next up previous
Next: About this document ...

From wrm1 Fri Nov 12 02:36:22 -0500 2004
From: Walter Mebane <wrm1@macht.arts.cornell.edu>
To: kathy@truthisbetter.org
CC: jasjeet_sekhon@harvard.edu, wand@stanford.edu
Subject: response to your responses

Hi Kathy,

Thanks for posting a link to the critique we circulated of your
posting of "Surprising Pattern of Florida's Election Results" at
http://ustogether.org/Florida_Election.htm

Here's a response to the responses (by Elizabeth Liddle and Marc
Sapir) to our critique that you also posted.  If you'd like also to
post a link to this response, I will post a copy of this message at
http://macht.arts.cornell.edu/wrm1/ustogether/ustogether.html

Best,
Walter Mebane


Elizabeth Liddle
http://ustogether.org/election04/liddle/liddle_response.html
states:

 > I have just repeated my original analysis of mid-size counties
 > counties, this time first of all omitting NW panhandle counties,
 > then omitting all N Florida counties. The machine effect remains
 > robust ((p<.01) despite loss of statistical power.

In the following display I have removed from the display presented in
my earlier message all op-scan counties at latitude at or north of
Alachua County.  Notice that all the remaining counties that have
DemChg04 less than -30 (or so) have Dem vote proportions roughly the
same (low) going back to 1992 and Dem registration proportions roughly
the same (high) going back to 1996.  This is again the "Dixiecrat"
phenomenon, and again the voting patterns haven't much changed going
back at least to 1992.  The Dem registration proportions in these
counties are not as high as in some of the panhandle counties, but
aside from that, omitting all the North FL counties doesn't change
anything.

                 DemChg04 DemV00 DemV96 DemV92 DemR04 DemR00 DemR96
Brevard (O)          13.7   0.45   0.41   0.31   0.37   0.38   0.40
Citrus (O)            8.3   0.45   0.44   0.36   0.39   0.41   0.46
DeSoto (O)          -30.6   0.43   0.43   0.36   0.59   0.64   0.70
Flagler (O)          26.6   0.51   0.48   0.41   0.38   0.40   0.42
Glades (O)          -35.6   0.43   0.45   0.39   0.65   0.69   0.76
Hardee (O)          -53.5   0.38   0.39   0.31   0.64   0.73   0.79
Hendry (O)          -28.3   0.40   0.44   0.34   0.57   0.63   0.66
Hernando (O)         19.0   0.50   0.49   0.39   0.39   0.40   0.42
Highlands (O)        -3.1   0.40   0.42   0.35   0.40   0.42   0.45
Levy (O)            -38.9   0.42   0.45   0.40   0.60   0.67   0.72
Manatee (O)          29.1   0.45   0.43   0.34   0.33   0.35   0.37
Marion (O)            3.2   0.43   0.41   0.36   0.40   0.42   0.45
Monroe (O)           37.6   0.49   0.47   0.36   0.36   0.39   0.44
Okeechobee (O)      -27.7   0.47   0.49   0.37   0.58   0.65   0.69
Orange (O)           23.9   0.50   0.46   0.35   0.40   0.41   0.41
Osceola (O)          18.8   0.51   0.47   0.33   0.40   0.42   0.44
Polk (O)             -4.2   0.45   0.44   0.35   0.43   0.46   0.50
St.Lucie (O)         26.5   0.53   0.49   0.35   0.41   0.41   0.44
Seminole (O)         27.9   0.43   0.39   0.30   0.32   0.33   0.35
Volusia (O)          25.7   0.53   0.49   0.42   0.41   0.43   0.46


Marc Sapir MD, MPH
http://ustogether.org/election04/sapir/sapir_response.html
states:

 > Many people, joined into a number of ad hoc groups of professionals
 > and academics, looking into possible errors in vote counts in the
 > 2004 election, have already concluded that the small counties in
 > Florida (the ones Mebane refers to as mostly panhandle states)
 > should be discounted in statistical analysis. Mebane et al merely
 > document these effects which were well known to our statistical and
 > survey research consultants who live in Florida. Moreover, further
 > analyses excluding those small counties, as posted in graphics
 > linked to this web page, have stood up well enough to continue to
 > take this effort seriously.

The first quartile of county population in FL in 2000 is 24,476.  The
counties with 2000 population at or below that level are 

Baker      22259
Calhoun    13017
Dixie      13827
Franklin   11057
Gilchrist  14437
Glades     10576
Gulf       13332
Hamilton   13327
Holmes     18564
Jefferson  12902
Lafayette  7022
Liberty    7021
Madison    18733
Taylor     19256
Union      13442
Wakulla    22863
Washington 20973

Hence Glades is the only small op-scan county in the preceding set.
Therefore my response to Sapir is the same as to Liddle.

