Monopoly Politics 2002: Full Report

Monopoly Politics 2002: How “No Choice” Elections Rule in a Competitive House
The Center for Voting and Democracy
September 2002

I. Overview

II. The Center’s U.S. House Election Projections, 1996 – 2002
A. Total projections
B. Particularly vulnerable incumbents
C. Incumbent vs. incumbent match ups favor Republicans in 2002
D. Projections in open seats

III. How Our Projection Model Works

IV. The Power of Partisanship

V. Why Partisanship Outweighs Campaign Spending in General Elections
A. Charts on campaign spending and results
B. Open seat elections, 1994-2000

VI. Profile: The Entrenched Incumbent

VII. Redistricting 2001-2002 and Incumbent Protection

VIII. Short Takes
A. Who “won” redistricting and the long-term partisan landscape
B. A surprising role for third parties and independents
C. Prospects for women and racial minorities
D. State legislatures – Incumbent projection may be even worse

IX. Appendices
A. Details of model
B. State-by-state data on the partisan outcome of redistricting
C. Monopoly Politics 2002 projections in all 435 districts

The Center for Voting and Democracy is a non-profit, non-partisan organization that promotes fair elections. For more information about this report and other work of the Center, please contact:

The Center for Voting and Democracy
6930 Carroll Avenue, Suite 610
Takoma Park, MD 20912
(301) 270-4616        
I.  Overview

Democrats and Republicans are as narrowly divided in the U.S. House of Representatives as they have been in decades. In each of the last three House elections, the total number of votes received by Republican and Democratic House candidates was nearly dead even. Most political observers see the question of which party will control the House after the November elections as essentially a toss-up, mirroring the close partisan division evidenced in the 2000 presidential elections, the battle for control of the U.S. Senate and state elections around the nation.
Yet this fierce competition for power should not disguise the fact that the outcomes in the great majority of U.S. House races are so predictable that few Americans have any chance to change their own representation in the House. In both the 1998 and 2000 general elections, for example, fewer than one in ten House races were won by competitive margins of less than 10%. More than 98% of incumbents were re-elected in both years, and of the twelve defeated incumbents, only one had been elected before 1992.

What’s going on and what does it tell us about who will control Congress in the future?

Starting in 1997, the Center for Voting and Democracy has released a biannual report, Monopoly Politics, that projects the outcome of most House races based on a simple, but powerful observation: the partisan division in most districts determines the winner in most elections. In most House races, we can project not only who will win but by what margin without knowing anything about the identify of the challenger, about the voting record or any other characteristic of the incumbent, about campaign spending in past or current elections or about polling data and organizational endorsements. All we need to know are the results from recent federal elections in the district and the incumbent’s party and seniority.

Of the 361 House winners which we projected winners in July 1997, for example fully 358 won, including 346 who won by comfortable margins of at least 10%. Of the 340 incumbents we projected would win, only one lost. In the 18 months between our 1997 projections and the 1998 elections, a great deal happened – the controversy over whether to impeach President Clinton, the launching of campaigns by well-funded challengers and more – but those developments had little impact on these “no choice” elections.

Similarly, nearly two years before the November 2000 elections, we projected 235 races to be won by more than “landslide” margins of more than 20%. We were wrong in only a single race, which was won by a margin of “merely” 18%.

This year, we have developed a new “one size fits all” projection model to underscore the power of our analysis. Our past projections considered some relatively subjective factors such as regional differences – the fact that Democrats run more strongly than one might expect in the northern Plains and Texas, for instance, and Republicans run more strongly than expected in New York State. But our new projection model relies exclusively on a few simple variables to make projections, with no variation by year of election and by region. To make it clear that our process of making projections is completely transparent, we are making our spreadsheets with data and calculations easily available to the public on the web.

In order to avoid errors, our simple model makes fewer projections than in previous years, but it is even more accurate and still narrows the competitive election playing field to fewer than one in four races. Applying our model to elections in 1996-2000, there is only a single error out of 930 total party projections. Out of 515 total projections of landslide victories of more than 20% (totaling well over a third of all House races taking place in this period), our model was accurate 497 (97%) of the time in its margin prediction. Of 18 landslide projection errors, 13 races were still won by comfortable margins of greater than 10%, and none of these seats was wrong in its partisan projection. Overall, our model is accurate on victory margin projection in 96% of races from 1996 to 2000.

This year, for November 2002 our model projects 332 U.S. House winners, including 195 by margins of at least 20% and another 100 by margins of at least 10%. This is a larger number of projections than even in the non-competitive elections at the end of the 1990’s. And of the remaining 104 districts, most in fact will not be competitive.

Opportunities for voters to define their own representation and for new groups to be represented in the halls of Congress thus are all too rare. Our findings suggest that for most Americans interested in influencing which party controls the House, it would be more effective to send a campaign contribution to candidates in a competitive race than vote in their own because most of us live in thoroughly uncompetitive districts.

We certainly don’t advocate Election Day boycotts, but the two best chances for most people to change their representation in “the people’s House” for the next decade have already passed – in this year’s primary’s election, where nominees were chosen to run in districts that are mostly safe for their party, and even more importantly, in the recent process of redistricting when the landscape of U.S. House elections for the next decade was largely determined. But voter turnout in primaries was at a near-historic low of less than 20%, according to the Committee for the Study and the American Electorate, while few citizens had a glimpse behind the closed doors that mark a redistricting process in which legislators choose their constituents before their constituents choose them.

This language might seem extreme, but the most entrenched Members of the U.S. House could tell you otherwise. Before retiring after the 2000 elections, Pennsylvania Republican Bud Shuster faced a general election opponent exactly once in his seven elections. Many other incumbents face at best token opposition. In fact, nearly one out of every ten incumbent House Members has won each of his or her last five elections by more than 40%. Every single one of these entrenched incumbents represents a district that leans heavily toward his or her party. In this year’s general election they essentially have zero chance to lose as well. Indeed, it is one thing to point out that the percentage of winning incumbents has not dropped below 90% in more than a quarter-century. It is another to add that most incumbents represent districts that even if open, would be safe for their party.

