Last week, I discussed “The Big Lie goes viral” in terms of the causes of the financial crisis.

But that commentary begged the issue: What exactly is “Causation“?

This is a precise term of art that has a very specific meaning. Many of the world’s greatest thinkers, from Aristotle to Kant to Hume, have examined exactly what causality is. It plays a major role in the fields of Law, Physics, Logic and Economics.

To me, Causality is the relationship between one event (“the cause”) and a second event (“the effect”). I want to focus on the nature of that relationship. How and when can we say that one event is a consequence of another?

To assess how blameworthy any factor is regarding the cause of a subsequent event, I look at the following:

1) Is that factor proximate?

2) Is it statistically valid? Asked another way, does any data eliminate that factor?

3) Last, is that factor Necessary to the outcome? Is it Sufficient?

All of these elements are not necessary to demonstrate a cause and effect relationship; however, a lack of these factors is quite damning to claims of causation.

~~~

Discuss . . .

Category: Bailouts, Legal, Philosophy, Really, really bad calls

Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data and lack of respect for scientific knowledge. Be sure to create straw men and argue against things I have neither said nor implied. If you could repeat previously discredited memes or steer the conversation into irrelevant, off topic discussions, it would be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.

58 Responses to “Open Thread: What is Causation?”

  1. Tim says:

    “…does any data eliminate that factor?”

    Scientists perenially argue that >”correlation does not prove causation”<, and many times coincident occurrences are misleadingly concluded to indicate that because of x, y occurred.

    In questionable cases, in the end, causation can only be concluded impressionistically (Big Picture, BR!)

  2. Tim — that is exactly what I mean by “statistically valid” — is it coincidental or random?
    Is it merely correlated without any causation?

    Having statistical validity means it is not merely correlated . . .

  3. NoMiddleInit says:

    Do we need proximity? If the correlation is positive, and if necessity and sufficiency are satisfied, I’m comfortable that some effects may be delayed.

  4. PeterR says:

    Causation is not “Post Hoc, Ergo Propter Hoc.” [How's that for my high school Latin?]

    Did you drink mother’s milk as a child, and later smoked marijuana?

    Ban mothers (and their milk) therefore to solve the marijuana/drug problem!

    ~~~

    BR: I actually used the George Carlin quote (Mother’s milk leads to EVERYTHING!) on XMSR Potus today

  5. Factualist says:

    With respect to economic variables, there is the well-known Granger Sims causality test. I believe it was taken from the engineering field to identify what is causing what electrical flow to go in what direction. In economics, the test has been applied to see if changes in monetary aggregates “Granger-Sims” cause changes in economic output. It is a statistical technique you may find helpful to your endeavor. Randy P

  6. drewburn says:

    A little niggles, Barry. But I get the jist. Proximate? Is that clear?

    I just think a much more common sense explaination in there somewhere. T00 much academics there for my comfort zone. You make a point about the financial crisis; and you have already made it brilliantly, and very compellingly. I don’t think you need to beg the point.

    F them if they want to obfuscate the simple facts…………

  7. Asymptosis says:

    I tend to think that assertions of causation need a good story explaining the mechanics. Those mechanics can be complex, and each part of/step in that complexity needs to be explained, with those sub-explanations supported with validated evidence.

    We believe that the sun rising is caused by the earth turning. It’s a good story, but only because the whole mechanism is well explained, with solid evidence for all its parts.

    re: Proximity.

    While it can be suggestive, it’s problematic. The effects of El Nino are not terribly proximate, geographically, or even temporally. Ultimate causes matter too. The U.S. beat the U.S.S.R. because we were richer, more successful. But there were a lot of causes for *that.*

    “does any data eliminate that factor?”

    I think this is huge, and greatly underappreciated. Repeated demonstrations of no correlation constitute strong proof of no causation, IMO.

    Big example: The failure of scholars and scholarly journals to publish “no correlation” results — desk-drawer bias — is a tragic flaw of the academic publishing machine.

  8. streeteye says:

    Event B happens if and only if event A happens first A causes B.

    Bah, metaphysics!

