A New Measure of Party Strength

• Author: Robert P. Saldin,James W. Ceaser
• Published date: 01 June 2005
• Date: 01 June 2005
• DOI: http://doi.org/10.1177/106591290505800205
• Subject Matter: Articles

A New Measure of Party Strength
JAMES W. CEASER, ROBERT P. SALDIN, UNIVERSITY OF VIRGINIA
In order to measure the strength of the parties in each state, the Major Party Index (MPI) was built by averag-
ing the results of the six major elections that take place in the fifty states. This index allows us to describe the
absolute and comparative partisan leaning of each state in each election and identify trends of party strength
over time within individual states, among regions, and within the nation as a whole. The MPI sheds consider-
able light on three general developments: (1) a national change from Democratic dominance in the 1980s to
a Republican edge by 2002, (2) significant regional realignments in the South and New England, and (3) a
strong trend toward greater consistency between partisan voting at the federal and state levels.
Party strength in the United States is generally measured the advantage of directly measuring the phenomenon: if
by public opinion polls in which respondents report
one is interested in party strength, it makes great sense to
their own partisan identification. There are well-known
look at how citizens actually vote. Still, there are weak-
advantages to this approach. It permits analyses that relate an
nesses in using electoral results. The statistics are tied to
individual’s party preference to attitudinal and demographic
aggregates (collective units), not to individuals, which
factors, thereby allowing for the generation and testing of
makes it impossible to connect the vote to attitudinal vari-
hypotheses about the causes of party attachment. Self-identifi-
ables and difficult to relate it to many demographic factors
cation can also be correlated with reported votes, which allows
(other, of course, than geography itself).3 Another problem
for inquiries into the relative importance of party preference
is that a vote for a particular candidate is by no means the
for electoral behavior. For these reasons, poll surveys have
same thing as an expression of support for a party; it may
become the preferred standard in academic research. But this
only reflect a preference for a particular individual, as when
method has admitted weaknesses. Polls are expensive and
millions of Democrats pulled the lever for the war hero
they often have a significant margin of error, which in the case
Dwight Eisenhower in 1952 and 1956. Election results
of general national polls increases greatly as one begins to
from any single election may therefore reveal little about
investigate particular sub-groups or geographical regions. Par-
“real” or “normal” party strength.
tisan self-identification by itself, moreover, does not tell us
All approaches are bound to have strengths and weak-
how people actually vote. For example, many white South-
nesses, and the question of a measurement’s worth should be
erners after 1950 identified themselves as Democrats or inde-
judged on the practical grounds of its helpfulness as an
pendents even as they had become reliable Republican voters
investigative tool. Despite the drawbacks just noted, the idea
at the presidential level. Finally, partisan self-identification
of measuring party strength on the basis of election results
itself says nothing about the habitual behavior—if it exists—
seems attractive enough to hold considerable interest for
of independents. Many analysts today suspect that most voters
political scientists—not as a substitute for, but as a supple-
who call themselves independents in fact have a fairly distinct
ment to, polling methods. Yet this approach is virtually
partisan leaning when it comes to casting their ballots.1
absent from contemporary political science. Frequently, of
Another approach to assessing party strength is to look
course, one comes across rule-of-thumb descriptions that
at election results and measure party standing by the votes
refer to election outcomes. Pick up any book on elections
each party’s candidates received. This method also has its
and you will likely read statements such as “Massachusetts
strong points. It is inexpensive (at least for scholars), as the
voted Democratic in the last four presidential elections,” thus
government picks up all of the costs of the research project
qualifying it as a solidly Democratic state. General historical
by holding elections. The margin of error is fairly low, with
studies also frequently rely on such characterizations as in
the inaccuracies being limited to fraud and the now well-
Black and Black’s (1992) characterization of the South: “For
documented difficulties encountered in counting and
more than six decades, from 1880-1944, the eleven states of
recording votes (hanging chads, undervotes, voting
the old Confederacy . . . regularly voted as a solid block in
machine failure, and the like).2 Election results also have
favor of the presidential candidate of the Democratic Party.”4
Such references, however, are completely ad hoc; they have
no common standard and offer no basis for systematic
1 Bruce E. Keith, et al., The Myth of the Independent Voter (Berkeley: Uni-
versity of California Press, 1992).
2
3
For a discussion of some of these problems, see Brady, Henry E., Justin
For a recent method that can account for ways of using aggregate data
Buchler, Matt Jarvis, and John McNulty. 2001. “Counting All the Votes:
for some of these purposes, see Gary King. 1997. A Solution to the Eco-
The Performance of Voting Technology in the United States.’’ Available
logical Inference Problem: Reconstructing Individual Behavior from Aggregate
at: http://ucdata.berkeley.edu/.
Data. Princeton, NJ: Princeton University Press.
4 Earl Black and Merle Black, The Vital South (Cambridge: Harvard Uni-
Political Research Quarterly, Vol. 58, No. 2 (June 2005): pp. 245-256
versity Press, 1992) 4.
245

