Oregonian: Journalism in a Bubble

Sunday morning’s Oregonian had a conventional wisdom cover article about how out-of-touch the city folk are from the real people of the country, based on the same faulty mathematics and map-reading that have characterized similar coverage. My response:

As Barbie once said: “Math is hard.” A prime example is in this morning’s cover
story “Voting in a Bubble” by Edward Walsh and Jeff Mapes, which posits that a
gap has opened between voters in Multnomah County and the rest of the state,
nay, the country. To bolster their argument, Walsh and Mapes tout the fact that
Multnomah County’s vote went for John Kerry by “nearly 24 percentage points
higher than his national showing” and about 21 points over the statewide

What Walsh and Mapes seem not to realize is that the state average includes
Multnomah County. An average is the middle ground of a set of numbers. For
every outlying data point in an averaged set (Multnomah County for the
Democrats), there has to be some sort of balance. While it’s true that
Multnomah is the only county in the state where the vote is so lopsidedly
Democratic, much of Eastern Oregon is just as out-of-whack with the average
as Multnomah — just the other direction. Baker, Crook, Grant, Harney, Klamath,
Lake, Malheur, Wallowa, and Wheeler counties all voted for George W. Bush at
rates more than 20% of the state average this year. All but two of those counties
did so at rates higher than the 21% deviation from average they claim separates
Multnomah County from the rest of the state. Grant County, where over 78% of
the voters chose Bush (according to state figures available Sunday morning), is
over 30 percentage points off the statewide tallies of Bush voters, making them
the most out-of-touch county in the state, if you accept the terms of the article.
Not only was their vote far off the state average, but according to the majority of
Oregonians, they chose the wrong candidate.

There have been a number of official-looking graphs from news organizations in
the past few days showing “a pattern of heavily Democratic cities surrounded by
a sea of pro-Bush voters” in this election. Most of those graphs are organized by
county. At first glance, they appear somewhat intimidating for Democrats. And it
might indeed be scary if acreage — rather than people — voted.

I’ve provided a couple of charts of my own, which show the deviation from both
state and national averages in this election. There is a disparity from the norm in
Multnomah County. But there’s an equally wide deviation in most of the
counties on the other side of the Cascades. In fact, the only counties where the
vote was within a couple of percentage points of the state average were
Columbia and Washington.

open chart in a separate window

I’m surprised that a 1,300-word article managed to get to print without someone
realizing that if Multnomah County’s numbers are far off in one direction that
there has to be something of equal size on the other side of the average pulling
the other way. It took me all of about 10 seconds to see the problem in the
article’s argument, and I haven’t taken a math class for over twenty years.
Perhaps it’s time, though, for the Oregonian to send a few of its reporters and
editors to some remedial math courses.

The article points out that without Multnomah County, Bush would have won the
state by 80,000 votes. It’s no real surprise that if you eliminate a fifth of the
electorate, the results of the vote might change. Multnomah County accounted
for 343,290 of the 1,754,873 votes in the state, just over 19.5% of the voters in
the presidential race. Ignoring a similarly-sized number of ballots from the
counties with the highest percentage of Bush voters would eliminate the counts
of 19 counties (the nine mentioned above plus Curry, Douglas, Gilliam,
Jefferson, Josephine, Linn, Morrow, Sherman, Umatilla, and Union counties),
288,434 voters — only 16.5% of the total, and would have given Kerry a victory
margin of 161,341 votes, over two-and-a-half times the actual statewide result.
Eliminating Multnomah County from that scenario so that the counties
containing the fifth of the votes most skewed toward each candidate aren’t
counted still gives Kerry a slim margin of 8,827. To someone with even a
layman’s understanding of mathematics, it’s really no surprise that as you
eliminate data that deviates from both sides of the norm, the numbers tend
toward the norm.

I’m looking forward to Walsh and Mapes’s follow-up article, showing how
Eastern Oregon is drifting away from the state and nation because of a more
pronounced shift away from the Oregon norm than they found in Multnomah

— Darrel Plant