  9. NoMiddleInit & drewburn

    The issue of Proximate cause is huge. How far are we willing to go back in time and space to assign a causal effect?

    If we are discussing global weather patterns/El Nino, then 1000s of miles and decades of weather cycles may not too far removed. If we are discussing foreclosures, I suspect we do not want to be nearly as terribly far removed.

  10. streeteye says:

    <=> got stripped (let’s see if it happens again)

    Event B happens if and only if event A happens first <=> ≡ A causes B.

  11. glengarry says:

    In the law, even in the case of fraud, proximity is necessary. Actors put a sequence of events in play, but we limit liability to what is proximate, or foreseeable, regardless of whether the initial action was the “but for” cause of the harm. That’s a societal decision, made and defined over time through the experience of many hard cases.

    We’re not in court, but I wouldn’t write an article assigning blame to actors for distant, unforeseeable results. Of course, knowledge and past experience help define what is foreseeable.

    The law also can apportion blame among many actors. So I might think that the CRA, Jim Johnson, Greenspan, et al do hold a lot of responsibility for the financial collapse. But that doesn’t dispute Barry’s ultimate point. Maybe that’s what his critics fail to understand? Or intentionally ignore.

  12. BusSchDean says:

    Often the problem is one of an “over-specified model” — meaning that multiplefactors are shown to positively and significantly correlate with an outcome. In this case teasing out the relative importance of any one causal factor becomes a challenge. The data Barry has presented thus far has a sequencing aspect (timing is always a component of causality). The presence of multiple factors happens a lot but of course some like to gravitate to their favorite factor (or boogieman).

  13. Observation: 5 elements that make causation a challenge for many of the Big Lie arguments:

    1) The Housing Boom & Bust was global, not just US based.
    2) Non Bank (non regulated) mortgage underwriting exploded from 2000-07
    3) Private label (non GSE) securitization market eclipsed F&F
    4) Private lenders utterly collapsed lending standards on their own
    5) The GOP controlled the Congress for 1994-2006 — they are not fans of banks giveawys to the poor.

    Can any of the usual arguments (F&F, CRA, etc.) seriously survive a challenge of these 5 bullet points? There are obviously many more, but — if a thesis cannot get past these 5 points, I think it will have trouble demonstrating causation.

  14. beaufou says:

    In this case (financial crisis), causality may be obsolete because of functional dependance. (low interests leading to bubble).
    The system has reached maturity (not psychologically) and operates in a closed circle incapable of fixing itself.

    I was actually thinking about this the other day; we could easily go back as far as Nixon in 1971 to find a cause.

  15. glengarry says:

    Asymptosis wrote: “I tend to think that assertions of causation need a good story explaining the mechanics. Those mechanics can be complex, and each part of/step in that complexity needs to be explained, with those sub-explanations supported with validated evidence.”

    I guess that’s the difference between science and belief (in the modern definition). I like the definition of science: a hypothesis which can be disproved through empirical testing. In a rough sense, the hypothesis that the gov’t or GSEs caused the subprime crisis has been largely disproved in the proximate sense through empirical data.

  16. theexpertisin says:

    I remain a skeptic on causation rationales, as the information bucket drawn from the well of all information leaves too many facts and other variables out of the container. This includes the massaging of statistics in that bucket of information.

    Does the presenter of a causation rationale include all relevant facts in an unbiased manner, acknowledge historical norms and human behavior as variables, and out personal agendas? In fact, there are likely unintended causation effects that override a concluded causation, because we cannot know everything. Can we?

    Good discussion post, BR.

    Carlin rules! What a comic genius he was.

  17. glengarry says:

    “5) The GOP controlled the Congress for 1994-2006 — they are not fans of banks giveawys to the poor.”

    I agree with The Big Lie but wasn’t a lot of Congressional fault the result of lack of oversight, as opposed to giveaways? Or worse, actively squashing attempts to regulate or even understand what was going on at F&F?

    ~~~

    BR: Yes, but that directly contradicts the arguement Bloomberg made, that Congress forced banks to make ill advised loans . . .

  18. RW says:

    Proximity is certainly important if you can but there are direct causes and those that work through intervening factors or by changing the state of a system so it could be difficult to demonstrate.

