One year ago I tried out an experiment where I plotted US Representatives' voting records against the presidential lean of their districts, in an effort to identify what representatives were not the most liberal or conservative, but who most overperformed or underperformed their districts. After some hemming and hawing, it was called the PVI/Vote Index. The point of the exercise was to give some clarity and focus to one of the most frequently heard refrains of the liberal blogosphere: "We're going to primary that ratfink so-and-so," usually delivered without much consideration as to what kind of candidate that district might actually support.
It's time to revisit the topic, partly because another year has gone by, and aggregators have released another year's worth of data, letting us look at the 110th Congress as a whole (instead of just 2007). Also, with the creation of the blogger/labor Accountability Now PAC for purposes of nudging (or primarying) recalcitrant Dems, with Progressive Punch adding a similar function to their website, and with even the Cook Political Report (subscription req'd) tipping a toe into this type of analysis, it seems like other people are starting to zero in on who is and isn't a good fit for his or her district.
As before, the Index is based on a pretty simple idea: rank every district from 1 to 435 in terms of how Democratic its presidential voting record is, rank every representative from 1 to 435 in terms of how liberal his or her voting record is, and find the difference, with a larger difference in one direction or the other meaning that representative is overperforming or underperforming the district's lean. (There are a host of methodological issues that go along with this assumption, and I'll discuss some of them over the fold. In the meantime, let's get right to the numbers.)
Let's start with Democrats who are underperforming their districts (in other words, Democrats whose voting records are less liberal than their district composition would ordinarily support):
Rep.
District
PVI
PVI rank
DW/N
Liberal rank
Difference
A. Davis
AL-07
D+17
65
- 0.286
183.5
- 118.5
Meeks
NY-06
D+38
6
- 0.397
122
- 116
Meek
FL-17
D+35
11
- 0.390
126
- 115
Jefferson
LA-02
D+28
28
- 0.371
139
- 111
Doyle
PA-14
D+22
42
- 0.363
142
- 100
Engel
NY-17
D+21
45
- 0.378
137
- 92
Brady
PA-01
D+36
9.5
- 0.439
96
- 86.5
Sires
NJ-13
D+23
39
- 0.398
121
- 82
Berman
CA-28
D+25
35.5
- 0.406
117.5
- 82
Fattah
PA-02
D+39
5
- 0.454
84.5
- 79.5
D. Scott
GA-13
D+10
112
- 0.257
191
- 79
Moran
VA-08
D+14
81
- 0.345
152.5
- 71.5
Crowley
NY-07
D+28
29
- 0.431
100
- 71
Rush
IL-01
D+35
12
- 0.455
83
- 71
Lipinski
IL-03
D+10
106.5
- 0.312
174
- 67.5
Reyes
TX-16
D+9
117.5
- 0.286
183.5
- 66
Towns
NY-10
D+41
3
- 0.492
69
- 66
Harman
CA-36
D+11
103.5
- 0.319
169
- 65.5
Rangel
NY-15
D+43
2
- 0.493
67
-65
Cooper
TN-05
D+6
144.5
- 0.211
208.5
- 64
Three of the top four underperformers here were also in the top four last year: Artur Davis, Kendrick Meek, and Bill Jefferson, which indicates that the pattern is pretty consistent. (The fourth, Greg Meeks, not coincidentally the only African-American member of the New Dems besides Davis and Meek, replaces Charlie Rangel.)
Notice something else interesting? We don't have to primary any of those three! Jefferson learned the hard way that the future is Cao, while Davis and Meek are doing us a solid by opening up their seats to run for higher office. (And if they somehow win, they'll immediately switch from goats to heroes in my book, since if they stay consistent policy-wise, they'll suddenly be vastly overperforming the lean of their states as a whole.)
But it does shine a spotlight on the open primaries in AL-07, FL-17, and LA-02. These primaries should be absolute top priority for blogosphere action: these are districts that can support progressives, not just centrists, and we have basically free shots at electing Better Democrats here. (These mostly-African-American districts may be a little outside the familiarity zone of the mostly-white blogosphere, but remember that one of our signature achievements is knocking off Al Wynn in MD-04, which is what can happen when the netroots and the local grassroots actually work in concert.)
