Via Pandagon I find an appallingly innumerate article on polls by Michael Barone:
Blogger Steven Den Beste has prepared an interesting chart. Den Beste charges that pollsters "deliberately gimmicked" the results, "in hopes of helping Kerry." I don't agree with that at all. But he has made another interesting observation. Eliminating some of the peaks and valleys of the Bush and Kerry percentages in realclearpolitics.com's average of recent polls, Den Beste shows that Bush's percentages have tended to rise over time while Kerry's have risen much less if at all.
He draws the Bush long-term trend line from a low point around 43 percent in May, when the media were full of stories about the Abu Ghraib prison scandal, to higher numbers around 45 percent in July and August, then up to the 49 percent level he has reached today. His long-term Kerry trend line runs through the 44 to 45 percent level in the spring to the 45 to 46 percent level in August, after the Democratic National Convention, to the same 45 to 46 percent level of today.
It seems curious that the percentages of the incumbent should rise while the percentages of the challenger have not risen much if at all. As a general proposition, you expect an incumbent's standing to change less, because voters already know much more about him than about his opponent. But that hasn't happened this time.
But what is actually rising is not support for the incumbent but the trend line Den Beste drew on his "interesting chart" (on left). Den Beste did not use any statistical procedure to work out his trend line, he just eyeballed the graph.
Underneath Den Beste's chart is one I just drew. My trend lines fit the data just as well as Den Beste's and show exactly the opposite---support for Bush has not risen much if at all, while Kerry's support has been rising. So whose support is increasing faster? To answer this you have to ignore the suggestive but misleading trend lines and look at the graph of the spread at the bottom of the chart. In March Bush was ahead by about two points, in October he was ahead by about two points. It follows that both Bush and Kerry have increased their support by about the same amount. Duh.
It gets worse. Barone explains the trends:
My tentative explanation is this. Bush's most effective opposition this year has come not from Kerry and the Democrats but from Old Media, the New York Times and the news pages of the Washington Post, along with the broadcast networks ABC, CBS, and NBC. Old Media gave very heavy coverage to stories that tended to hurt Bush---violence in Iraq, Abu Ghraib, the false charges of Richard Clarke and Joseph Wilson, etc.
Curse that Old Media and their Bush-hating facts!
And check out Barone's slant here:
The theory about the Democrats can be tested by looking at the 1996 and 2000 elections, when the Democrats were the incumbent presidential party in times of apparent peace and apparent prosperity
Gee, he can't even bring himself to admit that there was peace and prosperity under Clinton. It was only "apparent" peace? American soldiers were being killed by the score and nobody noticed? It was only "apparent" prosperity? All those people only "apparently" had jobs but they weren't real jobs?
Barone continues:
George W. Bush is not running this year as an incumbent in a time of apparent peace or, in public perceptions, a time of apparent prosperity. (Actually, the economic numbers are about where they were when Bill Clinton was running for re-election in 1996, but Old Media consistently report economic news more pessimistically when Republicans hold the White House than when Democrats do.)
First of all, "in public perceptions, a time of apparent prosperity"? Can anyone figure what this is supposed to mean? How does something perceived as apparent prosperity differ from something perceived as prosperity and how does it differ from apparent prosperity? Second, the economic numbers have been much worse than the first Clinton term, see here. And third, he seems to be channelling Lott's bogus study alleging bias against Republicans in reporting economic news.
It is disturbing that someone so innumerate as to not understand that a random line drawn on a graph does not make a trend should be interpreting polls for a major news magazine. I've never read anything by Michael Barone before. Perhaps his other columns are higher quality, but this piece was just dreadful.
My earlier post on Den Beste's chart is here.
**Update**: In comments and after the election, Den Beste asks who was right. Which trend line better predicted the polls up to the election? Look at the detail to the left.
More like this
You left a link on my blog to this post, because I linked and discussed Barone's article. I came over to see what you had to say. While I'm open to the possibility that Den Beste - whom I respect - and Barone - whom I respect - are both wrong, the narrative of your post is unconvincing. I see no methodology section explaining why Den Beste's line is a "random line" and yours is not. You said Den Beste did not use any statistical procedure, which implies that you did, but I don't see the narrative on your process. I have some statistical training, so I'm capable of assessing your method if you explain it. So please do so. As I said, I'm open to your correction, I just don't see yet any reason to believe you.
I did not say that I used a statistical procedure. I didn't and neither did Den Beste. I produced mine the same way he did his -- by eyeballing the graph. My trends fit the data just as well as Den Beste's but show the opposite result. My point is that trends produced this way are misleading and no-one should pay attention to them.
Michael Barone is the acknowledged historian of Presidential politics in the United States.
You'll be happy to know that my page with my chart has now been viewed more than 81,000 times.
1. An historian can still be innumerate, no matter how credentialed.
2. Does the number of page views establish the veracity or usefulness of the chart?
3. I can't believe I'm pointing this out. Am I missing something?
D
Tim, the points you chose to use for your trend lines don't balance. Den Beste's do -- or at least, better than yours.
"2. Does the number of page views establish the veracity or usefulness of the chart?"
Of course not. He's merely boasting that his (mis)information reaches far out. Which is the significant factor for some people, like think-tank people or their ilk.
McGehee, neither set of trend lines balances. But the secret to making one trend line increase while the other stays flat is to start them in different months. The shorter one can increase more steeply than the long one.
He's merely boasting that his (mis)information reaches far out. Which is the significant factor for some people, like think-tank people or their ilk.
Yes, sadly that was my implication, although I was thinking how sad it was that there are 81K rubes out there. And my 3. was more like a cry in the wilderness, which is why I appreciate Tim's efforts at identifying these people and their tactics.
Best,
D
Dano........"And my 3. was more like a cry in the wilderness"....... and you will continue to do so untill you understand that the foundations of the Left are slowly being consigned to the scrap heap of history. And pray tell, are you going to call 20% of the Black voters 'rubes' because they voted for President Bush ??
Who was closer to being right?
I was. Look at the graph on the left. This graph starts where your graph ends and shows what the results of the polls were in the two or so weeks after your post.