And it doesn’t stop at the U.S. House. All evidence suggests that state legislative elections are even less competitive than House races and that state legislators are all the more zealous in protecting their political fortunes in redistricting. In 1998 and 2000, more than two out of every five state legislative races were uncontested. That number likely will be similar in 2002. While some states like Arizona, Idaho, Iowa and Washington have adopted redistricting processes that are designed to protect the public interest, most states give legislators nearly unchecked power to draw their own lines. And in many states, there is simply no way to make certain areas competitive for more than one party as long as we use “winner-take-all” elections in which a simple majority of voters has the power to define representation for everyone.
Due to redistricting, election data for many districts was not available until recently, which explains why we are releasing our projections so close to the election. Immediately after the 2002 election, however, we will have nearly all of the pieces of information we need to be able to release our model’s projections for as many as four in five of the November 2004 House race. Between that time and November 2004, the only factor we will use to adjust our projections is whether a seat becomes an open seat. This year’s projections and those for 1996-2000 can be understood in the same way. They do not rely on “horserace” factors like campaign financing and tactics or external events such as the state of the economy.

Finally, we are quick to admit that there are anywhere from 40 to 100 House races in any given year in which campaign factors and quality of representation indeed do matter. While our model provides insight into what might happen in those races, we do not want to overstate its power. Indeed the latest campaign financing patterns indicate that those wielding campaign cash know where their money makes a difference. Where races are within our margin of error, the amount of spending is rising to astronomical levels, as in last year’s special U.S. House election in Virginia where more than $10 million was poured into a swing district. Where races are already decided, however, most money only flows to incumbents from friends and organized interests who want to help the incumbent fend off future primary challenges or gain more power to influence colleagues and political players in their district.

II.  The Center’s U.S. House Election Projections, 1996 – 2002

Our projection model represents cautious estimates about how poorly incumbent might do if they have a bad year in a year in which the national vote is closely divided between the major parties. Most incumbents will run significantly better than we project, even if they have weak campaigns, but our goal is to minimize the number of races in which incumbents do worse than projected. Our projections easily can be modified to gauge the impact of a shift in the two-party vote.

To test the model, we calibrated it against House elections from 1996 to 2000. Of 930 races between 1996 and 2000 in which we made predictions, there was only one error in winning party, and only 33 (3.5%) errors in the winning margin (see spreadsheet to view calculations).
A.  Total projections

Table 1.  Number of projections by category, 1996-2000 and 2002

Projection  1996  1998  2000  2002
Landslide  148  176  191  195
Comfortable  103  92  86  100
Win   37  44  53  37
Total wins  288  312  330  332

Most vulnerable    9    2    4  17
No projected win 138  121  101  86

Table 2.  Errors in projected party, 1996-2000

Projection  1996  1998  2000
Landslide  0  0  0
Comfortable  0  0  0
Win   1  0  0
Total   1  0  0

Table 3.  Accuracy of projected party, 1996-2000

Projection  1996  1998  2000
Landslide  100.0% 100.0% 100.0%
Comfortable  100.0% 100.0% 100.0%
Win   97.3%  100.0% 100.0%
Total   99.7%  100.0% 100.0%

Table 4.  Errors in projected range, 1996-2000
Projection  1996  1998  2000  
Landslide  6  8  4  
Comfortable  9  2  3  
Win   1  0  0  
Total   16  10  7  

Table 5.  Accuracy of projected range, 1996-2000
Projection  1996  1998  2000
Landslide  95.9%  95.5%  97.9%
Comfortable  91.3%  97.8%  96.5%
Win   97.3%  100.0% 100.0%
Total   94.4%  96.8%  97.9%

Table 6.  Results of Races in Districts with Errors in Landslide Projections
  Result (by margin)
Year Races Comfortable Win  Loss 
1996 6 5  1  0 
1998 8 5  3  0 
2000 4 3  1  0 
Total 18 13  5  0

Table 7.  Results of Races in Districts with Errors in Comfortable Projections
Year Races Win Loss  
1996 9 9 0  
1998 2 2 0  
2000 3 3 0 
Total 14 14 0

Out of 930 projections for wins, the one error was the 1996 defeat of Rep. Funderburk (R, NC-2). His cautious performance projection was 54%, just 1% outside the “no projected win” category. Of the errors in projected margin, all but five were wrong by just one category of victory margin (meaning, for instance, a projected landslide winner instead was victorious by a comfortable margin of between 10% and 20%).

The appendix contains a full listing of our projections, while our spreadsheets for the 1996-2000 trio of elections and this year’s elections allow users to see what variables we use and to experiment with adjusting the variables to see their impact. Our model also sheds light on some particularly high-profile races detailed below.

B. Particularly vulnerable incumbents

Our “one size fits all” model suggests the following incumbents had the potential to be particularly vulnerable. They are listed in the order of being most vulnerable to least vulnerable according to the model’s projections. Most of these candidates will win re-election, and some indeed will win quite easily – for example, Forbes is actually uncontested in Virginia – but the incumbents who were projected as vulnerable had a significantly higher defeat rate than other incumbents in 1996-2000.

Incumbent  State   CD Cautious Projection
Jim Matheson (D) Utah   2 37%
Connie Morella (R) Maryland  8 39%
Charles Stenholm (D) Texas   17 39%
Kenneth Lucas (D) Kentucky  4 41%
Chet Edwards (D) Texas   11 42%
Earl Pomeroy (D) North Dakota  1 42%
Ralph Hall (D)  Texas   4 43%
Dennis Moore (D) Kansas   3 43%
Max Sandlin (D) Texas   1 45%
Ted Strickland  (D) Ohio   6 45%
Bill Luther (D) Minnesota  2 46%
Jim Turner (D) Texas   2 46%
Anne Northup (R) Kentucky  3 46%
John Spratt (D) South Carolina 5 46%
Rob Simmons (R) Connecticut  2 47%
Baron Hill (D)  Indiana   9 47%

Of these 17 incumbents, 13 are Democrats, and only 4 are Republicans.  This does not mean necessarily that fewer Republican incumbents will lose in November. Rather, it suggests that the underlying political landscape of district partisanship holds more threats for Democrats than Republicans.