As with last year, the list is heavy on Congressional Black Caucus members, some of whom are also Progressive Caucus members. The latter may not be terribly fruitful targets (although, again, the primaries will be very important once they retire), who are being penalized a bit unfairly for living in some of the nation's most Democratic districts. They'd need to be in McDermott/Kucinich/Lee territory to be truly apt 'fits' for their districts.
Some better targets might be a little further down the list, including frequent netroots foils like Dan Lipinski and the newly-vulnerable Jane Harman. To my eye, one of the juiciest targets is Jim Cooper, about the only representative in a district with a solidly Dem PVI who's not just voting poorly around the margins but on some of the important stuff as well (like the stimulus). Pressure on Cooper is particularly important as the focus turns to health care, as his singular influence in the health care arena gives him unique power to obstruct progessive health care policy.
Now let's turn to the good news: the Democrats who are most overperforming their districts, and who are most deserving of our praise (or in the case of the bluest Dogs, our tolerance). As with last year, it's a mix of flat-out progressives in swing or light-blue districts, and Blue Doggish types who are entrenched in deep-red districts that would likely flip without them (or, in the sad cases of Lampson and Boyda, Blue Doggish types who failed to get entrenched):
Rep.
District
PVI
PVI rank
DW/N
Liberal rank
Difference
C. Edwards
TX-17
R+18
417
- 0.240
196
221
G. Taylor
MS-04
R+16
404.5
- 0.248
193
211.5
Matheson
UT-02
R+17
408
- 0.154
222
186
Pomeroy
ND-AL
R+13
379
- 0.245
194
185
DeFazio
OR-04
D+0
200
- 0.602
27
173
Lampson
TX-22
R+15
390
- 0.038
234
156
Doggett
TX-25
D+1
187.5
- 0.533
49
138.5
Herseth
SD-AL
R+10
337
- 0.234
199
138
Skelton
MO-04
R+11
347
- 0.203
212
135
Hinchey
NY-22
D+6
147
- 0.685
13
134
Stupak
MI-01
R+2
228.5
- 0.436
97
131.5
Filner
CA-51
D+7
137
- 0.723
9.5
127.5
Oberstar
MN-08
D+4
160
- 0.570
36
124
Kucinich
OH-10
D+8
125
- 0.791
3
122
Spratt
SC-05
R+6
283.5
- 0.325
165
118.5
Obey
WI-07
D+2
185
- 0.486
72
113
Chandler
KY-06
R+7
300.5
- 0.256
192
108.5
Rodriguez
TX-23
R+4
254.5
- 0.348
150
104.5
Boyda
KS-02
R+7
308
- 0.218
206
102
Boucher
VA-09
R+7
303
- 0.232
201
102
One advantage of the PVI/Vote Index is that, at the same time as shining a light on Democrats who are lagging their districts, it also illuminates right-wing Republicans camped out in moderate districts, who should theoretically be vulnerable a good Democratic challenger because of their poor fit with their districts. If there's any doubt, check out which of these nutjobs who've overperformed their districts got defeated in 2008, and how many more got a serious scare.
Rep.