    Statistical validity would certainly be important but if necessity and sufficiency could be clearly demonstrated it would not be essential.

    What is really critical I think is the establishment of a connection between data and a claim about it — what Stephen Toulmin referred to as the warrant — this could be a model, a set of equations or what have you but the point is there is a mechanism that connects one event or a sequence of events to a consequence.

    The real problem with those who get causality wrong — assuming they are not purposely engaged in the spreading of confusion and ignorance (which they often are unfortunately) — comes from their data error or or errors in the steps necessary to establish warrant …or both.

    Doesn’t matter if it is purging-is-good-for-the-economy, government caused the crisis, etc, the folks who argue these things sincerely either make serious mistakes in the interpretation of data or they make serious mistakes in mechanism …or both.

    As for those who are debating dishonestly, obfuscating and sowing doubt for other purposes, then drewburn has it right, “F them.”

    The real task in considering Big-Lie arguments, since they allo clearly fail in warrant, is figuring out who might profit from correction of their error and who might just need to be told to go F themselves.

  19. glengarry says:

    To add to your observations, a lot of other types of sub-prime lending from the 90s went bad, and brought down financial institutions, in the late 90s and early 00′s. Like trailers, manufactured housing, used car lending: see Conseco, Green Tree Financial, etc.

  20. whskyjack says:

    Having been 30 years since I took any soft science stat class, I may be off in the term I want but I think it is “universality” ( Please some one correct me ) or does your causal relationship apply everywhere.

    I live in one of those zipcodes where the CRA had impact. I can see how it and other HUD programs may have contributed to a bubble but once you leave my neighborhood, not so much.

    Jack

  21. ilsm says:

    The conclusion must follow from the premises. Valid.

    The conclusion must be true if the premises are true and follow from the valid logic. Sound.

  22. Gibbon says:

    Physics, Logic and even Law simply should not be grouped with Economics. Economics is based on the collective decisions made by a now global population of humans, i.e. behavioral and to grasp those dynamics the underlying motives that drive decisions must be considered. At the very root of individual conclusions dwells self interest and greed, I think. I biological terms success measured in term of the survival of the species. That is the continuance of the genetic line. This plan is flawed in that it is exclusively quantitative. Be it a mold on bread of flies on a banana the outcome is predictably always the same. Either the food supply is depleted or it becomes so fouled it will no longer support the ever increasing numbers. Make more of your kind and hope some will survive and locate a new substrate. In higher species social dynamics come into play where individual compete to reproduce to increase their own particular set of genetic combinations in hope of them being furthered. That at its most basic is the origin of greed. From that can be derived an almost infinite variety of things to make individuals seem more attractive. –Economic drivers- Logic and even Law be damned. Sadly only in a few cases did some individuals have a glimpse of quality of life. Alas, astronomical distances being as vast as they are, there, is very little chance of our finding another planetary banana. “There ain’t no Warp Speed, Scotty.”

  23. PeterR says:

    Re: mother’s milk —

    Thanks Barry, George Carlin was way ahead of his time. Sorry for his struggle to wake us all up.

    I fear we face much rougher times, “teat-less,” if you would forgive a bad.

  24. hotei13 says:

    Another great George Carlin quote…… “Death is caused by swallowing small amounts of saliva over a long period of time.”

  25. formerlawyer says:

    I would think that the nature of the enquiry – explanatory, descriptive, statistical or blame affixtion would be part of the larger metaphysical model of assessing action/reaction/causation in various actors.

    “These uses of causation by historians, moralists and lawyers raise the question, adumbrated by Collingwood, of whether the attribution of responsibility requires a different conception of cause from that employed for prediction or explanation. In the legal theory of causation this problem is of central importance.”

    From: Stanford Encylopaedia of Philosophy (there are 348 entries when you search “causation”!)
    http://plato.stanford.edu/entries/causation-law/

    The basis of this exercise is to attribute blame and the granularity of the traditional plaintiff v. defendant enquiry contest of causation is probably not accurate. Further, the the proximity/foreseeability requirement which is as glengarry noted above is a policy choice to attribute consequences rooted in experience would complicate the use of trial models of causation.