C. Incumbent versus incumbent match ups favor Republicans in 2002

Redistricting has thrown several incumbents into the same districts. Four incumbents lost to colleagues in primaries this year, and another four will lose to colleagues in the general elections.
The landscape of three of the races (IL-19, MS-3 and PA-17) clearly favors the Republicans, while one race (CT-5) is a toss up.

A review of the partisanship landscape in these races shows the districts the incumbents held from 2000 to 2002, each incumbent’s most recent two general election results and the partisanship of the districts in 2000 and in 2002. Note that the partisanship scale goes from 0% (completely Republican) to 100% (completely Democratic).

Connecticut (CT-5 in 2002)

   Results Partisanship  
   1998 2000 2000 2002 Cautious Projection
CT-5 Maloney (D) 50% 54% 53% 54% Dem 50%, “no projection”
CT-6 Johnson (R) 58% 63% 54% 54% Rep 54%, “win”

If Democrat Maloney were not running against another incumbent, the model would project a cautious result of 50% due to his previous close races. If Republican Johnson were not running against an incumbent, the model would project her as a “win” projection. Given that she is facing a tested incumbent in a district that leans toward his Democrat, however, we see this race as a toss up.

Illinois (IL-19 in 2002)

  Results Partisanship 
  1998 2000 2000 2002 Cautious Projection
IL-19 Phelps (D) 58% 65% 44% 42% Dem 48%, “no projection”
IL-20 Shimkus (R) 61% 63% 47% 42% Rep 63%, “landslide”

Even if facing a non-incumbent, Democrat Phelps would be on the edge of being listed as “most vulnerable” because the strong Republican leanings of the district outweigh his strong performances. If facing a non-incumbent, on the other hand, Shimkus would be projected as a landslide winner. With two incumbents, the Republican is projected to win.

Mississippi (MS-3 in 2002)

  Results Partisanship 
  1998 2000 2000 2002 Cautious Projection
MS-3 Pickering (R) 85% 73% 35% 38%* Rep 65%, “landslide”
MS-4 Shows (D) 53% 58% 47% 38%* Dem 42%, “vulnerable”

We lacked presidential results in the new Mississippi 4, but do have its racial composition, which is an important number given how closely the vote for Al Gore in Mississippi’s districts 3 and 4 tracked the African American population in 2000. With an African American population of 33%, the new district much more closely resembles Pickering’s old district than Shows. Based on this proxy for calculating the 2002 partisanship, our model projects a comfortable Republican win.

Pennsylvania (PA-17 in 2002)

  Results Partisanship
  1998 2000 2000 2002 Cautious Projection
PA-17 Gekas (R) 100% 72% 39% 43% Rep 61%, “landslide”
PA-6  Holden (D)  61% 66% 45% 43% Dem 49%, “no projection”

Gekas’ projection would indicate a landslide win, while Holden’s would be no projection. In the head-to-head race, our model suggests a Republican victory. Given Holden’s strong performance in a similarly Republican-leaning district in 2000, however, he cannot be completed counted out.

D.  Projections in Open Seats

Although incumbency in itself is not as powerful a factor as many might assume, open seats still are certainly where changes in party control are disproportionately likely. This year, of 49 open seats, our model projects 9 seats going to the Democrats, including one with a vacating Republican incumbent, and 11 seats going to Republicans. Overall, the Republicans have an edge in political demography in 28 of 49 seats, but 17 of those 28 seats are ones where Democrats have a fighting chance.  (See Table 9, p. 14)

Because open seats are less predictable than other seats, our model uses a generous cushion. In 29 open seats, we do not project an outcome, although our model would suggest clear leanings in many of those races. At the same time, the history of a strong correlation between district partisanship and open seat success in 1994-2002 detailed in our discussion on page 14 would suggest that many open seats being termed as potentially competitive in fact are unlikely to be so.

III.  How Our Model Works

The model uses four pieces of information about each district: 

1. For incumbents, their previous two election results,
2. The number of terms served by the incumbent,
3. The partisanship of the district, which is based only on the votes in the most recent presidential election received by the Democratic candidate in that district, and
4. Whether an incumbent is running for re-election or whether the seat is open.

We determine the district partisanship by a simple calculation that we may refine to make more precise in the future. The Democratic presidential candidate’s performance in the district is compared to that candidate’s national average. If Al Gore won 52.5% in a congressional district in 2000, for example, he would have run 4% ahead of his national 48.5% national average. That would make that particular district a “Democratic 54%” district.

For open seats, we use partisanship alone to make the projection. The model is based on our observation that winners in open seats almost never run more than 5% below the partisanship of their party in the district – in fact, it is relatively rare for that deviation to be more than 2% or 3%.

In races with incumbents, the model selects the incumbent’s weaker performance from the past two elections, and then adjusts that number based on the district partisanship to establish our cautious projection. We know that district partisanship has the potential to have an impact on an incumbent:  those incumbents who ran ahead of their partisan projection may do worse than their past results would suggest, and those running behind their projection may do better.

The model contains several parameters. Users of our spreadsheet can vary them to see how they affect the accuracy of past predictions and then apply those same parameters to the 2002 predictions. We provide more details about our algorithm in the Appendix and in the Read Me file on the CD.

IV.  The Power of Partisanship

Does district partisanship matter? There is no doubt.

For example, of the 100 most Democratic districts based on the 2000 presidential results, 97 are represented by Democrats. The three Republicans in these districts – Quinn (NY-30), Morella (MD-8) and Horn (CA-38) – were elected before 1994 in districts that were substantially less Democratic than they are now. Between 1992 and 2000, Quinn’s district became 11% more Democratic, Horn’s 10% more Democratic and Morella’s 2%. In 1996, none of these incumbents’ districts were in the 100 most Democratic districts.