District
PVI
PVI rank
DW/N
Liberal rank
Difference
Ryan
WI-01
R+2
224
0.690
397
- 173
Feeney
FL-24
R+3
241
0.744
409
- 168
Chabot
OH-01
R+1
205.5
0.626
372
- 166.5
Garrett
NJ-05
R+4
261
0.771
417
- 156
Shadegg
AZ-03
R+6
288.5
0.903
429
- 140.5
Rohrabacher
CA-46
R+6
291
0.826
424.5
- 133.5
Kline
MN-02
R+3
233.5
0.616
365
- 131.5
Bilbray
CA-50
R+5
264
0.684
394
- 130
Fossella
NY-13
D+1
191
0.507
317
- 126
Walberg
MI-07
R+3
230.5
0.589
356.5
- 126
Weldon
FL-15
R+4
251.5
0.622
367.5
- 116
Campbell
CA-48
R+8
311
0.826
424.5
- 113.5
Bachmann
MN-06
R+5
273.5
0.663
385.5
- 112
Manzullo
IL-16
R+5
263
0.630
374
- 111
Franks
AZ-02
R+9
322
0.910
431
- 109
Tiberi
OH-12
R+1
210
0.508
318
- 108
Royce
CA-40
R+8
315
0.794
421
- 106
Roskam
IL-06
R+3
236.5
0.552
341
- 104.5
Mica
FL-07
R+4
251.5
0.583
355
- 103.5
Castle
DE-AL
D+7
142
0.291
245
- 103
Finally, one last table: the Republicans who are "underperforming" their very conservative districts. While there are a few moderates here (like the primaried-out Wayne Gilchrest), mostly it's semi-sane conservatives in some of the darkest-red districts in the nation. I'm keeping this list to 10, as either way, there's not much we can do about these guys, other than sit back and watch as the Club for Growth goes after them with chainsaws. (Note that Jerry Moran, who's vacating his seat to run for Senate, is one of them. His moderation, relatively speaking, may be an asset for him when running statewide.)
Rep.
District
PVI
PVI rank
DW/N
Liberal rank
Difference
W. Jones
NC-03
R+15
395
0.279
242
153
Simpson
ID-02
R+19
421
0.397
271
150
Moran
KS-01
R+20
427
0.442
286
141
Platts
PA-19
R+12
366
0.327
255.5
110.5
D. Young
AK-AL
R+14
387
0.420
278.5
108.5
Lucas
OK-03
R+18
414
0.493
310
104
Crenshaw
FL-04
R+16
407
0.489
308
99
Bachus
AL-06
R+25
433
0.538
335.5
97.5
Gilchrest
MD-01
R+10
335
0.254
238
97
Aderholt
AL-04
R+16
399
0.476
303
96
Much more discussion of the methodology and what this all may mean, over the flip.
We need to talk methodology briefly. I'll do this as a Q&A in order to make it a little livelier.
What the heck is DW/N? As my primary vote-aggregating resource, I'm using DW/Nominate scores, which are a tool I used in a number of vote-scoring-themed diaries last summer. The main advantage DW/N has over other scores is that they aggregate absolutely every vote, instead of cherry-picking. (ADA ratings and CQ party unity scores, for instance, pick so few votes that it's terribly insufficient gradation among representatives; nearly all Dems have an ADA score of 90, 95, or 100, while nearly all have a CQ score in the 96-98 range... which is why I don't use either of those metrics.) In each case, I'm using the DW/N score of whatever representative ended the session holding the seat, even if someone else held it the majority of the term.
On the other hand, most everyone else (Progressive Punch, National Journal, CQ, the ADA) uses a 0-100 score, with 100 being most liberal, which is easy for people to mentally convert to the A-B-C-D-F grading scale. By comparison, DW/Nominate scores are difficult to interpret. The scores generally run from - 1 (most liberal) to 1 (least liberal). The scoring algorithm seems to measure similarities between voting records among representatives; a number further away from 0 indicates a greater amount of distance between your record and those of other reps. In fact, if your voting record doesn't look anything like anybody's elses, you can exceed the 1 to -1 range (as with Ron Paul's 1.4).
You may recall last year, to do this project, I created one averaged-out liberal rating using both Progressive Punch and National Journal scores. While I'd very much like to use Progressive Punch scores again -- I think they do the best job of the "just right" amount of vote cherry-picking and turning it into an easy-to-understand score -- they've already turned their attention to the 111th Congress now in session and their old scores from the 110th have already vanished from public view.
And rather than try to average out DW/Nominate and National Journal scores, I've just decided there's too much apples and oranges going on there. This is partly because of the different scoring techniques, which results in some odd discrepancies... National Journal's method is insensitive to 'purity' votes (i.e. voting against something not because you disagree but because it doesn't go far enough) so the furthest-left or right members of the caucuses (like McDermott, Kucinich, Stark, DeFazio, Woolsey, Waters, or Capuano for the Dems, or Paul for the GOP) tend to get buried in mid-caucus or even treated as centrists.
