Can outcomes be predicted before even a single vote is cast?
By Christian Grose, assistant professor of government
Lawrence Today magazine, Fall 2004
How
can we predict the outcome of a presidential election before it even occurs?
In recent decades, political scientists and economists have been able
to forecast, with relative precision, the exact vote percentage that
an
incumbent presidential candidate or the candidate of the incumbent’s
party will receive, based on past election results.
By using statistical models to examine the impact of theoretically relevant
economic and political variables on past presidential election outcomes,
scholars have, for the most part, been able to predict the winner with
quite a bit of accuracy — although there have been notable embarrassments.
Forecasting elections is part science and part art. One of Lawrence’s
key contributions to the field of government and political science was
Professor William H. Riker, a
member of the Lawrence faculty from 1948 to 1962, who reshaped the field
to be more scientific in nature. Specifically, he was
a proponent
of pairing theory with rigorous analysis to examine predictable patterns
observed in politics.
The “science” of forecasting elections lies in knowing that there
are reliable patterns of behavior on the part of both voters and candidates.
Voters are usually predictable, and there are certain factors that regularly
explain
electoral outcomes. In an election with two
major-party candidates, the candidates will also behave predictably, taking
stances and engaging in campaign activities designed to give them the most
complete advantage. These activities often have the effect of canceling each
other out. Thus, at the end of the election, even though the day-to-day battles
of the campaign are important, other factors, some outside of their control,
are likely to contribute to the election’s final outcome.
The “art” part of forecasting elections has to do with determining
just what those “other factors” are and tweaking the statistical
models to lead to the best prediction. Based upon scholars’ statistical
models, I will detail four key factors that are able to predict which
presidential candidate is likely to win. This will provide a guide for
what you might
expect this November in the contest between George W. Bush and John Kerry.
There are four key indicators that work well in forecasting presidential
elections:
(1) economic growth,
(2) presidential approval,
(3) the extent of U.S. international involvement, and
(4) the comparative political experience of the two major-party candidates.
It is important to note that these models predict only the two-party popular
vote between Republicans and Democrats. Of course, as we know from both
the 2000 election and from the work of the late Professor Lawrence
Longley, a Lawrence
government professor from 1965 to 2002, the design of our presidential
election system, with its electoral college, can occasionally lead to
a popular-vote winner who is not the electoral-vote winner. As these
forecasting models
are based on aggregate national data, the prediction is for the two-party
popular vote. In addition, only the two-party vote is predicted for the
sake of simplicity.
Factor 1: Economic growth rate
Of the four factors, two have since World War II been the best predictors:
(1) the economic growth rate (the second-quarter Gross Domestic Product
growth rate) between six and nine months before the November election
and (2) the
presidential approval rating between four and six months before the November
election.
Even more than (3) the extent of U.S. international involvement and (4)
the candidates’ political experience, these are the most critical
indicators, although international involvement and political experience
are also relevant.
If the economic growth rate is strong in the second quarter of 2004 and
the incumbent’s approval numbers are around 50 or higher, then
Bush is in a strong position. If the economic growth rate in the second
quarter is
relatively low or Bush’s approval ratings are hovering in the mid-
to low-40s, Kerry is in a strong position for victory.
The simplest forecasting model posits that second-quarter growth in the
gross domestic product (GDP) affects the vote received by the candidate
of the
president’s party in November. Forecasters take the second-quarter
GDP in the year before the election and in the year of the election and
examine the change from one year to the next (2003 second quarter compared
to 2004
second quarter).
Political scientists James Campbell, Tom Rice, and Michael-Lewis Beck,
among others, use second-quarter growth rates in their models of election
forecasting.
Ray Fair, a Yale economist, uses first-, second-, and third-quarter GDP
growth in his. The second-quarter growth rate is the best indicator because
there
is a time lag for positive or negative economic growth to trickle down
to the average voter. Economic growth in the second quarter is likely
to have
reached the pocketbooks of Americans in tangible ways before ballots
are cast in November.
More specifically, political scientist Randall Jones, Jr., has summarized
a variety of forecasting models and notes the levels at which economic
growth spells trouble or signals likely victory for an incumbent president
or candidate
from an incumbent president’s party. If GDP growth is 2.6 or higher,
the incumbent party’s candidate is likely to win the popular vote.
If GDP growth is 1.5 or lower, the incumbent party’s candidate
is likely to lose the popular vote. And if GDP growth is between 1.6
and 2.5, the prediction
is not clear.
This factor has done the best job in correctly predicting the winner
of every presidential election since 1956 except one (based on a second-quarter
growth
rate of 6.5 percent in 1968, Hubert Humphrey, and not Richard Nixon,
should have won the presidency). However, in all other years, this predictor
worked.
For instance, in 1980, economic growth in the second quarter was at an
abysmal -10.2, and Ronald Reagan defeated incumbent Jimmy Carter. Four
years later,
in 1996, the economic growth rate was 5.3 percent, and Ronald Reagan
sailed to reelection. Based solely on economic growth, which was 3.0
percent in
the second quarter of 2004, George W. Bush appears to be at an advantage
for this election.
