November 11, 2002
The Trouble With the Star-Trib Poll
by Scott W. Johnson
The Minneapolis Star
Tribune dominates the newspaper market in the upper Midwest. As
the region's leading newspaper, it vows to exercise the power
it wields in covering the news with a sense of responsibility
regardless of the paper's editorial views. In the election just
concluded, however, the Star Tribune's Minnesota Poll failed the
paper's most basic responsibility to its readers.
The
Minnesota Poll has a long and inglorious history in Minnesota.
Most famously, in 1978 the Minneapolis Tribune (as it then was)
called all three major statewide races wrong by a wide margin
on the basis of its Minnesota Poll. According to the Tribune on
the Sunday before the election, the Democratic candidates were
about to sweep the gubernatorial and two senate races.
Instead,
1978 was the year of the "Minnesota massacre." The Democrats were
routed; Republican Al Quie was elected governor, and Republicans
Dave Durenberger and Rudy Boschwitz were elected senators.
The
Tribune immediately acknowledged the gravity of its errors and
promised to set things right. In 1987 the Star Tribune hired Rob
Daves to run the Minnesota Poll and the poll was returned to the
newsroom. Daves has continued to direct the poll since that time.
In
the past two elections, the Minnesota Poll's final pre-election
poll results have proved wildly misleading in comparison with
the actual electoral results. In each case, the final poll results
have dramatically understated Republican support. The discrepancy
between the Minnesota Poll results and actual electoral results
does not appear to be random; it has consistently disfavored Republicans.
Let's review recent history.
In
the year 2000 election cycle, the Star Tribune published its final
Minnesota Poll on November 5, 2000, two days before the election.
The story summarizing the poll results ran on page one and dramatically
reported that in a race that had been neck-and-neck, Gore had
opened a 10-point lead over Bush, 47 percent to 37 percent.
The
story reported that the race was "still-volatile" and quoted University
of Minnesota political science professor Steve Smith as saying,
"Gore's in the driver's seat in Minnesota. It appears a number
of Minnesotans came back to Gore-where a lot of people expected
them to be all along."
On
election day, however, the race was in fact neck-and-neck. Gore
edged Bush in Minnesota by only 60,000 votes out of 2,450,000
cast, 47.9 to 45.5 percent.
The
poll cannot have been accurate, and its effect on Republican voters
can only have been demoralizing. The remarkable fact about the
2000 presidential election is that Bush's pre-election lead, measured
in every national poll, evaporated in the days before the election.
In
their post-election recap in the Weekly Standard (November 27,
2000), Jeffrey Bell and Frank Cannon wrote, "National pollsters
are nearly unanimous in believing that a George W. Bush lead of
5 percentage points at the end of October turned into the dead
heat in the popular vote that was cast on November 7." The article
reviewed final shifts in voter sentiment in detail, showing that
Gore's closing surge varied in size around the country; his gains
were widespread but not uniform.
I thought
at the time, and still do, that the Star Tribune's final pre-election
poll was wrong and probably affected the election result in Minnesota.
I called Rob Daves to say as much and to complain about it. I
also summarized the Bell and Cannon article that belied the poll.
With no evidence other than his own poll, Daves stated that Minnesota
was an exception to the national trend; in Minnesota, according
to Daves, Bush had a closing surge.
This
year's Minnesota Poll performed miserably as well. When the Star
Tribune published its final Wellstone/Coleman poll in mid-October,
showing Coleman trailing far behind Senator Wellstone, I called
Daves to ask about the peculiarities of that particular poll.
I mentioned
my earlier, November 2000 conversation with him and concluded
by offering to bet Daves dinner for two at a restaurant of his
choice that Coleman would do five points better than the Star
Tribune's final pre-election poll, whatever it showed. Daves (wisely)
declined the bet.
The
Star Tribune's final pre-election poll was published on November
3, two days before the election. It showed Mondale leading Coleman
46 percent to 41 percent. In the actual election results, of course,
Coleman beat Mondale 50 to 47 percent. The Minnesota Poll understated
Coleman's strength as measured in the actual election results
by 9 points and missed the margin between them by roughly the
same amount.
Again
Daves has attributed the discrepancy to a volatile electorate.
However, it is a mysterious kind of "volatility" that somehow
manages to disfavor only the Republicans.
There
appears to be a problem here that has less to do with a volatile
electorate than with the Rube Goldberg methodology of the Minnesota
Poll. Traditional electoral polling methods call for the identification
of "likely voters" and the tabulation of their preferences. These
are the methods used, for example, by the Mason-Dixon polling
organization that conducts polls for the St. Paul Pioneer Press.
However,
this is not the methodology employed by the Minnesota Poll. The
Minnesota Poll takes into account the preferences of all respondents,
but it "adjusts" the survey results; it "weights" the preferences
of poll respondents according to "formulas verified in past elections."
At
the City Center shopping mall in downtown Minneapolis, the fire
alarm occasionally goes off accidentally. When it does so, City
Center security staff deactivates the alarm and announces that
the "alarm has been verified as false." That is the sense in which
it appears the Minnesota Poll's formulas have been verified in
past elections.
If
I were the editor or publisher of the Star Tribune, I would be
seriously concerned about, if not mightily embarrassed by, the
quality of my product. If the Star Tribune's poll product were
edible instead of legible, it would long ago have been recalled
as dangerous to human health, or it would have killed off its
customers. We can only hope that some day the Star Tribune cares
as much about the quality of its news product as McDonald's does
about the quality of its hamburgers.
Scott
W. Johnson is a Minneapolis attorney and an adjunct fellow of
the Claremont Institute for the Study of Statesmanship. He comments
on the events of the day for the Web log "The
Power Line."