More importantly, though, there are 19 seats for which there is no National Journal composite score for both 2007 and 2008, mostly because the seat changed hands in a special election (or because of a lot of absences, either for sickness or leadership duties). As a result, what I've decided to do is run entirely separate tables based solely on National Journal numbers. As you can see, many of the same people appear relatively in the same places. Members for whom there are scores, and the PVI of their districts, are rated 1-416 instead of 1-435. (The missing parties are Pelosi, Lantos/Speier, Millender-McDonald/Richardson, Crenshaw, Norwood/Broun, Rush, Hastert/Foster, Carson/Carson, Jindal/Scalise, Baker/Cazayoux, Wynn/Edwards, Meehan/Tsongas, Wicker/Childers, Andrews, Gillibrand, Gilmoor/Latta, Pryce, Davis/Wittman, and Cubin.)
Here are the tables based on National Journal composite scores instead. (There is a rating for both 2007 and 2008, so I averaged the two to get one score for each. Again, representatives and districts are ranked only 1 to 416 in this series, because scores aren't available for 19 seats.) Here are the underperforming Dems:
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
A. Davis
AL-07
D+17
60
58.15
181
- 121
Jefferson
LA-02
D+28
25
74.35
119
- 94
Meek
FL-17
D+35
10
78.25
95.5
- 85.5
Capuano
MA-08
D+33
17
77.85
101
- 84
Stark
CA-13
D+21
38
74
120
- 82
Serrano
NY-16
D+43
1
80.75
80
- 79
G. Green
TX-29
D+8
124.5
54.1
198
- 73.5
Lipinski
IL-03
D+10
99.5
61.95
172
- 72.5
Emanuel
IL-05
D+18
56
72.8
128
- 72
C. Brown
FL-03
D+16
63
71.6
134.5
- 71.5
Ryan
OH-17
D+14
71
70.05
142
- 71
Maloney
NY-14
D+26
27
78.15
97
- 70
Meeks
NY-06
D+38
6
81.05
75.5
- 69.5
M. Udall
CO-02
D+8
118
57
186.5
- 68.5
Engel
NY-17
D+21
40
77
108
- 68
Woolsey
CA-06
D+21
42
76.75
109
- 67
Reyes
TX-16
D+9
110.5
61
176
- 65.5
Berkley
NV-01
D+9
113
60.55
177
- 64
Waters
CA-35
D+33
18
80.25
81.5
- 63.5
Cooper
TN-05
D+6
136.5
53.95
199
- 62.5
Here are overperforming Dems:
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
C. Edwards
TX-17
R+18
400
55.7
193
207
Pomeroy
ND-AL
R+13
363
61.55
175
188
Matheson
UT-03
R+17
391
48.85
218
173
G. Taylor
MS-04
R+16
388.5
48.65
219
169.5
Skelton
MO-04
R+11
332
58.55
180
152
Lampson
TX-22
R+15
374
45.4
227
147
Obey
WI-07
D+2
177
85.15
42
135
Herseth
SD-AL
R+10
323.5
52.6
203
120.5
Spratt
SC-05
R+6
272.5
65.75
153
119.5
Price
NC-04
D+6
138
89.1
22
116
Mollohan
WV-01
R+6
275.5
63.75
163
112.5
Delahunt
MA-10
D+9
116
91.4
11
105
Hinchey
NY-22
D+6
139
86.4
35
104
Boyda
KS-02
R+7
296
55.4
194
102
Boucher
VA-09
R+7
291
56.55
189
102
Hooley
OR-05
D+1
185
80.1
85.5
99.5
Holden
PA-17
R+7
288.5
56.1
191
97.5
Chandler
KY-06
R+7
288.5
55.15
195
93.5
Capps
CA-23
D+9
108
90.35
16
92
Grijalva
AZ-07
D+10
103
91.2
12
91
Here are the overperforming Republicans:
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
Kline
MN-02
R+3
223.5
10
398
- 174.5
Mica
FL-07
R+4
241.5
11.75
386
- 144.5
Feeney
FL-24
R+3
231
14.1
372.5
- 141.5
Ryan
WI-01
R+2
215
16.4
352
- 137
Shadegg
AZ-03
R+6
277.5
5.5
412
- 134.5
Bachmann
MN-06
R+5
262.5
10.55
396
- 133.5
Weldon
FL-15
R+4
241.5
14.35
370.5
- 129
Saxton
NJ-03
D+3
162.5
30.95
280
- 117.5
H. Wilson
NM-01
D+2
170
28.7
283
- 113
Garrett
NJ-05
R+4
251
15.45
360
- 109
Walberg
MI-07
R+3
221
20.6
328
- 107
Putnam
FL-12
R+5
267.5
14
374
- 106.5
Franks
AZ-02
R+9
310
4.7
416
- 106
Chabot
OH-01
R+1