Factor 2: Presidential approval rating
Economic growth is not the only key explanatory factor in predicting
elections. Presidential approval approximately four to six months before
the election
is a critical predictor of the final outcome.
While a handful of truly “swing” voters do not decide until just
before Election Day, most undecided voters or leaners solidify their views
in favor or against the president and the candidates in mid-summer. If the
incumbent president’s approval rating is 50 percent or greater, then
the candidate of the party of the president is in strong shape to win the
election in November. However, if this approval rating falls below 50, especially
if it is significantly below 50, then it is likely that the candidate of
the president’s party will lose the popular two-party vote.
The president’s approval rating is measured by the percentage of Americans
who respond that they approve of the president’s job performance when
asked the following question: “Do you approve or disapprove of
the job George W. Bush is doing as president?”
Thus, the formation of public opinion regarding the job that Bush is
doing as president is critical for his reelection chances. Bush’s
approval ratings have been relatively erratic (at press time), and this
does not provide
the clearest prediction of his likelihood of winning. The most successful
candidates in post-WWII elections have typically been held in high esteem
by the electorate in the summer before Election Day. For instance, Ronald
Reagan’s approval rating hovered in the mid-to-high 50s through
the summer of 1984, and never dropped below 52 percent during this period.
In contrast, when George H.W. Bush sought reelection in 1992, his approval
rating
was below 40 percent throughout most of the summer months, sealing his
fate in November.
According to a Roper Center poll taken at the end of July, George W.
Bush’s
approval rating was 47 percent.
Factor 3: International involvement
The third key factor in forecasting presidential election outcomes is
the extent of U.S. involvement in overseas affairs.
In the wake of September 11, 2001, when America found itself under terrorist
attack, the political dynamics of Washington — and perhaps presidential
elections — changed.
Generally, during times of foreign-policy crises, Americans “rally
around the flag” and their president. Specifically, some political
scientists and economists who statistically model election outcomes suggest
that when U.S. involvement in foreign affairs is high, the incumbent
president or the candidate of the president’s party is more likely
to win. Some measure this by the number of troops deployed overseas,
while others
simply
consider the number of recent U.S. international conflicts.
This indicator would seem to suggest that President Bush’s reelection
chances are enhanced by the relatively high U.S. involvement overseas
in the multiple venues of Afghanistan and Iraq following the 9-11 attacks.
However, Kerry fans should take note. The one year in which the economic-growth
predictions mentioned earlier were unsuccessful was in 1968. Hubert Humphrey,
the Democratic candidate, was predicted to win simply based on an economic
growth rate of 6.5 percent and the fact that the incumbent president,
Lyndon Johnson, was also a Democrat. However, the U.S. was involved heavily
in the
Vietnam conflict at this point, which was growing more unpopular as each
day progressed closer toward November.
It is mere speculation at this point, but 2004 may be an anomalous case
where the extent of international involvement does not necessarily help
the incumbent
president. If Americans ultimately determine that extensive involvement
in Iraq was not a good idea, then this is likely to be reflected in Bush’s
approval rating, the second predictive factor. However, if recent history — except
1968 — is our guide, then the high level of U.S. involvement in
overseas affairs will pay off electorally for President Bush in November.
Factor 4: Candidates’ political experience
The fourth and final factor that is useful in predicting presidential
electoral outcomes is the relative experience of the two major-party
candidates.
Typically, incumbent presidents have a slight advantage here, though
most candidates have extensive experience in government if they become
a major
party’s nominee. In 2004, since Kerry has nearly two decades of
experience in the U.S. Senate and Bush obviously has experience as president,
there
is not a huge imbalance on this factor and it is unlikely to be of consequence.
Generally, though, if there is a serious imbalance in experience, the
more experienced candidate does better.
What does this all mean?
While there are clear patterns that are able to
help us predict the likely winner in a presidential election before voting
occurs, do not sit at home on Election Day. It is still normatively important
to participate in democracy and cast your vote.
Also, the practice of forecasting elections is similar to the ability of
meteorologists to predict the weather with a reasonable degree of accuracy.
While they often predict correctly, sometimes the forecast is for sun,
and it actually ends up raining. Take heart: if you do not think these indicators
line up for your candidate of choice in the 2004 November elections,
then
hope for rain! Or, if they do line up for your candidate, remember that
even forecasters cannot be 100 percent correct 100 percent of the time. Nevertheless,
by harnessing the predictive power of observable patterns in repeated presidential
elections, we are able to make an educated prediction going beyond mere speculation.
Christian Grose, who joined the Lawrence faculty in
2002, is interested in American government and politics, with specialties
in congressional
representation,
parties, elections, voting behavior and public opinion, voting rights,
racial politics, research methods, and the empirical testing of formal
models. He
holds a bachelor’s degree from Duke University and a doctorate
from the University of Rochester.
He is the 2004 recipient of the American
Political
Science Association’s Carl Albert Dissertation Award for the
best doctoral dissertation in the area of legislative studies. The
dissertation, “Beyond
the Vote: A Theory of Black Representation in Congress,” examines
the effect of electoral structures and the election of black legislators
on the
representation of black constituencies in Congress. Professor Grose
is the first faculty member at a liberal arts college to receive the
award.