This analysis has shown that the quantity of regular news and campaign
news that voters consumed helped explain variation in their retrospective
evaluations of the national economy in 1992.
The more news voters consumed and the closer they followed the campaign through the media, the worse their retrospective assessments of the economy were. The 1992 election was the only election of the last three exhibiting a relationship between media consumption and national retrospective economic evaluations.
This was the case for several related reasons-the unchallenged importance of the economy during the campaign season, the news media's almost exclusively negative accounts of its performance, Clinton and Perot's constant attention to the issue, and the fact that the recovery from the 1990-91 recession was not as pronounced as 1984.
Conservatives may be tempted to use these results as confirmation of a liberal bias in the media, but President Clinton's experience over the first three years of his presidency should undermine that thesis. Rather, the institutional bias of the media towards controversy and negative information (Patterson 1993; Robinson and Sheehan 1983), coupled with the skill of the challengers' campaigns in enlisting the media to project their massages explains the 1992 findings.
Similarly in 1994, a vigorously improving economy did not bring Democratic congressional candidates many votes.
While many possible explanations exist, one of the most compelling is the gap between economic perception and reality.
At the time of the 1994 election, a significant portion of the electorate believed that the United States' economy was still mired in recession.
This gap may provide a plausible explanation for the uncharacteristically poor performance of the economic forecasting models in 1992 as well (e.g., Campbell and Wink 1990; Campbell 1992; Erikson 1989, Fackler and Lin 1995; Fair 1978, 1988; Lewis-Beck and Rice 1992; Rosenstone 1983).
These models have been remarkably successful in forecasting presidential
elections over time. For example, Lewis-Beck and Rice (1992) note that their model had correctly predicted 10 of the last II presidential contests
before the Bush-Clinton-Perot race.
The Fair model has enjoyed similar success. From 1916 and 1988, it correctly called 16 of the 19 races.
In 1992, however, both predicted comfortable Bush victories. Only Abramowitz's
(1988) model, which estimated Clinton's share of the two party vote to be 53.7%, was close to predicting both his victory and margin of victory correctly.
Perceptions of the economy tainted by relentlessly negative reporting
of its condition may offer some insights.
To account for this potential gap, perhaps economic voting analyses and forecasting models should pay increased attention to some perceptual variable such as consumer confidence.
During 1992, consumer confidence
and GNP growth were often negatively related (Stanley and Niemi 1994).
Inclusion of some mediated variable seems particularly important today for two reasons.
First, in the post Cold War world, it stands to reason that issues of economic performance and domestic policy will play a larger role in the discussion of public affairs.
While discussion of foreign affairs will not cease, it will not reflect the consistently high stakes that it did when the Soviet Union was a constant military threat.
The results presented above suggest that the economy must dominate the agenda for the media to have an effect.
Second, the data presented in this study also suggest that divergences between perceptions of economic activity and actual performance may be more likely when growth is positive but not dramatic. With the premature end of the "'American Century" (e.g., Kennedy 1987),
the United States' economy may be more inclined in the future to produce
statistics similar to those of 1992.
Therefore, capturing the public's perception of economic activity may become all the more important.