The following chart illustrates the relationship between our projections based on district partisanship and past elections and the actual winning percentages in those districts.

V.  Why Partisanship Outweighs Campaign Spending in General Elections

Because winning House races is strongly correlated with campaign spending and with outspending one’s opponent, some observers are quick to mistake cause for effect when it comes to the effect of money on general elections to the U.S. House. But money in fact flows for a variety of reasons that have nothing to do with helping candidates win elections – in fact, certain donors prefer to give to candidates whom they expect to win, as certain winners are a better investment for seeking future access to address policy concerns. Of course in today’s heated battle for control of Congress, competitive races draw more attention from donors on both sides of the partisan divide. The result is that winning percentages in House races in fact are negatively correlated with the campaign spending by both winners and losers. In other words, as a candidate spends more and more money, he or she tends to win by smaller and smaller amounts.
A. Charts on campaign spending and results


For indications that much money isn’t being given to sway elections, take Pennsylvania-9 (Bud Shuster), who spent more than $1.1 million in 2000, and Arkansas-3 (Asa Hutchison), who spent more than $800,000. Their spending levels were above average but otherwise unremarkable – except that both candidates were uncontested and won election with 100% of the vote, safely ensconced in districts secure for their party. Clearly, in these and many other cases, their contributors were betting on a sure thing.

On the other hand, there are many races where the candidates spent similar amounts, but the winner won by more than 20%. Examples from 2000 include Steve Buyer (IN-5), whose opponent spent over $400K compared to Buyer’s $330K; Patsy Mink (HI-2), where both candidates spent around $200K; and Tom Allen (ME-1), where Allen and his opponent spent a littler more than $350K.

There were three House three races in which both candidates spent more than $2 million. The candidate who spent more money lost two out of three. Four out of the top five biggest spending losers outspent their opponents. Ten candidates spent more than $2 million in 2000.  Five won, and five lost.  Of the five losers, four of them outspent their opponents but lost anyway.

If spending really determined the outcome of every race – not just who won but by how much – then one would expect to find competitive races when candidates spend similar amounts and lopsided races when one candidate outspends the other. This turns out not to be the case.  Instead, the winner and winning percentage is closely correlated with the partisanship of the district, which is determined long before candidates start raising and spending money.  For example:

· NC-3:  Republican Walter Jones, Jr. and his opponent both spent about $1.1 million, yet Jones crushed his opponent with 61% of the vote. Without using any information on campaign finance, our model projected that Jones would win with at least 60%.

· IN-3:  Democrat Tim Roemer was outspent by his Republican opponent by more than $250,000 won with but won with 52% of the vote, within 1% of our projection.

· OK-6:  Republican Frank Lucas was outspent by his Democratic opponent yet ended up within 2% of our projection and won by over 18%.

If money doesn’t determine the outcome of races as much as the fundamental political landscape of voters, perhaps media coverage does. To show that this is not the case in most Congressional elections, consider two adjacent districts with very different partisan compositions. Two such districts are AL-6 (Spencer Bachus) and AL-7 (Earl Hilliard). Voters in these two districts were exposed to identical media coverage of the presidential election, yet Bill Clinton in 1996 received 45% more of the vote in Hilliard’s district than in Bachus’ district. Four years later, with two very different candidates and presumably different strategies in the state, Al Gore won 44% more of the vote in Hilliard’s district than in Bachus’. Just like their party standard-bearers, in each of their last 4 elections, both Bachus and Hilliard won with landslide margins of over 40%, just as our model predicts based on the partisanship of the district.

Finally, although outspending one’s opponent is certainly associated with winning elections and winning percentage, the association between winning percentage and partisan prediction is much tighter than between winning percentage and the winner’s spending ratio in the 2000 open seat races for which we were able to obtain campaign finance data for both candidates.

Ratio of winner’s spending to loser’s spending, projection and winning percentage in US House races, 2002
 B. Open seats elections, 1994-2000

Given that incumbents have such a high re-election rate, open seats lead to a disproportionate number of seat changes in House races. As a result, elections for open seats are often extremely competitive and candidates and parties tend to pour lots of resources into these races.  In many cases, these efforts are futile.

Between 1994 and 2000, in districts with at least an 8% edge in partisanship for one party, Democrats went 30-0 and Republicans went 39-1. In races where the partisan advantage smaller (between 0% and 8%), the GOP run 82% of the races and the Democrats won 68%.

The GOP had a 63-37% overall edge in winning open seats from 1994-2000, but the real advantage for the Republicans was that two-thirds of these open seats races were in districts that had a partisanship favoring Republicans. The Republicans had their biggest year in 1994, of course, but not in open seats favoring Democrats, where they won zero of 18. The following table shows that the winning party was closely correlated to the partisanship of the district.


Table 8.  Open seat results by district partisanship, 1994-2000 
    Total >8% Rep 0 to 8% Rep 0 to 8% Dem >8% Dem
1994 GOP win 29 11 18 0 0
  Dem win 24 0 6 8 10
1996 GOP win 39 13 22 4 0
  Dem win 13 0 2 4 7
1998 GOP win 16 4 11 1 0
  Dem win 18 0 5 5 8
2000 GOP win 26 11 11 4 0
  Dem win 9 1 1 2 5
Total GOP win 110 39 62 9 0
Total Dem win 64 1 14 19 30
  GOP win % 63% 98% 82% 32% 0%
  Dem win % 37% 3% 18% 68% 100%

As in 1994-2000, more open seats favor Republicans than Democrats in 2002. The best hopes for the Democrats lie in the 17 open seats with a moderate Republican advantage, but based on recent history, Democrats are unlikely to win more than 4 or 5 of them, while Republican are unlikely to win more than 4 or 5 of the 12 seats that lean Democratic.