To get rid of the counties remaining in the set (after deleting
Glades) that are exhibiting substantial "Dixiecrat" behavior, let's
try removing all counties that in 1996 had Dem registration proportion
greater than .60 and Dem vote.proportion less than .45.  Doing that
and putting the e-voting machine counties back in the display gives:

                 DemChg04 DemV00 DemV96 DemV92 DemR04 DemR00 DemR96
Broward (E)          27.5   0.67   0.64   0.52   0.50   0.51   0.53
Charlotte (E)        34.6   0.44   0.43   0.37   0.32   0.34   0.36
Collier (E)          40.0   0.32   0.32   0.26   0.24   0.24   0.25
Miami-Dade (E)       25.4   0.53   0.57   0.47   0.43   0.44   0.48
Hillsborough (E)     11.0   0.47   0.47   0.37   0.42   0.44   0.48
Indian River (E)     30.0   0.40   0.37   0.28   0.30   0.31   0.33
Lake (E)             13.6   0.41   0.40   0.33   0.34   0.36   0.37
Lee (E)              33.7   0.40   0.40   0.32   0.30   0.32   0.34
Martin (E)           51.5   0.43   0.38   0.28   0.28   0.27   0.28
Nassau (E)          -28.9   0.29   0.34   0.30   0.37   0.48   0.56
Palm Beach (E)       34.8   0.62   0.58   0.46   0.45   0.45   0.46
Pasco (E)            18.9   0.49   0.50   0.39   0.37   0.40   0.43
Pinellas (E)         30.8   0.50   0.49   0.38   0.38   0.37   0.39
Sarasota (E)         45.0   0.45   0.43   0.35   0.31   0.31   0.31
Sumter (E)          -10.9   0.43   0.46   0.41   0.41   0.48   0.57
Brevard (O)          13.7   0.45   0.41   0.31   0.37   0.38   0.40
Citrus (O)            8.3   0.45   0.44   0.36   0.39   0.41   0.46
Flagler (O)          26.6   0.51   0.48   0.41   0.38   0.40   0.42
Hernando (O)         19.0   0.50   0.49   0.39   0.39   0.40   0.42
Highlands (O)        -3.1   0.40   0.42   0.35   0.40   0.42   0.45
Manatee (O)          29.1   0.45   0.43   0.34   0.33   0.35   0.37
Marion (O)            3.2   0.43   0.41   0.36   0.40   0.42   0.45
Monroe (O)           37.6   0.49   0.47   0.36   0.36   0.39   0.44
Okeechobee (O)      -27.7   0.47   0.49   0.37   0.58   0.65   0.69
Orange (O)           23.9   0.50   0.46   0.35   0.40   0.41   0.41
Osceola (O)          18.8   0.51   0.47   0.33   0.40   0.42   0.44
Polk (O)             -4.2   0.45   0.44   0.35   0.43   0.46   0.50
St.Lucie (O)         26.5   0.53   0.49   0.35   0.41   0.41   0.44
Seminole (O)         27.9   0.43   0.39   0.30   0.32   0.33   0.35
Volusia (O)          25.7   0.53   0.49   0.42   0.41   0.43   0.46

Now among the op-scan counties only Okeechobee has a very negative
DemChg04 value.  So does one e-voting county (Nassau).  This is
nowhere near enough to support worry about vote fraud among a high
proportion of the op-scan counties.

-- 
* - * - * - * - * - * - * - * - * - * - * - * - * - * - * - * - * - *
Walter R. Mebane, Jr.                        email:  wrm1@cornell.edu
Professor                             office voice:  607/255-3868    
Department of Government                      cell:  607/592-0546
Cornell University                             fax:  607/255-4530    
217 White Hall              WWW:  http://macht.arts.cornell.edu/wrm1/
Ithaca, NY 14853-7901
* - * - * - * - * - * - * - * - * - * - * - * - * - * - * - * - * - *

My original letter (sent to the Editor of Common Dreams) is here

The responses this message is responding to are posted here specifically Liddle and Sapir

For two technical but also more rigorous statistical analyses that support our conclusion that there was no effect of machine type (Op-Scan versus E-Touch), see an analysis using a matching estimator, by Jasjeet Sekhon and an analysis using a robust binomial regression estimator, by Jonathan Wand. The data used in those papers is available here.





Walter Mebane 2004-11-16