontario working families             798            799        -1
4:     Schuyler       democratic            2814           2817        -3
5:       Seneca             <NA>               1             NA        NA
6:      Steuben             <NA>            1313             NA        NA
7:      Steuben       democratic           10718          11982     -1264
8:      Steuben women's equality             136            155       -19
9:      Steuben working families             359            389       -30

The totals in Steuben County are correct, but 1,313 absentee votes in the precinct data aren't associated with a party. In the state canvass, they've been assigned parties. So the discrepancies are in Dutchess, Ontario, Schuyler, and Seneca.

[dwillis]

Looked at Dutchess recanvass and the Clinton Dem total is now 58933, which is 23 more than state's total for the county.

[jamesdunham]

Dutchess looks good now. With 5a482c4, we also have 58,933.

   jurisdiction            party votes.precincts votes.counties vote_diff
1:     Dutchess       democratic           58933          58910        23
2:      Ontario       democratic           21055          21062        -7
3:      Ontario working families             798            799        -1
4:     Schuyler       democratic            2814           2817        -3
5:       Seneca             <NA>               1             NA        NA
6:      Steuben             <NA>            1313             NA        NA
7:      Steuben       democratic           10718          11982     -1264
8:      Steuben women's equality             136            155       -19
9:      Steuben working families             359            389       -30

[jamesdunham]

Here's another pass through the NY returns more broadly, using 5a482c4: gist.

# Public Counter
d[candidate %=% 'Public Counter', unique(county)]
# [1] "Bronx"    "New York" "Queens"   "Richmond"

# `total` in candidate field
unique(d[candidate %=% 'total', .(county, candidate)])
#         county             candidate
#  1:     Cayuga   Total Special Votes
#  2:     Cayuga           Total Votes
#  3:    Chemung                 Total
#  4:   Chenango                 Total
#  5:   Delaware                 Total
#  6:       Erie                 Total
#  7:   Franklin                 Total
#  8:    Genesee                 Total
#  9:  Jefferson                 Total
# 10:      Lewis                 Total
# 11: Livingston                 Total
# 12:     Monroe                 Total
# 13:     Orange      Total Votes Cast
# 14:    Orleans                 Total
# 15:     Putnam                 Total
# 16:  Schoharie   Total Special Votes
# 17:  Schoharie           Total Votes
# 18:     Seneca                 Total
# 19:    Steuben Total Enrolled Voters
# 20:   Sullivan                 Total
# 21:    Wyoming                 Total
# 22:      Yates                 Total
#         county             candidate

# In Niagara, we have "Wilson/000/1", "Wilson/000/2", ... "Wilson/000"
d[(county == 'Niagara' & str_count(precinct, '\\/') == 1), unique(precinct)]
#  [1] "Cambria/000"          "City of Lockport/001" "City of Lockport/002"
#  [4] "City of Lockport/003" "City of Lockport/004" "City of Lockport/005"
#  [7] "Hartland/000"         "Lewiston/000"         "N Tonawanda/001"     
# [10] "N Tonawanda/002"      "N Tonawanda/003"      "Newfane/000"         
# [13] "Niagara Falls/003"    "Niagara Falls/004"    "Niagara Falls/005"   
# [16] "Niagara Falls/006"    "Pendleton/000"        "Porter/000"          
# [19] "Royalton/000"         "Somerset/000"         "Lockport/000"        
# [22] "Niagara/000"          "Wheatfield/000"       "Wilson/000"          

# In Columbia, mode "votes" = "AbsenteeAffidavit" + "ElectionDayVotes"
d[county == 'Columbia', unique(mode)]
# [1] "votes"        "election_day" "absentee"    

# Same in Herkimer, but "votes" = "absentee_affidavit" + "polling_place"
d[county == 'Herkimer', unique(mode)]
# [1] "votes"        "election_day" "absentee"    

# In Lewis and Seneca, "absentee" + "election_day" = "votes"
d[county %in% c('Lewis', 'Seneca'), unique(mode)]
# [1] "votes"        "election_day" "absentee"    

# In St. Lawrence, mode "votes" = "machine_votes" + "absentee" + "affidavit" + "absentee_hc"
d[county == 'St. Lawrence', unique(mode)]
# [1] "votes"         "absentee"      "machine_votes" "absentee_hc"  
# [5] "affidavit"    

unique(d[precinct %=% 'cumulative|total', .(county, precinct)]) 
#         county            precinct
#  1:      Lewis              TOTALS
#  2:    Madison               TOTAL
#  3: Montgomery               Total
#  4:     Nassau               Total
#  5:     Oneida               Total
#  6:   Onondaga      Syracuse Total
#  7:   Onondaga Onondaga Town Total
#  8:   Onondaga               Total
#  9:    Ontario               Total
# 10:    Orleans               TOTAL
# 11:     Oswego               Total
# 12:     Otsego               TOTAL
# 13:   Rockland               Total
# 14:   Saratoga               Total
# 15:   Schuyler          Cumulative
# 16:   Schuyler               Total
# 17:     Seneca          Cumulative
# 18:     Seneca               TOTAL
# 19:   Tompkins               Total
# 20:     Ulster               Total

unique(d[precinct == '', .(county, precinct)]) 
#     county precinct
# 1:   Wayne         
# 2: Wyoming         

# Ballots cast...
d[candidate %=% 'ballots cast', unique(county)]
# [1] "Niagara"      "Schenectady"  "St. Lawrence"

[dwillis]

I've gone through and removed the total rows from the counties mentioned above. I'm not sure what the issue is with Steuben - it looks like the county total file is incorrect.