Slashdot Mirror
Don't Read My Lips
scitex writes "Two economists, Rasa Karapandza and Milos Bozovic, from University Pompu Fabra, Barcelona wrote a very interesting paper "You Can Fool Some People Sometimes..." They say that on all US Presidential Elections from 1960, when presidential debates were held, won the candidate that used less future tense in the debate. First debate was televised in 1960. They predict that these elections will be won by George W. Bush. They used a similar approach to analyze performance of US companies and found that companies that use less future tense in their annual reports perform better. Will this change the way companies write they reports and presidential candidates speak?"
This story conjures up some clever side-conjectures, such as the notion that you can create statistical probability for virtually any scenario you desire if you keep researching until you find the data set that substantiates your claim. I wonder how many sets of criteria these guys churned through before they identified a pattern that jived with something significant? It sounds like a great gimmick to use to get grant money. 1. Pick an issue, formulate a premise, and then 2. start mining historical data until a substantive pattern arises which substantiates your claim. 3. Profit.
In this case, the future-tense references are kind of obvious if you ask me. People that talk about the future are often trying to detract attention from the present, and in those scenarios whether we're talking about a presidential election or a corporate report, tend to be reflective of whether or not the status quo is desireable or change is needed. Who needs a report to recognize this? If a company is doing badly, of course they're going to focus on the future. If a company is doing well, they'll be talking about their present circumstances. DUH.
As a result, I'm inclined to believe the outcome of this upcoming election is going to shatter these analysts predictions. In the business world, you can't really get away with lying about the present to the degree Bush has managed politically, and he's invented a new approach towards attempting to re-invent reality, which I'm not quite sure the majority of the public has bought, so his present tense propaganda is a completely different monster than any historical data these guys might have used for comparison.
I'd like to post this to suprnova but the upload page has timed out for me in the past 2 weeks. Maybe someone can try to put it on suprnova?
I ripped these older Presidential debates from various websites. It was a pain to get them. Dowload the torrent here or here or here
Low quality Real Video files ripped from streamings from the web.
If you have other debates, 1988, 1992, 1996, 2000, Presidential and/or VP, or higher-quality of the above, PLEASE POST!
is likely to give you many seemingly uncanny correlations between multiple events. The outcome of Washington Redskins home games prior to the election has predicted all the presidential elections since 1936. The sale of Halloween maks of the presidential candidates have predicted the elections since 1980. Does this bear any significance to this election? No. However, on the surface, this new prediction seems interesting. A person's word usage is a reflection of his character, education, and background. And there's been plenty of analysis of the word used by Bush and Kerry in this year's debates.
Yeah, well whatever. The thing that bugs me most about presidential candidates is that they always say "when I am president" and not "if I am elected president", even the ones who know they have no chance of winning (like Kucinich). I know their speech coaches or whoever tell them it makes them sound more assertive and confident and blah blah blah, but shut up already. NBC did this too, with their "MUST SEE THURSDAY" crap. It always sounded like a threat to me. Oh. Also, have they done an analysis of whether the person who repeats the same damn phrases over and over again the most is more likely to win? Because that seems to be the strategy of both camps.
And they all predict different things.
I've lived through several elections by now. There is always a tiny New Hampshire which always predicts the election since 1918 (until this time); always a sports team that wins when the Democrats win (until this time); etc, etc.
In the modern world, we have a lot of things we can keep statics on; therefore, we would expect to find impressive simple coincidences. If you line up 10,000 people and they all start rolling die, and leave as soon as they get the first non-pair, a day later their is going to be one sleepy and exhausted guy who just kept rolling pairs through the whole thing. Don't bet your savings that his next roll is pair, however.
You can use people's interest in coincidences to rip them off.
Get a list of 16,000 investors. Send them all a letter saying you have developed a super-secret algorthim that predicts the Dow; in 8,000 of the letters, offer the free prediction that it is going up next week, and in 8,000 other ones, offer the free prediction that it is going down.
Next week, toss out whichever 8,000 got the bum prediction. Send out 4,000 letters elaborating a little on your algorithm and predicting a bull market, and 4,000 predicting a bear market.
Soon you will be down to a handful very excited morons. Offer to sell them the next prediction for $10k a piece. Be sure to give half them one prediction and half another, so that their are some rich people who whorship you who will fund your defense if you get caught.
There are ways to do this that give you a pretty plausible defense.
For example, sell a neural network machine to stock traders. Tell them you sell a subscription to weekly updates to the neural weights, that you are using your Cray with it's secret sauce to compute. The manner in which you get away with giving different predictions to different customers is that you calculate a different set a weights according what stocks they are invested in and what amount of risk they are willing to take.
Then, if you get busted, you can plausibly claim you believed it yourself.
The above are true stories. The guy who did it with letters in the 1920s was knowingly committing fraud, and fled the country with a lot of other people's money. The neural network guys who did it in the 1990s may well have believed it themselves, or most of them did. I don't think anyone ever even publically called them on it.
Notice the recent scandal about drug companies commissioning dozens of trials of the effectiveness of a drug, and then only reporting the ones that said the drug worked ? Same principle. That's why you can't allow as evidence of effectiveness any trial that was not publicized before it was started; and you also must force them to pre-disclose the "protocol" or method of analyzing the data, because otherwise they will try out several different ways to divide up the data until they get what they want.
Statistics are getting to people again. Specifically with regard to the reference to corporations, the relation is probably inverse: successful corporations tend to use the past or present more, rather than that doing so will make them successful.
Marxist evolution is just N generations away!