|
 |
Numbers For The Modern Man
by Paul McGoldrick
In this year of a US General Election (2008), we are being bombarded with numbers. The inability of the Democratic Party to decide on a nominee is well up there as a primary source of such volumes of statistics to hurl at us. Polls disagree with one another, and frequently with themselves, and the level of unknowns in the margins of error seems to go up and up. It would be a relief to know that people are simply lying to pollsters just to get rid of them, but there seems instead to have been a blind acceptance that polling has some kind of scientific logic.
That is rot to anybody who has been through the rigorous mathematics of statistics. Back in 1936 the Literary Digest made a prediction, based on their polling, that Alf Landon (how could any population elect an Alf?) would be the next President; they were totally wrong and FDR was re-elected. Their numbers were totally skewed by their distribution of survey ballots to telephone customers and registered vehicle owners – nearly 100% Republican voters. George Gallup took advantage of that gross mistake by starting his own polling company the next year.
But even Gallup couldn’t get it right some of the time: in 1948 the company classically predicted, and newspapers believed, that Harry Truman would lose the election to Thomas E Dewey. He certainly did not.
“Lies, damned lies and statistics” has been a quote attributed to a number of people. In the autobiography of Mark Twain (one of those said to have uttered those words) he clearly attributes them to British Prime Minister Disraeli.
More modernly, and how appropriate for an election year, the expression has been modified by one blog to "lies, damned lies and political statistics." This lays out, quite well, how different sites present their numbers mostly for their own best benefit.
We understand the conundrum of web statistics very well here at EN-Genius. Relating one set of honest numbers to another set is ridiculously difficult. We used to be able to see the bias of bingo card returns in a print publication and be able to justify why a reader in Romania would want information from every advertiser in the book. We cannot see the bias in why the busy design engineer, who actually specifies components, is not going to click through on a banner for a product that he does not need today. The metric has to change – but how?
This digital era is supposed to be the end-all message of how to count statistics. Every month I look at the numbers and marvel at the inconsistency of what they say. Most people who invest in the web have drawn lines, rather arbitrarily, of the value that advertising sponsorship will give them. Fall below those numbers and you are toast.
You can find, as politicians do every day, a set of statistics that will make your position acceptable, clear, tenable. But can you go back and show that it was so? I don’t think so. As the goal posts move to suit your position you have to eventually realize that individuals don’t always do what is statistically expected of them. Pollsters will say that their numbers are an instant in time, a photograph of what those who were asked thought of their candidates at that moment. It is to me more like a movie shot at a very low shutter speed – but someone drew the conclusions based on an average they could not see because they missed some important intra-frame changes.
Do you know anyone called by a pollster? Have you been called? Or maybe you missed that all important call because you were chatting up the young lady from Hyderabad who wanted to talk to you about getting more miles on our credit card? Yeah, go the extra mile, it always counts…
|
|
|
|
|
