At voters are capable of evaluating economic circumstances at

At first glance, Fair’s data seems to
prove that voters are capable of evaluating
economic circumstances at the aggregate
level. He admits that voters do not know the
“exact numbers,” but he believes they can
sense the direction of the economy through
their daily interactions. Fair also suggests
that voters are strongly influenced by the
media, and the “good news quarters”
variable supposes that they can remember
the number of quarters in which the media
has praised the economy’s growth.9
He rests
his theory on the statistical accuracy of his
model, which has come within one or two
percentage points of the incumbent’s vote
share in most elections, and he insists that
these results are due more to marginal┬ávoters’ retrospective macroeconomic
evaluations than to campaigns effects.11
Fair’s model also makes dangerous
assumptions. One keystone of his argument
is that marginal voters rely on retrospective
judgments to guide their vote choice, but
recent data suggests that this behavior is not
as prevalent as he claims. During the 1992
campaign, Marion Just and her colleagues
extensively interviewed voters of all
backgrounds, and found that their
retrospective considerations of candidates
declined with each interview. Respondents
were uncomfortable with the “up-or-down
vote on the incumbent” that Fair’s model
assumes, and they looked to the campaigns
for information about challengers. Many
respondents also indicated that candidates’,
even incumbents’, past actions were not
sufficient to predict future actions, and so
prospective considerations were also used.17
This data seems to have been reaffirmed in
2000, when Al Gore should have held a
solid edge in retrospective considerations
due to the economic prosperity of the
Clinton era. Gore’s campaign, however, cut
almost all ties with the President because of
his moral failings, and voters did not
associate Gore with the past success of
Clinton’s administration.18 Instead, voters
seemed to rely heavily on the current events
of the campaign, as seen through media
coverage and television advertising.19
While Fair shows impressive
statistical correlations, and makes clear that
economic conditions play some role in
electoral outcomes, his model has technical
flaws. For one thing, he includes no margin
of error. Jay Greene of the American
Prospect estimates Fair’s error to be around
8% for the 1992 election, which, if true,
means that a victory by either candidate
could have satisfied his model.12 Also, the
fact that his data is historically accurate does
not prove its predictive capabilities. Green
points out that constructing the model after
the fact allows Fair to fit the variables to
past conditions, and until the model is tested
extensively in the future, it cannot be
considered a reliable forecast.13 In fact, the
model has failed to accurately choose the
winner of the two most recent elections, and
this trend may be likely to continue because
of two glaring omissions. First, Fair’s
model only predicts the two-party vote
share, despite the fact that two of the past
three elections have been heavily influenced
by third party candidates. Not surprisingly,
his model was the least accurate in 1992,
when a third party candidate got a
significant portion of the vote share.14 Also,
his model discounts the role of the Electoral
College, a point which he openly admits and
dismisses as immaterial.15 Rebecca Morton,
however, has shown that the Electoral
College encourages targeted campaigning,
which affects voter preferences in key states
and can give candidates an electoral victory
without winning Fair’s prized two-party
vote.16