197.5
24.9
301.5
- 104
Fossella
NY-13
D+1
183
28.2
286
- 103
S. King
IA-05
R+8
305
6.5
407.5
- 102.5
Latham
IA-04
D+0
188
28.25
285
- 97
M. Rogers
MI-08
R+2
212
24.1
308.5
- 96.5
Royce
CA-40
R+8
303
9.9
399
- 96
Akin
MO-02
R+9
311
6.6
406
- 95
And finally, here are the underperforming Republicans:
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
Moran
KS-01
R+20
408
33.45
268
140
W. Jones
NC-03
R+15
379
38.5
247
132
Simpson
ID-02
R+19
403
31.5
275
128
D. Young
AK-AL
R+14
371
36.25
257
114
Paul
TX-14
R+14
373
35.3
261
112
Gilchrest
MD-01
R+10
322
49.4
217
105
Bachus
AL-06
R+25
414
23.2
314
100
Platts
PA-19
R+12
351
36.75
254.5
96.5
Inglis
SC-04
R+15
375
31.1
279
96
Emerson
MO-08
R+11
335
40.55
240
95
Is this old or new PVI? This is Classic PVI, calculated using the Cook formula and based on 2000-2004. My rationale is that their 110th Congress votes all predate the 2008 election, so if representatives actually were taking their districts' lean into consideration, it would be based on the previous elections.
Still, this raises the interesting question of whether the 2008 election results have changed the dynamic for representatives in those few districts that changed dramatically one way or the other (for instance, are Marion Berry or Mark Kirk more endangered now?). Perhaps some of them might change their records in the 111th Congress, for better or worse, to reflect what they can see is happening on the ground in their districts. Accordingly, I'm creating yet more tables... this time, based on the newly released Cook PVIs reflecting the 2004 and 2008 elections. (I'm leaving out defeated or retired representatives from these lists, which, for instance, removes Jefferson from the underperforming Dems list, or Chabot and Feeney from the overperforming GOPers list.)
Here are the underperforming Dems. Not that many dramatic changes, but note that David Scott has zoomed up to near the top of the list, as his previously D+10 district in the Atlanta exurbs became D+15 on the strength of a huge influx of African-American residents. He's one of only two black Blue Dogs (Sanford Bishop in rural GA-02 is the other one, and is a fine fit), and might want to rethink that.
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
A. Davis
AL-07
D+18
60
- 0.286
183.5
- 123.5
D. Scott
GA-13
D+15
71
- 0.257
191
- 120
Meeks
NY-06
D+36
7
- 0.397
122
- 115
Meek
FL-17
D+34
13
- 0.390
126
- 113
Moran
VA-08
D+16
64
- 0.345
152.5
- 88.5
Doyle
PA-14
D+19
55
- 0.363
142
- 87
Brady
PA-01
D+35
10
- 0.439
96
- 86
Fattah
PA-02
D+38
4
- 0.454
84.5
- 80.5
Berman
CA-28
D+23
37
- 0.406
117.5
- 80.5
Engel
NY-17
D+18
58
- 0.378
137
- 79
Schiff
CA-29
D+14
75
- 0.347
151
- 76
Reyes
TX-16
D+10
109
- 0.286
183.5
- 74.5
Sires
NJ-13
D+21
48
- 0.398
121
- 73
Harman
CA-36
D+12
98
- 0.319
169
- 71
Rush
IL-01
D+34
14
- 0.455
83
- 69
S. Davis
CA-53
D+14
74
- 0.362
143
- 69
Crowley
NY-07
D+26
32
- 0.431
100
- 68
Lipinski
IL-03
D+11
106
- 0.312
174
- 68
Towns
NY-10
D+38
3
- 0.492
69
- 66
Rangel
NY-15
D+41
2
- 0.492
67
- 65
Here are the overperforming Dems. Gene Taylor overtakes Chet Edwards at the top, based on the different direction their districts are going. There are definitely more Blue Dogs and their ilk on the new list than the old list, thanks to a number of southern uplands districts plunging from GOP-leaning to dark-red (Boren, Berry, Lincoln Davis, Gordon, Mollohan, Rahall).