---

246
POLITICAL RESEARCH QUARTERLY
communication among those studying party strength. What
totals for federal elections only.5 By contrast, Ranney (1965)
is needed, clearly, is a general index to which all can refer.
and King (1989) measured the level of party competition
The idea behind an index—think for example of the Dow
only in state-level contests.6 A few efforts combined results
Jones index—is to capture in one number a measure that
from both levels, as we shall do here (Hofferbert 1964;
supplies a broad picture and that can provide a baseline for
David 1972; Bullock 1988).7
charting trends and movements in party strength.
Another point of difference revolved around the type of
A few political scientists in the period between 1960 and
data employed and the frequency of observation. Some
1989 sought to devise such an index. These efforts did not
indexes employed the raw vote totals or percentages of the
gain wide currency at the time, and none of them has been
vote each party won (Hofferbert 1964; Ranney 1965;
updated or remains in use today. Part of the reason is no
David1972; King 1989). Others relied solely on the number
doubt the aforementioned appeal of survey data. But
of seats each party held (Bullock 1988). The frequency of
another reason, we would argue, is to be found in the flaws
observation also varied. Some indexes were based on long
in these indexes. With improvements, we believe it is possi-
periods of aggregate analysis (Hofferbert, Ranney, King),
ble to construct a measure that overcomes some of the past
while others took new measurements every two years
difficulties and that can serve today as a helpful tool for
(David, Bullock). The Cox Index used a combination of
political analysis. The objective is to construct a measure
these approaches. These indexes were also used for different
that captures the underlying strength of the parties while
purposes. Hofferbert and King sought to determine the level
managing to absorb and smooth out distortions of particu-
of competition in each state and ascertain which party was
lar elections.
in control, while others were more concerned with observ-
Any measurement using electoral results must, as noted,
ing partisanship over time and in analyzing the reasons for
proceed on the basis of geographical units. This unit could
change (Cox, Ranney, David, Bullock).
be as small as the precinct. But for the purposes of a general
These previous attempts to examine party competition
index for the whole of American politics, the state is the
demonstrate the diversity of opinion over what should be
proper starting point. States serve as the unit for selecting
measured and how such measurements should be carried
presidential electors, senators, and governors. Additionally,
out. They also reveal, however, some methodological draw-
state majorities generally determine the majority party in
backs, even when taking account of the purposes for which
state legislatures. In order to measure the underlying
they were intended. Three of these indexes deserve further
strength of the parties in each state, we have built an index
attention because their successes and failures informed the
that averages the results of the elections for the six major
development of our new measure.
offices that take place in each of the fifty states: president,
The Cox Index was an early effort to measure party com-
U.S. senator, U.S. representative, governor, state senator,
petition in and between the states. Relying solely on vote
and state representative.
totals for federal offices, Cox labeled each state as either
Our Major Party Index (MPI) allows one to describe the
Democratic or...