Table 9.  Partisanship of 2002 Open seats 

Range Number of Seats
>8% Rep  11
0 to 8% Rep 17
0 to 8% Dem 12
>8% Dem 9
Total 49
These 49 open seats will likely have a minimum of 17 Democratic winners and 23 Republican winners, with the nine remaining seats potentially going either way. To maintain their 223 seat majority, the Republicans would need to win 27 out of 49 open seats, all 194 of their incumbents running for re-election and 2 of the 4 incumbent versus incumbent match ups.  If the Democrats win more than a combination of 22 of 49 open seats, 2 of 4 incumbent versus incumbent match ups and 188 incumbents, the Democrats will pick up seats.

VI.  Profile:  The Entrenched Incumbent

Very few House races end up being truly competitive, and 2002 is sure to continue this pattern.  In each of the last two elections, more than 300 races out of 435 races were won by landslide margins of more than 20%, and fewer than 1 in 10 races was won by less than 10%.  The average U.S. House winner in 2000 took more than two-thirds of the vote.

Given the dramatic re-election success of incumbents, incumbency could be seen as the overriding factor for determining electoral outcomes. From 1996 to 2000, of more than 1,150 House incumbents running for re-election, only 33 incumbents lost in general elections during this period. Most of these defeated incumbents did have something in common, however: lack of seniority. Eighteen were freshmen and 10 more had served no more than three terms. Only one incumbent elected before 1990 lost between 1996 and 2000.

Incumbency certainly helps, but one reason incumbents win is they usually represent districts that match their partisanship. Take the Republican freshman class of 1994. Of the 34 Republicans who defeated Democratic incumbents, eight were defeated in 1996-2000. Most of those eight losers represented districts that clearly would favor Democrats in an open seat. None of the remaining victorious 1994 challengers in more safely Republican districts have been defeated.

Certainly the most entrenched incumbents are particularly settled in districts that would be safe for any candidate from their party. Consider the following chart. It shows the most entrenched incumbent winners after the 2000 elections. The “worst” percentage indicates the lowest vote share they won in the entire 1992-2000 period.

    2000     Worst % 2002
State ST Dist Incumbent Party  (‘92-2000) Project.
New York NY 15 Charles Rangel D 91% 91%
New York NY 16 Jose Serrano D 91% 94%
New York NY 10 Ed Towns D 89% 90%
Florida FL 17 Carrie Meek (retiring) D 89% 86%
New York NY 11 Major Owens D 87% 88%
California CA 8 Nancy Pelosi D 82% 79%
Michigan MI 14 John Conyers D 82% 83%
New York NY 8 Jerrold Nadler D 81% 78%
New York NY 17 Eliot Engel D 78% 72%
California CA 32 Julian Dixon (died) D 78% 85%
California CA 35 Maxine Waters D 78% 83%
New York NY 12 Nydia Velazquez D 77% 81%
Texas TX 19 Larry Combest R 77% 74%
Illinois IL 1 Bobby Rush D 76% 83%
Louisiana LA 3 Billy Tauzin D 76% 53%
Virginia VA 3 Robert Scott D 76% 68%
New Jersey NJ 10 Donald Payne D 76% 86%
Illinois IL 4 Luis Gutierrez D 75% 77%
Maryland MD 4 Albert Wynn D 75% 78%
Florida FL 21 Lincoln Diaz-Balart R 75% 56%
Ohio OH 9 Marcy Kaptur D 74% 54%
Pennsylvania PA 9 Bud Shuster (retired) R 74% 64%
Louisiana LA 2 William Jefferson D 73% 77%
Florida FL 14 Porter Goss R 73% 60%
Washington WA 7 Jim McDermott D 73% 73%
Texas TX 26 Dick Armey (retiring) R 72% 72%
Texas TX 6 Joe Barton R 72% 65%
Texas TX 21 Lamar Smith R 72% 71%
Texas TX 3 Sam Johnson R 72% 71%
Tennessee TN 2 Jimmy Duncan R 71% 57%
Colorado CO 5 Joel Hefley R 71% 64%
North Carolina NC 6 Howard Coble R 71% 67%
Indiana IN 6 Dan Burton R 70% 69%
West Virginia WV 1 Alan Mollohan D 70% 61%
Alabama AL 7 Hilliard (primary loss) D 70% 72%
California CA 7 George Miller D 70% 68%
Ohio OH 8 John Boehner R 70% 61%
Wisconsin WI 9 J Sensenbrenner R 70% 66%
Pennsylvania PA 17 George Gekas R 70% 57%

These incumbents are far more likely to retire (Carrie Meek), lose in the primary (Earl Hilliard) or die (Dixon) than they are to lose in a general election, except perhaps when they are matched up against another incumbent during redistricting (potentially Gekas).

VII.  Redistricting 2001-2002 and Incumbent Protection

In reviewing our projections, what is particularly striking in addition to their overall accuracy is how the political landscape in 2002 is actually less favorable to challengers than in the years at the end of last decade’s redistricting cycle. Despite modifying our model to be even more cautious this year (first lowering the projections for long-term incumbents based on past performance, given that incumbents may be in districts with new constituents, and second providing cautious estimates about district partisanship in the 60 districts for which we lacked presidential election data), there still is a sharp increase in the number of races in which we make predictions. The reason? Redistricting.

Redistricting usually results in more competition, not less. As congressional analyst Charlie Cook wrote in his March 19 2002 Off to the Races column, “Perhaps most alarming about this decline in competition is that, typically, greater competition and turnover characterize the first couple of congressional elections after redistricting. Then legislators settled into their new districts and the level of competition goes down until new maps are drawn. If the competition is this low in the first election after a redistricting, imagine what it will be like in 2008 or 2010.”

California is the poster child for efforts to shield incumbents. In 2000, our model made no predictions in nine races with incumbents, including three who lost, and only a “win” prediction for five additional incumbents. In 2002, however, all 51 incumbents are predicted to win, and all but one by either a comfortable or landslide margin. How did they do it? By methodically dividing voters so that potentially vulnerable incumbents received more partisan votes from neighboring districts whose Members either didn’t need them to win (safe colleagues of the same party) or who didn’t want them (colleagues from the other party). Of the six Democrats in competitive districts in 2000, their average increase in Democratic partisanship was 5% in their favor, while the remaining 25 Democrats on average had a very slight decrease in partisanship. Similarly, of the nine Republicans in competitive districts in 2000, their average Republican partisanship increased by 3.5% even as the remaining ten Republicans in safer districts became on average just slightly more Democratic.

