Lying in Ponds

The write-up about Lying in Ponds in The Atlantic is now available online (”The Other CPI”).

Alert readers may have noticed that while I
usually post the day’s boxscore early in the morning, the list
is frequently updated with additional columns hours or even days
later. The problem is that some of the web sites I download from seem
to be updated erratically. The New York
Times and Washington
Post are updated fairly consistently around midnight Eastern Time, and the WSJ OpinionJournal posts new
columns each day at exactly 12:01 a.m. But TownHall.com has been shaky
lately, and Creators
Syndicate always lags behind by at least a day or two.

Reader Jacob Matthew Petre has a concern
about the partisanship index (PI) methodology:

The real problem with the partisanship index as it stands now lies in
how it treats those columnists who make partisan references most
rarely. Let us say, for example, that a pundit writes 10 columns of
1000 words each. Nine of these columns contain zero partisan
references. The other column contains one positive Democratic
reference. This pundit gets a partisanship index of D-50, and is
considered even more partisan than most pundits who produce “balanced”
partisan screeds, e.g. Frank Rich. I
propose that you consider as a “neutral” reference 100 words without a
partisan reference, to distinguish between partisan
screeds and occasional partisan references.

Good question. I’ve always been concerned with that scenario, and
if you look at
Collin Levey’s columns so far this year,
you’ll see that her anomalous score is an excellent example of the
problem. But I’m less worried now for a couple of reasons.

When I selected a set of columnists to evaluate for 2003,
I tried to choose those who cover mainly national politics
rather than cultural issues or international or local politics. So I
think it’s unlikely that one of the current Lying in Ponds
columnists will make only one partisan reference in ten columns.
Although Ms. Levey’s score is very anomalous so far this year, it’s
only February. In her 2002
columns, she made enough partisan
references (26 negative and no positive Democratic references) to
convince me that her high score was justified.

I really don’t understand the
relationships between syndicated columnists and the newspapers in
which they appear. I’ve switched the affiliation of Robert Scheer from the Los
Angeles Times to Creators Syndicate,
because he is listed there instead of with Tribune
Media Services International (parent company of the former Los
Angeles Times Syndicate International). Since I’m not calculating
team scores this year, it’s not that important but I want to be
accurate.

A couple of weeks ago I proposed a change to the
calculation of the Partisanship Index (PI), which is the foundation of
the scoring system. The goal is to find a way to reduce the
suspected elevation of Democratic scores caused merely by greater
opportunities naturally present during a Republican administration.
Regular reader John Berner has concerns:

I disgaree with the “solution” of weighting R references and D
references at 50%.

Most columnists in 2002 found a way to have at least have a 35%/65%
split of their references to one party vs another.
8 of the top 10 did in 2002 anyway.

Krugman had one of the largest republican/democrat reference ratios
in 2002 of almost anyone in the whole list. Isn’t this signifcant in
itself?
Aren’t you destroying data by ignoring his ratio reference difference
vs other columnists?

I have to agree that if a columnist chooses to overwhelmingly write
about their least favorite party, it’s probably meaningful and should be
reflected somehow in the scores. I need to think about this and look
more closely at the numbers before making a change. As always,
opinions will be welcomed.

Well, I finally picked up a copy of The
Atlantic’s March issue to read a very sympathetic little write-up
(”The Other CPI” on page 38) about Lying in Ponds in the
“Primary Sources” section. That issue is
now online, but for some reason that section is not. Anyway, thanks
to The Atlantic and also to Media Minded (and another
reader), who called
my attention to it, and to Dean Esmay. Those guys have been
embarrassingly nice to me lately!

A couple of weeks ago I noticed that the
Washington Post had added Anne Applebaum to its roster
of political columnists. She seems to cover mainly international
affairs, so I added her to the “Inactive” list — that means that her
columns will appear in the daily boxscores, but will not be analyzed
for partisanship.

