Monday, December 12, 2011

How to Lie with Statistics and Graphs

Correlation does not mean causality. Here's proof, unless you think Staten Island Cakes is behind Michelle Bachmann's decline in the polls and that Facebook is driving the Greek debt.

Writes Vali Chandrasekaran of Businessweek:
Need to prove something you already believe? Statistics are easy: All you need are two graphs and a leading question.

Click on chart for larger view.


It should be noted that Austrian economists have always been solidly in the camp of understanding that economics is a deductive science and that econometricains, who use complex equations to "prove" correlations, can be way off.

Wrote Ludwig von Mises:
There are, in the field of economics, no constant relations, and consequently no measurement is possible. If a statistician determines that a rise of 10 percent in the supply of potatoes in Atlantis at a definite time was followed by a fall of 8 percent in the price, he does not establish anything about what happened or may happen with a change in the supply of potatoes in another country or in another time. He has not "measured" the "elasticity of demand" of potatoes. He has established a unique individual historical fact. No intelligent man can doubt that the behavior of men with regard to potatoes and every other commodity is variable. Different individuals value the same things in a different way, and valuations change with the same individuals with changing conditions. . . .

The impracticability of measurement is not due to the lack of technical methods for the establishment of measure. It is due to the absence of constant relations. . . . Economics is not, as . . . positivists repeat again and again, backward because it is not "quantitative." It is not quantitative and does not measure because there are no constants. Statistical figures referring to economic events are historical data. They tell us what happened in a nonrepeatable historical case. Physical events can be interpreted on the ground of our knowledge concerning constant relations established by experiments. Historical events are not open to such an interpretation. . . .

Experience of economic history is always experience of complex phenomena. It can never convey knowledge of the kind the experimenter abstracts from a laboratory experiment. Statistics is a method for the presentation of historical facts. . . . The statistics of prices is economic history. The insight that, ceteris paribus, an increase in demand must result in an increase in prices is not derived from experience. Nobody ever was or ever will be in a position to observe a change in one of the market data ceteris paribus. There is no such thing as quantitative economics. All economic quantities we know about are data of economic history. . . . Nobody is so bold as to maintain that a rise of A percent in the supply of any commodity must always – in every country and at any time – result in a fall of B percent in price. But as no quantitative economist ever ventured to define precisely on the ground of statistical experience the special conditions producing a definite deviation from the ratio A:B, the futility of his endeavors is manifest.

12 comments:

  1. I see an inverse correlation between the number of pirates and the amount of global debt. So, if you ask me, the best way to stop the increasing debt is to enact a policy that increases the number of pirates. Arrrrr!

    ReplyDelete
  2. Any two monotonic time series are correlated - either positively or negatively. If you allow time shifting and/or time scaling, the list of available correlates (noun) greatly increases. The smoothing that you do to make visual graphs greatly improves the measures of correlation.
    Regards,
    Bill Drissel
    Grand Prairie, TX, USA

    ReplyDelete
  3. So "time shifting" explains how increased CO2 increases global temperatures even though the CO2 rises 30 years AFTER the warming.
    Hmmm. Guess those warmers have figured out a way to control time. They're so gosh-darn smart!

    ReplyDelete
  4. Wow, that was by far the funniest thing I have seen today (yeah, I need to get out more). Proof that the average American has no statistical education; one of the first things you learn in a statistics course is how to recognize bias in stats, and yet we have a huge part of our country that just believes whatever numbers and premise the media throws out there. "Correlation is not causation", one of my favorite phrases, because it is so simple yet it flies over so many peoples heads.

    ReplyDelete
  5. This doesn't work as easily in reverse.

    How many white swans do we need to observe before we know, "all swans are white" is true?

    1? 2? 100? 1,000,000?

    How many black swans do we need to observe before we know, "all swans are white" is false?

    1

    We should always be searching for reverse correlation, not positive correlation. That is we should *test* our ideas, not seek to confirm them.

    I've seen many arguments about why this doesn't apply to Austrian economics.

    I think the idea that we are building up from *a priori* axioms is particularly pernicious -- and even authoritarian.

    First, consider that a lot of ideas that goes into physics can't be explained at all. What is a *force* or a *field* precisely. We don't really know. Kant suggested that some of this stuff was built into our heads, and we couldn't think otherwise -- but note Einstein changed the rules here -- because he did revise at least some of this stuff, and it turns out it wasn't built into our heads after all.

    Second, consider that we need to know Einstein's theory of relativity *first* before we can even know what we are looking for. This isn't immediately obvious. There were "black swans" as far as Newton's theories of gravity were concerned -- and Einstein's theory accounted for them. But Einstein's theory also lead us to look places we would have never *observed* on our own without the theory. Who would have thought to look for a gravitational redshift:
    http://en.wikipedia.org/wiki/Pound–Rebka_experiment

    Einstein's theory of relativity said, look here, and you'll see this. We looked and indeed it was there. The point being -- in this sense, Einstein's theory was *a priori* to the observation. The observation was not even possible without it. All theories are *a priori* in this sense.

    What are generally regarded as *a priori* "axioms" by the Austrians gives us the sense that this is some authoritarian bedrock we must build all our assumptions on. But one can maintain the "axioms" while dismissing the authoritarianism.

    Instead, of saying these are self-evident truths, it would be better to look at them the way scientists do their basic assumptions -- as something that allows us to build up empirical theories.

    From the Austrian "axioms" we can build up theories that do *say* things about the world. If these theories were out and out wrong -- wouldn't we know it? So until the "axioms" just utterly fail to reveal to us anything about the world -- we accept them. So long as they continue to allow us to build up theories that show us where to look -- and when we do look we don't find black swans, then those "axioms" are showing us their mettle.

    That is, instead of saying we're building up from some unquestionable, authoritarian bedrock, we ask people to show us where we could possibly be wrong. If they can't, then we presume we're right for now.

    It's worth noting F. A. Hayek favored Karl Popper's methodology -- though certainly not Popper's early economic views.

    ReplyDelete
  6. Looking for empirical "proofs" in human action is a non-starter.
    I think Austrians accept "a priori" truths, because we have NO CHOICE. Ya gotta start somewhere.
    I think some really smart guy back in the 1700's said "We hold these truths to be self-evident". Maybe he was on to something...
    As a non-economist, I'm just sayin'...

    ReplyDelete
  7. @Capn Mike:

    The point is what happens if someone disagrees. Do you presume to be right and the other person wrong? Or do you discuss it to try and find the point of disagreement -- with the possibility in mind that you might be wrong and they might be right?

    We're each responsible for the truth. But if we presume to have finally gotten the truth because it's self-evident, then why bother to tolerate someone with a different opinion?

    I think if there's a disagreement -- you either can insist the other person accept your viewpoint or engage them in a dialog --- in which case, you ask, what is the problem you see with my viewpoint? What is your alternative?

    I see what are usually regarded by "axioms" by many Austrians as being really robust, it's not as if they're going to suddenly melt away unless we insist they are self-evident. I mean they are obviously not self-evident to everyone.

    I think at the time the declaration of independence was written, there was a sense that rights needed some type of justification -- and there was a loss as to what that justification might be -- so they resorted to we hold these truths to be self-evident.

    But the important point was that rights are not something that anyone can give or take away from us -- we have them independent of any authority. I would challenge anyone to try and find a flaw in that idea.

    But I'm sure no end of trouble has resulted from people thinking really bad ideas were self-evident, when in fact, they were just bad ideas.

    Interestingly, I think the Christian idea of brotherhood along with the impetus for Reason (with a capital R) is that we generally do see things the same way -- at least on some deeper level. So when there's a disagreement, if we work at it in earnest we can find it -- but at the same time it's wrong to presume we have the truth already because it's self-evident. It strikes me as authoritarian.

    W. W. Barley, III has a good essay on this here:
    http://www.the-rathouse.com/2008/Bartley1964CCR.html

    Also, for people who still think Popper was a positivist, I recommend this paper:
    http://fs1.law.keio.ac.jp/~popper/v7n1jarvie.html

    I think Rafe Champion is basically on to something here:
    http://www.the-rathouse.com/rc_popperpaper.html

    ReplyDelete
  8. @MAtt
    Wow, thanks for the generous reply. Even in disagreement.

    But I sense that YOU sense an element of coercion in praxeological methods. Hey, if it don't float your boat, let it be. It's descriptive, NOT prescriptive.

    Without SOME grounding assumptions, any dialog is meaningless. It would be like talkin' baseball and one of the group says: "Well, since the ball is made of cheese...". Oops, gotta call up that axiom about horsehide.

    P.S. And I WILL read your suggested material. Thanks.

    ReplyDelete
  9. Great blog entry, Wenzel. I have forgotten how many times I have had to state that "correlation does not equal causation" in the last several years in person and online...

    ReplyDelete
  10. @Matt

    We know we are dealing with an axiomatic truth when in the act of denying it, one confirms it.

    For instance, take the axiom of human action: "humans act," the premise from which the whole body of economic theory can be deduced. The idea here is that each individual acts with a specific end in mind. Human action consists in the choice of certain means to achieve a certain end -- in short, to substitute a more satisfactory state of affairs for a less satisfactory one.

    Now if it is your wish to deny this axiom as such, then it appears you do have a certain end in mind. To achieve this particular end, however, you must deny you have an end in mind. In which case, you slip into self-contradiction.

    So there's your *proof* of the axiomatic status of the human action proposition. You cannot refute it without contradicting yourself.

    I see nothing unquestionable or authoritarian in this. Anyone is free to question or deny the proposition. Of course, to do so is to confirm the very proposition you wish to deny.

    ReplyDelete
  11. As a great economist (who introduced me to statistical analysis) told me, "there are no straight lines, and no fixed relationships". Meaning, no matter how linear or correlated a relationship may appear to be, the lines will diverge over time.

    A relationship may APPEAR linear or correlated, just as the earth may appear flat to some observers. But given sufficient perspective and a clear view of the horizon, it is evident the earth is curved.

    Another example is the well-established relationship between incomes and housing. Before the 30-year mortgage, when people had to pay cash for homes the relationship between price and income was mostly linear. After introduction of 30-year mortgages, it became curved for a while after which a new plateau was reached. And when the toxic mortgage products became prevalent in the 90's and 00's, it went almost parabolic for a number of years. Then it crashed back without plateauing.

    The relationship remained between income and price remained, but the rate of change between the two variables did not.

    ReplyDelete
  12. Hahaha, the last one is just too funny.

    ReplyDelete