California is not alone. As John Fund wrote in the Wall Street Journal, on March 13, “Incumbents are using high-powered computers to create lifetime sinecures for themselves.” Nationally, the average partisanship of incumbents of both parties in safe districts became slightly less partisan as a means to help boost the percentages of others from their party, who saw an average increase in their party’s partisanship (see chart in appendix). Sometimes a partisan gerrymander in states like Georgia, Maryland, Michigan and Utah saw efforts to oust an incumbent, but the norm was to boost vulnerable incumbents odds in the next election.

VIII.  Short Takes

A. Who “won” redistricting and the long-term partisan landscape

Our first answer to this question is to point out that, once again, the incumbents were the big winners over voters seeking competitive choices. As with many experts, we would suggest that neither party received a great boost, with Democrats being fortunate to escape with less harm in Ohio and Texas and Republicans being pleased with the results in California. We would point to two important numbers. First, in 2000, George Bush won 228 districts to Al Gore’s 207. Even though Gore won the popular vote by more than a half million votes, his vote was more concentrated, as the famous color-coded maps demonstrate with particular power when done based on results in counties. Given the power of district partisanship, as defined by the vote in the presidential race, the 2000 results means that the median district in the country is one that favors a Republican. Looking at 2000 districts, that median district was one that we would categorize as one with a Republican performance of 52%.

B. A surprising role for third parties and independents

The U.S. House currently has two Members elected as independents: Bernie Sanders from Vermont and Virgil Goode from Virginia. If and when he is elected, Goode will become a Republican in 2003, leaving the House with just one independent. There are no immediate prospects of others being elected as independents or minor party members in the near future, although former Congressman Tim Penny’s possible election in the Minnesota gubernatorial race as a member of the Independence Party, following in the heels of Jesse Ventura, could make Minnesota a base for future strong candidacies for Congress by the Independence Party.

But short of winning, third parties and independents may gain a different kind of attention and potential leverage because of the current extreme narrowness in partisan control of the House (and Senate and presidential race, for that matter) and the relatively few close races that will decide it. Even winning 2%-3% of the vote can have a major impact, as the Green Party’s Ralph Nader demonstrated in the 2000 presidential race. The Greens and the Libertarian Party also affected several congressional races in 2000, and promise to do so again in 2002. Their potential affect may lead to even more serious attention to instant runoff voting, the ranked-choice voting method recently adopted in San Francisco that ensures minor party candidates are not by definition put into the spoiler role by their candidacy.

C. Prospects for women and racial minorities

Following the last cycle of redistricting, the 1992 elections marked a remarkable rise in the number of women and racial minorities in the U.S. House. The number of women increased from 28 to 47, by far their biggest increase ever, and the number of racial minorities increased from 37 to 59, also a historic increase.

This year’s elections suggest nothing remotely on this scale is likely to happen. African Americans gained 13 new seats in 1992, for example, but are likely at best to gain three seats in 2002, with one or two gains for black Democrats in open seat elections in Georgia and one potential black Republican win in Nevada. At the same time, because Julia Carson faces a competitive bid to remain in office and J.C. Watts is stepping down as the only black Republican in Congress, it is quite possible that the number of black House members will not increase at all. Latinos are making a few inroads in open seat races, but are seeing fewer new opportunities than anticipated two years ago. Women also appear poised to pick up new seats, but are also losing Members due to retirement and the recent death of Patsy Mink. Certainly 2002 will bring nothing comparable to “the year of the woman.”

There is no single reason for this change from 1992, but certainly two can be fingered. One, with far fewer competitive seats, there are fewer chances for women and racial minorities to break in – for Latinos, for example, the degree of incumbent protection built into U.S. House maps in California and Texas makes it hard for their growing numbers in those states to translate into new representatives. Two, the racial composition of districts is a major factor in racial; minorities’ chances of electoral success, and there was less legal and political pressure – and simply fewer opportunities – for states to create new “minority opportunity” districts. It is quite possible that for any future major increases in representation of racial minorities and women along the lines of 1992, some non-winner-take-all method such as cumulative voting and choice voting will need to be adopted.

D. State legislatures – Incumbent protection may be even worse

Our report provides disturbing evidence that most voters are locked out of chances to change their representation in U.S. House elections and that this is unlikely to change without dramatic changes in how we draw congressional district lines, or, more fundamentally, challenge the exclusive reliance on winner-take-elections where political majorities represent everyone in a given area. But limited evidence from states like Minnesota (where FairVote Minnesota is releasing a report similar to Monopoly Politics) and North Carolina suggests that this trend is even worse in state legislative races. State legislators draw their own districts in most states, and with increased technological tools, were often able to create safer districts for themselves. Already more than two in five state legislative races were not even contested by one of the major parties in 1998-2000. We suspect that number may well increase to more than 50% during this decade without elected officials and reformers taking more notice of the critical importance of districting and voting methods.

IX.  Appendices

Appendix A:  Details of model

We first establish the winning range of our categories:

Range  Value  Meaning:  winning percentage of
No prediction 3%  under 53%
Competitive 5%  53% to 55%
Comfortable 10%  55% to 60%
Landslide >10%  over 60%

We then use a combination of past performance and partisanship to make separate calculations for open seats, for first-year Members, for two-term incumbents and for incumbents who have served three or more. We give more weight to past performance for incumbents who have served longer.

We use a variety of parameters to make the actual adjust of past election results based on partisanship.  The user can modify any of these variables to see their effect on projections for 1996-2000 as well as 2002.

Meaning     Value  Applies to
Buffer for open seats    11%  Open seats
Increase for underachievers   33%  Freshman
Decrease for overachievers   67%  Freshman, 2 term
Reduction for underachievers   2%  2 or more term
Previous race uncontested   35%  2 term
Decrease for long term overachievers  50%  3 or more term
Adjustment to better 2nd election  33%  2 term

For a description of the actual algorithm used for each calculation category, please see the file, algorithm.doc, on the CD that contains our spreadsheets.

Appendix B:  State-by-state data on the partisan outcome of redistricting

See below for notes.  Data on 2002 district partisanship was lacking for AL, HI, KS, KY, MA, MS, NE, NH, OH and SC.

       Average Change by Seat Safeness Number   
  Average change Number   R  D   R  D 
  R D Overall R D Overall Safe Comp. Safe Comp. Safe Comp. Safe Comp.
Alabama -- -- -- -- -- -- -- -- -- -- -- -- -- --
Alaska 0.0% -- 0.0% 1 0 1 0.0% - - - 1 0 0 0
Arizona 2.3% 1.1% 2.0% 4 1 5 -0.3% 3.1% 1.1% - 1 3 1 0
Arkansas 0.0% -0.6% -0.5% 1 3 4 0.0% - - -0.6% 1 0 0 3
California 1.6% 0.3% 0.8% 19 31 50 -0.2% 3.5% -0.8% 4.8% 10 9 25 6
Colorado 2.6% -0.2% 1.5% 3 2 5 0.3% 7.3% -0.9% 0.6% 2 1 1 1
Connecticut 1.9% 0.0% 0.7% 2 3 5 - 1.9% -0.6% 1.1% 0 2 2 1
Delaware 0.0% -- 0.0% 1 0 1 - 0.0% - - 0 1 0 0
Florida 0.4% 1.3% 0.7% 14 7 21 -2.0% 2.2% 3.5% -4.3% 6 8 5 2
Georgia 7.3% 0.1% 5.5% 6 2 8 4.3% 13.3% -3.7% 3.9% 4 2 1 1
Hawaii -- -- -- -- -- -- -- -- -- -- -- -- -- --
Idaho -0.2% -- -0.2% 2 0 2 -0.2% - - - 2 0 0 0
Illinois 1.0% -1.7% -0.4% 9 9 18 -0.2% 2.4% -2.2% 0.2% 5 4 7 2
Indiana 1.5% -2.2% 0.1% 5 3 8 1.5% - -4.0% 1.3% 5 0 2 1
Iowa -2.1% 1.5% -1.2% 3 1 4 -5.9% -0.3% - 1.5% 1 2 0 1
Kansas -- -- -- -- -- -- -- -- -- -- -- -- -- --
Kentucky -- -- -- -- -- -- -- -- -- -- -- -- -- --
Louisiana 0.5% -1.2% -0.1% 4 2 6 0.9% 0.3% -2.1% -0.2% 1 3 1 1
Maine -- -3.0% -3.0% 0 1 1 - - - -3.0% 0 0 0 1
Maryland 0.6% -6.2% -3.3% 3 4 7 2.4% -0.3% -6.2% - 1 2 4 0
Massachusetts -- -- -- -- -- -- -- -- -- -- -- -- -- --
       Average Change by Seat Safeness Number   
  Average change Number   R  D   R  D 
  R D Overall R D Overall Safe Comp. Safe Comp. Safe Comp. Safe Comp.
Michigan 1.0% 1.3% 1.2% 7 6 13 -0.1% 1.5% 1.2% 2.1% 2 5 5 1
Minnesota 0.5% 0.1% 0.2% 2 4 6 - 0.5% -0.4% 0.5% 0 2 2 2
Mississippi -- -- -- -- -- -- -- -- -- -- -- -- -- --
Missouri 1.0% -1.0% 0.1% 5 4 9 1.2% 0.2% -3.6% 1.7% 4 1 2 2
Montana 0.0% -- 0.0% 1 0 1 0.0% - - - 1 0 0 0
Nebraska -- -- -- -- -- -- -- -- -- -- -- -- -- --
Nevada 3.7% 1.5% 2.6% 1 1 2 3.7% - 1.5% - 1 0 1 0
New Hampshire -- -- -- -- -- -- -- -- -- -- -- -- -- --
New Jersey 1.9% 0.9% 1.4% 5 7 12 - 1.9% 0.2% 5.5% 0 5 6 1
New Mexico 0.1% 0.1% 0.1% 1 1 2 - 0.1% - 0.1% 0 1 0 1
New York 1.0% -0.1% 0.3% 11 18 29 -1.0% 1.4% -0.1% 0.2% 2 9 17 1
North Carolina 1.8% 0.6% 1.3% 7 4 11 2.4% -2.0% -0.7% 1.0% 6 1 1 3
North Dakota -- 0.0% 0.0% 0 1 1 - - - 0.0% 0 0 0 1
Ohio -- -- -- -- -- -- -- -- -- -- -- -- -- --
Oklahoma -1.1% 0.6% -0.7% 3 1 4 -1.1% - - 0.6% 3 0 0 1
Oregon 0.0% 0.2% 0.2% 1 4 5 0.0% - 0.3% 0.2% 1 0 1 3
Pennsylvania 0.6% 4.1% 1.8% 11 6 17 -0.5% 1.5% 2.8% 6.7% 5 6 4 2
Rhode Island -- -0.1% -0.1% 0 2 2 - - -0.1% - 0 0 2 0
South Carolina -- -- -- -- -- -- -- -- -- -- -- -- -- --
South Dakota -- -- -- 0 0 0 - - - - 0 0 0 0
Tennessee 0.9% -0.1% 0.4% 3 3 6 -0.2% 3.0% -5.3% 2.5% 2 1 1 2
       Average Change by Seat Safeness Number   
  Average change Number   R  D   R  D 
  R D Overall R D Overall Safe Comp. Safe Comp. Safe Comp. Safe Comp.
Texas 1.5% 0.6% 1.0% 12 16 28 1.5% - 0.2% 0.8% 12 0 7 9
Utah 0.1% -4.7% -2.3% 1 1 2 0.1% - - -4.7% 1 0 0 1
Vermont -- -- 0.0% 0 0 1 - - - - 0 0 0 0
Virginia 0.3% -0.4% 0.0% 7 4 11 -0.7% 3.0% 1.5% -2.4% 5 2 2 2
Washington 0.2% -0.4% -0.2% 3 6 9 0.1% 0.5% -0.7% 0.0% 2 1 3 3
West Virginia 0.1% 0.0% 0.0% 1 2 3 - 0.1% - 0.0% 0 1 0 2
Wisconsin -0.4% 7.6% 3.6% 4 4 8 -2.2% 0.2% 3.3% 9.0% 1 3 1 3
Wyoming 0.0% -- 0.0% 1 0 1 0.0% - - - 1 0 0 0
TOTALS 0.6% -0.2% 0.2% 164 164 329 -0.3% 1.6% -0.8% 0.9% 89 75 104 60


This table shows the average change in partisanship for the 329 districts in which incumbents are running for re-election in 2002 and for which we have data on the 2000 and 2002 district partisanship.  The table shows the average change in partisanship for all seats held by Democrats and Republicans in each state and the average change for safe and competitive seats held by each party.  Safe seats are defined as seats whose leaning suggests at least a 55%-45% margin for the incumbent party.  Competitive seats are all those seats than have less than a 10% advantage for the incumbent party. All changes are from the perspective of the incumbent party:  positive changes means that the seats became safer for the incumbent; negative changes mean the seats became less safe for the incumbent party.

Example:  California

California’s 19 Republican seats became on average 1.6% more Republican and the 31 Democratic seats became 0.3% more Democratic.  The 10 safe R seats and the 25 safe D seats actually became less safe for the incumbents, while the 9 competitive R seats and the 6 competitive D seats all became safer for the incumbents.  In other words, incumbents were protected.
Appendix C:  Monopoly Politics 2002 Projections in All 435 Districts

    2002 2002 2000  Date 2002 Democratic   Projection (cautious) 
Codes State CD Incumb Party Elect P-ship Projection Party Result Range
O * Alabama 1 Open R 1984 43% 48.7% No proj win 51% 
* Alabama 2 Terry Everett R 1992 41% 38.4% R 62% Landslide
O * Alabama 3 Open R 1996 47% 50.0% No proj win 50% 
U * Alabama 4 Robert Aderholt R 1996 45% 48.0% No proj win 52% 
* Alabama 5 Bud Cramer D 1990 42% 52.1% No proj win 52% 
U * Alabama 6 Spencer Bachus R 1992 34% 32.8% R 67% Landslide
O D U * Alabama 7 Open D 1992 72% 66.5% D 67% Landslide
  Alaska 1 Don Young R 1973 29% 38.0% R 62% Landslide
O N Arizona 1 Open  2002 47% 50.0% No proj win 50% 
O Arizona 2 Open R 1976 42% 47.6% No proj win 52% 
  Arizona 3 John Shadegg R 1994 44% 40.1% R 60% Comfortable
  Arizona 4 Ed Pastor D 1991 63% 65.5% D 66% Landslide
  Arizona 5 J.D. Hayworth R 1994 44% 48.3% No proj win 52% 
  Arizona 6 Jeff Flake R 2000 38% 38.8% R 61% Landslide
O N Arizona 7 Open  2002 58% 52.2% No proj win 52% 
  Arizona 8 Jim Kolbe R 1984 48% 47.9% No proj win 52% 
  Arkansas 1 Marion Berry D 1996 51% 54.5% D 54% Win
  Arkansas 2 Vic Snyder D 1996 49% 52.7% No proj win 53% 
U Arkansas 3 John Boozman R 2001 39% 42.3% R 58% Comfortable
  Arkansas 4 Mike Ross D 2000 51% 50.2% No proj win 50% 
  California 1 Mike Thompson D 1998 54% 57.7% D 58% Comfortable
  California 2 Wally Herger R 1986 35% 38.1% R 62% Landslide
  California 3 Douglas Ose R 1998 43% 42.9% R 57% Comfortable
  California 4 John Doolittle R 1990 38% 36.9% R 63% Landslide
  California 5 Robert Matsui D 1978 62% 65.5% D 66% Landslide
  California 6 Lynn Woolsey D 1992 64% 64.1% D 64% Landslide
  California 7 George Miller D 1974 68% 70.0% D 70% Landslide
  California 8 Nancy Pelosi D 1987 79% 81.0% D 81% Landslide
  California 9 Barbara Lee D 1998 80% 81.2% D 81% Landslide
U California 10 Ellen Tauscher D 1996 57% 55.0% D 55% Comfortable
  California 11 Richard Pombo R 1992 46% 43.7% R 56% Comfortable
  California 12 Tom Lantos D 1980 68% 70.3% D 70% Landslide
  California 13 Pete Stark D 1972 69% 69.3% D 69% Landslide
  California 14 Anna Eshoo D 1992 63% 64.9% D 65% Landslide
  California 15 Michael Honda D 2000 62% 58.7% D 59% Comfortable
  California 16 Zoe Lofgren D 1994 66% 67.6% D 68% Landslide
  California 17 Sam Farr D 1993 62% 63.2% D 63% Landslide
O D California 18 Open D 1988 55% 50.0% No proj win 50% 
  California 19 George Radanovich R 1994 41% 38.6% R 61% Landslide
  California 20 Calvin Dooley D 1990 57% 56.8% D 57% Comfortable
O N California 21 Open R 2002 39% 44.1% R 56% Comfortable
  California 22 Bill Thomas R 1978 35% 32.1% R 68% Landslide
  California 23 Lois Capps D 1998 55% 57.3% D 57% Comfortable
  California 24 Elton Gallegly R 1986 44% 42.6% R 57% Comfortable
  California 25 Buck McKeon R 1992 44% 39.9% R 60% Landslide
  California 26 David Dreier R 1980 45% 42.2% R 58% Comfortable
  California 27 Brad Sherman D 1996 62% 57.7