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
G. Taylor
MS-04
R+20
415
- 0.248
193
222
C. Edwards
TX-17
R+20
417
- 0.240
196
221
Matheson
UT-02
R+15
386
- 0.154
222
164
Skelton
MO-04
R+14
374
- 0.203
212
162
DeFazio
OR-04
D+2
183
- 0.602
27
156
Boren
OK-02
R+14
368
- 0.128
224
144
Pomeroy
ND-AL
R+10
336
- 0.245
194
142
Berry
AR-01
R+8
300
- 0.338
159
141
Boucher
VA-09
R+11
342
- 0.232
201
141
L. Davis
TN-04
R+13
359
- 0.177
218.5
140.5
Melancon
LA-03
R+12
344
- 0.220
205
139
Oberstar
MN-08
D+3
174
- 0.570
36
138
Childers
MS-01
R+14
369
- 0.010
236
133
Mollohan
WV-01
R+9
321
- 0.268
189
132
Stupak
MI-01
R+3
228
- 0.436
97
131
Gordon
TN-06
R+13
350
- 0.171
220
130
Hinchey
NY-22
D+6
140
- 0.685
13
127
Spratt
SC-05
R+7
289
- 0.325
165
124
Rahall
WV-03
R+6
286
- 0.331
163.5
122.5
Grijalva
AZ-07
D+6
138
- 0.655
18
120
Here are the overperforming Republicans. There's a lot of turnover on this list from the old PVI list, but that has more to do with defeats and retirements than vastly changed districts (Chabot, Feeney, Walberg, Tancredo, Musgrave, Keller, Tom Davis, Heather Wilson, Weldon, and Knollenberg would all clock in higher than Dave Camp). However, note the sudden appearance of a lot of Illinois, Michigan, and California districts on the new list, based on Obama's strong performance in those states.
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
Ryan
WI-01
R+2
218
0.690
397
- 179
Rohrabacher
CA-46
R+6
262
0.826
424.5
- 162.5
Bilbray
CA-50
R+3
232
0.684
394
- 162
Campbell
CA-48
R+6
263
0.826
424.5
- 161.5
Manzullo
IL-16
R+2
222
0.630
374
- 152
Roskam
IL-06
Even
193
0.552
341
- 148
Tiberi
OH-12
D+1
192
0.508
318
- 126
Garrett
NJ-05
R+7
291
0.771
417
- 126
Kline
MN-02
R+4
239
0.616
365
- 126
Royce
CA-40
R+8
303
0.794
421
- 118
Lungren
CA-03
R+6
261
0.641
377
- 116
Castle
DE-AL
D+7
133
0.291
245
- 112
Shadegg
AZ-03
R+9
317
0.903
429
- 112
Biggert
IL-13
R+1
210
0.512
321
- 111
Kirk
IL-10
D+6
142
0.320
251.5
- 109.5
Petri
WI-06
R+4
246
0.568
350.5
- 104.5
Sessions
TX-32
R+8
302
0.727
403.5
- 101.5
Rogers
MI-08
R+2
212
0.498
313.5
- 101.5
Pitts
PA-16
R+8
304
0.728
405
- 101
Camp
MI-04
R+3
236
0.538
335.5
- 99.5
And here are the underperforming Republicans...
Rep.
