Translate

Tuesday, August 28, 2018

Gender and the Vote in WA

This piece is under construction as of 8/30/2018. Voter Histories for Primary 2018 are due first week in September and will allow for Primary 2018 to be incorporated into this post. The tables below want a wide screen. Gender is a such a widespread and binary classifier that understanding it's significance in terms of blue/red, young/old, white/POC appears difficult. That women register and vote more than men is an easy insight. Who women vote for in what counties and why they vote is more difficult. To give a simple example relevant to the 2018 midterms: It is possible that the new stampless ballot could increase conservative votes among women of low motility and busy schedules in red counties. There is some evidence that partisanship of gender  is more clearly defined when gender is considered as a function of location, age, religion, race, class, income, party allegiance and education levels. Peering into a wide scale 'binary classifier' like gender and gender's relationship to other electoral covariants is proving to be more that a weekend's task.  The tables below want a wide screen.

WA voter database for previous five years. All totals from end of year except 2018 (July):

   Year  Female    Male Unknown   Total pctF pctM pctU
1: 2014 2048184 1872884    2277 3923345 52.2 47.7  0.1
2: 2015 2072658 1897010    2651 3972319 52.2 47.8  0.1
3: 2016 2224698 2042803    8990 4276491 52.0 47.8  0.2
4: 2017 2215647 2039814    9314 4264775 52.0 47.8  0.2
5: 2018 2232141 2056350   10904 4299395 51.9 47.8  0.3


Taken from Collection of ~15M voter histories from ~4.8M voters from 2014 - 2018:

      Election  Female    Male Unknown FminusM FdivMF FdiffMF
 1: 2015-02-10  153839  133103     251   20736   53.6     7.2
 2: 2015-04-06      78      65       1      13   54.5     9.0
 3: 2015-04-28  233624  206687     153   26937   53.1     6.2
 4: 2015-08-04  407243  352293     392   54950   53.6     7.2
 5: 2015-11-03  755260  682315     880   72945   52.5     5.0
 6: 2016-02-09  372580  324207     606   48373   53.5     7.0
 7: 2016-04-04      55      43       0      12   56.1    12.2
 8: 2016-04-26  162157  141572     128   20585   53.4     6.8
 9: 2016-05-24  735603  608070    1543  127533   54.7     9.4
10: 2016-08-02  722975  633278    1289   89697   53.3     6.6
11: 2016-11-08 1677072 1485433    5270  191639   53.0     6.0
12: 2017-02-14  162439  140334     565   22105   53.7     7.4
13: 2017-04-17     178     139       0      39   56.2    12.4
14: 2017-04-25   65404   55448     373    9956   54.1     8.2
15: 2017-08-01  538272  468729    1327   69543   53.5     7.0
16: 2017-11-07  821959  741396    2110   80563   52.6     5.2
17: 2018-02-13  473456  416577    1483   56879   53.2     6.4
18: No History  449410  468259    4790  -18849   49.0    -2.0

Current WA voter  database by 'Generations':

   GTEAge LTAge   Male Female MaleMeanAge FemaleMeanAge MaleMedianAge FemaleMedianAge
1:     18    38 620563 656464    28.64927      28.70344            29              29
2:     38    58 690152 729129    47.70612      47.69196            48              48
3:     58    78 621679 685044    66.27935      66.31080            66              66
4:     78    98 122893 158515    83.43901      84.22610            82              83
5:     98   118   1059   2987    99.73088      99.75226            99              99

GE 2016 County based M/F Correlations from Precinct Results. These are actually Clinton-Stein/Trump-Johnson Female/Male correlations. The GE2016 Results are merged with a current VRDB (July 2018) for gender and precincts.

     sc Clinton-Female Clinton-Male Trump-Female Trump-Male
 1: AD          0.948        0.758        0.902      0.826
 2: AS          0.926        0.873        0.894      0.930
 3: BE          0.918        0.948        0.900      0.961
 4: CH          0.910        0.936        0.893      0.940
 5: CM          0.924        0.876        0.880      0.903
 6: CR          0.760        0.692        0.729      0.736
 7: CU          0.837        0.897        0.791      0.934
 8: CZ          0.864        0.704        0.819      0.775
 9: DG          0.945        0.950        0.933      0.968
10: FE          0.742        0.794        0.565      0.927
11: FR          0.681        0.676        0.666      0.709
12: GA          0.955        0.965        0.949      0.973
13: GR          0.864        0.871        0.860      0.899
14: GY          0.937        0.937        0.928      0.942
15: IS          0.760        0.577        0.640      0.715
16: JE          0.884        0.612        0.812      0.723
17: KI          0.661        0.557        0.674      0.525
18: KP          0.677        0.775        0.614      0.825
19: KS          0.865        0.907        0.807      0.942
20: KT          0.831        0.896        0.829      0.900
21: LE          0.949        0.944        0.911      0.968
22: LI          0.976        0.995        0.979      0.995
23: MA          0.932        0.938        0.923      0.975
24: OK          0.718        0.517        0.691      0.629
25: PA          0.983        0.971        0.974      0.984
26: PE          0.906        0.940        0.877      0.968
27: PI          0.850        0.737        0.808      0.789
28: SJ          0.985        0.944        0.989      0.932
29: SK          0.892        0.889        0.865      0.920
30: SM          0.865        0.908        0.860      0.947
31: SN          0.820        0.747        0.781      0.799
32: SP          0.680        0.569        0.612      0.641
33: ST          0.866        0.936        0.827      0.964
34: TH          0.778        0.786        0.731      0.828
35: WK          0.839        0.917        0.772      0.944
36: WL          0.897        0.897        0.872      0.919
37: WM          0.619        0.412        0.578      0.474
38: WT          0.923        0.820        0.926      0.816
39: YA          0.853        0.668        0.788      0.730
    sc Clinton-Female Clinton-Male Trump-Female Trump-Male