    In these circumstances, I would suggest you could phrase the discussion as to causation as follows:
    (again from Stanford Encylopaedia of Philosophy & wikipedia)

    1. “[T]he identification of the causally relevant conditions of an outcome, or, in the language of causal minimalists, ‘cause-in-fact’.
    a. a necessary condition – “but for A then no B”;
    b. a sufficient condition – “A implies B”; (Note a sufficient condition is a subset of the following) or
    c. a necessary member of a set of conditions (A and B implies C).

    2. Relative causal strength of a condition – a measure of the strength of the implication
    a. Necessary condition – 100%;
    b. Sufficient condition – 0% to 99%; or
    c. Necessary member of a set of conditions – 0% to 99%.

    These measures of probability with 0% being statistically co-related but not casual and with 99% being extremely probable. In short this would involve a measure of “likelihood”, that is the implication is required for a condition, unlikely to affect the outcome or be more likely than not (ie. above 50%) to cause the result.

    3. Relative measure of conditions – a measure of responsibility.
    a. Necessary – but for condition X then consequence;
    b. Sufficient – given condition X more than likely consequence; and
    c. Necessary member of a set of conditions – given condition X, Y and Z then likely consequence.

    Without having societies to experiment upon, one can nevertheless derive useful information from a negative correlation ie. other countries did not have a CRA; timing eg. Fannie and Freddie were prohibited from purchase sub-prime loans until late in the game, market share ie. most of the sub-prime loans were taken by private lenders not subject to CRA and other measures.

  26. ToNYC says:

    Causation is most always the result of a Law of Nature working, whether you notice it or not.

  27. Bruman says:

    In a statistical sense, you generally say that concluding that A causes B requires:

    1) A happens before B (incidentally, this is the reason that information cannot travel faster than light, because the logical consequence is that this would permit A to cause B even if A happens after B).

    2) A correlation between A and B (although intervening variables/processes can magnify or suppress the correlation), so correlation – ironically – is not absolutely necessary (though it is necessary after controlling for relevant variables).

    3) a plausible mechanism linking A and B. Good research will go beyond simply finding a correlation and a plausible mechanism and actually go through something called “process tracing” which is an attempt to document the process in action.

    A complicating factor is that something can be caused by more than one thing. Did my coming home late cause my wife to cry? Maybe. It certainly did one time ago when she was expecting me for something. But maybe she cried for a different reason this time (like she got into an argument at work).

    In large scale economic and social phenomena, causality is difficult, because there are almost always competing causes and effects for any phenomenon, and they can be true simultaneously. The harder science types prefer to build explanations out of “causal chains”: A causes B causes C causes D, therefore A causes D. The problem is that the linkages between A and D depend on the strength of the correlation between A and B and B and C, etc., and even a little uncertainty in any one link starts to compound to pollute all the other links. And if you can break any one of the causal links, the explanation doesn’t hold.

    That’s why social scientists tend to build explanations that are not so much causal chains as they are causal “walls,” many factors pointing in the same direction. You can challenge or knock out one of the links, but the others still hold. Many traditional logicians feel that this is nothing more than throwing every argument but the kitchen sink at something, and there are certainly novices that do do it that way, but – appropriately constructed – these arguments turn out to be much more robust: based on ensembles of evidence, rather than a single causal chain.

  28. leveut says:

    1) Is that factor proximate?

    Comment: Proximity in time is an important factor for deterimining if something is a cause–the more remote in time, the less likely a thing is likely to be related. But foreseeability, known or should have known, is also an aspect of proximity. Something can be put in place, and yet be a cause of something that it is not proximate in time to. One could observe, for example, that the Glass Steagal Act was a proximate cause, in terms of foreseeability, of not having a crash likt 2007-8 until 2007-8. Similarly, something happening in the early 1990s could be a proximate cause, in terms of putting in place conditions, for a 2007-8 crash, even though it is not as proximate in time as things happening in the 2000s.

    2) Is it statistically valid? Asked another way, does any data eliminate that factor?

    That is certainly a reasonable thing to look at, but the problem with looking at that alone is to ignore the psychological and political components of causation that data, or, “data” don’t capture very well. One might look at any number of “analytical” studies using “data” that showed the banking system was fine in 2006 and the future was so bright we should wear shades.

    3) Last, is that factor Necessary to the outcome? Is it Sufficient?

    As a cause, necessary is certainly important. Sufficient is irrelevant unless one is arguing something is the sole cause. But, I think, one must be careful about what “necessary” means. For example, the realization that e = mc^2 was a cause of the bombs dropped on Japan in WW2–without that realization the bombs could never have been conceived of much less made. It was necessary. But it was not proximate in time, and although remote in time was necessary to the outcome, regardless of the brilliant physics and engineering that were still required.

    ——————————————————–

    1) The Housing Boom & Bust was global, not just US based.

    That certainly suggests that there were common causes, or a few causes that were common to the various global country busts. But to argue that all countries had the same causes is, I think, a difficult thing to sustain given the various governments and regulatory regimes. I think there is also a separate question that is overlooked, and that is timing. If the global causes were all the same, did they all occur contemporaneously? If not, why not? Did activity in one country “lead the way” and were there things in that country that were causes of the path initiated, with only the later acts adopted by other countries?

    2) Non Bank (non regulated) mortgage underwriting exploded from 2000-07

    Was there none before 2000? Was there not a Crash 1.0 in the 1990s in the US having many similar characteristics to the 07 disaster, aka Crash 2.0, Bigger and Better?

    3) Private label (non GSE) securitization market eclipsed F&F

    It did. But one must ask: were there actions by F&F that were part of “leading the way” for…poorly performed underwriting?

    4) Private lenders utterly collapsed lending standards on their own

    Sure. F&F were right in there with them collapsing lending standards on their own. Do you deny that?

    5) The GOP controlled the Congress for 1994-2006 — they are not fans of banks giveawys to the poor.

    The problem with that is that it ignores who controlled the Congress from 1987-1993, and who controlled the Executive Branch from 1993-2001. With control of the Executive Branch comes the power to appoint members of overlooking oversight agencies, to set policies that encourage or discourage bad things through regulation, to jawbone and browbeat the suckbutts who want something for their businesses that in return gives campaign contributions. In short , it is to ignore the power of the Presidency.

  29. victor says:

    BR: how in the heck do you manage to keep up with all of our comments and even respond to those -I suppose- more worthy of a rebuttal: some FAIL, some “Squishy Thinking” even a few “good point”? All the while going about your other professional and personal endeavors? Truly amazing!

    ~~~

    BR: A combination of cloning and Red Bull

  30. ShakyShot says:

    The “causal chain” discussed by Bruman (10:42pm) is the lens through which I view the guilty parties. The bottom of the chain is the root cause. The root cause is the bribing bankster bastards that through many years of effort captured (bought and paid for with assets or psychological influences) our Government and Mr. Praiseme Greenspan. Thus, the banksters are not only responsible for their own dastardly deeds, but, through exploitation, are also partially responsible for the bad deeds of this “second link” of characterless actors that should be removed from office. There were additional links and other independent or loosely related factors, but he banksters should be assigned the majority of the “cause”. They influenced themselves into enough rope to greedily hang themselves, and us with them. (In auto accidents, the drivers are some times judged to share the fault, with a determined percentage assigned to each. I assign the banksters 90% of the fault.)

  31. Julia Chestnut says:

    Jesus, I hated torts. I still can’t figure out if it was the teacher, the book, or all that navel gazing about water subsidence and cattle getting loose. But one important thing I learned is that there is a fundamental difference between whom we assign as the cause of the event and to whom we assign the fault. Sometimes they line up, but they need not: the importance of proximate cause in the law is a nod to the fact that an event has many creators – many circumstances have to work together, only some will be blamed.

    And that is a value judgment that societies make. I believe it also accounts for some of the heat behind the arguments regarding the “cause” of the crisis – denying causation is a perversion of where the fault lies. But frankly, I am less interested in the “cause” than the “crime.” Yes, a mark has to take the bait, with the lure flapping around on his lip, before the fraud can occur: but in most civilized communities, we don’t “fault” the mark. We fault the guy who went out with the intent to commit fraud.

    So they can always play logic games with what factors had to come together in lunar alignment for a small group of people to crash the whole financial planet: it’s a subterfuge, of sorts. The real question is who among the players broke society’s rules to the point where blame should be assigned. And if what they did was wrong in the eyes of most, and did harm – but DIDN’T break society’s laws – then the laws need to change if we’re not pleased with the outcome and the mindset that went into creating it.

  32. rootless says:

    Bruman, you wrote:

    1) A happens before B (incidentally, this is the reason that information cannot travel faster than light, because the logical consequence is that this would permit A to cause B even if A happens after B).

    Shouldn’t this be phrased the other way around? Our current understanding of the physical world according to which information can’t travel faster than light (pending the results from double checking the Neutrino experiments in Switzerland) is reason for our understanding of causality according to which the effect can’t precede the cause. Otherwise you postulate that the real world has to follow some philosophical normative about causality. BTW: There is a time-symmetric interpretation of quantum mechanics where future quantum states can determine preceding quantum states (http://dx.doi.org/10.1063/1.3518209)

    (So, contrary to the widely accepted meme perpetuated also by Barry Ritholtz, the interest rate policy of the Federal Reserve under Alan Greenspan can’t have been the main cause for the low interest rates in the mortgage loan market in the years before the real estate bust, fueling the real estate bubble by lowering the Fed Funds Target rate, if the decrease in the mortgage interest rates preceded the lowering of the Fed Funds target rate.)

    2) A correlation between A and B (although intervening variables/processes can magnify or suppress the correlation), so correlation – ironically – is not absolutely necessary (though it is necessary after controlling for relevant variables).

    The presence of a correlation between A and B is a necessary condition to infer causality between A and B, although it is not sufficient. The correlation doesn’t need to be 1, but Zero correlation implies no causality. To say the presence of a correlation wasn’t “absolutely necessary” is meaningless. There is no relative necessity here. If the presence of X wasn’t required in all instances, then X just wouldn’t be a necessary condition.

  33. BusSchDean says:

    Another telling point about identifying causality — once you nail it critics tend throw verbal tomatoes and eggs but no real data analysis (back to that old thing about attacking the messenger). While I haven’t monitored closely the comments to Barry’s “The Big Lie” I have not thus far read a thoughtful data analysis supporting a counterargument.

    It would be so nice to have Da Mayor publicly admit that upon re-examination of the available information (his bread and butter) he cannot help but place a great deal of responsibility at the feet of a credit-driven banking and mortgage industry. Or is a person in his position and with his reputation unable to do that?

  34. rootless says:

    ShakyShot, you wrote:

    The “causal chain” discussed by Bruman (10:42pm) is the lens through which I view the guilty parties. The bottom of the chain is the root cause. The root cause is the bribing bankster bastards that through many years of effort captured (bought and paid for with assets or psychological influences) our Government and Mr. Praiseme Greenspan.

    If “bribing bankster bastards” were the “root cause”, they just appeared out of thin air or from outer space and usurped the financial system? I don’t think that attributing the root cause of financial and economic crises, booms and busts, which have occurred again and again since capitalism came into existence a few hundred years ago, to the character flaws of a group of people which is declared the “guilty” one really contributes to an analytically based understanding of what is going on in the system. What caused it is the better question than who caused it. Behavioral patterns of people under given circumstances are just a symptom.

  35. budhak0n says:

    Ah , always the statistician. You know there comes a time when even the greatest Physicist needs to step outside his equations and reflect upon his conclusions.

    Absence of causal connections through a series of well detailed and thought out arguments is convenient.

    And gives us all a fresh start every now and again.

  36. phb says:

    Damned butterflies.

  37. To truly determine if a causal effect is valid, does the removal of the cause reverse or negate the effect? This is a back-test we use to determine if, for example, rising oil prices cause airline stocks to sell off. The same can be applied (albeit with more difficulty) to a question like “Did unregulated underwriting practices cause the housing bubble?” Isolating other factors is the tricky part but if you look at a trend up to the date of a change in regulation, for example, and see an otherwise unaccounted for deviation from the underlying trend – that’s pretty good evidence. Then the real trick is going back in history and finding other examples to back you up.