After some prodding from readers,
I’ve started adding nicknames to the partisan names lists, so
that they can be evaluated along with proper names (I mentioned
the issue last month). So I’ll add nicknames as they come up in
columns from now on. The names I’ve added so far are “Rummy” and
“Bushies”. I don’t expect it to have much of an impact on the scores,
Maureen Dowd may get a little bump, but then
Peggy Noonan used “Dubya” more than any
other columnist last year.

Salon’s Laura McClure does an interview today
with syndicated columnist Molly Ivins.
To read the entire article, you must be a subscriber or get a “day
pass” by viewing some ads.

I made a “2002″ link
underneath the banner a while back. You can click there to access all
of the 2002 results, including the final rankings.

In a column last week, Paul Krugman found a way to criticize
the Bush administration in an otherwise non-political column about the
space
program:

The sad truth is that for many years NASA has struggled to invent
reasons to put people into space — sort of the way the Bush
administration struggles to invent reasons to . . . but let’s not get
into that today. It’s an open secret that the only real purpose of the
International Space Station is to give us a reason to keep flying
space shuttles.

This merits inclusion in what we could call the Partisan Non Sequitur Hall
of Fame. Earlier inductees would be Thomas
J. Bray, for accusing Bill Clinton of fraud in a column about
golf, and Mary
McGrory for taking time in a column about gardening to charge
George W. Bush with hating international treaties. I don’t think
it’s coincidental that each of these columnists have ranked at or
near the top of the partisanship list.

Reader Joel Garcia recently offered some
thoughtful criticism:

I was a bit surprised at the #5 ranking for Ann Coulter, expecting her
to rank much higher. I therefore took a quick look at the first of
the columns used in determining her rank. Not surpisingly, three of
the her four positive references to Democrats appear in a quote by
John Kerrey.

According to your methodololgy:

“When political figures are themselves quoted, it usually is evaluated
as a positive reference unless the statement is clearly being
ridiculed.”

After quoting Kerry, Coulter writes:

“There’s a rallying cry to unite the Democrats! If there has been a
material breach “by everybody’s standard,” then and only then, we can
boldly … go to the United Nations! ”

Isn’t that clearly ridiculing Kerry’s stance (or at the very least,
detectable sarcasm), and aren’t the positive references to Stevenson
and the Kennedy Administration being made by Kerry and not Coulter?
Or did I misunderstand the methodology?

I’ve found evaluating quotes to be pretty difficult. Here are
three examples of quotes from the Ann
Coulter column Mr. Garcia refers to:

1. All three of the following are evaluated as negative references to
   Democrats:

   John Kerry says he thought a resolution authorizing the president to
   use force against Iraq meant that the United Nations would have to
   approve. Dianne Feinstein said she voted for the resolution assuming
   it meant we would invade only if “our allies” approved. Joe Biden made
   the terrific argument that if we don’t wait for U.N. approval, it
   would “make a mockery of the efficacy of the U.N.”

   John Kerry is paraphrased rather than quoted. Dianne
   Feinstein gets only a two-word quote (”our allies”), not enough to be
   meaningful. Joe Biden gets a longer quote, but it is introduced in a
   hostile way (”made the terrific argument”).

2. In the following, John Kerry is allowed a quote long enough to get
   his point across, but a negative attribution verb (”claims”) and a
   sarcastic word (”foursquare”) is used to undercut the quote, so I
   split the difference and evaluated it as neutral:

   Kerry claims he is still foursquare behind disarming Saddam Hussein,
   but not “until we have exhausted the remedies available, built
   legitimacy and earned the consent of the American people, absent, of
   course, an imminent threat requiring urgent action.”