Added a Correlation Chart

Click on this chart to enlarge.  The 10 (top) counties in the chart make up ~80% or more of WA's state registered voters. It appears that the more blue your county, the tighter and higher the correlations between presidential candidates and men and women in GE2016. The converse appears to be true for red.  I am picking up a rural/urban split here: the more denser counties have bluer votes and less significant difference between gender correlations.  The redder counties have male/female votes with a greater spread between the correlations of the number of votes by gender and blue/red candidates. To be honest,  I don't have the math chops to understand if my data has any real significance or whether I am just seeing irrelevant patterns. Gender may not be a strong macro scale predictor in WA state. It may have significance inside smaller segmentations (e.g. precinct groups).


Contrary 2016 Precincts

Table 1 contains precincts dominated by female voters, but votes overwhelmingly for Trump or Johnson. Table 2 contains precincts dominated by male voters, but votes overwhelmingly for Clinton or Stein.

Table 1
      StateCode  CS  TJ   F   M      pctF     pctCS
 1:      OK_203  25  73  97  44 0.6879433 0.2551020
 2:       WL_68  24  63  63  39 0.6176471 0.2758621
 3:      CU_103  62 137 147  97 0.6024590 0.3115578
 4:       FR_47  10  20  26  13 0.6666667 0.3333333
 5:     BE_1676 196 378 421 272 0.6075036 0.3414634
 6:      LE_206 187 347 428 280 0.6045198 0.3501873
 7:      LE_112 226 403 532 310 0.6318290 0.3593005
 8:      CH_100   4   7  10   6 0.6250000 0.3636364
 9:      YA_142 141 227 303 181 0.6260331 0.3831522
10:      YA_307  18  27  44  29 0.6027397 0.4000000
11:      OK_214  27  40  60  35 0.6315789 0.4029851
12:     SP_7036 230 309 413 250 0.6229261 0.4267161
13:     KI_3517 100 133 300 172 0.6355932 0.4291845
14: SN_23811572 131 173 489 287 0.6301546 0.4309211
15:      OK_147  25  33  45  28 0.6164384 0.4310345
16:      YA_139 312 405 614 328 0.6518047 0.4351464
17:      OK_107  24  31  43  23 0.6515152 0.4363636
18:      OK_108  35  42  64  37 0.6336634 0.4545455


Table 2
    StateCode  CS TJ   F   M      pctF     pctCS
 1:   KI_3416 300 74 133 421 0.2400722 0.8021390
 2:   KI_2987 208 51 205 330 0.3831776 0.8030888
 3:   KI_1823  83 19  59 116 0.3371429 0.8137255
 4:    KI_568  57 13  36  56 0.3913043 0.8142857
 5:   KI_2081 161 32 139 250 0.3573265 0.8341969
 6:   KI_3497 389 75 224 394 0.3624595 0.8383621
 7:   KI_3665 417 78 265 419 0.3874269 0.8424242
 8:   KI_1975 120 22 126 195 0.3925234 0.8450704
 9:   KI_2089 170 29 108 211 0.3385580 0.8542714
10:   KI_3647 229 31 152 402 0.2743682 0.8807692
11:   KI_1928 229 26 144 221 0.3945205 0.8980392
12:   KI_3571 286 29 209 341 0.3800000 0.9079365
13:   KI_2549 558 50 296 471 0.3859192 0.9177632
14:   KI_2852 444 37 219 331 0.3981818 0.9230769
15:   KI_1860 786 51 314 493 0.3890954 0.9390681
16:   KI_1847 511 27 256 398 0.3914373 0.9498141
17:   KI_2541 260 11 122 208 0.3696970 0.9594096
18:   KI_1859 497 20 244 385 0.3879173 0.9613153


APPENDICES

Statistical Concepts


A 'Correlation Coefficient' is a measure of the strength of the relationship between two sets of data. The 'Pearson correlation coefficient'  is defined as the covariance of X,Y over the product of the standard deviations of X,Y. The base R language correlation coefficient function is 'cor'.   'Cor' returns a data point between 1 : -1.  For practical usage:

1 =  absolute correlation
0 = no correlation
-1 = absolute negative correlations

There are no universally agreed upon definitions for the strength of correlations, only generally accepted guidelines.  Correlation coefficient functions produces a  statistical measure that shows the mathematical strength of a numerical relationship. An oft repeated warning is "Correlation does not imply Causation". All that being said, covariance and correlation are some the best measures of relationships between linear data. The Pearson correlation coefficient normalizes covariance with standard deviation .

No comments: