a». .V cw »

(I

. x .153». . V V M
first: , V V , . . . ,
A . . _ V V A .31.?
V V . «it .w. ._. 1
$1 . afifimnu

. WM. . .dlfi,

3?.
V51 4: 3:!
. .. ‘

33.

«.951an
V . $2.: . V...
V .32 fizugfifiil
.. a . 3avaniflfiw?‘

u
3

.WMwa

II

we...” . “thin I

II ”a...

3‘“ II.
a...
z.
. :5 id."
:1 V Xi”. .abfll
:..v.mnW.fi.y3¥...i #4“; f...
Al 1 I, ll‘kii‘
$3. flush”! ”tilt!!!"
.V .57....)4:

. . , . fluent} ‘
havfiéghu
(‘3‘, till :3. \
staufiifihflggl
3; 1.
L‘Dgikixx V

.54.};
. .

 

v... T“. lil‘l

(ti ' cl. ('3‘.

gases?“

‘3 .

 

This is to certify that the
dissertation entitled

Redistricting and Representation

presented by

Michael H. Crespin

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Political Science

 

 

WWW»

Major Professor’s Signature

W5-

Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

‘ UBRARY

Michigan State
University

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

112008

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 plelRC/DateDuejnddpJ

REDISTRICTING AND REPRESENTATION
By

Michael H. Crespin

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY

Department of Political Science

2005

ABSTRACT
REDISTRICTING AND REPRESENTATION
By

Michael H. Crespin

In this dissertation, I use redistricting to test the bonds of representation between
members of Congress and their constituents. Throughout the dissertation I use
geographic information systems (GIS), a tool that has yet to be fully exploited in
political science. First, I demonstrate how redistricting can make a district more or less
competitive depending on how much the district changes in terms of the number of new
constituents, and their general political leanings. Then, I show how a member’s
behavior changes to adapt to her new constituency, but only when voters are paying
attention. Finally, I illustrate how redistricting influences fundraising from individual

contributors who reside in a different district after the boundaries change.

Copyright by
MICHAEL H. CRESPIN
2005

This dissertation is dedicated to the Political Institutions and Public Choice program -—
an institution that matters.

ACKNOWLEDGMENTS

I would first like to acknowledge my friends and family for sticking with me through
this long process. Thank you for your support and understanding. I would also like to
thank my faithful companion Sadie, who provided both a necessary and sufficient
distraction from my research. Next, I thank my PIPC colleagues, especially Jamie
Carson, Chuck F inocchairo, Mark Hurwitz, Jeff Jenkins, Bryan Marshall, Tim Nokken,
and Mark Souva, who were all invaluable contributors to my training. Other graduate
students who contributed to this dissertation were Amon Key, Frederick Wood, K0
Maeda and Carl Snook. Ashton Shortridge from the Geography Department helped me
with my struggles with GIS and geography in general. Finally, I would like to thank
three professors who contributed to my intellectual advancement. First, Nathan Monroe
helped by bringing a little bit of the corporation to the Midwest. Second, Burt Monroe
taught me to think outside the proverbial box. Finally, I would like to thank Dave Rohde
for allowing and encouraging me to pursue my own agenda and whose invisible hand

provided constant guidance and support throughout my graduate career.

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................... VIII
LIST OF FIGURES ......................................................................................................... IX
CHAPTER 1 - INTRODUCTION .................................................................................... 1
CHAPTER 2 - USING GEOGRAPHIC INFORMATION SYSTEMS TO MEASURE
DISTRICT CHANGE, 2000-02 ........................................................................................ 3
Introduction ............................................................................................................ 3
Previous Measures of Redistricting ....................................................................... 3
Data and Method .................................................................................................... 4
Technical Details ................................................................................................... 6
Uses of New Measure ............................................................................................ 8
Does Change Matter? .......................................................................................... IO
Advantages of New Measure ............................................................................... 13
Conclusion ........................................................................................................... 14

CHAPTER 3 - SOME BOOTS ARE MADE FOR WALKING: CONGRESSIONAL
VOTING BEHAVIOR IN RESPONSE TO CHANGING DISTRICT BOUNDARIES 21

Introduction .......................................................................................................... 21
Electoral Accountability ...................................................................................... 22
Theories of Congressional Organization ............................................................. 25
District Change and Behavior Change ................................................................ 28
Data ...................................................................................................................... 30
Results .................................................................................................................. 35
Conclusion and Discussion .................................................................................. 38

CHAPTER 4 - GEOGRAPHY AND INDIVIDUAL CAMPAIGN CONTRIBUTIONS

TO CONGRESSIONAL CANDIDATES ....................................................................... 45
Introduction .......................................................................................................... 45
Previous Research ................................................................................................ 46
Geography and F undraising ................................................................................. 49
Campaign Finance Data ....................................................................................... 54
The 2000 Congressional Election ........................................................................ 58
Results .................................................................................................................. 62
Redistricting and Contributing ............................................................................ 66
District change data ............................................................................................. 68
Redistricting Results ............................................................................................ 69
Motivations for Giving ........................................................................................ 7|
Conclusion and Future Research ......................................................................... 74

CHAPTER 5 - CONCLUSION ....................................................................................... 87

vi

Pitfalls and Practicalities ...................................................................................... 87

Concluding Remarks ........................................................................................... 90
APPENDIX ..................................................................................................................... 93
REFERENCES ................................................................................................................ 94

vii

LIST OF TABLES

Chapter 2

Table 1A Displacement of Iowa’s 107th Congressional Districts into 108th .............. 16
Table IB Continuity of Iowa’s 107th Congressional Districts into 108th....................l6
Table 2A Displacement of New Hampshire’s 107th Congressional Districts into

108th ................................................................................................................................. 17
Table 23 Continuity of New Hampshire’s 108th Congressional Districts from

107th ................................................................................................. 17
Table 3 Finding “New” Districts in Arizona .................................................... 18
Table 4 Influences on Change in Incumbent Victory Margins, 2000-2002 ................ 19
Chapter 3

Table I — Hypotheses and Predictions ........................................................... 40
Table 2 — Change in Member Voting Behavior as a result of Redistricting. . . . . . . . ..........41
Table 3 - Substantive Results ..................................................................... 42
Chapter 4

Table l - Percent of Money to Incumbent Candidates, 1999-2000 Election Cycle ....... 77
Table 2 — Percent Contribution from Geographic Areas ...................................... 78
Table 3 - Contributions from Areas that are Continuous, New or Left the District ...... 79

Table 4 — Contributions from Areas Drawn Out of the District (In-Out) by Party ........ 80

viii

LIST OF FIGURES

Chapter 2

Figure 1 Example of Spatial Analysis — Iowa .................................................. 20
Chapter 3

Figure I - Spatial Representation of District Change ......................................... 43
Figure 2 — Marginal Effect of District Change on Voting Behavior Change ............... 44
Chapter 4

Figure l — Individual Congressional Campaign Contributions, 2001-02 (Percent by
Party) ................................................................................................. 8]
Figure 2 — Individual Congressional Campaign Contributions, 2001-02 (total per

ZCTA) ................................................................................................ 82
Figure 3 — Area Typology ......................................................................... 83
Figure 4 — Effect of Seniority on Contribution Area .......................................... 84
Figure 5 — Effect of Percent New in the District on Contributions ......................... 85
Figure 6 — Effect of Percent new in the District on Contributions by Party ................ 86

Chapter 1 - Introduction
In this dissertation, I seek to better understand the link between constituents and their

elected representatives. Since elected officials in the United States represent a specific
geographic area such as a congressional district, this variable will play an important
role in understanding the relationship. In many cases, geography is not a variable
precisely because it does not vary, hence it cannot explain much. However, and more
specific to this dissertation, there are times when the boundaries change and this
provides a unique opportunity to study members of Congress and their constituents. In
this dissertation, I will study how redistricting influences members and constituents in
the electorate, in government and in organization.

In three essays, I will explore this topic. In the first essay, I introduce a better
measure of district change using geographic information systems (GIS) and study the
effect of redistricting on electoral competition. The new measure will provide details
of where the new population in a district came from and how the old population is
distributed within new districts. 1 will then use the measure to demonstrate the
independent and interactive influence of district change on competition in
congressional elections.

In the second essay, I will use partisan theories of legislative government (Rohde
1991, Cox and McCubbins 1993) to predict changes in members’ voting behavior after
their districts have changed. I will show how member's behavior changes aflcr
redistricting. Downsian models of representation tell us that members of Congress
should modify their voting behavior in the direction of a changing district median.

Other partisan theories of representation, such as cartel theory, suggest that members

take more than district preferences into account before casting a vote. The theory
assumes that party can pressure members to vote against their district if it is important
to the party. More specifically, the party pressures members more on procedural
motions compared to votes on substance. If Downsian theory is correct, then it follows
that as a member’s district median moves, she should modify her behavior on final
passage votes in the direction of the change. If the party cartel model is accurate and
party exerts influence over procedural matters, then it also follows that changes in the
district should not be related to shifts in voting on partisan procedural motions.

In the final essay, I will use redistricting as a quasi-experiment to uncover
geographic patterns of campaign contributions across different types of contributors. I
will Show that a member’s ability to raise funds is a function of the supply and demand
of contributors. Then I will demonstrate that after redistricting, members will earn less
from areas that have been drawn out of their district, but they will see an increase in
contributions from new areas of the district. Finally, I provide evidence that

Democratic and Republican incumbents are influenced differently after redistricting.

Chapter 2 - Using Geographic Information Systems to Measure District Change,
2000-02

Introduction
How can we improve on our current understanding of the redistricting process? One

way is by using the same tools that states use to draw district boundaries to create better
measures of district change. In this brief research note, I outline how to use geographic
information systems (GIS) to create a continuous measure of district continuity and
change.1 The primary goal of this analysis is to track the changes in the populations of
congressional districts in the pre- (2000) and post-redistricting (2002) years using basic
GIS techniques. The measure allows us to precisely state the fraction of constituents in
district d in election year y has candidate c as the incumbent.
Previous Measures of Redistricting

As mandated by the US. Supreme Court following Baker v. Carr (1962), Wesberry v.
Sanders (1964) and Reynolds v. Simms (1964), states must redraw district boundaries
every ten years to ensure equality of representation. Soon after the initial rounds of
redistricting, political scientists have tried to determine the influence of district changes
on competition (Erikson 1972 and Tufte I973), patterns of turnover (Bullock 1975),
and more recently, challenger entry and incumbent exit decisions (Cox and Katz 2002
and Hetherington, Larson, and Globetti 2003).

Previous research has measured redistricting in different ways. The recent accounts
of redistricting by Cox and Katz (2002)2, Hetherington, Larson and Globetti (2003) and

others measure redistricting with a simple dichotomous variable when the boundaries

 

lIn particular, I use Arc/Info 8.3 workstation for spatial analysis and ArcMap 8.3 to
display the resulting shape files.
2 Cox and Katz (2002:15-18) discuss a similar continuous measure but only use it

sparingly.

are changed. In contrast, some are concerned with "major" redistricting as defined by
some professional source such as Congressional Quarterly (see, e.g., Jacobson and
DimoCk 1994). Finally, Carson, Crespin, Finocchairo and Rohde (2004) go one step
further and rely on county level maps to classify districts as "continuous" if 50 percent
or more of the population is shared between the old and the redrawn district and "new"
otherwise.

While each of these measures are valid, and have provided us with many useful
insights into the effects of redistricting, it is now possible to move beyond simple
dichotomous measures of redistricting and adapt the same techniques used by the states
to draw district boundaries. This is important since, as scholars have pointed out, there
are advantages in paying attention to measurement when trying to capture abstract
concepts (Zeller and Carmines 1980, Jacoby 1991). In this case, the concept is
changing district boundaries, but the measures to date have not indicated to what extent
the districts have actually changed. It is now feasible to use geographic information
systems (GIS) programs such as ARC/Info in combination with boundary shape files
and census data to produce continuous measures of district continuity and change"
Data and Methods

Creating a continuous measure of district change requires only simple spatial
techniques and readily available census data.4 The data include district and census tract

boundary files that are then combined with population tables for the tracts within each

 

3 ARC/Info is one of several programs that can be used to manipulate, explore and
display data that is spatial in nature.

4 Creating a continuous measure is so simple, in fact, that the only ARC/Info
commands used to create the measure are intersect and union . Sec Bolstad
(2002) for an introduction to GIS and spatial analysis.

state.5 Along with the coordinates in each file is a polygon attribute table that provides
data for each shape (i.e. district or tract) including the area and perimeter of each
polygon as well as an identification code that matches up polygons with congressional
district and census tracts.

The steps to create the measure are demonstrated in Figure 1, using Iowa as an
example.6 The first step in the process is to spatially intersect the district boundary files
for the 107th (labeled A in Figure I) and 108th (B) Congresses. This new shape file is
then overlaid on top of the census tract file (C).7

The result (D) includes three types of census tracts.8 The first types of tracts deemed
“continuous” appear in both the new and the old districts. Second, I define "new" tracts
as those that have been drawn into the 108th congressional district and, finally, "old"

tracts that have been drawn out of the 107th districts after redistricting.9 The new

 

5 The 2000 census data used in this paper are available at www.ccnsus.gov. A

cartographic boundary file is a series of coordinates that ARC/Info can read to produce

a visual depiction of the shape.

6 This state provides an interesting case study because of the degree of district change
roduced by the non-partisan redistricting commission.

According to the census, tracts are small, relatively permanent statistical subdivisions
of a county with populations ranging from 1,500 to 8,000. The tracts grow smaller in
dense urban areas making it possible to measure change in all districts.

8 Since some tracts do cross district boundaries, some post-processing of the data is
necessary. Further, spatial analyses such as intersect and overlay frequently create
small Spurious polygons called “slivers” which are an artifact of imprecise overlay
(Bolstad 2002:398). The result in the attribute tables are the repeated listing of tracts
with the same tract id (and population), but with different areas. To erase these slivers
and place tracts in the appropriate district if there was an overlap, I dropped all of the
repeated tracts except for the one with the largest area. This technique produced a final
attribute table that had only one unique entry for each of the original census tracts.

9 This means that a tract can be simultaneously new for one district and old for another.
Terms like new, old and continuous are really a mental short-cut for ways to think
about tracts. An alternative way to think about tracts would be to identify each tract
with a double (i,j) with 1' identifying the district number in the old scheme and j

 

attribute table, which is automatically updated by Arc/Info, makes it relatively easy to
identify continuous, new and old tracts since each tract has an identifier that labels
where it resided in both the 107th and 108th congressional districts. To produce a
measure of change (continuity), simply divide the population for the new (continuous)
tracts by the total population of the district. The final results are actually two measures
of change or continuity. The first tells us what happened to the population of the 107th
districts while the second reports where the population came from for the new 108th

districts.

Technical Details

In this section, I will provide a bit of a technical description for creating the data
within the GIS software. The initial step is to download the district boundary files and
the census tract files from the United State census web-page at http://www.census.gov.
The easiest version to start with is the one with the .eOO file extension. In this analysis,
I performed the Spatial intersections separately for each state; however, one could use
the mapj oin command to combine each of the states at any point of the process. The
mapj oin command is similar to most of the basic Arc/Info commands in that it
requires the user to “tell” the program the names of the two input files (the files that are
to be joined) and then give a new name for the output file (the file that was created by
the particular command).

In order to get the data into Arc/Info, one must first use the import: command. The

syntax for this command is import AUTO <input file name> <output

 

identifying the district number in the new scheme. In fact, this is essentially how they
are treated in the dataset.

f i le name>.10 This command needs to be run for each of the district and tract
boundary files. Then, the next step is to intersect the files for each state. I initially
intersected the pre-redistricting (107th Congress) files with the post-redistricting (108th
Congress) boundary file. Similar to the other commands, the syntax is intersect
<pre—redistricting file name> <post-redistricting file
name> <output fi le name>. Then, this new boundary file is intersected with
the census tract file with similar syntax."

Once the spatial analysis is completed the post-processing mentioned in footnote
eight must be performed. This is due to the fact that census tract boundaries cross
district boundaries and intersections in general tend to produce small areas called
slivers. Slivers are defined as by Boldstad (20022398) “small, spurious polygons at the
margins or boundaries of feature polygons that are an artifact of imprecise digitizing or
overlay.” To remove the slivers and other extraneous polygons, I exported the polygon
attribute table (PAT) into the statistical package Stata as it is better suited for this task.
Essentially, this file needs to be edited so each census tract is identified as residing in
only one district at each point in time. Of course, the tract could be in the same district
during both time periods, or reside in different districts before and after redistricting. In
order to edit the tracts, I created a Stata .do file that, for any tract listed more than once,

kept the listing with the largest area.12 Now, each tract should be only listed once for

 

'0 Text inside < > indicates the user provides the name of the files.

H Since each of the three files, old districts, new districts and census tract files are
contained within the state boundaries; intersection and union perform similar tasks.
Spatial intersection and union are analogous to the same techniques used in Venn
diagrams.

‘2 When polygons are edited in Arc/Info, the PAT records and updates the new areas
and perimeters.

each time-period. Of course, an alternative approach would be to determine the percent,
in terms of area, that each tract overlaps with different districts. This percentage can
then be multiplied by the population in each tract and a weighted population average
can be computed. For the purposes of this study, however, I decided this step was not
necessary.

Once this editing is completed, the boundary files need to be merged with the
population files to create the measures of continuity and change. The population data
can also be downloaded from the census web-page. It can also be obtained in DVD
format from most libraries with a census depository. Similar to most merges, a
common identifier across files is necessary. In this case, the census tract number is
sufficient. Once the merge is completed, I used the population data for each district to
determine what percent of the population came from which district to make up the new
district. Alternatively, I can determine the percentage of the population that was drawn
out of the district and the new districts where they now resides.

Uses of New Measure

In order to better understand the different measures of district change, refer to Tables
IA, 18, 2A and ZB that display the results of the spatial analysis for Iowa and New
Hampshire. Each of the A tables reads from lefi to right across the rows and shows the
first measure of district change, the distribution of population from the 107th
Congressional districts to the new districts in the 108th Congress.‘3 The B tables read

from the top down in the columns and display where the population came from for each

 

'3 While it may rarely be the case that we would want to know where voters have gone,
this technique can be used to examine where contributors reside in two different
election cycles (see e.g. Crespin 2005).

of the new districts. For example, table 1A shows that 37.93 percent of the population
in Iowa’s first district in the 107th Congress remained in the first district for the 108th
Congress. Meanwhile, the same district lost 62.07 percent of its population to the
second district in the 108th Congress. The same population movements can be tracked
for each of the other four districts.

Table 1B shows that new district one is made up of 39.13 percent of old district one
and 60.87 percent of old district two. This measure also makes it relatively easy to
match up new districts with their “parent” districts (see Cox and Katz 2002:107fn 1).”
Clearly, new district one’s parent district is not old district one, but district two since
old two and new one share the largest percentage of the population.

In some states such as New Hampshire, redistricting only produces minor changes,
as Table 2A and B demonstrate. In fact, the 100 percent in the row for district two in
Table 2A means that all of the population in that district remained there during
redistricting. Meanwhile, 98.73 percent remained in district one while it gave 1.27
percent away to district two. This demonstrates that district two lost population relative
to district one between the 1990 and 2000 census.

As a final descriptive example, the measure can also be used to determine which
districts are added to a state as a result of reapportionment. Table 3, read from top to
bottom, describes where the population came from to make up the 108th congressional
districts for Arizona, a State that gained two seats prior to the 2002 elections. It is

possible to identify districts one and four as “new” since they share the smallest

 

’4 Some districts will not have parents as a result of reapportionment. Depending on
the research question, it is also possible for a district to have one or several “children”
if districts are added to a state.

population with parent districts from the 107th Congress. Although these examples
only make use of basic population counts, they hopefully show the worth of the new

measure. 15

Does Change Matter?

The above examples are interesting but they do not mean much if change does not
matter. To show the importance of change on election outcomes, I add the new
measure of district changes to a basic model of congressional elections (Jacobson 2004).
The dependent variable is the change in the margin of victory measured as the
difference between the incumbents vote in the 2002 congressional elections to the
108th Congress and the 2000 election, only for contested races in both years.16 The
main independent variable of interest is the percent new (% new 108th) for each district
before the 2002 election.'7 To be clear, this variable is measured using the same
incumbent’s district for both elections. Although this measure does not take into
account the types of voters who are being shuffled in and out of districts, we do know
one thing about them for certain; they have not been represented by the incumbent.18 If

part of the ineumbency advantage lies in the fact that all an incumbent has to do to win

 

'5 The district change data are available on the author’s web-page,
www.msu.edu/~crespinm.

'6 To be clear, this means that for a district to be included in the regression there must
be an incumbent facing a major party challenger in both the 2000 and 2002 elections.
Conversely, open seat races, or races without a major party challenger in either year
were excluded.

'7 For example, referring back to table 18, the value of the variable would be 100-
60.87=39.l3 for district I, lOO-64.04=35.96 for district 2, and 100-23.79=76.21 for
district 3 Since Leonard Boswell decided to remain in district three, even though the
bulk of his old district is in new district 4.

'8 See F enno 1977 for a discussion of how members represent different parts of their
geographic constituency.

reelection is get all of the voters who voted for her the last time to do it again, then
some of the advantage may be taken way if after redistricting, some of those voters no
longer reside in the district. In other words, the ineumbency advantage is not present
for incumbents if the voters are new. For this reason, it is expected that the greater
percent new for a district, the greater the decline in the incumbents vote share, all else
equal. This similar to the “personal vote” measure developed by Ansolabehere, Snyder
and Stewart (2000).

It is also necessary to control for changes in the underlying political make-up of the
district. This is measured by the change in the district presidential vote for the
incumbent party’s presidential vote in the district (A Presidential Vote). This variable
is operationalized, for Democratic incumbents, by subtracting Al Gore’s vote in the
new 108th congressional districts from his vote share in the old 107th districts as
reported by Politics in America. For Republican incumbents, the measure is based on
the change in George W. Bush’s vote share. Including this control variable allows us
to test for the independent influence of district change while controlling for the political
nature of the change. As a district becomes more Democratic (Republican) then
Democratic (Republican) incumbents should see an increase in their margin of victory.
I also include an interactive variable between the percent new in the district and the
change in district partisanship. This variable shows not only how much the district
changed in terms of population, but also the ideological direction of the change.

The other control variables are standard measures taken from the congressional
election’s literature (see Jacobson 2004 for an example). First, I include a dummy

variable for quality challenger coded one in the presence of a quality challenger (where,

I]

as Jacobson 1980 suggests, previously electoral experience is used as a proxy for
“quality”), zero otherwise. There are also controls for the challenger’s and incumbent’s
spending. In each instance, the natural log is used (In Incumbent spending and In
Challenger spending).19 There is also a dichotomous variable that records if the
incumbent is up for reelection for the first time. This variable, Freshman incumbent
will control for the any presence of the sophomore surge and party (coded 1 for
Democrats) will control changes in national trends. Finally I include a dummy variable
controlling for South defined as the eleven states of the Confederacy. Since the
dependent variable is theoretically continuous, OLS is used and Huber-White robust
standard errors are used to correct for heteroskedasticity.

Table 4 displays the results of the regression. The negative and Significant
coefficient on percent new 108th is evidence that the more a district changes, the
incumbent’s vote share declines. Although the coefficient on the change in presidential
vote is not significant, the interaction between percent new and presidential vote is
significant and in the expected direction. If the district changes in terms of population
and that change is in the direction of the incumbent’s party, then the incumbent’s vote
share increases.20

In terms of the control variables, the more a challenger spends, the greater the

decline in the incumbent’s vote Share (Jacobson 2004). The party variable suggests that

 

’9 I employ the convention adopted by Jacobson (1980) in assuming a minimum of
$5000 spent by each candidate.

20 A joint hypotheses test between the three variables is significant at p<.001. I also
examined scatter-plots and added-variable plot to check for non-linearity and did not
uncoverany.

12

the Republicans did better in 2002 compared to 2000. Freshman incumbents also did

better in there second election, as expected.

Advantages of New Measure

The most obvious advantage of this new measure is that it provides more information
than previous measures of district change. As mentioned above, previous measures
could at best tell us if the district was 50 percent new, or not. The measure developed
here was also created using equipment specially designed for the task. Previous
scholars within political science who attempted similar types of analyses used counties,
instead of census tracts, as the smallest geographic unit. Since tracts are much small
than counties, this measure is likely more accurate. It is also possible to obtain
measures for urban areas, something that is difficult when counties are used since
multiple congressional districts are frequently contained within a single country. If the
urban districts are not included in any subsequent statistical analysis, the results of the
analysis may be biased. The bias may stem from districts being excluded in a
deterministic, rather than random fashion.”

Another use of the new change, and one that is exploited throughout this dissertation,
is to use it in a “quasi-experimental” fashion. That is, we can take a set of actors and
expose them to some sort of treatment. In this case the treatment is a change in their
congressional districts. By taking measurements before and after the treatment, the
influence of the changing district boundaries can be determined. If the only available

measure was the dichotomous rcdistricted yes/no, then this type of analysis would be

 

2 ' Of course, this is an empirical question and it is possible that results may not change
by excluding these districts.

l3

essentially meaningless because at most, only nine districts would be coded as not
having undergone any change and there would be little variation within the data.22 In
the chapters to follow, I will use the new measure to determine how members of
Congress change their behavior in response to changes in the district. By measuring
how much the district changes in terms of population, I can better capture the degree of
uncertainty a member may have about the political change in a district. That is, if a
member’s district changed a great deal in terms of population, it may be more difficult
for the member to determine the new political make-up of her district. However, if the
district is only slightly new, it may be easy for the member to know if the new district
moved in a liberal or conservative direction.

In the final empirical section, I use the measure and techniques more directly by
tracking the geographic areas of individual campaign contributions. That is, I can
determine if a contribution came from inside the member’s district, inside the
member’s state, or outside the state. I can then test to see if members are able to earn
contributions from areas that are new to the district, or areas that have lefl the district.
The more a district changes in terms of population, the more a member should lose
from areas that used to reside in the district. In contrast, the more a district changes, the
more chances a member has to earn contributions from new constituents. In sum, this
new measure of district change appears to have many uses within political science.
Conclusion

This chapter provides evidence that not all redistricting should be treated equally and

offers an example to better measure district change. As illustrated above, the results

 

22 The nine districts are the seven single states plus the two Maine districts as Maine
waits an additional election cycle to change its district boundaries.

l4

suggest that this change can have an influence on the margins of victory in
congressional races. If members wish to remain in office, they will need to adjust their
homestyle accordingly (F enno 1977).

More importantly, this note provided a new way to measure redistricting that can be
used to answer other interesting questions. For example, the measure can be used in
future research to track changes in member behavior before and after redistricting. As
members pick up different types of voters, we would expect them to change their
voting behavior accordingly. This measure allows us to learn where the new voters
came from, and with the addition of other demographic census data, make predictions
about how members will react to their new constituents. We can also move beyond
voting and learn how potential campaign contributors will behave in response to their
new representatives. Finally, we can also determine if different types of plans, i.e.
Democratic plans or judicial plans differ in there propensity for change. Using this new

measure, these questions may now be easier to answer.

Table 2A Displacement of Iowa’s 107th Congressional Districts into 108th

 

108th Congressional Districts

 

 

1 2 3 4 5 Total
1 37.93 62.07 100
Con 1I're7stsiional 2 62'63 12-56 24-31 100
Dfismcts 3 36.68 24.27 27.89 1 1.16 100
4 60.28 8.81 30.90 100
5 41.07 58.93 100

 

 

 

 

Table 18 Continuity of Iowa’s 107th Congressional Districts into 108th

 

108th Congressional Districts

 

 

 

1 2 3 4 5
1 39.13 64.04
Conéfzggonal 2 60.87 12.21 24.1 1
Districts 3 35.96 23.79 27.33 10.94
4 64.00 9.36 32.81
5 39.20 56.25
Total 100 100 100 100 100

 

 

 

l6

Table 2A Displacement of New Hampshire’s 107th Congressional Districts into
108th

 

108th Congessional Districts

 

 

107th l 2 Total
Congressional l 98.73 1.27 100
Districts 2 l 00 100

 

 

 

 

Table ZB Continuity of New Hampshire’s 108th Congressional Districts from
107th

 

108th Congressional

 

 

 

Districts
107th 1 2
Congressional 1 100 1.29
Districts 2 98.71
Total 100 100

 

 

 

l7

Table 3 Finding “New” Districts in Arizona

 

108th Congressional Districts

 

1

107th 2
Congressional 3
Districts 4

5

6

 

 

 

1 2 3 4 5 6 7 8
5.77 17.89 72.65 33.72
0.77 48.25 72.24 2.40
31.41 90.51 8.63 13.80 9.57
8.57 74.54 20.06 8.72
8.14 15.03 97.60
60.45 0.14 11.07 18.62 66.28 3.16
100 100 100 100 100 100 100 100

 

 

Table 4 Influences on Change in Incumbent Victory Margins, 2000-2002

 

 

Variable Coefficient
(Robust Standard Error)
% New 108th -.037*
(019)
A Presidential Vote .060
(170)
%New 108th X A Presidential Vote .019*
(005)
Quality Challenger 1.40
(1.04)
Ln Incumbent Spending .160
(527)
Ln Challenger Spending -.532*
(.185)
Party -4.03*
(.631)
Freshman Incumbent 5.22"“
(1.40)
South .503
(.840)
N 274
Adj. R2 .39
Root MSE 5.08
F-stat 2010*
*p<.05

Dependent Variable - Incumbent 2002 vote minus Incumbent 2000 vote.

19

Figure 1 Example of Spatial Analysis - Iowa23

A - 107th Districts

 

05 02

l

B - 108th Districts
\5

 

 

 

 

 

C - Census Tracts

 

 

D - Intersection and Union

fi

I

1

1

1

1

I

1

I

1

‘1

I

L-

9
1

' i
K .
I
1 1
1

 

 

 

 

 

2" This map of Iowa may appear to be “squished” to some readers. However, it is only
due to the choice of projection system from the round earth to the flat paper. GC S
North American 1983 is the default system applied here.

20

Chapter 3 - Some Boots are Made for Walking: Congressional Voting Behavior in
Response to Changing District Boundaries

Introduction
What are the implications for representative democracy after a redistricting? The

Downsian model of elections predicts that elected representatives will change their
behavior to correspond with the new district median (Downs 1957). Previous scholars,
(Glazer and Robbins 1983, 1985; Stratmann 2000; Boatright 2004) have demonstrated
that members of congress will modify their voting patterns in predictable ways in
response to changes in district boundaries. Does this result apply to all types of votes
or does it only pertain to votes that are visible to the district? The congressional
organization literature contends that parties can control legislative outcomes through
agenda control by pressuring members to vote with the party on procedural matters
(Cox and McCubbins 1993, 2002, 2005). This is an effective technique because it is
difficult to trace voting on procedure to changes in real policy (Arnold 1990). If
members in altered districts are not also adopting new positions on procedural votes,
then representation at the district level may be present, but at a sub-optimal level.

In this essay, I use a quasi-experimental research design to determine how
redistricting affects democratic representation by examining how members respond to
competing pressure from their district and party. By developing hypotheses based on
theories from the congressional organizations literature and then testing them
empirically, I build upon previous research to assess whether members change their
behavior in response to shifts in district boundaries after the 2000-02 round of

congressional redistricting. The results indicate that representatives adjust their roll call

 

‘ Apologies to Nancy Sinatra and now apparently Jessica Simpson

21

behavior to fit their new districts on votes that are visible to their constituents.
However, when it comes to votes that are important to the party for controlling the
agenda, movements in district boundaries are not related to members’ roll call decisions.

In the following section, I briefly review some of the congressional elections and
legislative organization literature to reveal different expectations regarding the
motivations for members’ voting behavior in the US. House of Representatives. Then,
based on that literature, I develop testable hypotheses that predict why members may
change their ideal points after redistricting on some votes but not on others. Next, I
empirically test the hypotheses using basic econometric techniques and in the final
section, I discuss the implications for the findings.
Electoral Accountability

A basic principle of representative democracy is that elected officials must pay
homage to their districts if they wish to secure reelection. Edmund Burke, of course,
would have done well to heed this advice. As the well known parable goes, Burke was
not reelected to Parliament following his famous speech to the electors at Bristol when
he proclaimed his own views to be superior to those of his constituents. The idea of
electoral accountability is prevalent in many theories of representation and elections.
The Downsian model, in its simplest form, predicts that candidates representing the
parties will converge to the district median in order to win elected office (Downs
1957)“. Although more nuanced versions of the model (e.g. Aaronson and Ordershook
1972; Calvert 1985; Palfrey 1984) predict some degree of divergence depending on the

presence of primary elections, motivation of the candidates, the completeness of

 

3“ See also Hotelling 1929 and Black 1958.

22

information and the wish to deter potential third party challengers, they all depend on
district preferences.

In a less formal fashion, F roman (1963) argued that constituency is the most
important determinant of member behavior and that members may even need to
exercise some independence from the party to increase their chances of reelection. If
there is a conflict between party and district, the member should choose the district
since the costs of party defection surely is less than losing the next election (9).

F iorina (1974), Mayhew (1974) and Arnold (1990) all theorized that members of
Congress will pay attention to their district when they are deciding how to vote on the
floor and subsequently how it will influence their reelection chances. F enno (1978)
posited reelection, along with institutional and policy goals contribute to the decision
making calculus. During his travels with representatives in their districts, members
frequently remarked that their vote choice was often dependent on how certain parts of
the constituency would react.25 When a larger group of members were asked questions
about representation on surveys, they also indicated that district plays a part in roll call
decision making (Clausen 1973; Kingdon 1981; Sullivan et al. 1993).

Canes-Wrone, Brady and Cogan (2002) argued that members are held accountable at
the ballot box based on their legislative history. Specifically, they found that the more
often a member supports her party, the lower her vote Share and the probability of
retaining office decreases. Ansolabehere, Snyder, and Stewart (2001), Bovitz and
Carson (2000) and Erikson and Wright (2005) also found support for this notion.

Meanwhile, Carson (2005) demonstrated that members are more likely to face a quality

 

25 Bishin (2000) later found empirical support for F enno’s earlier anecdotal evidence
involving sub-constituencies.

23

challenger the more often they vote with their party on key issues.26 In sum, members
say they care about the ideology of their districts when they are deciding how to vote
and the empirical evidence suggests that representatives also act as if it is important.
Additionally, evidence suggests that voters are paying attention (or at least acting as if
they are) since there is a possibility that when members begin to ignore their districts
and vote in too extreme a fashion, they are punished at the polls.

Although the electoral accountability literature finds that members who are a good
match for their district do better at the polls, it does not directly test if representatives
are able to (or chose to) adapt to changes in their district. While Poole and Rosenthal
(1998) have argued that members “die with their ideological boots on,” that is,
members do not change their voting behavior over their careers, one exception to this
rule has been when a member’s district changes as a result of redistricting (Glazer and
Robbins 1983, 1985; Stratmann 2000, Boatright 2004). These scholars found that
members are indeed responsive to changes in their district.27 Through similar research
designs but various measures of behavior (conservative coalition scores, ADA scores
and, Dw-NOMlNATE) they all concluded that as a district becomes more liberal or
conservative so does the representative. Based on these results, it appears that some
members do not just happen to “fit” their district, rather many are willing to actively

adjust their behavior if their district changes.

 

2" These results hold even though surveys indicate that voters are frequently not well
informed about their representatives. See Stokes and Miller 1966, Mann 1978, or Delli
Carpini and Keeter 1996 for more on the public’s knowledge of candidates. For more
research on how member’s voting behavior can influence their election results see
Sehoenberger 1969, Erikson 1971, Johannes and McAdams 1981, Brady et a1. 1996 and
Jacobson 1996.

27 Sec Poole 1998 for a critique of Glazer and Robbins 1983.

24

Theories of Congressional Organization
In comparison to the literature on electoral accountability, a substantial segment of

the congressional organization literature argues that party plays a consequential role in
deciding legislative outcomes.28 If this body of work is to be believed, then party must
surely play a part in helping members to decide how to vote. One sub-section of the
literature which falls under the label “conditional party government,” argues that party
strength is conditional on internal party homogeneity and external heterogeneity
between the two parties (Rohde 1991; Aldrich and Rohde 1998, 2000, 2001). When
both conditions hold, rank—and-filc members give up power to the leadership so the
party can “encourage” members to act in ways to further the party’s goals. One way to
stay in favor with the leadership is to vote with the party when needed on important
legislation.

Another related theory, cartel theory, tells us that the key to legislative success in
Congress is not necessarily trying to enforce party discipline on substantive votes, but
lies in controlling the agenda (Cox and McCubbins 1993, 2002, 2005).29 Here, we can
think of party success in terms of positive and negative agenda control. Positive
agenda control is associated with carrying legislation forward in the legislative process
and negative agenda control is the ability to stop legislation from coming to a vote.
Parties act as procedural cartels by exercising negative agenda control to assure that no

legislation reaches the floor that could possibly split the party or move the status quo in

 

28 For a thorough review of the congressional organization literature see Cox and
McCubbins 2005. Of course, partisan theories are not without their critics; sec e.g.
Krehbiel 1993, 1999, and 2000.

29 Conditional Party Government and Cartel theory are not necessarily contradictory.
See F inocchiaro and Rohde 2002 for a discussion of the similarities and differences
between the two theories.

25

ways that are unfavorable to the party. To make certain that such legislation (or
amendments that would move policy in an adverse direction) does not receive a floor
vote, rank-and-file members are expected to support the party on procedural votes in
exchange for the possibility of securing a more powerful position in the institution and
increase the probability of maintaining (or achieving) majority status. This idea is
supported by Oleszek (2004:141) who argues that the majority party demands support
on rules votes and any defectors could be punished.

Although positive agenda control is conditional and hence variable, negative agenda
control, Cox and McCubbins (2005) argue, is not conditional but constant. They argue
that, “party pressure can affect members’ decisions on procedure more than their
decisions on substance,” (66). This gives members more freedom to vote their district
on substantive votes but not on procedure. Of course, this does not mean that party
does not try to pressure members on final passage votes, only that the party is relatively
more successful when it comes to keeping members in line when it comes to voting on
procedure.30 This point is demonstrated by the high percentage of votes that are party
unity votes on procedural motions, 89.1% versus final passage votes, 38.7% (Crespin,
Rohde and Vander Wielen 2002).3 '

Although previous research (Glazer and Robbins 1983, 1985; Stratmann 2000,

Boatright 2004) found that members are adapting to meet their new districts, their

 

30 Sometimes the party can pressure members on final passage. In the 108th Congress,
for example, House Republicans held open the vote on the Medicare bill for three hours
while the leadership successfully pressured C.L. "Butch" Otter (ID) and Trent Franks
(AZ) to vote yes in order to secure passage. Jackie Koszczuk and Jonathan Allen
“Late-Night Medicare Vote Drama Triggers Some Unexpected Alliances,” C Q Weekly
(1 1/29/2003) p. 2958

3 I Where a party unity vote is defined as a majority of one party voting against a
majority of the other.

26

research designs did not account for the fact that members may be more responsive to
their district on some votes and more responsive their party on others. In other words,
representatives may vote with their district on substance, but by voting with the party
on procedure they can stop legislation from coming to a vote. If legislation that the
district wants never comes to a vote because of procedure, then the district, one could
argue, is not fully represented. If we truly want to know if members are adapting to
their new constituencies, and providing complete representation, then I argue that it is
necessary to look separately at votes that are visible to the district, namely final passage
votes, and votes that are important to the party but largely invisible to the district,
procedural votes.

If cartel theory is correct and the party applies strong pressure on procedural votes
but members are relatively free to vote their district on final passage, then it follows
that as a member’s district median moves, she should modify her behavior on final
passage votes in the direction of the change. However, if the party demands discipline
and is able to force members to vote against their district on procedural matters, then it
should also follow that changes in the district should not be related to shifis in voting
on partisan procedural motions. Rather changes in the party median should explain any
shift in voting behavior on procedural votes. We can think of members having two
masters to serve, but each puts emphasis on different matters. Members have to please
their district on substantive matters and their party on procedure. The task, however, is
not as troublesome as it may appear since procedural voting, which can stop any

potentially troublesome issues from coming to the floor, is not visible to the district.

27

This way, members will not be forced to choose between their district and the party
when it comes to roll calls that are noticeable by the district.
District Change and Behavior Change

In order to test these hypotheses, I take advantage of the same quasi-experiment
presented by the shifting of district boundaries as a result of redistricting that Glazer
and Robbins (1983, 1985), Stratmann (2000), and Boatright (2004) used.32 In this
“experiment,” a member’s district -and the ideological characteristics therein- is
allowed to vary while the party affiliation is held constant.33

I use the spatial model depicted in Figure 1 to provide a visual representation of my
argument. Each of the two lines represents the ideological spectrum before and after
redistricting with liberal on the left and conservative on the right. Each member has
two ideal points, one for reelection that is largely a function of district characteristics
and is based on votes that are visible to the district and one for the party that is a
function of party pressure and is realized through voting on procedure. Based on the
theoretical expectations, I assume that a) procedural votes are largely invisible to the
district but final passage votes are visible, and b) the party is more successful at
pressuring members on procedural votes.

In the figure, XRL. represents the district reelection ideal point, for member i at time t,

where t is some period before redistricting and XRi = f (district characteristics, party

 

32 This is a quasi, rather than a natural experiment, because the changes in the district
boundaries are not random nor are they under control of the researcher. Instead, they
are the result of actions other actors. See Campbell and Stanley 1963 for a discussion of
Quasi—experiments.

3‘ This experiment is, in effect, the mirror image of Nokkcn (2000, 2005) and Nokkcn
and Poole (2004) who hold districts constant but allow party to move for party
switchers. Their findings confirmed the notion that a member’s voting behavior will
change if he switches parties, even as the district remains constant.

28

pressure and personal preferences). Xpm represent the individual party ideal point that
member i at time t must meet in order to avoid punishment from the member’s party.
Note that there is a party ideal point for each member (although it need not be unique),
not one point for the entire party since some defection may still occur. It is further
assumed that XP, = f (party pressure).34

Between time t and t+1, the district boundaries for each member’s district may
change as a result of redistricting and in many cases, this will cause a shift in district
characteristics.” Further, since the make-up of the House also changes, the individual
party ideal points can also shift.36 The new ideal points will then be inw and XP n+1-
Based on the spatial model and theoretical expectations, I will test the following
hypotheses over two sets of votes, final passage and procedural:

H1: As a member’s reelection ideal point shifts left (right) the member’s
behavior should shift left (right),

H10: As a member’s reelection ideal point shifts left (right) the member’s
behavior should not shift lefi (right),

and

 

3" It is not necessary to assume that XP, is more extreme than, XR; and the theoretical
expectations will not change if it fell to the left or right of XR,. These ideal points do not
necessarily have to be the medians, but rather some point that allows the member to
secure reelection for the member. Research by Ansolabehere, Snyder and Stewart 2001
demonstrates that members do not converge to the district median, rather Democrats
are to the left and Republicans are to the right. This may be related to personal policy
preferences, the threat of a primary opponent or other reasons. Sec F iorina 1999 for a
discussion of why members may diverge from the district median. Similar logic can be
applied to the party ideal points. It is only necessary to have some ideal point that is
representative of the party. It could be the median, but could also be represented by the
“leadership”.

35 For simplicity, I assume that a member’s personal preference will not shift between
the two periods.

36 I am purposely being agnostic about why the individual party ideal point may change.
It could be a change in the agenda, leadership, or rules.

29

H2: As a member’s party procedural vote ideal point shifts left (right) the
member’s behavior should shift left (right),

H20: As a member’s party procedural vote ideal point shifts left (right) the
member’s behavior should not shifi left (right),

If the reelection ideal point is a function of district characteristics, as the electoral
accountability literature suggests, then as a member’s district changes, their behavior
should shift accordingly between periods t and t+l. This means that H10 should be
rejected on votes that are visible to the district (final passage) and can influence a
member’s reelection chances. Conversely, since the individual party ideal point is not a
function of individual district characteristics, then I expect to fail to reject the null
hypothesis that changes in the district should produce shifts in X: between periods t
and t+1. This hypothesis provides a critical test of the theoretical expectations. If
changes in the districts predict changes in an individuals member’s party ideal point,
then this would be evidence against the theory.

Both the electoral accountability literature and the congressional organization
literature predict that party can sometimes pressure members on votes that are visible to
the district. This suggests the secondary hypothesis, H2. If party pressure is strong for
visible votes, then H20 should be rejected if the party ideal point changes after
redistricting, but not if party is weak. Finally, cartel theory predicts that party can
pressure members on procedural votes, therefore, I expect to reject H20 on these types
of votes, if the party ideal point Shifts.

Data
In order to measure members’ voting behavior on votes that are visible to the district

and procedural votes, I use a variation of Poole and Rosenthal’s (1997) w-NOMINATE

30

procedure to estimate ideal points on both subsets of votes.37 More specifically, I use a
“bridging” technique (see Rivers 2003, and Bailey and Chang 2001). The technique
calls for two members from the wings of each party to be used as glue to bridge the two
sessions together in order to insure comparability. In this case, I used Tom Delay who
has been described by Politics in America as possessing a “conservative zeal,” and Jose
Serrano the most liberal member of the House according to National Journal.38 Both
sessions are then estimated using one roll-call matrix to obtain ideal points for each
member for the first sessions of the 107th and the 108th Congress with the same
procedure used for both vote-type categories. The coding for the vote-types is based
on a categorization by Crespin, Rohde and Vander Wielen (2002) and applied to the
Rohde (2004) dataset for subsets of the 107th and 108th Congresses. All bills and
resolutions that had a recorded final passage vote are included in the visible vote
category since they are easily linked to policy outcomes. Meanwhile procedural
motions that enable partisan advantage such as voting on special rules, the motion to

recommit or moving the previous question make up the procedural grouping.39

 

37NOMINATE procedures in general are based on the theoretical assumptions from the
spatial theory of voting used to develop my hypotheses (Poole and Rosenthal 1985,
Lewis and King 1999). Alternative measures such as ADA scores cannot be used since
they do not differentiate between substantive and procedural votes and are also based
on a relatively small subset of votes. Dw-NOMINATE also cannot be used because it
needs n > 2 congresses to calculate a score and is based on all votes, not just different
sub-sets of votes. Finally, since others have used different measures and found
significant changes in behavior after redistricting, I think it is reasonable to assume that
any significant results are not a function of the scaling procedure.

38 Note that in an alternative model, I estimated the ideal points separately for each
session and the results were similar, with the exception of the change in party median
variable where 1 did not have theoretical expectations in the final passage votes
regression. It was significant using the alternative technique.

39 The appendix lists the specific types of votes in each of the two categories. In the
first session of the 107’h Congress, there were 230 final passage votes and 85 partisan

31

Similar to previous work by Rothenberg and Sanders (2000) and Boatright (2004), it
is necessary to calculate the ideal points for two time periods, one before the exogenous
shock (redistricting) and one afterwards. I chose the first session of the 107th Congress
(2001) as the pre-redistricting period and the first session of the 108th (2003) as the
post-redistricting stage. In the first session of the 107th, states were beginning to
formulate redistricting plans while by the first session of the 108th Congress, nearly
every multi-district state with the exception of Maine had enacted their new plan."0 If
the entire 107th Congress were compared to both sessions of the 108th, it is likely that
some of the behavioral shift will be missed since members will have begun to change
their behavior in the second session of the 107th in time for the 2002 elections."I
Further, I only included members who were incumbents and ran for reelection to the
108th Congress since retirees and higher office seekers may not change their behavior
in ways to suit their new congressional district (Rothenberg and Sanders 2000)."2 In
order to measure changes in member behavior, I subtract the member’s NOMINATE
score for the 108th Congress from the score for the 107th for both subsets of votes to
create two separate dependent variables, Final passage and Partisan procedural."3

Since NOMINATE ranges from -1 (liberal) to l (conservative) this means that a positive

 

procedural votes. For the first session of the 108th Congress, there were 283 final
passage votes and 176 partisan procedural votes.

40 Maine rcdistricted in 2003.

4' On this point, see Boatright 2004.

42 Including retirees and higher office seekers does not change the results.

4" These variables correspond to (XRL, - XRLH.) and (XPL, - XPLH.) respectively.

32

change in scores corresponds to a shift to the left and a negative difference means the
representative moved to the right.“

To gauge district change I use two measures, one political measure which is the
standard change in the presidential vote in the district (Glazer and Robbins 1983, 1985;
Stratmann 2000; Boatright 2004) and another variable that more generally captures the
degree of district continuity after redistricting based on new population. The first
measure, change in Presidential vote, is taken from C Q ’s Politics in America and is the
difference between the two-party Democratic share of the vote in 2000 and the share of
the vote recalculated to the new district boundaries for the 2002 elections. Since this
measure spans a redistricting, it was necessary to match up the correct districts
corresponding to where members stood for election before and after redistricting.
Based on the theoretical expectations, the sign on the coefficient for this variable
Should be negative and significant for final passage votes but not different from zero for
procedural votes.

The second measure of district change, detailed in Crespin (2005), is created by
using geographic information systems to spatially intersect district boundaries for the
2000 and 2002 congressional elections. These new areas are then overlaid on top of
census tract boundary files. After this step, the census tracts are deemed new (drawn
into the incumbent’s district), removed (drawn into a new district and no longer
remaining in the incumbent’s district) or continuous (remain in the district pre and post-
redistricting). After this step, the population of the tracts that are new to the

incumbent’s district is divided by the total district population and the result is the

 

4" For example, John Dingell (D-MI) was at -.326 in the 108’h Congress and -.532 in the
107th Congress, therefore -.532--.326 = -.206 means he shifted to the right.

33

percent of the population that is new to a member’s district. This measure, percent New
population, is included in the model separately and also interacted with the change in
presidential vote in order to capture political and population change in the district. The
interaction allows for a variable that measures both direction and magnitude of change.
As districts receive a greater percentage of new constituents, I expect it to be more
difficult for members to gauge the political make-up of the new geographic
constituency. These three variables are used to test hypothesis one. In order to test
hypothesis two, I created a variable that measures the change in the Party median from
the first session of the 107th Congress to the first session 108th for each sub-set of
votes to gauge the effect that changes in the party ideal point has on a member’s
behavior."5 If members are moving with their parties, then this variable should be
positively signed.

The theoretical expectations and hypotheses are displayed in Table I. First, the
effect of district change should be negative and significant for the final passage votes
model, but not diflerentfiom zero for procedural votes. Again, this is the key test of
the paper. Rejecting the null will provide strong evidence against the partisan theory.
Second, the change in party median variable may be significant in the final passage
model if the party is able to exert pressure on these types of votes but it will have a
Significant effect for voting on procedure since party the theory argues that party
applies pressure on procedural votes. Since the dependent variables (the change in a
member’s ideal point from time t and t+1) are theoretically continuous, ordinary least

squares regressions are used to estimate the results.

 

45 Bernie Sanders (l-VT) is coded as a Democrat Since he votes with them to organize
the House.

34

Results
Table 2 displays the results from both regressions.46 In the final passage vote

regression, the change in presidential vote is negative and significant, suggesting that
member’s of Congress change their voting behavior in accord to changes in the district.
If a district becomes more conservative (or liberal) so does the member and is
consistent with the previous research (Glazer and Robbins 1983, 1985; Stratmann 2000;
Boatright 2004). This result suggests that the elections literature is correct, members
act in accordance with their district’s preferences in order to secure their reelection goal.
The coefficient on change in the party median, however, is not significant, suggesting
that although there is anecdotal evidence of parties influencing voting on substance; in
this case I do not find any systematic evidence to support the claim.

The results from the partisan procedure regression show that as a member’s district
changes, they do not change their behavior accordingly. Instead, the results suggest
that shifts in the parties’ ideal points and the resulting pressure to conform influences
changes in their voting behavior on procedural votes. This is evidenced by the positive

and significant coefficient on the party variable. Further since this variable and the

 

46 In order to test the robustness of the findings, 1 ran diagnostics and estimated several
other regressions. Since both models suffer from heteroskedasticity, the
Huber/White/sandwich estimator of variance is used to obtain robust standard errors. A
test for outliers revealed that none were substantially influencing the results. I also
estimated models with state fixed effects and another regression that corrects for
multiplicative heterosccdasticity to determine if individual states were influencing the
results. In either case, they were not. The difference in the number of cases between
the two models is a result of not being able to estimate ideal points for all members in
the procedural vote category. Again, this does not substantively change the results
since they were similar if the models were run on the same set of members. Finally,
various other measures of party such as a simple dichotomous variable or shifts in party
leadership ideal points were tried and the conclusions drawn were once again quite
similar.

35

dependent variable are measured on the same scale, the coefficient of .93 means that as
the party moves, individual members move nearly the same distance. This result gives
support for the partisan theories of legislative organization, especially cartel theory
which argues that the party will use procedural motions to exert agenda control. Since
the engine for agenda control is procedure, this is where we would expect to find the
strongest party influence. This is also substantiated by the relatively large adjusted R2
(.84) for a model with party as the only significant variable. Further, the procedural
votes model does a much better job at explaining the variance in changes in voting
behavior since the R2 for the second model is quite large relative to the first."7

In order to interpret the district change terms correctly, I turn to a graphical
procedure suggested by Brambor, Clark and Golder (2005). Their procedure calculates
substantively meaningful marginal effects and correct standard errors for interaction
terms. Figure 2 displays the marginal effect of district political change on voting
behavior change, conditional upon the percent new population in the district for both
models. The solid line in each of the graphs is the marginal effect while the curved
dashed lines indicate the 95 percent confidence intervals. If the confidence intervals
bound the dotted zero line, then the variable is not significant at that point. In the final
passage model, we see that the marginal effect of district political change is negative
until percent new in the district reaches just over 32 percent when the variable is no
longer significant."8 The increasing slope of the line indicates that the effect dissipates

the more the district changes. The decreasing effect is likely indicative of growing

 

47 Although the R2 for the final passage vote model is quite small at .03, it is consistent
with other similar studies.
"’8 This range represents nearly 75 percent of the cases.

36

uncertainty about district characteristics. F enno (1978) found support for this notion in

his district travels as one urban congressman remarked about his rcdistricted territory:
“Let me pull over to the curb and look at the map. I can’t tell whether this area
is in the district or not. It could be, but I’m not sure. That must have been some
redistricting when the congressman who represents the district can’t even tell
whether he’s in it or not,”(6).

If the members elected to serve in a district do not know who is in and out of their

district, then this implies that there may be a learning curve for members becoming

acquainted with new parts of their district when the boundaries change by large

amounts.

In the partisan procedural model, the confidence intervals continuously bound zero
and the solid marginal effect line actually lies above the zero line signifying that no
matter how much the district changes, members do not change their behavior
accordingly, providing strong evidence in support of the key hypothesis. In sum, (see
table 3 for a summary of the findings) representatives change their behavior in
accordance with theoretical expectations on votes that their constituents (or their
potential opponents) are paying attention to. This give support to the simple Downsian
hypothesis that elected officials will alter their own ideal points in order to secure
reelection or other goals. In contrast, shifts on yeas and nays on procedural motions are
not related to movements in the district median, rather party explains much of this
change. Finally, we only see evidence of members moving with the party on

procedural matters, not final passage.

37

Conclusion and Discussion
The results demonstrate that when a member’s district changes, they shift their

voting behavior on votes that their constituents are paying attention to, specifically,
votes on final passage. However, I fail to find a relationship between district changes
and how a member behaves on votes on which the party exerts influence, namely
procedural votes. Although we generally need to be careful about drawing conclusions
from null hypotheses, the chosen research design alleviates several of the problems.
First, since both hypotheses are tested over identical models that only allow the
dependent variable to vary, we can be more confident that the null results are not just a
function of chance. Second, the null result of no effect is the prediction of the theory,
not just the usual hypothesis test that assumes a null of zero. Finally, providing
evidence in support of a theory with null results is not without precedent (See e. g.
Krehbiel 1992 and Cox and McCubbins 2002).

I also find that changes in party medians explain a large share of the shift in voting
on procedural votes, but not on final passage. In sum, members appear to be wearing
two pairs of boots, only one of which is made for walking. Although this paper does
provide evidence that “parties matter,” it also implicitly shows how they matter. While
previous research has demonstrated that members vote in accordance to their new
districts, the findings may obscure the effects of party. If negative agenda control can
keep issues off the floor, and members are not representing their new district on these
types of votes, then representative democracy may not be working as well as previous
research implied. Rather, it appears that elected officials are only representing their
districts when they are paying attention and not voting with their new constituents

when they can get away with it on procedural matters. Based on these results and what

38

the literature says about the importance of procedural control, I can conclude that a
changing constituency may not be getting everything it wants from their members of
congress.

In future work, I hope to better link member’s voting behavior on various vote-
types to their election results. Although previous research has demonstrated that
members are punished or rewarded based on voting behavior, it does not test this
hypothesis over different types of votes. If constituents are not paying attention to how
a member votes on procedure, then we should not expect to find a link between election

outcomes and these types of votes.

Table 1 — Hypotheses and Predictions

 

Party-Influence Predictions
on Voting Behavior

 

 

 

Final Passage Vote Procedural Vote
Regression Regression
A Reelection Ideal Point Negative and Not Different from
Significant Zero
A Party Median No Prediction Positive and Significant

 

Note — Directions are a function of measurement

 

40

Table 2 — Change in Member VotingBehavior as a result of Redistricting

OLS Coefficient
(Standard Error)

 

Partisan Procedural

 

 

 

 

Variable Final Passage Votes Votes
District C hangg
A Presidential Vote ,,
(2000 Dem. vote in old 111$) ('83:)
district — new district) ' '
o/o New Population (833;) ((0333?
A Presidential Vote X .0003* -.00006
% New Population (.0001) (.0002)
Party
. .138 934*
A Party Median (.112) (.022)
-.434* -.014
Constant (.057) (.011)
N 356 353
Adj. R2 .03 .84
F 285* 517.96*
*p < .05

Dependent Variable — 107th Congress first session NOMINATE minus 108th Congress
first session NOMINATE. Negative coefficient corresponds to a shift to the right. Only
members serving in both Congresses and not retiring or running for higher office in the
2004 elections are included.

41

Table 3 - Substantive Results

 

Variable Final Passage Votes Partisan Procedural

 

Votes
A Presidential Vote
(2000 Dem. vote in old
(“8100‘ — new diSlTiCt) As the Democratic As the Democratic

presidential vote in the presidential vote in the
district increases, the district increases,
member becomes more behavior does not
liberal change

(See Figure 2)

°/o New Population

A Presidential Vote X
% New Population

 

Members do not move Members move in the
A Party Median in the direction of the direction of the party
party median median

 

42

Figure 1 — Spatial Representation of District Change

107th A l 1

Congress f I I

P R
X 1,1 X 1,1

V

District change through redistricting between 107‘h and 108th Congresses

 

108‘“ - l 1 -
Congress‘ I I '
9 R 9 7 9
<— —-> <— R——>
X i,t+l X i,t+1

- Where XRiJ is the reelection ideal point for member i at time t and XPiJ is the
party ideal point for member i at time t.

' x11” may shift as a result of redistricting, members should respond on final
passage voting, but not on procedural votes.

0 XPt may shift as well, members should respond on procedural votes and
possibly on final passage votes.

43

Marginal Effect ofDistrict Political Change

Figure 2 - Marginal Effect of District Change on Voting Behavior Change

 

 

 

 

 

 

 

 

Final Passage _ Partisan Procedural
8 "‘ \
v
o. -i
' f T I I I I I I I I I I
0 10 20 30 40 50 0 IO 20 30 40 50

Percent New Population in the District

 

 

Marginal Effect of Change in Democratic Presidential Vote
----- 95% Confidence Interval

 

 

 

G raphs by Vote-type

44

 

Chapter 4 - Geography and Individual Campaign Contributions to Congressional
Candidates

Introduction
Individual contributions represent over half of the total money given to congressional

candidates (Jacobson 2004), yet relatively few scholars focus on the contributions that
originate from this important group of political activists. One exception to this lack of
research is F rancia, Green, Herrnson, Powell and Wilcox (2003: 105) who argued that,
“Geography is the most important factor influencing contribution decisions.” F rancia
et al. (2003) found that contributors are 33 percent more likely to give to a candidate if
they reside in that representative’s district. On the surface, this would seem to run
counter to Grenzke (1988) which reports that over half of individual donations come
from outside a member’s district. However, if the roughly 434 possible contributors
outside the district for every one inside the district are taken into account, then this
outcome becomes empirically plausible.

Although the numbers “add up,” a set of interesting questions still remains. What
enables a member of Congress to collect a substantial portion of her contributions from
outside her district? Are different forces influencing contributors based on their
geographic location relative to the candidate? Finally, what does it say about
representation if members’ money constituencies lie outside their geographic
constituency? In this paper, I seek to understand which factors contribute to incumbent
members of Congress receiving money from areas inside and outside the district. 1 will
argue that a member’s ability to fundraise from different geographic areas is
constrained by both supply and demand of the donor pool and both of these factors are

a function of incumbent, election, and district characteristics.

45

In the following section of the paper, I examine the influence of redistricting on a
member’s fundraising totals. Although most of a district remains intact after
redistricting, some parts of the district may be removed while other areas are added.
Consequently, redistricting allows for a quasi-experimental research design where the
bonds of representation are severed for some individuals who find they are living in
new districts after the boundaries have changed. By way of a new continuous measure
of district change, I can test a member’s ability to earn contributions from constant,
new, and old areas of the district. This will shed light on whether members form
representational bonds with new constituencies as well as whether they hold onto their
link with old donors who have left the district.

Next, I review redistricting and campaign finance literature in order to develop a
theory and test hypotheses with respect to individual campaign contributions to
incumbent members of Congress residing from different geographic locations.
Building on this, I am able to determine the influence of redistricting on a member’s
ability to fundraise and show that there is a difference between the two parties. Finally,
I conclude and offer suggestions for future research.

Previous Research
Redistricting

The literature on congressional redistricting is quite extensive. Butler and Cain
(1992) wrote one of the most comprehensive examinations of this topic. They cover
the court decisions that led to the revival of nationwide redistricting in the 19605, how
the process worked during the 19805 and 19908 as well as comparing the efforts by

different states.

46

A portion of this research has been devoted to the influence of redistricting on
congressional stability (see, e.g., Bullock 1975; Gelman and King 1994a), minority
representation (see, e.g., Lublin 1997; Lublin and Voss 1997; Canon 1999; Epstein and
O'Halloran 1999; Shotts 2001) and responsiveness (Tufte 1973; F erejohn 1977;
Gelman and King 1994b; Ansolabehere, Snyder, and Stewart 2000, 2001).
Redistricting has also influenced the competitiveness of individual House races over
time. Prior scholars examining this issue have linked the decline in competitiveness to
the grth in the ineumbency advantage (Mayhew 1971, Tufte 1973).49

Beyond the ineumbency advantage, scholars have also focused on the aggregate
effects of redistricting in terms of the partisan balance of seats in the House
(Abramowitz I983; Cain 1985; Butler and Cain 1992; Niemi and Winsky 1992; Niemi
and Abramowitz 1994).50 In the best recent account of the consequences of the
redistricting process, Cox and Katz (2002) showed how the courts influenced the
drawing of new district boundaries in the 19608. In addition, Carson and Crespin (2004)
showed that competition can vary when courts or commissions become involved with
the redistricting process.

Scholars have also focused some attention on changes in member’s voting behavior
as a result of redistricting (Glazer and Robbins 1986; Stratmann 2000; Boatright 2004).
Finally, some have shown that members may concentrate work directed at different
parts of the district based on changes in sub-constituencies after redistricting (Bickers

and Potoski 2000, Grose 2000, Ansolabehere, Gerber and Snyder 2002).

 

49 For a response and critique of the Tufte argument, see Bumham 1974, F crcjohn 1977
and Cox and Katz 2002.

50 On this topic see also Squire I985, Cain and Campagna 1987, Basehart and Comer
1991, Lyons and Galderisi 1995 and Swain, Borrelli and Reed 1998.

47

Until recently, redistricting students have not paid much attention to how altering
district boundaries can influence campaign fundraising. Boatright (2004) demonstrated
that candidates increase their fundraising in response to partisan and racial changes in
the district. Boatright’s piece, unlike this paper, focused on total fundraising numbers
(PAC and individual) by the candidates and did not directly account for movement of
new and old constituents.

Campaign Fundraising

Similar to the redistricting literature, there is quite a lot of work dealing with
congressional campaign fundraising. Most of the research focuses on political action
committees (see e.g. Jacobson 1980; Smith 1985; Wright 1985; Hall and Wayman 1990;
Cigler and Loomis 1991; Romer and Snyder 1994; Stratmann 1995; Wright 1996;
Hermson 1998; Herrnson, Shaiko and Wilcox 1998; Box-Steffensmeier and Radcliffe
2001; Jacobson 2004). However, few scholars focus on individual campaign
contributions, even though they make up over half of all contributions to congressional
elections (Hermson 1998; Jacobson 2004).

In the words of the some of the leading scholars in the field of campaign finance,
“There is little systematic research on the motivations and strategies of individual
congressional donors,” (Francia et al. 2003:12). Some of the earliest to study
individual contributors were Jones and Miller (1985) who used questions from the 1980
National Election Study to examine determinants of the donation. They found that
voters, people who were interested in politics and respondents who felt a high degree of
civic duty contributed to candidates. Another early look at individual donations by

Grenzke (1988) demonstrated that over 50 percent of individual and nearly all PAC

48

contributions come from outside of the district. However, most of the out of district
money did come from inside the member’s state.5 '

Recently, F rancia et al. (2003:16) undertook a survey of contributors to the 1996
congressional election and provided a “thick description” of individual contributions.
They also relied on a smaller longitudinal survey of “big” donors as well as interviews
with donors and campaign officials. They found that contributors were, for the most
part, rich, highly-educated white males. However, reasons for contributing were quite
varied. They stylize three types of contributors, “investors” who gave for financial gain;
“ideologues” who gave for more broad societal reasons and finally “intimates” who
gave for reasons of solidarity, such as a personal relationship with the candidate or their
fundraising surrogates.52 Although this study has helped to lay the groundwork for
studying contributions from individuals, there are still many questions left unanswered.
Geography and Fundraising

Many accounts that offer a description of individual fundraising (Grenzke 1998,
Hermson 1998) point out that a significant portion of funds come from outside a
member’s district with a non-trivial fraction coming in from outside of the state. Since
money is a necessary factor to win elections (Davidson and Oleszek 2004), then
candidates will go wherever the money is to get it. A member’s ability to fundraise
from different geographic regions - district, state, and out of state - should be based on

their demand for resources, which is constrained by the supply. In congressional

 

5|Grenzke’s study included the 1981-82 election cycle that encompassed 1982
redistricting but she did not take this into account.

52 See Snyder 1990 and Baron 1994 for more discussion on investors and Roscnstone
and Hanson 1993 for ideologues.

49

elections, supply and demand is a function of three factors, those specific to the
incumbent, the election, and the characteristics of their district.
Incumbency

Previous research (Franica et al. 2003) suggests that a candidate’s ideology is an
important factor in the decision making calculus for individuals who contribute to
congressional campaigns. Indeed, Grenzke (1988) found that as members become
more liberal, they are more likely to earn a greater share of their campaign funds from
outside the district. The political parties like to claim that extremists from both ends of
the spectrum are improperly influencing members of Congress because the supply of
potential donors increases as they become more extreme. Citizens who are moderates,
and do have strong opinions on many issues, will be less likely to make a contribution.
The Republicans like to blame the Hollywood liberals while the Democrats favorite
whipping boys are the Texas oil tycoons. In addition, both parties like to point to Wall
Street fat cats as contributing to problems with government. If these relatively small
groups from distinct geographic areas are contributing to polarization in politics, then
extreme members should get a greater share of their contributions from out of their
district and state, relative to more moderate members.

The size of a candidate’s donation network should have a significant effect on a
member’s ability to raise funds from outside of the district by increasing the potential
money supply. Once members have tapped the bulk of their district for funds, they will
have to look elsewhere for money. If members want to increase the size of their
network, they will have to look outside of their district. Therefore members with large

networks will have an easier time earning contributions from out of state compared to

50

members with small networks. In view of the fact that incumbents are unlikely to give
up their lists of potential donors, I argue that a representative’s seniority will be a good
proxy for the size of their donor network. According to F enno (1978), members spend
the early part of their career in what he called an “expansionist phase.” During the
phase, members work to increase their base of support. Although F enno’s work
focused on the member in the district, it is not much of a stretch to argue that members
will also work to expand their donor list during this stage of their career. The longer
members serve and the larger their networks, then they are more likely to earn a greater
portion of their contributions from outside of the district and less from inside the
district.

Members who serve on important committees receive more contributions from PACs
(Herrmson 1995), but what about from individuals? Prestige committees such as Ways
and Means and Budget tend to produce bills that target broad constituencies from
around the country. Therefore, members on prestige committees should attract more
money from outside their district. In contrast, constituency committees produce bills
that are more geographically focused. Further, members frequently try to serve on
constituent committees that are relevant to their constituency. If this is indeed the case,
then representatives on constituency committees should collect a smaller percentage of
their funds from out of state, and more from inside the district. Finally, a member’s
party affiliation should affect their ability to raise funds from different areas. Since
Democrats tend to represent less affluent districts and thus have a smaller supply of

money in the district, they are forced to go outside of their district to look for funds.

51

Thus, they should collect less from inside the district and more from other areas of the
state compared to Republicans.
Election

Election factors that will influence where members raise funds fall largely on the
demand side. The factors include the total amount of spending in the race and whether
the incumbent is facing a quality challenger. Both of these election characteristics send
a signal that the incumbent is in a competitive race and have a greater demand for funds.
In terms of spending, the more money that is spent, the more that the incumbent needs
to raise. Also, if donors see that a race is competitive, they may be more likely to
contribute since their contribution may make a difference in the outcome. Given that
the donor pool within the district is finite relative to the pool outside the district,
members who are in high spending races will need to look outside the district for more
money. Therefore, as total spending increases, members should earn a smaller
percentage of their contributions from inside the district and more from areas outside
the district and state. Another signal that the incumbent is in a competitive race is
whether or not their opponent is a quality challenger, defined as someone who has held
previous elected office (Jacobson 2004). These candidates should also have a greater
demand for more money and they too should collect more money from outside the
district and state. Further, if it appears that an incumbent is vulnerable, then political
activists from around the country may contribute in order to defeat the sitting member

of the House.

52

District Factors

Finally, there are supply and demand factors relevant to the member’s district that
will influence their fundraising. An incumbent who resides in a district that has more
wealth, and thus a greater supply of money, should collect more money from inside
their district. In contrast, members in poorer districts will have to look beyond their
district and possibly outside the state for contribution.

Since fundraising can be costly, accounting for nearly 12 percent of a House
candidate’s budget (Hermson 2000), then members will likely try to minimize costs.
One way to keep costs low is to solicit funds from people that live in or near the district.
F enno (197 8) argued that those closest to the member, inside the personal constituency,
are often a source of campaign funds. However, any candidate who needed to raise
enough money to win a congressional election would quickly bankrupt that circle if he
repeatedly relied on them as the only source of campaign funds. Further, the
limitations placed on the size of donations from individuals would also make it difficult
to run a successful campaign if the personal constituency was the only donor pool. As
F rancia et al. (2003) argue, this group does makes for a good source of seed money, but
it is probably not enough to win. The next logical step then, is to look for the next
closest set of donors. Donors who live near the incumbent’s district make for relatively
less costly solicitations and are more likely to have a connection (personal or business)
with the incumbent or his chief fundraisers. In this respect, members of the personal
fundraising constituency need not be inside the geographic district. Previous research
(Grenzke 1988) is consistent with this idea and found that most of the money comes

from inside the state or district and contributors are more likely to give if they live in

53

the candidate’s state or district (F rancia et al. 2003). If this is true, then the size of the
state will be important in determining how much money candidates receive from
different geographic areas. Incumbents who live in larger states should be less likely to
collect money from outside the state. However, since the state’s donor pool is large
relative to the districts, these members will be able to earn more from the state than
their districts.

Members who live in districts where they are a good political match to the
underlying political conditions should demand less money from their district because
they are relatively safe and insulated from a strong challenger from the other party.
Owing to their safety, they may also be able to spend less time in the district and more
time elsewhere raising funds. In sum then, a member’s ability to raise funds is
dependent on their demand for money and where they can find it. Members who need
more money, will go where they have to in order to get it.

Campaign Finance Data

In order to test the hypotheses developed above and throughout the paper, I turn to
data for incumbent members of Congress who ran for reelection in the 2000 or 2002
congressional elections. The 2000 election to the 107th Congress was the last held with
the old district boundaries while the 2002 election immediately followed the national
round of reapportionmcnt and redistricting. After the passage of the 1974 Federal
Election Campaign Act (FECA) any contribution over $200 must be reported to the
Federal Election Commission and is then made public.53 Attached to each contribution

is the contributor’s name, ZIP code and the name of the candidate who received the

 

53 Contributions over $200 represent approximately 75 percent of all individual
contributions (Lioz 2003).

contribution.“ By manipulating these data with geographic information systems and
using ZIP codes as the key link between different geographic areas, I can determine
who the contributor gave to and where they lived at the time of the contribution.55

The accounting of ZIP codes and their corresponding congressional districts requires
only simple spatial techniques and data readily available from the US. Census. The
necessary data include the 2000 and 2002 district and 2000 ZIP code boundary files.
They consist of a series of XY coordinates that GIS programs can translate into
polygons that represent the shapes for the districts and ZIP codes.56

The first step in the process is to merge the F EC data with the shape files that
represent the postal ZIP codes. A piece of information about ZIP codes is worth
mentioning here. Technically, a ZIP code is not a geographic area because it is really
just a list of addresses that follow a postal route and lines, by definition, do not have an
area. Instead, the shape files used here are zip code tabulation areas (ZCTAS), which
are a statistical geographic entity that approximates the delivery area for a US. Postal
Service five-digit zip-code. Further, not all ZIP codes are included in the geographic
boundary files. According to the US. census bureau, ZIP codes for businesses that
have their own ZIP codes, post office boxes and general delivery addresses primarily
located in areas otherwise served by a rural route or city style mail delivery are not

represented by ZCTAS. Approximately, 7 percent of the data were lost due to the

 

5" The contributor’s full address is not made public as this would likely invite unwanted
solicitations from other candidates and interest groups.

55 Sec Bolstad 2002 for an introduction to GIS and spatial analysis.

56 The census data and boundary files used in this paper are available at
www.ccnsusgov.

55

inability to match the ZIP code listed in the F EC file with the ZCTAS.57 While this
may appear problematic, assuming that ZIP codes and the ZCTAS are similar is a
necessary assumption to answer the questions posed in this paper.

After merging the two files, I initially summed the total amount of contributions
from each ZIP code to incumbent members of Congress running for reelection.58 I then
overlaid the ZCTA shape file with the boundary files for the respective congressional
districts, the 107th and the 108th. The new attribute table, which is automatically
updated by the GIS program after the intersection, makes it relatively easy to identify in
which district the ZCTAs fall for each election.59

Figure 1 displays these data for each ZCT A for the 2001-02 election cycle. The
shaded areas denote the percentage of the total for each ZCTA that recorded a
contribution that went to Democratic incumbents running for reelection and the black
lines are the 2002 district boundaries. An initial look at the map, and the attached table,
suggests that the Republican Party gets money from more areas of the country

compared to the Democrats. For instance, 6,241 ZIP codes gave 100% of their

 

’7 This unmatched data could also include cases where the ZIP code was incorrectly
reported by the candidates committee.

58 I excluded members not up for reelection since they are likely to change their
fundraising behavior.

59 There are 33,178 total ZCTAS. Since some ZIP codes cross district boundaries, some
post-processing of the data is necessary. Further, spatial analyses frequently create
small spurious polygons called “slivers” which are an artifact of imprecise overlay
(Bolstad 2002:398). The result in the attribute tables are the repeated listing of ZIP
codes with the same number, but with different weights based on the size of the area in
each district. For example, if 25 percent of the area resides in one district and 75
percent in another, the same ZIP code would be listed twice with weight of .75 and .25,
respectively. To erase the slivers and place ZIP codes in the appropriate district if there
was an overlap, I dropped all of the repeated ZIP codes except for the ones with a
weight greater than 50 percent. This technique produced a final attribute table that had
only one unique entry for each of the original ZIP code areas.

56

contributions to Republicans while only 3,752 gave only to the Democrats. The map
also clearly shows the Republicans dominating in the West while the Democrats do
well on the coast and in cities. Both California and New York provide interesting
juxtapositions between contributions from the two parties. Democrats do well on the
California Coast and in New York City, but fair poorly in the Central Valley and in
upstate New York. This is, of course, consistent with how well they do at the polls.
Arkansas appears to be an anomaly in the South with most of the state giving to
Democrats. However, three of the four members of the House from this state are
Democrats, with the exception of John Boozman who represents the lightly shaded area
in the Northwest corner of the state. Generally speaking, this map is consistent with the
electoral success of the two parties. That is, contributions tend to come from areas
where the parties win.

In order to asses the how much money is coming from different areas, instead of just
which party is receiving the contribution, I created Figure 2. In this figure, separated
by party, the shaded areas represent the total dollar amount from each ZIP code, with
darker shading meaning more money. AS one might expect, the Democrats received
large contributions from areas such as Los Angeles and the California coast, Seattle,
Minneapolis, Chicago, Dallas/Forth Worth, Houston, Detroit, Atlanta and the east coast
cities of Boston, New York and Washington DC. The largest ZIP code contributor to
the Democrats was Manhattan, donating over $500,000. The Republicans also did well
in big cities, but closer inspection reveals that they collect quite a bit of money from the
suburbs too. For example, while most of the Democratic contributions from Illinois

appear clustered around Chicago, the Republican contributions extend much further out

57

from the center city. Of course, Republicans also collect quite a bit of money from
more rural areas such as Montana, Wyoming and Idaho. However, the largest
Republican ZIP code contributor is just outside of Washington DC. in Potomac, MD.
The 2000 Congressional Election

In order to determine why incumbents receive contributions from different areas, I
need to measure if the contribution came from inside the incumbent’s district, inside the
incumbent’s state, or came from out of state. These measures are easily created from a
dataset similar to the one used to create Figures 1 and 2, but for the 1999-2000 election
cycle. Data from this cycle will be free from any contamination as a result of
redistricting for the 2002 election. As such, I created three variables; the first is the
percent of money from areas inside each member’s district, the second is the percent
from outside the district but in the state and the final is the percent from out of state.
According to Table 1, Republicans garner 14 percent more of their contributions from

inside the district compared to Democrats (significantly different at 01 < .05 one-tailed,
t333 = 6.38). This is not too surprising since Democrats traditionally are elected in
districts with lower incomes.60 In contrast, Democratic incumbents get more money
outside of their district than Republicans. They earn 4 percent more from in the state
and 10 percent more outside it (both significantly different at a < .05 one-tailed; t333 = -
1.95 and -5.02 respectively). Since contributors will presumably vote for the candidate

that they give to, the fact that Republicans have an additional advantage over

Democrats in in-district contributions might give them an advantage on Election Day.

 

(’0 This is also an empirical fact as the median household income for districts held by
Democrats is about $2600 less than that of Republicans.

58

Measures that may influence a member’s fundraising include factors that are specific
to the incumbent, the particular election and the district. Factors that are specific to the
incumbent include measures of their behavior, their seniority, the committees they
serve on and their party affiliation. Most of these variables are relatively standard in
the congressional literature. In order to test if more extreme members are receiving
contributions, I use the absolute value of the incumbent ’s w-nominate score where
members become more extreme as this variable increases.6| If extreme members are
receiving more contributions from any of the three geographic regions, then the sign on
the coefficient for this variable should be positive. A member’s Seniority is measured
as the number of terms that an incumbent has served. I expect that the more senior the
member, the larger their contribution network will be. Therefore, as seniority increases,
they should get more of their contributions from outside of the district.

Another member factor that may influence incumbents’ fundraising abilities is their
committee assignments. Members on “constituent committees” should do a better job
at providing particularized benefits to their constituents so individuals should be more
likely to give to their own representative. Representatives serving on a “prestige
committees” deliver goods that are not as geographically concentrated. Therefore, they
should earn more money from beyond the district’s borders. If a member served on a
constituency committee during the 107th Congress, then the dichotomous variable

Constituency Committee is coded one, zero otherwise. Prestige Committee is also

 

6’ Scores are available from Keith Poole at http://www.voteview.com. See Poole and
Rosenthal 1997 for more information.

59

coded one if the member served on a prestige committee.62 If either of these variables
contributes to a member’s ability to raise funds from a particular area, then they should
be positive and significant. Since Table 1 suggests that there is a difference between
the parties in terms of in vs. out of district fundraising, I also include a variable Party
that is equal to one if the member is a Democrat, zero otherwise.

Factors particular to the specific election may also have an effect on fundraising.
Members who are in a costly race must by definition raise substantial sums of money.
This is also a signal that the race is competitive. Therefore, I created a variable Total
Spending, that is the sum of the incumbent and challenger’s spending, divided by
10,000. Since there is relatively less money available in a member’s district compared
to the rest of the donor pool, I expect this variable to be negative for fundraising in the
district and positive outside the district. If a member is facing a quality challenger, a
cue that the incumbent is vulnerable (Jacobson 2004), then they may be more likely to
raise funds outside of the district. To test this, I include a variable, Quality Challenger
that is coded one if the incumbent is facing a challenger that has previously held
elective office.63

There are factors related to the member’s district that may influence where a
candidate raises funds. If a member is from a relatively wealthy district, then they
should be less likely to look outside the district for money. To control for this, I

included District Median Income which is the median household income for the

 

62 Following the guidance of Smith and Deering 1990 and Alvarez and Saving 1997 I
coded Appropriations, Budget, Rules and Ways and Means as prestige committees and
Agriculture, Armed Services, Resources, Transportation and Infrastructure, Science,
Small Business and Veterans’ Affairs as constituency committees. Committee and
seniority data are from the 2002 and 2004 issues of CQ’s Politics in America.

63 Spending and quality challenger data were provided by Gary Jacobson.

60

member’s district, again divided by 10,000”4 Members who do not match up well with
their district politically may also need to look elsewhere for funds to bolster their
reelection changes. This possibility is tested with a variable that measures the two-
party share of the District Presidential Vote for the incumbent’s party from the 2000
presidential election as reported by CQ ’5 Politics in America. Finally, I include a
variable that measures the number of districts in a member’s state. Since previous
research (Grenzke 1988) shows that most money comes from inside the member’s state,
then representatives who live in more populous states have a greater number of
potential contributors. Thus, if a member is in a large state, they should get more of
their contributions from inside that state and less from outside the state or from the
district.
Due to the nature of the data, i.e. compositional, OLS is not the appropriate
estimation technique. Compositional data analysis,
“. . .refers to analysis of non-negative data referring to mutually exclusive and
exhaustive categories, where the data sum to one. The constraints thus imposed
on the data mean that analysis of any one category must take into account the
balance in the other categories: an in-crease in one category must be
compensated by a decrease distributed across all the other categories (Brehm,
Gates and Gomez l998:1).
In mathematical terms, this constitutes a simplex and therefore, observations across
each unit of analysis (an incumbent) are not independent. In order to correctly test my

hypotheses, I employ a technique developed by Katz and King (1997) based on work

 

(’4 Data obtained from the US Census www.census.org.

61

by Aitchison (1986). The first step in the analysis is to select the percent of
contributions from outside the district as the reference category and transforrn the
percent contribution shares from the other two areas (in the district and outside the
district but inside the state) into log ratios with respect to the percentage of
contributions from the base category (Tomz, Wittenberg and King 2003). More
specifically, the formulas for creating the dependent variables are ln(percent in the
district/percent out of state) and ln(percent in state, out of district/percent out of state).
Then, the estimation technique is to run a seemingly unrelated regression model with
the new dependent variables. In terms of interpretation, the Sign of the coefficient is in
comparison to the base category.
Results

The results from the seemingly unrelated regression model are displayed in Table 2.
In the both models more extreme members do not receive a higher level of
contributions from any particular area relative to incumbents who are more moderate.
Thus, there is no evidence that more extreme representatives are collecting more money
from individuals who live outside the district and could be influencing their voting
behavior. This also runs contrary to the results by Grenzke (1988:91) who found that
incumbents with liberal voting records receive a higher proportion of contributions
from outside the district. However, Grenzke did not control for party affiliation and
speculated that conservatives may come from wealthier districts.

A member’s seniority does have a discemable influence on where they earn
contributions. The negative sign on the seniority coefficient in both of the models

means that as members accrue seniority, they cam a greater percentage of their

62

contributions from individual contributors outside of their district and their state. This
is consistent with F enno’s (197 8) notion of career expansion where he argued that
during a course of a member’s career, they work to expand their base of support.
Although more speculative, it may be that as members enter into the protectionist phase
of their career also described by F enno, they have more time to attend fundraisers
outside of the district since they do not have to spend all of their time meeting with
constituents.

Figure 4 provides a substantive interpretation of this result for both Democratic and
Republican incumbents. The solid lines in each of the figures represent the percent of
contributions coming from the district while the dashed lines represent the percent
coming from out of the state. 65 The x-axis measures the number of terms served while
the y-axis is the percent of contributions. Republican incumbents early in their career
have an advantage over Democrats in terms of the percent contributions from inside
their district. First term Republicans collect about 50 percent of their individual
contributions from inside the district compared to about 40 percent for Democrats.
However, they each collect similar amounts from outside their state at similar points in
their career, suggesting Democrats collect a greater percentage of their individual
contributions from outside their district, but inside the state.

Over the course of their career, there are similarities and differences between
Republicans and Democrats. Incumbents in both parties see an increase in

contributions from outside the state and a decrease from contributions from inside the

 

65 CLARIFY (Tomz, Wittenberge and King 2003) was used to obtain the values in the figure. All
continuous variables were set to their mean, constituency was set to one, prestige was set to zero and
quality challenger was also set to zero. A member’s seniority was allowed to vary over the range of the
data.

63

district. However, for Republicans, it is not until their 10th term when the out of state
contributions pass the in district contributions. For Democrats, this event takes place
earlier, around their 7th term. As such, it appears that seniority has a definite
substantive influence on campaign contributions from individuals.

Referring back to Table 2, it appears that a member’s committee assignment seems
to have an effect over where they collect their contributions, but not necessarily in the
expected direction for both types of committees. Representatives on prestige
committees collect more of their funds from inside the district compared to the rest of
their state. This runs contrary to the predictions. It may be that members on prestige
committees such as appropriations are able to selectively target their constituencies and
this is contributing to the result. However, members on constituency committees act as
expected. They bring in more from inside the district, probably because they are less
likely to help individuals from outside the district. Finally, in terms of incumbent or
member factors, Democrats collect less of their money from inside the district
compared to the rest of the member’s state. However, neither party collected
significantly more money from outside the state. This result combined with the null
finding on the extremity variable suggests that liberal Democrats are at least as
answerable to Hollywood liberals as conservative Republicans are to Texas oil tycoons.

When it comes to election characteristics, it appears that as the total cost of the race
increases, incumbents collect less money from inside that district and are forced to look
outside the district into other parts of the state for more funds. However, candidates
who face a quality challenger do not receive a greater percentage of contributions from

any of the three areas.

64

District characteristics also prove to be influential in determining where members
collect their campaign funds. As the underlying partisanship of the district moves away
from the representative, they earn less money from inside the district and more from
contributors in other areas of their state. The partisanship of the district does not seem
to force members to look outside the state though for contributions. Again, this is
evidence against the argument that money is pouring in from outsiders and adversely
affecting democratic representation. Finally, the size of the state does not have the
hypothesized effect. If an incumbent is serving in a large state, they collect a smaller
percentage of their campaign funds from other areas of the state compared to outside
the state.

In sum, member, election and district characteristics are influential in determining
where members of congress collect their contributions from individuals. As
representatives serve longer, they appear to increase their donor networks and collect a
greater percentage of their contributions outside of the state. Democrats, who represent
less affluent districts, are forced to seek out contributions from beyond the boundaries
of the geographic constituency. AS such, is looks as if they make up that deficit by
collecting from inside their state. This is also apparent for representatives who serve in
a district that is a poor match politically. They too receive less from inside the district
and more from their state. Members who are in costly races also earn less of their
funds from inside the district. However, in this situation, members search outside of

their state for more money.

65

Redistricting and Contributing
Although the above analysis explains where members collect contributions from

individual contributors, it rests on an important assumption. The assumption is that the
geographic areas remain constant over time. In actuality, the assumption does not
always hold as every ten years states are forced to redraw district boundaries to
accommodate for changes in population. After the boundaries change, some
contributors may find themselves living in a new district. This event provides a unique
opportunity to test the link between the representative and the contributor. If a
member’s ability to fundraise is not effected by redistricting, then it suggests a weak
link between being elected to represent a contributor and earning their donation.

After each round of redistricting, members of Congress find that some of their old
constituents no longer remain in their districts while new ones take their place. In
F enno’s (1978) language, this represents a change in the geographic constituency.
Figure 3 presents a stylized depiction of how a member’s district can change. The
upper portion of the figure, pre-redistricting, shows two districts, one and two,
separated by the solid vertical line. It also shows four areas, A, B, C and D, separated
by both the solid vertical line and the horizontal dashed line. In this simplified version
of redistricting, each of the areas contains an equal number of constituents. The lower
portion of the figure shows the new districts post-redistricting, now separated by the
solid horizontal line. The table depicts an exhaustive but non-mutually exclusive
typology of areas. For example, area A was part of district one before and after
redistricting so it is type in-in for district one but out-out for district two. Area B,
however, left district one after redistricting and is now new to district two so it is

labeled type in-out for district one and out-in for district two. Area C is out-in for

66

district one and in-out for district two while area D is out-out for district one and in-in
for district two. Since each of the areas contains equal populations, both district one
and district 2 are 50 percent continuous, or conversely, 50 percent new.

This section of the paper seeks to determine what happens to the contribution
patterns from areas that are continuous to a district (in-in), new to the district (out-in)
and areas that no longer remain inside the boundaries (in-out). In their study, F rancia et
al. (2003:105) find a strong link between geography and the contribution. If there is a
solid link between contributing to a representative and living in that person’s district
then I expect that the more a district changes in terms of population, the less an
incumbent will earn from the area that remains within the district (in-in) before and
after redistricting, ceteris paribus, compared to members who represent districts that do
not change.66 Members earn less the more the district changes because they have a
smaller district continuous core compared to a district that does not change. If
members collect less money as a result of district change, then this implies that the
bond between a member and contributor can be broken.

Keeping that in mind, individuals who are in districts that undergo a large amount of
change will have more of their old constituents in other districts and should be able to
continue to collect from those individuals. As such, the more a district changes, the
more a member should collect from areas that have left the district (in-out) compared to
representatives who have few of their previous constituents living outside the district.

Members in large changing districts will also have the opportunity to earn contributions

 

66 There is some evidence that redistricting may not matter. For example, F rancia et al.
(2003:23) point out that 45 percent of donors who gave to a member in 1978 continue
to give to that member in 1996, even though they are now represented by someone else.
Yet, this still leaves 55 percent who did not continue to give.

67

from individuals that are new to the district. Therefore, the more a district changes, the
more a member can collect from areas that are new to the district (out-in)”7
District change data

To test these hypotheses, I use a model similar to the one used in the first set of
regressions with a few modifications. First, as the hypotheses suggest, the dependent
variable is the amount of money collected from each of the three rcdistricted areas for
the 2002 election divided by 10,000. I estimate three separate models, one for each of
the areas.

In order to gauge district change I use a measure, detailed in Crespin (2005) which 1
will briefly describe. First, the district boundaries for the 2000 and 2002 congressional
elections are spatially intersected. Then, these areas are overlaid on top of census tract
boundary files.68 After this step, the census tracts are deemed new (drawn into the
incumbent’s district), removed (drawn into a new district and no longer remaining in
the incumbent’s district) or continuous (remain in the district pre and post-redistricting).
Next, the population of the tracts that are continuous to the incumbent’s district is
divided by the total district population. This number is then subtracted from 100
percent and the result is the percent of the population that is new to a member’s district.
Refening back to Figure 3, the percent new in district one would be the population
from area A divided by the sum of the population from areas A and C. I expect that the
Sign for the coefficient for this variable, % New population, to be negative in the

continuous area regression and positive for the other two, out-in and in-out. Percent

 

(’7 Since the out-out category remains constant before and after redistricting, it is not
included in the analysis.

68 According to the census, tracts are small subdivisions of a county. They are designed
to be relatively homogeneous units and average about 4,000 inhabitants.

68

new population is also interacted with another variable, A Presidential vote to create a
variable that measures both direction and magnitude of change.

Change in presidential vote is the standard measure used to gauge political change in
the district (Ansolabehere, Snyder and Stewart 2000, 2001). This measure is taken
from CQ ’s Politics in America and is the difference between the two-party share of the
vote in 2000 and the share of the vote recalculated to the new district boundaries for the
2002 elections. Al Gore’s vote share is used for Democrats and George W. Bush’s for
Republicans so positive numbers mean the district is moving in the incumbent’s
direction.69 In addition to these two variables and the interaction term, each model has
a variable that measures the median household income for the respective areas. As the
median income increases, I expect the incumbent to earn more contributions.
Redistricting Results

Table 3 displays the results of the ordinary least squares regressions used to test the
influence of redistricting on individual campaign contributions. In terms of the control
variables, members with more seniority collect less from the core of their district and
from constituents that left the district. This is consistent with the earlier results that
found as members serve additional terms, they collect more money from out of the state.
This suggests that members may be trying to insulate themselves from redistricting
during the course of their career. Next, the results indicate that Democratic incumbents
lose money from the core of their district (in-in) and earn fewer contributions from new

constituents compared to Republicans. Similar to the previous model, the more

 

69 For example, if Bush received 63 percent in 2000 which was then calculated to be 61

percent in a new Republican district in 2002, the change in presidential vote would be
61—63=-2.

69

spending in a race, the more money members cam. Finally, in each of the three areas,
as median incomes increase, members are more successful at fundraising.

Because the main variable of interest (percent new population) is interacted, the
marginal effects and standard errors from the table are sufficient for interpretation since
they vary depending on the values of the two interacted terms.70 Figure 4 displays
meaningful marginal effects and standard errors for percent new in the district, however,
the dependent variable is rescaled to actual dollars instead of tens of thousands of
dollars. In the figures, the solid line is the marginal effect of one percent new in the
district over the range of change in presidential vote while the dashed lines represent 95
percent confidence intervals. The Y axis is total dollars and the X axis represents the
change in presidential vote. Recall that negative numbers mean that the district is
moving away from the incumbent. In the first column are the results for continuous
areas, the second column shows results for areas that are new to a district and the final
column represents the marginal effect for areas that were drawn out of the district. As
long as the dashed lines do not bound zero (the dotted line), then percent new in the
district is significant for that particular value of change in presidential vote.

When it comes to continuous areas, the marginal effect for each percent new in the
district is negative and significant until change in presidential vote is equal to positive

six. Substantively, this means that if the district moves away from the incumbent or

 

A

70 For example, if Y = ,60 + Ax + ,BZZ + fl3XZ then the marginal effect,

BY . .
5— = ,6] + fl3Z and the variance used to calculate standard errors is,
X

A2 A A A A

a 37: = var(,Bl ) + 22 var(,B3) + 22 cov(fll [33) (see Bambor, Clark and Golder 2005).

(

70

slightly in their favor, they collect less money from individuals compared to a district
that did not change. If the district moves more in favor of the incumbent, then the
amount of population change has no effect on the amount of funds raised. When
change in the district is politically neutral, members lose $2,034 for every percent of
the population that is new. In terms of dollars then, this means that at the average level
of percent new, 21.4 percent, and no political change, members lose about $43,500.
This corresponds to 35 percent of the total collected from continuous areas and about
13 percent of the total amount of contribution from individuals.

Although members lose money from their core district after a redistricting, they can
still hope to earn contributions from their new constituents and those that have left the
district. However, as the results indicate, they do not take in enough new money to
make up for the money that they lose. For districts that have neutral political change,
members earn $576 from new constituents and $634 from constituents that have left the
district for every one percent new population. At the average amount of district change
then, this amounts to a net loss of about $17,600 or about 5 percent of the total
collected from individuals as a result of redistricting. While this result is not
overwhelming, it is non-trivial since members who are losing voters are the most in
need of additional campaign funds.

Motivations for Giving

The results so far indicate that geography is an important determinant of the
contribution, but the process does not take place in a vacuum. Individuals do not give
simply because they reside in the same district as the member. Many give because they

expect something in exchange for the contribution (Snyder 1990). Some contributors

7l

are looking for tangible rewards and are deemed “investors” by F rancia et al. (2003).
These types of contributors are in search of narrow benefits for their particular industry,
company or even their personal financial situation. However, it may not be correct to
assume that all donors are seeking a monetary reward. The study also found that others,
deemed “intimates”, give for solidary or social motives. Similar to the interest group
literature (e. g. Olsen 1971), these contributors donate for the psychological rewards that
come from the social aspect of giving. Unlike investors, solidary contributors do not
give seeking tangible rewards. Consistent with solidary motivations, this type of
contributor will give when invited to a fundraiser or directly asked by the candidate or a
close friend. Some intimates stated “that friendship with a candidate or a solicitor was
always important in helping them to decide to contribute,” (47). Others acknowledged
that local community ties played a major role in the decision calculus. In other words,
there needs to be a social or local connection between the contributor and the campaign
for an intimate to donate.

These two types of contributors, the study determined, were not evenly distributed
across the two major parties. Democrats gave largely based on solidary reasons while
Republican focused on material gains. Since 75 percent of contributions did not cross
party lines (F rancia et al. 2003: l 05), it is reasonable to assume that Democratic
members get most of their individual contributions from people who give for solidary
reasons while Republicans receive donations based on material rewards. Below, I
extend their theory to incorporate changes as a result of redistricting.

Given the different types of contributors, it may not be clear how redistricting will

influence their behavior. Based on the reasons for giving, it may be the case that

72

redistricting does not influence the types equally. However, it seems reasonable to
assume that since investors, i.e. Republicans, are giving for tangible rewards, they
should be more likely to abandon their previous representative. This should hold
especially if the member in the old district no longer has the ability to exchange
particularized benefits for contributions.71 Solidary contributors, i.e. Democrats, are
likely to give if their social group is connected to a candidate or local party. Since
contributors who live in a different district after redistricting physically reside in the
same location, their social networks are unlikely to have changed. In view of this detail,
Democrats who have been drawn into new districts should be less likely to abandon
their former member of Congress.

To test this hypothesis, I reran the regression on areas that were drawn out of the
district separately for the Democrats and Republicans. These results are presented in
Table 4 and are quite similar across the two parties, except for one key difference. The
sign on the interaction term is negative for Democratic incumbents and positive for
Republicans. Substantively, this means that as a Democrat’s district becomes more
Republican, he collects more money from areas that have left the district. In contrast,
as a Republican’s district becomes more Democratic, she earns less money from those
areas. These results are made more apparent in F igure 5, a figure analogous to Figure 4.
In the first column, the marginal effect of each percent new in the district is positive
and significant for Democrats until the district moves five percentage points in the

Democratic direction. At this point, the effect is no longer significant. Further, the

 

7' While there is no evidence of one to one exchanges between contributors and
representatives, it is likely that a member of Congress will target rewards in their
district versus outside of the district if given a chance.

73

slope of the line is negative indicating that Democrats collect less money the more their
district moves away from them. This implies that Democratic contributors are willing
to help their old representative out if they are in trouble but if the old member is safe,
they will no longer contribute.72 Almost the exact opposite is evident for Republicans.
Republicans will collect more money from their old constituents only if it looks like
they have a good chance of returning to office. As the district moves in favor of the
Republican incumbent they pick up more money from areas that previously were in
their district. However, if the change in presidential vote is negative, then the effect is
not statistically significant. Both of these pieces of evidence are consistent with

F rancia et al.’s theory that Democratic and Republican donors have different
motivations for giving. Since Democrats receive contributions if they are more likely
to lose, then it is unlikely that Democratic contributors are seeking some sort of
tangible reward. On the other hand, Republican incumbents receive more contributions
from their old constituents only if they have a good shot at returning to office. This
means that Republican contributors are more likely to give if they can get something in

return.

Conclusion and Future Research
In this paper, I have shown that representatives’ abilities to earn money from

different geographic regions are dependent on their demand for funds and the available
supply of contributions in each area. Although members of Congress frequently have
to look outside of the district for funds, there was no evidence that more extreme

incumbents were more successful at caming outside donations. When it comes to

 

72 One could also argue that members who are in trouble would be more likely to ask
former constituents for a donation.

74

factors that are specific to each member, more senior incumbents take home a greater
percentage of their individual contributions from outside the state, and less from inside
their home state or district. Democrats also collect more from outside the district, but
largely from inside their state. However, there is no evidence that more extreme
members collect a greater percentage of money from outside their states.

When total spending in a race is high, and members demand for money increases,
they are forced to look outside the district to other states for money. Members also
look outside the district, although within the state, when they are not a good political
match for the district.

When it comes to redistricting, the more a member’s district changes, the less money
they earn from individuals compared to districts that do not change. On average,
representatives lose about $17,600 as a result of redistricting. This amount includes
money lost when the district core shrinks, and money gained from new constituents.
When it comes to holding onto old contributors, there is a difference between the two
parties. Democrats earn money from the former parts of the district if their new district
is more Republican. In contrast, Republicans receive money from their old
contributors when their new district is more favorable politically. This helps to
highlight the differences between contributors who give for solidary or tangible reasons.
In terms of representation, this paper showed that the relationships between a member
and contributor are not static; rather they can change depending on where the
contributor resides.

In future research dealing with individual contributors and geographic areas I hope to

move on to a more micro analysis at the individual donor level. By matching up

75

contributors across elections, I will better be able to test the influence that redistricting
has on constituent behavior. I anticipate that individuals moving from districts that
were in the minority to the majority and vice versa may also change their donation
behavior in different ways. Although this paper has answered some basic questions, it
could not answer all of them and in turn, has raised others. By moving beyond the

aggregate level, I hope to find answers to some of those questions.

76

Table 1 - Percent of Money to Incumbent Candidates, 1999-2000 Election Cycle

 

 

Party In District Out of District/ Out of State
In State

Democrat .30 .27 .43

Republican .44 .23 .33

 

77

Table 2 — Percent Contribution from Geographic Areas

 

Seemingly Unrelated Regression

 

 

Coefficient
(Standard Error)
Variable In the District Out District of
in State
Member
Absolute Value W-NOMINATE .082 .125
(373) (340)
Seniority -. 172* -. 124*
(.016) (.014)
Prestige Committee .325* .035
(142) (129)
Constituency Committee .351* .213
(139) (126)
Party (I=Dem.) -.477* .237
(.155) (.141)
Election
Total Spending -.002* -.0006
(.0006) (.0005)
Quality Challenger -.081 -.086
(.160) (.146)
District
District Median Income .070 .046
(.055) (.050)
District Presidential Vote -.015* .001
(.006) (.006)
Number of Districts .0007 .024*
(.004) (.004)
Constant 1.92* .136
(.480) (.436)
388 388
x2 185.31 * 134.77*
R2 .323 .258

 

Note - Dependent variable for “In the District” is the log-ratio of the percent of in-

district contributions with respect to the percent of out-of-state contributions.

Dependent variable for “Out of District in State” is the log-ratio of the percent of
out-of-district-in-state contributions with respect to the percent of out-of-state

contributions.

78

Table 3 — Contributions from Areas that are Continuous, New or Left the District

 

 

OLS Coefficient
(Standard Error)
Variable Continuous New to the Left the
(In-In) District District
(Out-In) (In-Out)
Member
Absolute Value W-NOMINATE -6.66 -1.66 .589
(4.24) (.952) (.926)
Seniority -.692* 0.040 -. 147*
(.134) (.030) (.036)
Prestige Committee 1.92 .349 .191
(1.42) (.292) (.425)
Constituency Committee -1 .18 .501 -.092
(1.34) (.303) (.393)
Party (1=Dem.) -6.74* -.707* .221
(1.47) (.303) (.411)
Election
Total Spending .059* .008* .007*
(.014) (.002) (.003)
Quality Challenger 1.64 .068 -.073
(2.63) (.437) (.562)
District
Median Income 1.79*
Continuous Areas (.602) ___ ---
Median Income .299*
New Areas --- (.091) ---
Median Income .334*
Old Areas --- --- (.084)
District Presidential Vote -.072 .026 -.009
(.057) (.016) (.017)
% New Population -.203* .058* .063*
(.036) (.011) (.011)
A Presidential Vote -.757* .114 .112
(.335) (.076) (.094)
% New Population x .020* -.002 -0.0004
A Presidential Vote (.009) (.002) (.002)
Number of Districts -.041 -.008 -.001
(.028) (.008) (.008)
Constant 1835* -2.36* -l.28
(4.37) (1.03) (1.16)
N 384 283 284
F 11.05* 8.78* 7.74*
Adj. R2 .42 .33 .33

 

Dependent Variable — Amount of Contributions divided by 10,000.

*p<.05

79

Table 4 — Contributions from Areas Drawn Out of the District (In-OuQ by Party

 

 

OLS coefficient
(standard error)
Variable Democrat Republican
Incumbent Incumbent
Law
Absolute Value W-NOMINATE .273 .970
(1.13) (1.74)
Seniority -.156* -.136*
(.058) (.039)
Prestige Committee .365 -.289
(.700) (.340)
Constituency Committee .475 -.278
(.726) (.331)
Election
Total Spending .020* .003*
(.006) (.001)
Quality Challenger .349 -.188
(1.05) (.367)
District
Median Income .369* .319*
Old Areas (.136) (.110)
District Presidential Vote .013 -.0001
(.025) (.025)
% New Population .067* .048*
(.022) (.008)
A Presidential Vote .189 .023
(.137) (.118)
% New Population X -.004 .004
A Presidential Vote (.003) (.003)
Number of Districts -.012 -.003
(.013) (.008)
Constant -3.97 -1 . 12
(2.02) (1.54)
N 122 162
F 3.83* 10.47*
Adj. R2 .41 .37

 

Dependent Variable — Amount of Contributions divided by 10,000. All income and
spending variables divided by 10,000

*p<.05

80

 

 

Figure l - Individual Congressional Campaign Contributions, 2001-02
Percent by party per ZCTA* - Incumbents only

 

  

Percent Democratic Contribution

0% (N = 6,241)
- 1} >019% (N =1,429)
123,? 203W. (N = 1.353)
M 40-59% (N =1.228)
- (50-79% (N = 1,223)
- 80—99% (N = 1,242)
’ZCT A = Zip Code Tabulation Area

- 100% (N = 3,752)

 

 

81

 

 

 

Figure 2 - Total Individual Congressional Campaign Contributions, 2001—02

Contributions by party per ZCTA” — Incumbents only

 

 

 

 

Total Democratic Contributions

[:1 No Contribution

' : si - 1.000 (N - 4.043)

11.001- 5.000 (N = 3.609)

m $5.001 - 10,000 (N = 1.107)

- $10,001 - 100,000 (N - 1.414)

- 3100.001 - Max ($526,040) Mannoilon. NY (N = 54)

 

 
 

 

 

 

 

Total Republican Contributions
1:] No Contribution

” $1-1_000(N= 4.954)

$3.37,! $1.001 - 5.000 (N = 4.569)

- 55.001 - 10.000 (N = 1.478)

- $10,001 - 100000 (N = 1,652)
- 3100.001 - Max (3439.260) Potomac. MD (N = 63)

 
  

'ZCTA - Zip Code Tabulation Are-

 

82

 

 

Figure 3 — Area Typology

 

Pre-Redistricting

 

 

 

 

. .h.. 1""; j
greener. l

  

 

 

 

 

 

Post-Redistricting

(A Districtclt. c

 

 

’B , District2: "D _.

 

 

 

 

District
One

—|
fb
E

Type
ln-In
ln-Out
Out-In
Out-Out

 

 

Out-Out
Out-In
ln-Out
ln-In

Two

UGW>UOW>>

 

83

 

% of Total Contributions

Figure 4 - Effect of Seniority On Contribution Area

 

Republican Democrat

 

 

 

 

 

 

 

I r
0 5 10 15 20 O 5 10 15 20
Number of Terms Served

 

—— % in District -'-"'"'" °/o out of State

 

 

 

Graphs by party

84

Dollars

Figure 5 - Effect of Percent New in the District On Contributions
Dependent Variable - Contributions from ZCT As

 

Continuous in District

New to District

Drawn out of District

 

0 5000 l 0000
I I I

-5000

 

 

-10000
I

 

 

 

 

 

 

If I I I T

I I j
~20-15-10-5 0 5 101520

I I I I

I I I I T

T T r 1 I I I r I
-20-15-10 -5 0 5 101520 -20-15-10-5 0 5 101520

Change in Presidential Vote

 

 

 

Marginal Effect of Each Percent New in the District
----- 95% Confidence Interval

 

 

ZCTAs - Zip Code Tabulation Areas

85

 

Figure 6 - Effect of Percent New in the District On Contributions by Party
Dependent Variable - Contributions from Old ZCTAs

 

 

 

 

 

 

 

 

 

Democrat Republican

8_
V)
N
O
O-
2
3%-
5 %4
O
O
2-
O
O
m-

‘7‘ I I fl I T I I I I I I I I I I I I T

-20 -lS -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 IO 15 2

Change in Presidential Vote

 

 

Marginal Effect of Each Percent New in the District
----- 95% Confidence Interval

ZCTAs - Zip Code Tabulation Areas

 

 

 

86

Chapter 5 - Conclusion
Pitfalls and Practicalities

Throughout this dissertation, I have largely ignored two potential threats to validity
that plague all social science research — endogeneity and spatial-autocorrelation. In this
section, 1 will first describe the two terms and then discuss how they may threaten my
results.

Spatial autocorrelation stems from Tobler’s (1970:236) first law of geography, that is,
“everything is related to everything else, but near things are more related than distant
things.” This is clearly true in the social sciences as all one has to do is imagine the red
and blue map of county election returns. Red counties dominate most of the space in
the map and blue counties are clustered near each other in cities.

Spatial autocorrelation is defined as the “correlation among values of a single
variable strictly attributable to the proximity of those values in geographic space,
introducing a deviation from the independent observations assumption in classical
statistics,” (Griffith 200313). The most common form of spatial autocorrelation is
positive spatial autocorrelation. This form is present when high values are proximate to
high values and low values are proximate to low values. For example, wealthy city
blocks tend to be near other wealthy city blocks. The less common type of this
phenomenon is negative spatial autocorrelation and is present when high values are
next to low values and medium values are next to other medium values.

Essentially, the problem of spatial autocorrelation is .one of interdependence of

observations. When data are spatially autocorrelated, the value of one observation can

87

be predicted based on adjacent observations. Spatial autocorrelation violates the
assumption of independence of observations which is a concern for hypothesis testing.
If the observations are spatially clustered in some way, the estimates obtained from the
OLS estimator will be biased and overly precise due to the lack of independence of
observations.

One way to test for spatial autocorrelation is an indicator called Moran’s I. This test
compares the value of the variable at any one location with the value at all other
locations (Moran 1950). The statistic is similar to a correlation coefficient and varies
between negative one and positive one. Positive numbers correspond to a high degree
of positive autocorrelation and negative numbers indicate negative autocorrelation.

In this dissertation, since the unit of the analysis is measured at the district (or
member) level, it could be the case that measures from one district are not independent
from neighboring districts. For example, are districts that undergo significant change
more likely to reside near each other? Alternatively, districts that undergo small
amounts of change may also me spatially close. If this is indeed the case, then the
results from the regression analyses may be biased.

Unfortunately, testing for spatial autocorrelation using the data in this dissertation is
not altogether straightforward. The district boundary files as created by the census
bureau do not contain one shape for each congressional district. Many districts are
made up from several smaller polygons and there is an average of two polygons for
each district. This does not pose a problem for spatial intersection but does for testing
for spatial autocorrelation. For example, if district A was made up of three separate

polygons and each of those polygons contained the exact same value on some measure,

88

then a test for autocorrelation is likely to report that it is present. However, it may
prove to be a false positive since there are multiple measures for each unit. In order to
correctly test for autocorrelation then the polygons for each district need to be merged
in some fashion so that there is only one polygon for each district. This task proved
quite difficult for this user within the Arc/Info software package and l was not able to
successfully test for spatial autocorrelation.

Since I was unable to test for spatial autocorrelation, then the reader must exercise
caution in terms of making causal inference. However, and perhaps more disconcerting,
is the fact that almost all the research within political science ignores this potential
problem.73 In future research, I hope to better directly account for the effects of spatial
autocorrelation.

Another potential threat to validity from this and all other research is the issue of
endogeneity. More specifically, endogeneity exists when an “explanatory variable in a
multiple regression model is correlated with the error term, either because of an omitted
variable, measurement error, or simultaneity,” (Wooldridge 20032835). If endogeneity
is a problem, then the results are biased and possible inconsistent. One way to control
for endogeneity is by using instrumental variables and/or two stage least squares.

In terms of the analyses presented here, endogeneity may be a problem since district
boundaries are not drawn in a vacuum. That is, there is a great deal of strategy behind
the drawing of district lines. When Republicans are in charge of drawing district

boundaries, they try to draw seats that can elect more Republicans and if possible, hurt

 

73 For an exception, see Cho 2003.

89

Democratic incumbents.74 The opposite hold if Democrats are drawing the lines.
Alternatively, when there is divided government, the result is generally an incumbent
protection plan. A final group of line drawers are commissions and courts. When they
are in charge of redistricting, districts tend to be more competitive compared to when
political parties are involved (Carson and Crespin 2004).

In future research, I hope to not only control for this potential problem, but exploit it.
For example, it would be interesting to see if members are more or less responsive to
their district depending on who draws the boundaries. If commissions or courts
increase competition, but decrease responsiveness, then perhaps policy makers may
want to rethink this type of map making. However, if members who represent
commission drawn districts do a better job of representing their districts, then this is
another argument that one could use to change districting methods. As such, I hope to

test for this in other research projects.

Concluding Remarks

In this dissertation, I have presented three essays that investigate the link between
members of Congress and their constituents. 1 studied representation in each part of the
party trifecta, in-the-electorate, in-govemment and in-organization (Key 1964, Beck
and Sorauf 1991). In each essay, I used a new continuous measure of district change
that is an improvement over the previous, dichotomous measure. In the first essay, the
results suggested that as a district changes, a member’s margin of victory varies. As

more people of the incumbent’s party enter the district, their margin of victory

 

74 See Butler and Cain 1992 for a discussion of districting strategy.

90

increases. In contrast, if the district moves in the other direction, members do worse in
the first election after redistricting. However, if the political change was neutral, but
many new constituents entered the district, members saw a decline in their margins
because the new constituents did not have time to form a representational bond with the
incumbent.

In the second essay, I tracked changes in member behavior before and after
redistricting. As members picked up different types of voters, they changed their
voting behavior accordingly. However, members only changed their voting patterns in
response to changes in the district on votes that were visible to their coefficients,
namely final passage votes. Alternatively, any changes in a member’s voting behavior
on procedural votes — votes that the district does not understand — were not related to
changes in the district. Instead, members moved to the left or right as their party
median changes. Contrary to previous research, these results demonstrate that after
redistricting, members of Congress do not adapt all of their behavior to fit their new
district.

In the third and final essay, I provide evidence that members are able to earn
campaign contributions from individuals from different geographic areas depending on
the amount of money they demand, and the supply of money from each area. Then, I
demonstrate that when districts change, members earn less as their core shrinks. They
can earn some, but not all, of the money back from their new constituents, and those
that have left the district. This means that the bonds of representation can be broken

after redistricting, but can be formed with new constituents. In sum, I have shown that

91

redistricting is able to influence representation in all aspects of the political process

when we look in the right places.

92

Appendix
Final Passage Votes

Passage of a Bill

Final Passage of Conference Report

Final Passage of Joint Resolution

Passage over Presidential Veto

Suspension of Rules for a Bill

Suspension of Rules for a Joint Resolution
Suspension of Rules for Conference Report
Motion to Suspend the Rules and Concur
Amendments to the Constitution

Final Passage of Resolution

Adoption of Concurrent Resolution
Suspension of Rules for Concurrent Resolution
Suspension of Rules for a Resolution
Adoption of First Part of Resolution
Adoption of Second Part of Resolution

Partisan Procedural

Budget Waivers

Motion to End Debate

Motion to Rise from the Committee of the Whole
Motion to Disagree

Passage of Rules (Special Rule)
Motion to Recede

Motion to Order Previous Question
Election of Speaker

Motion to Recommit

Motion to Instruct Conferees
Motion to Recede and Concur
Previous Question on Special Rules

93

REFERENCES

Abramowitz, Alan 1. 1983. “Partisan Redistricting and the 1982 Congressional
Elections.” The Journal of Politics 45: 767-770.

Aitchison, J. 1986. The Statistical Analysis of Compositional Data. New York: John
Wiley.

Aldrich, John H., and David W. Rohde. 1998. “The Transition to Republican Rule in the
House: Implications for Theories of Congressional Politics.” Political Science
Quarterly 112: 541-67.

Aldrich, John H., and David W. Rohde. 2000. “The Consequences of Party Organization
in the House: The Role of the Majority and Minority Parties in Conditional Party
Government.” In Jon R. Bond and Richard F leisher, eds. Polarized Politics:
Congress and the Presidential in a Partisan Era. Washington: CQ Press.

Aldrich, John H., and David W. Rohde. 2001 "The Logic of Conditional Party
Government: Revisiting the Electoral Connection." In Lawrence Dodd and Bruce
Oppenheimer, eds., Congress Reconsidered, 7th ed.

Alvarez, R. Michael and Jason L. Saving. 1997. “Congressional Committees and the
Political Economy of Federal Outlays,” Public Choice 92:55-73.

Ansolabehere, Stephen, Alan Gerber and James M. Snyder. 2002. “Equal Votes, Equal
Money: Court-Ordered Redistricting and the Distribution of Public Expenditures
in the American States.” American Political Science Review 96:779-94.

Ansolabehere, Stephen, James M. Snyder, Jr., and Charles Stewart, Ill. 2000. “Old
Voters, New Voters, and the Personal Vote: Using Redistricting to Measure the
Incumbency Advantage” American Journal of Political Science, 44: 17-34.

Ansolabehere, Stephen, James M. Snyder, Jr., and Charles Stewart III. 2001. "Candidate

Positioning in U.S. House Elections." American Journal of Political Science 45:
136-159.

Aranson, Peter H. Peter C. Ordeshook. 1972. “Spatial Strategies for Sequential
Elections.” In Probability Models of Collective Decision Making, ed. Richard G.
Niemi, Herbert F. Weisberg. Columbus, Ohio: Charles E. Merrill.

Arnold, R. Douglas. 1990. The Logic of Congressional Action. New Haven: Yale
University Press.

94

Baker v. Carr. 1962. 369 U.S. 186.

Bailey, Michael and Kelly H. Chang. 2001. “Comparing Presidents, Senators, and
Justices: Interinstitutional Preference Estimation,” Journal of Law, Economics
and Organization 17(2): 477-506.

Basehart, Harry and John Comer. I991. “Partisan and Incumbent Effects in State
Legislative Redistricting.” Legislative Studies Quarterly 16:65-79.

Beck, Paul Allen, and Frank J. Sorauf. 1991. Party Politics in America. 7th ed. New
York: HarperCollins

Bickers, Kenneth, and Matthew Potoski. 2000. “District Lines and Battle Lines:
Redistricting and the Politics of Distributive Programs Awards.” Presented at the
Annual Meeting of the Midwest Political Science Association, Chicago.

Bishin, Benjamin G. 2000. “Constituency Influence in Congress: Do
Subconstituencies Matter?” Legislative Studies Quarterly. 25: 389-413

Black, Duncan. 1958. The Theory of Committees and Elections. Cambridge:
Cambridge University Press.

Boatright, Robert G. 2004. “Static Ambition in a Changing World: Legislators
Preparations for, and Responses, to, Redistricting,” State Politics and Policy
4:436-454.

Boldstad, Paul. 2002. GIS Fundamentals. White Bear Lake, Minnesota: Elder
Press.Bullock, Charles S. 1975. “Redistricting and Congressional Stability,
1962-1972.” The Journal of Politics 37: 569-575.

Born, Richard. 1985. “Partisan Intentions and Election Day Realities in the

Congressional Redistricting Process.” The American Political Science Review 79:
305-319.

Bovitz, Gregory and Jamie Carson. 2000. “The Electoral Consequences of Position-
Taking in Congress: Exploring the Relationship Between Roll Call Behavior and
House Election Results.” Paper presented at the Annual Meeting of the American
Political Science Association, Washington, DC.

Box-Steffensmeier, Janet M. and Peter M. Radcliffe. 2001. "The Timing of Campaign
Contributions." Typescript. Ohio State University.

Brady, David W., John F. Cogan, Brian Gaines, and R. Douglas Rivers. 1996. “The

Perils of Presidential Support: How the Republicans Captured the House.”
Political Behavior 18:345-68.

95

Brambor, Thomas, William Clark and Matt Golder. 2005 "Understanding Interaction
Models: Improving Empirical Analyses" Forthcoming in Political Analysis 13.

Brehm, John., Scott Gates and Brad Gomez. 1998. “A Monte Carlo Comparison of
Methods for Compositional Data Analysis,” Paper prepared for presentation at
the annual meeting of the Society for Political Methodology, San Diego, CA.

Bullock, Charles S. 1975. "Redistricting and Congressional Stability, 1962-1972." The
Journal of Politics 37: 569-575.

Burke, Edmund. Select Works of Edmund Burke, and Miscellaneous Writings. Library of
Economics and Liberty. Retrieved April 21, 2005 from the World Wide Web:
http://www.econlib.org/library/LFBooks/Burke/brkSWv4c1 .html

Bumham, Walter Dean. 1974. “Theory and Voting Research: Some Reflections on
Converse’s ‘Change in the American Electorate’” The American Political
Science Review 68: 1002- I 023.

Butler, David and Bruce Cain. 1992. Congressional Redistricting: Comparative and
Theoretical Perspectives. New York: Macmillan Publishing Company.

Cain, Bruce E. 1985. “Assessing the Partisan Effects of Redistricting.” The American
Political Science Review 79: 320-333.

Cain, Bruce E. and Janet Campagna. “Predicting Partisan Redistricting Disputes.”
Legislative Studies Quarterly 12:265-274.

Calvert, Randall. 1985. "Robustness of the Multidimensional Voting Model: Candidate
Motivations, Uncertainty, and Convergence." American Journal of Political
Science vol. 29, pp. 69-95.

Campbell, Donald T. and Julian C. Stanley. 1963. Experimental and Quasi-
Experimental Design for Research, Boston: Houghton Mifflin Company.

Cancs-Wrone, Brandice, David W. Brady and John F. Cogan. 2002. “Out of Step, Out

of Office: Electoral Accountability and House Members’ Voting,” American
Political Science Review, 96: 127-140.

Canon, David T. 1999. Race, Redistricting, and Representation: The Unintended
Consequences of Black Majority Districts. Chicago: University of Chicago Press.

Carson, Jamie L. 2005. “Strategy, Selection, and Candidate Competition in U.S. House
and Senate Elections.” The Journal of Politics. 67: 1-28.

96

Carson, Jamie L. and Michael H. Crespin. 2004. “The Effect of State Redistricting
Methods on Electoral Competition in United States House Races,” State Politics
and Policy Quarterly 4:455-469.

Carson, Jamie L., Eric Engstrom and Jason Roberts. 2004. “Congressional Redistricting
and Candidate Entry Decisions in the 19th Century” Midwestern Political
Science Association, Chicago, IL.

Carson, Jamie L., Michael H. Crespin, Charles J. F inocchiaro and David W. Rohde.
2004. "Linking Congressional Districts Across Time: Redistricting and Party
Polarization in Congress." Paper presented at the annual Meeting of the
American Political Science Association, Chicago, IL.

Cigler, Allan J. and Burdett A. Loomis. 1991. Interest Group Politics. Washington
DC: CQ Press.

Cho, Wendy Tam. 2003. “Contagion Effects and Ethnic Contribution Networks,”
American Journal of Political Science 47(2): 368-3 87

Clausen, Aage. 1973. How Congressmen Decide: A Policy Focus. New York: St.
Martin’s Press.

Congressional Quarterly. 2001. Politics In America 2002: The 107th Congress
Washington DC: Congressional Quarterly.

Congressional Quarterly. 2003. Politics In America: 2004 The 108th Congress.
Washington DC: Congressional Quarterly Press.

Cox, Gary W. and Jonathan N. Katz. 2002. Elbridge Gerry ’s Salamander: The
Electoral Consequences of the Reapportionment Revolution. New York:
Cambridge University Press.

Cox, Gary W., and Mathew D. McCubbins. 1993. Legislative Leviathan: Party
Government in the House. Berkeley: University of California Press.

Cox, Gary W., and Mathew D. McCubbins. 2002. “Agenda Power in the U.S. House of
Representatives, 1877-1986.” In David W. Brady and Mathew D. McCubbins,
eds. Party, Process, and Political Change in Congress: New Perspectives on the
History of Congress. Stanford: Stanford University Press.

Cox, Gary W., and Mathew D. McCubbins, 2004. Setting the Agenda: Responsible
Party Government in the U.S. House of Representatives. New York: Cambridge
University Press.

Crespin, Michael H. 2005a. “Using Geographic Information Systems to Measure
District Change, 2000-02,” Political Analysis 13:253-260.

97

Crespin, Michael H. 2005b. “Out with the Old, In with the New: Redistricting and
Individual Campaign Contributions to Congressional Candidates,” Paper

presented at the Annual meeting of the Southern Political Science Association,
New Orleans, LA.

Crespin, Michael H., David W. Rohde, and Ryan J. Vander Wielen. 2002. “Variations in
Party Voting in the House of Representatives, 1953-2000.” Paper presented at the
Annual Meeting of the Southern Political Science Association.

Davis, Bruce E. 2001. GIS: A Visual Approach. United States: Onward Press.

Delli Carpini, Michael X. and Scott Keeter. 1996. What Americans Know about
Politics and Why it Matters. New Haven: Yale University Press.

Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and
Row.

Engstrom, Erik. 2003. “Stacking the States, Stacking the House: The Partisan
Consequences of Congressional Redistricting, 1870-1900.” Paper presented at
the Annual Meeting of the American Political Science Association.

Epstein, David and Sharyn O’Halloran. 1999. “A Social Science Approach to Race,
Redistricting, and Representation.” The American Political Science Review 93:
187-191.

Erikson, Robert S. 1971. “The Electoral Impact of Congressional Roll Call Voting.”
American Political Science Review 65: 1018—32.

Erikson, Robert S. 1972. “Malapportionment, Gerrymandering, and Party Fortunes.”
The American Political Science Review 66: 1234-1245.

Erikson, Robert S. and Gerald C. Wright. 2005. “Voters, Candidates, and Issues in
Congressional Elections.” In Lawrence C. Dodd and Bruce I. Oppenheimer eds.

Congress Reconsidered 8th ed.. Washington DC: CQ Press.

F enno, Richard F. Jr. 1977. “U.S. House Members in Their Constituencies: An
Exploration,” The American Political Science Review, 71: 883-917.

Fenno, Richard F. 1978. Homestyle: House Members and their Districts, Boston: Little
Brown.

Ferejohn, John. 1977. “On the Decline of Competition in Congressional Elections.”
The American Political Science Review 71 :166-176.

98

F inocchiaro, Charles J. and David W. Rohde. 2002. "War for the Floor: Agenda Control
and the Relationship Between Conditional Party Government and Cartel
Theory." Presented at the 2002 meeting of the Midwest Political Science
Association.

F iorina, Morris P. 1974. Representatives, Roll Calls, and Constituencies. Lexington,
MA: D.C. Heath.

F iorina, Morris P. 1999. "Whatever Happened to the Median Voter?" A paper delivered
at the Annual Meeting of the Midwest Political Science Association.

F rancia, Peter L., John C. Green, Paul S. Herrnson, Lynda W. Powell and Clyde Wilcox.
2003. The F inanciers of Congressional Elections New York: Columbia
University Press.

F roman, Lewis A. 1963. Congressman and Their Constituencies. Chicago: Rand
McNally.

Gelman, Andrew and Gary King. 1994a. “A Unified Method of Evaluating Electoral
Systems and Redistricting Plans.” American Journal of Political Science 38:
514-554.

Gelman, Andrew and Gary King. 1994b. “Enhancing Democracy Through Legislative
Redistricting.” The American Political Science Review 88:541-559.

Glazer, Amiahi, and Marc Robbins. 1983. “Voters and Roll Call Voting: The Effect of
Congressional Elections.” Political Behavior 5:377-90

Glazer, Amiahi and Marc Robbins. 1985. “Congressional Responsiveness to
Constituency Change,” American Journal of Political Science, 29: 259-73.

Green, Donald Philip, and Jonathan S. Krasno. 1988. "Salvation for the Spendthrifi
Incumbent: Reestimating the Effects of Campaign Spending in House Elections."
American Journal of Political Science 32: 884-907.

Green, Donald Philip, and Jonathan S. Krasno. 1990. "Rebuttal to Jacobson’s New
Evidence for Old Arguments." American Journal of Political Science 34: 363-
372.

Grcnzke, Janet. 1988. “Comparing Contributions to U.S. House Members from Outside
Their Districts,” Legislative Studies Quarterly 13283-103.

Griffith, Daniel A. 2003. Spatial Autocorrelation and Spatial Filtering. Springer: New
York.

99

Grose, Christian. 2000. “The Effect of Court-Ordered Redistricting on Pork Project
Allocation in Congress.” Paper presented at the Annual Meeting of the Midwest
Political Science Association.

Hall, Richard L. and Frank W. Wayman. 1990. “Buying Time: Moneyed Interests and
the Mobilization of Bias in Congressional Committees,” The American Political
Science Review, 84:797-820.

Herrnson, Paul S. 1998. Congressional Elections. 2"‘1 ed. Washington DC: CQ Press

Herrnson, Paul S., Ronald G. Shaiko and Clyde Wilcox. 1998. The Interest Group
Connection, New Jersey: Chatham House.

Hetherington, Marc J ., Bruce A. Larson and Suzanne Globetti. 2003. “The Redistricting

Cycle and Strategic Candidate Decisions in U.S. House Races” Journal of
Politics 65:1221-1234.

Hirsch, Sam. 2003. “The United States House of Unrepresentatives: What Went Wrong

in the Latest Round of Congressional Redistricting.” Election Law Journal
2: 179-216.

Hotelling, Harold. 1929. "Stability in Competition," Economic Journal 39: 41-57.

Jacobson, Gary C. 1980. Money in Congressional Elections. New Haven: Yale
University Press.

Jacobson, Gary C. 1990. “The Effects of Campaign Spending in House Elections: New
Evidence for Old Arguments,” American Journal of Political Science, 34: 334-
362.Jacobson,

Jacobson, Gary C. 1996. “The 1994 House Elections in Perspective.” Political Science
Quarterly 111: 203-23.

Jacobson, Gary C. 2004. The Politics of Congressional Elections. New York: Addison
Wesley Longman.

Jacobson Gary C. and Michael A. Dimock. 1994. “Checking Out: The Effects of Bank
Overdrafts on the 1992 House Elections” American Journal of Political Science
38:601-624.

Jacoby, William G. 1991. Data Theory and Dimensional Analysis. Newbury Park, CA.:
Sage Publications.

Johannes, John R., and John C. McAdams. 1981. “The Congressional Incumbency
Effect: Is It Casework, Policy Compatibility, or Something Else: An Examination
of the 1978 Election.” American Journal of Political Science 25:512-42.

100

Jones, Ruth S. and Warren E. Miller. 1985. “Financing Campaigns: Macro Level
Innovation and Micor Level Response,” Western Political Quarterly 38:187-
210.

Katz, Jonathan and Gary King. 1997. “A Statistical Model of Multiparty Electoral Data,"
Paper presented at the annual meetings of the Midwest Political Science
Association, Chicago, IL.

Key, V0. 1964. Politics, Parties, and Pressure Groups. 5th ed. New York City:
Crowell

King, Gary, Michael Tomz, and Jason Wittenberg 2000. “Making the Most of Statistical
Analyses: Improving Interpretation and Presentation.” American Journal of
Political Science 44: 347-61.

Kingdon, John W. 1981. Congressmen ’s Voting Decisions. New York: Harper and
Row.

Krehbiel, Keith. 1992. Information and Legislative Organization. Ann Arbor:
University of Michigan Press.

Krehbiel, Keith. 1993. “Where’s the Party?” British Journal of Political Science 23:
235-66.

Krehbiel, Keith. 1999. “Paradoxes of Parties in Congress.” Legislative Studies Quarterly
24: 31-64.

Krehbiel, Keith. 2000. “Party Discipline and Measures of Partisanship.” American
Journal of Political Science 44: 212-27.

Lewis, Jeffrey B. and Gary King. 1999. “No Evidence on Directional vs. Proximity
Voting,” Political Analysis 8:21-33.

Lioz, Adam. 2003. “The Role of Money in the 2002 Congressional Elections,” US
PIR G Education Fund.

Lublin, David. 1997. The Paradox of Representation: Racial Gerrymandering and
Minority Interests in Congress. Princeton: Princeton University Press.

Lublin, David and D. Stephen Voss. 1997. “Racial Redistricting and Realignment in
Southern State Legislatures.” American Journal of Political Science 44: 792-810.

Lyons, Michael. 2003. “Win, Lose or Draw: Reapportionment, Redistricting Strategy,

and Electoral Outcomes for the U.S. House Elections, 1972-2002.” Presented at
the Annual Meeting of the Midwest Political Science Association, Chicago.

101

Lyons, Michael and Peter F. Galderisi. 1995. “Incumbency, Reapportionment, and U.S.
House Redistricting” Political Research Quarterly 48:857-871.

Mann, Thomas. 1978. Unsafe at Any Margin: Interpreting Congressional Elections.
Washington DC: American Enterprise Institute.

Mayhew, David W. 1971. “Congressional Representation: Theory and Practice in
Drawing the Districts.” In Reapportionment in the I 970 '5, ed. Nelson W. Polsby.
Berkeley: University of California Press.

Mayhew, David. 1974. Congress: The Electoral Connection. New Haven: Yale
University Press.

Moran, P.A.P. 1950. Notes on continuous stochastic phenomena, Biometrika 37:17-23.

Niemi, Richard G. and Laura R. Winsky. 1992. “The Persistence of Partisan
Redistricting Effects in Congressional Elections in the 19703 and 19805.” The
Journal of Politics 54:565-572.

Niemi, Richard G. and Alan I. Abramowitz. 1994. “Partisan Redistricting and the 1992
Congressional Elections.” The Journal of Politics 56:81 1-817.

Nokkcn, Timothy P., 2000 “Dynamics of Congressional Loyalty: Party Defection and
Roll-Call Behavior 1947-97,” Legislative Studies Quarterly, 25: 417-44.

Nokkcn, Timothy. 2005 “Party Switching and the Procedural Party Agenda in the US
House of Representatives, 1953 — 2002” Paper prepared for presentation at the
Annual Meetings of the Midwest Political Science Association, April 7 — 10,
2005, Chicago, IL.

Nokkcn, Timothy P., and Keith T. Poole. 2004. “Congressional Party Defection in
American History,” Legislative Studies Quarterly 29: 545-68.

Olsen, Mancur. 1971. The Logic of Collective Action: Public Goods and the Theory of
Groups Cambridge, Mass: Harvard University Press.

Oleszek, Walter J. 2004. Congressional Procedures and the Policy Process 6th ed.
Washington DC: CQ Press.

Palfrey, Thomas, 1984. Spatial Equilibrium with Entry. Review of Economic Studies:
51: 139-156.

Poole, Keith T. 1998. “Changing Minds? Not in Congress!” GSIA Working Paper
#1997-22. Camegie-Mellon University.

102

Poole, Keith T., and Howard Rosenthal. 1985. “A Spatial Model for Legislative Roll-
Call Analysis.” American Journal of Political Science 29(2):373—399.

Poole, Keith T., and Howard Rosenthal. 1997. Congress: A Political-Economic History
of Roll Call Voting. New York: Oxford University Press.

Ramsey, 1B. 1969. “Tests for Specification Errors in Classical Least Squares Linear
Regression Analysis,” Journal of the Royal Statistical Society, Series B, 31: 350-
371.

Reynolds v. Sims. 1964. 377 U.S. 533

Rivers, Douglas. 2003. “Identification of Multidimensional Spatial Voting Models,”
Stanford University typescript.

Rohde, David W. 1991. Parties and Leaders in the Postreform House. Chicago:
University of Chicago Press.

Rohde, David W. Roll Call Voting Data for the United States House of Representatives,
1953-2004. Compiled by the Political Institutions and Public Choice Program,
Michigan State University, East Lansing, MI, 2004.

Romer, Thomas and James M. Snyder, Jr. 1994. “An Empirical Investigation
of the Dynamics of PAC Contributions,” American Journal of Political Science
38: 745-769.

Rothenberg, Lawrence S., and Mitchell S. Sanders. 2000. “Severing the Electoral
Connection: Shirking in the Contemporary Congress.” American Journal of
Political Science 44: 316-325.

Sehoenberger, Robert A. 1969. “Campaign Strategy and Party Loyalty: The Electoral
Relevance of Candidate Decision Making in the 1964 Congressional Elections.”

American Political Science Review 63: 515-20.

Shotts, Kenneth W. 2001. “The Effect of Majority-Minority Mandates on Partisan
Gerrymandering.” American Journal of Political Science 45: 120-135.

Smith, Richard A. 1995. “Interest Group Influence in the U. S. Congress,” Legislative
Studies Quarterly, 20: 89- 1 39.

Smith, Steven S. and Christopher J. Deering. 1990. Committees in Congress 2ed.
Washington DC: CQ Press.

Snyder, James M. 1990. “Campaign Contributions as Investments,” Journal of Political
Economy 98:1195-1227.

103

Squire, Peverill. 1985. “Results of Partisan Redistricting in Seven U.S. States During
the 19708.” Legislative Studies Quarterly 10:259-266.

Stokes, Donald E. and Warren E. Miller. 1966. “Party Government and the Salience of
Congress,” Elections and the Political Order, Angus Campbell, Philip E.
Converse, Warren E. Miller, and Donald E. Stokes, eds. New York: Wiley.

Stratmann, Thomas. 1995. "Campaign Contributions and Congressional Voting:
Does the Timing of Contributions Matter?" The Review of Economics and
Statistics. V(February): 127-36.

Stratmann, Thomas. 2000. “Congressional Voting over Legislative Careers: Shifting
Positions and Changing Constraints,” The American Political Science Review, 94:
665-676.

Sullivan, John L., L. Earl Shaw, Gregory E. McAvoy, and David G. Barnum. 1993.
“The Dimensions of Cue-Taking in the House of Representatives: Variations by
Issue Area.” Journal of Politics 55: 975-97.

Swain, John W., Stephen A. Borrelli and Brian C. Reed. 1998. “Partisan Consequences
of the Post-1990 Redistricting for the U.S. House of Representatives.” Political
Research Quarterly 51 :945-967.

Tobler, W. 1970. “A computer movie simulating urban growth in the Detroit Region,”
Economic Geography, 46(supplemental): 234-240.

Tomz, Michael, Jason Wittenberg, and Gary King. 2003. CLARIF Y: Software for
Interpreting and Presenting Statistical Results. Version 2.1. Stanford University,
University of Wisconsin, and Harvard University. January 5. Available at
http://gking.harvard.edu/

Tufte, Edward R. 1973. “The Relationship between Seats and Votes in Two-Party
Systems.” The American Political Science Review 67:540-554.

Wesberry v. Sanders. 1964. 376 U.S. 1

Wooldridge, Jeffrey M. 2003. Econometric Analysis of Cross Section and Panel Data.
Cambridge, MA: The MIT Press.

Wright, John R. 1985. “PACs, Contributions, and Roll Calls: An Organizational
Perspective,” The American Political Science Review, 79: 400-414.

Wright, John R. 1996. Interest Groups and Congress: Lobbying, Contributions,
and Influence. Needham Heights, MA: Allyn and Bacon.

104

Zeller, Richard A. and Edward G. Carmines. 1980. Measurement in the Social Sciences:
The Link Between Theory and Data. New York: Cambridge University Press.

105