District
PVI
PVI rank
NJ
Liberal rank
Difference
W. Jones
NC-03
R+16
397
0.279
242
155
Moran
KS-01
R+23
424
0.442
286
138
Simpson
ID-02
R+17
406
0.397
271
135
Emerson
MO-08
R+15
385
0.329
257
128
Aderholt
AL-04
R+26
430
0.476
303
127
Lucas
OK-03
R+24
425
0.493
310
115
R. Hall
TX-04
R+21
422
0.503
315
107
Whitfield
KY-01
R+15
381
0.430
281
100
Bachus
AL-06
R+29
435
0.538
335.5
99.5
Crenshaw
FL-04
R+17
407
0.489
308
99
Why aren't you using regression instead of discrete ranks? Good question. Last year, I used discrete ranks because that's all I knew how to do. I've gotten a lot more familiar since then with some of the more advanced tools in Excel, so when I set out to re-do this project this year, my first attempts tried turning this into a regression exercise. The results, however, weren't as satisfactory, so I went back to what I knew worked.
To illustrate this, take a look at the results graphed as a scatterplot (DW/Nominate data on the vertical axis, PVI on the horizontal axis).
For the most part, you can see very clear correlations, as the two parties cluster tightly but also have nice slopes indicating the relationship between voting record and PVI. (And the gap between the two parties shows how even the worst Dem is still much better than the best GOPer.) However, notice that the tight cluster of Dems start to flatten out and then turn into more of a spray as you get into the districts with super-high PVIs.
The representatives in the highest-PVI districts aren't especially more progressive than the ones in the lower but still solid-blue districts. A regression line, however, would predict that because of the great distance between say, the D+43 standing of NY-15 and NY-16 and the other districts, therefore the reps from these districts should also be much, much more progressive than anyone else. This is particularly a problem when using National Journal scores, which max out at 100; even if we somehow elected Angela Davis in those districts, she still would be underperforming because the "TREND" function on Excel says that to be a good fit, you have to bend the laws of mathematics and get a National Journal score of 105.
Just for the sake of comparison, here's the Top 10 most underperforming Dems according to a true regression model (based on residuals, which are the differences between the projected voting records according to the trendline and actual voting records): Serrano, Rangel, Meeks, Towns, Meek, Fattah, Brady, Rush, Pelosi, and Clarke. As you can see, there are a number of commonalities between this list and the list above... but a perceptive reader will also notice that these are basically just the people in the districts with the top 10 highest PVIs, in approximately the right order (although Kendrick Meek is still somewhat out-of-whack). On the Republican side, the method also views Ron Paul as the most overperforming and potentially vulnerable Republican (although after that it settles down to the usual suspects: Franks, Rohrabacher, Ryan, etc.).
So, I discarded the method that just tells me that our juiciest primary targets are the representatives who are in our bluest districts. Switching back to discrete ranks comes with its own problems: while it flattens out the distortion at the margins, it may create some distortion in the middle, where it may place more importance than should be accorded on small DW/Nominate score differences among members who are clustered near the median of their caucuses. So, as always, I welcome any thoughts on the methodology here and how to make this the fairest-possible assessment of House members' fits.
One other idea I've mulled over is the possibility of ranking representatives only against a cohort of the representatives in 5 or 10 most similar districts, similar not just in terms of presidential voting but other demographic characteristics. For example, let's take a look at AL-07, of interest to us not just because it's where the most out-of-whack Dem (Artur Davis) is, but because it's a district that points to the flaws of using only PVI as an indicator of who's a good fit, as it's a good bit more socially conservative than most other districts in the D+18 range.
One would want to compare Davis to the representatives in MS-03 and SC-06, similar not just in terms of their PVI but also their racial composition. Beyond that, TN-09, NC-12, FL-03, and VA-03 are also similar in region, PVI, and race, though somewhat less similar, in terms of being less impoverished and more urban. If you average out the DW/N scores for Bennie Thompson, Jim Clyburn, Steve Cohen, Mel Watt, Corrine Brown, and Bobby Scott, you get - 0.462 (ranging from - 0.544 for Watt to - 0.403 for Clyburn), so Davis's - 0.286 still points to a significant under-performance. Of course, I'd need to develop a statistical method for analyzing which districts truly are the most similar to each other using a similar technique as Nate Silver's state similarity index (rather than simply saying "Hmm, these ones seem similar"), so while this method seems the most promising to me, it'll still need a lot of work too.
