Examining the entrails
Haruspicy is the practice of divination through the entrails of animals. The Roman priest who conducted the rite was an haruspex.
This is, of course, by no means the sole method by which omens of the future are to be descried. A look at references on the –mancy words will show a multitude: ailuromancy, the observation of how cats jump; bibliomancy, opening certain texts, such as the Bible, at random for guidance; nephelomancy, the observation of cloud formations; scapulamancy, interpretation of the cracks in the charred shoulder bone of a sheep — you get the drift.
As our interminable presidential election lumbers to its conclusion, I have come to suspect that that the network and cable news programs and newspapers might as well have employed haruspices from the start.
Item: The repeated assertion that one candidate is ahead of another by one or two percentage points, on the basis of a poll in which the margin of error is three and a half percentage points. And often, in a poll in which we are not told what questions we asked to which people.
Item: I hear cryptic references to a “poll of polls” on CNN and, turning to cnn.com, find that the “poll of polls” is an average of various polls. That anyone can take a set of polls conducted at different times, among different population samples and with differently worded questions to arrive at an “average” more reliable than the observation of a flight of birds astonishes me.
Of course, as a journalist woefully uneducated in mathematics, it may be my ignorance of the manipulation of these magic numbers that leads to skepticism. Anyone better informed about statistics and statistical analysis is welcome to chime in with comments.
Item: The perception-distorting electronic maps in use. All the gee-whiz special-effect maps still represent the states with their geographical boundaries. Redrawing those maps by their population sizes, or even Electoral College votes, would give the viewer a much more realistic sense of the voting dynamic. As I pointed out previously, New York City, with 8 million people, has a population greater than the states of Montana, Nevada, North Dakota, South Dakota, Utah and Wyoming combined. And while New York state gets two electoral votes for its senators, and the other six states get 12.
I am an American by birth, I’ve been following presidential campaigns since 1964 and voting in them since 1972. I understand how relentlessly they depend on slogans and catchphrases — the 1840 log-cabin-and-hard-cider campaign that elected William Henry Harrison being the dumbest of the lot and the pattern for many later ones.* I am aware that candidates of all parties and persuasions will stand in public and mechanically repeat misrepresentations of fact — and outright falsehoods.
But I am also a journalist, and I would like to see my colleagues across the media bring a little more knowledge and sophistication to coverage of these carnival events than is required to observe how cats jump.
* Before anyone starts questioning this allusion, I did not settle on the Harrison campaign because it involved an elderly retired military officer running as an “outsider” against an administration that had presided over an economic collapse. That is purely coincidental.
Comments
a “poll of polls”
Completely worthless garbage.
Posted by: owl meat gravy | October 22, 2008 11:16 AM
I have woefully little math myself, but one guy who does is Nate Silver. (He's one of the guys who revolutionized baseball through statistics like on-base percentage.)
He runs a blog, fivethirtyeight.com that has some stunning (and mind-numbing) collections of polls. He weights them by credibility, and he's building a very strong case that one CAN treat polls to predict outcomes--IF you do it the way he does.
Don't know if it's worth your time to check out, but I'm intrigued by it.
Posted by: Ben Kolstad | October 22, 2008 11:32 AM
"And while New York state gets two electoral votes for its senators, and the other six states get 12."
But then add in the Electoral Votes for the corresponding Congresscritters...
and yes, you can aggregate polls. Clinicians aggregate medical studies all the time.
Posted by: c o'donnell | October 22, 2008 3:11 PM
I am uncertain of the impact, primarily out of tiredness and laziness, but we must also consider that all of the states mentioned also get electoral votes for their Representatives in the House. In this respect, New York should at least balance the scale, if not take the lead.
We can look at New York and its problems another way. New York City has a predominance of the state's population and those city dwellers can tilt elections to the will of the dense population.
All of the focus on the national electorate is nothing more than a way to attempt to effect the election through turnout. The electoral result is the only game. It would be nice if the media would limit their reporting and projections to that.
Early voting, which has traditionally favored Republicans while being promoted by Democrats is another toll for effecting the outcome of the election. Prior to the historic nature of this year's election and the historic nature of the early voting turnout (before this year it was never reported that 31 states have it) we didn't hear that it was a Republican activity. now we know that it is a Democratic activity. The more it is reported, the more likely it is that Republican voters will forgo a trip to the polls. in fact, here in Maryland, the Election Administrator for Life has announced that turnout will be 85% and may be 90%. The clear message is that voting will be onerous, and more so for those who find standing in long lines difficult. So long elderly voters.
Of all of the unfair aspects of voting, the dishonesty, the attempts to ensure a result by those with the power or means to do so with money or voice (or television, or radio, or newspaper) the most unfair is the use of means to suppress voters from casting their vote, whether by a physical threat, as seen in Jim Crow, or by instilling despair.
Posted by: Bruce Robinson | October 22, 2008 3:36 PM
Alright. i overcame the laziness, but am still tired.
New York has 31 Electoral votes.
Montana has 3, Nevada 5, North and South Dakota each have 3, Utah has 5 and Wyoming brings 3 Electoral votes. [Source - http://www.electoral-vote.com/ ] (Confusingly drafted with purpose.)
New York, with its superior number of citizens has 10 more Electoral votes than the other states combined. Fair is, after all, fair.
Posted by: Bruce Robinson | October 22, 2008 3:41 PM
and yes, you can aggregate polls. Clinicians aggregate medical studies all the time.
Yes, they do. Meta-analyses, (the formal name for these aggregations of medical studies), can be a very useful tool for trying to determine if a finding is significant or not. But, as is true for all things, there are good meta-analyses, and there are meta-analyses not worth the paper they’re published on, (unless we’re talking about resume paper, that is). The issues John raises regarding sample population, study/poll timing, and the difference in the way the questions are asked are very serious concerns when evaluating such studies in medicine, just as they should be when looking at political polls. That last in particular is perhaps the greatest weakness in the very concept of a meta-analysis. Hence Mr. McIntyre’s italics.
So, can you average political polls and get a worthwhile result? It’s hard to say without knowing the details of each poll in the average and the statistical methods used to arrive at the aggregate result. But my admittedly cynical hunch is that many organizations are aggregating bananas, pears, strawberries and grapes and arriving at an average circumference of fruit.
Posted by: Abigail Carlson | October 22, 2008 4:49 PM