3. In the final example, Mr. Kerry is given an entire paragraph to
   state his case, which could be convincing to the reader despite
   Ms. Coulter’s subsequent disagreement with it. So the reference to
   Kerry is evaluated as positive, along with his praise of other
   Democrats in the quote itself:

   That’s the short version. The long version of Kerry’s position is
   this:

   “[I]f you have a breach that, by everybody’s standard, at least in
   the United States, those of us in the House and Senate, and the
   president, join together and make a judgment, this is indeed a
   material breach, and then others — some of them can’t be persuaded –
   if we have evidence, sufficient to show the materiality of the breach,
   we should be able to do what Adlai Stevenson did on behalf of the
   administration, Kennedy administration, and sit in front of the
   Security Council and say, ‘Here is the evidence. It’s time for all of
   you to put up. We need to all do this together.’ And that’s what I
   think the resolution that was passed suggests.”

While the column as a whole was very negative toward Democrats, the
use of quotes resulted in some individual positive references. My
point is just that a columnist has the ability to use quotes in a
variety of ways, ranging from ridicule to respect.

Ben Fritz of Spinsanity criticized Maureen Dowd in a recent
piece. Ms. Dowd used an online CNN poll as supporting evidence,
generating this response from Spinsanity:

In citing this poll, though, Dowd should also have cited a crucial
piece of information that accompanies it. Just below the poll results
on its website, CNN cautions the following: “This QuickVote is not
scientific and reflects the opinions only of those Internet users who
have chosen to participate. The results cannot be assumed to represent
the opinions of Internet users in general, nor the public as a whole.”

In other words, the numbers are essentially meaningless.

CNN anchor Wolf Blitzer was careful to acknowledge this when he cited
the poll on his show, “CNN Wolf Blitzer Reports,” on Friday, stating,
“Remember, of course, this is not — repeat, not — a scientific
poll.”

Dowd, however, leads her readers astray by making them think there’s
some evidence for her contention about what “Most Americans”
think. With the proliferation of unscientific polls online, the public
should be able to expect that when polls are cited in the media
without qualification, they really mean something.

From the beginning, I’ve been bothered by what I
think is a tendency for Democratic pundits to have systematically higher
partisanship scores just because they will naturally have more
opportunities for criticism with a Republican administration in office
(the target-rich
environment issue). I may have found a reasonable way to reduce or
eliminate the problem. The current equation for Partisanship Index (PI) is:

               (D+ + R-) - (R+ + D-)
 PI = 100 * ---------------------------
            D+ + D- + D= + R+ + R- + R=
 

where D+, D- and D= are the number of positive, negative and neutral
Democratic references, and R+, R- and R= are positive, negative and
neutral Republican references. If we define the total number of
Democratic references as Dt = D+ + D- + D=, and the total number of
Republican references as Rt = R+ + R- + R=, we can define the
“Democratic Index” to be DI = (D+ - D-)/Dt and the “Republican Index”
to be RI = (R+ - R-)/Rt. These express how a pundit treats only one
party; they will range from +1 for only praise to -1 for only
criticism.

The equation for PI can then be rearranged to form:

                     Dt             Rt
 PI = 100 * ( DI --------- - RI --------- )
                 (Dt + Rt)      (Dt + Rt)
 

The second part of each term consists of each
party’s fraction of the total references. That’s where the problem
comes in, because there are currently far more references to
Republicans. That makes a columnist’s final PI mostly reflect how
they treat Republicans. So a columnist like Frank Rich is ranked as a
strong partisan despite being mostly negative toward
Democrats, because his Democratic component is overwhelmed by the
greater frequency of his criticism of the Republican administration in power.

My idea is simply to equally weight the Democratic and Republican
components, thus awarding the highest partisanship scores to pundits
who most strongly support one party while simultaneously opposing the
other:

 PI = 100 * 0.5 * ( DI - RI )
 

When I recalculate last year’s results using this new formula, there are
significant changes:

Paul Krugman would still have finished first, but weighting his
Democratic and Republican components equally would lower his partisanship
score by 24 points, bringing Claudia Rosett and Collin Levey into
range. Frank Rich would have benefitted even more from the change,
dropping from 28 to 1 (!), because of his strongly negative treatment of
both Democrats and Republicans.

I’m going to think about it for a few days and then decide whether
to switch to the new method. Drop me some e-mail if you have any comments.