    With stocks it’s easy, things go up and down all the time which affect each other – working with socio-economic data is much trickier but that, Barry, is why I read your column every monring – I love your work on these topics.

    - Phil

  38. rootless says:

    Philstockworld, you wrote:

    To truly determine if a causal effect is valid, does the removal of the cause reverse or negate the effect? This is a back-test we use to determine if, for example, rising oil prices cause airline stocks to sell off.

    Only if everything else was kept equal, or if there is only one cause that could have the specific effect to be explained. So, if removal of the cause eliminates the effect it validates the causal relationship. However, if removal of the cause doesn’t eliminate the effect it doesn’t necessarily invalidates the causal relationship. Besides rising oil prices, there are other possible causes that could lead to a drop in airline stock prices.

  39. BusSchDean says:

    Barry, you know your argument holds water when your critics can only take vague shots at it. While I have not followed all the comments to “The Big Lie,” I have yet to see a solid counter-argument based on data analysis.

    It would be great if Da Mayor could step down from what my mother would call “his high horse” and publicly state that under closer examination of the data (his presumed specialty) the banks did indeed play a critical role independent of gov’t action in creating the credit/mortgage bubble.

  40. FrViper says:

    Believe the logic term was “modus poniens (SP on memory)” where A implies B and B implies C, therefore, A implies C.

  41. budhak0n says:

    @BusSchDean. A solid counter argument based on “data analysis” isn’t necessary. This is America.

    Hey Wyatt, why’d you shoot that no good calf rustling whiney little sum bich ?

    Wyatt’s response, …

  42. moody says:

    Another way to look at causation is that you need a model of the system that has the causal linkage in question. Things get a bit complicated with such a model. To the best of our knowledge, we are dealing here with both a complex and chaotic system. In that case, the issue of proximate is not relevant ala the butterfly effect.

    And in a complex model, multiple causes can result in the same behavior. So the issue of necessary and or sufficient becomes questionable. Furthermore, in such models, the behavior in question can be the result of interacting causes rather than a single cause.

    If you are applying something like a System Dynamics approach you have ways of getting at causality in a whole different way.

  43. efrltd says:

    John Stuart Mill wrote the definitive discussion on causality, “On Causality,” I believe the section I recall was called. Your standards are far from his principles. Secondly, statistics cannot establish causality. Statistics is most post hoc, ergo proctor hoc reasoning without the precursor, establishing a priori causality. Statistics help sort the wheat from the chaff to confirm the causality with the data. Statistics scrap the plague from the data to leave the skeleton that links the cause to the effect. However statistics only work well in carefully controlled situations. Almost no economic or financial statistics can fit a carefully controlled situation. Most statistics one sees “in the wild” are simple number crunching to map a terrain, but not the paths and roads from one point to another. And such casual empiricism is fraught with myths, old wives tales, prejudice and hunches, often lacking much insight. I find it astounding the logic of recent arguments about banking–little changed from Jefferson, Jackson, Wm. Jennings Bryan and many others. All that is stripped from the days of the German and American Progressives and Populists of the 19th Century are the ethnic pejoratives. Still to be explained is why bankers could so easily enable “the righteous consumer,” or the “hardy homesteader,” had so little sense so as to fail to live within their income and to, with a bright smile, sign a Faustian mortgage on his life, if not his soul. Without insight that the debt problem has bidirectional causality, there will not be an understanding of the economics of depression. The borrower, as well as the banker, are both codependent, enabling one another to reach for the gold ring. Ranting that some banker should be jailed ignores the borrowers standing eagerly at the door. Some years ago in Iowa, a grain elevator operator came up with a deal too good to be true for some local farmers, a few cents a bushel more but you have to wait for your money and not get a warehouse receipt. Grain prices collapsed, the elevator couldn’t pay, a suicide and a big scandal. And the farmers, co-conspirators, suddenly became the victims. Co-conspirators who forgotten the one rule–don’t hold anyone’s paper. Someone else might get to the bank first to cash it and you’re out, back to the back of the line. The point–the causality flows from the banker (elevator) to the borrower ( farmer); as well as from the eager borrower (farmer) to the banker (elevator.) Nobody made home buyers flip houses, buying bigger and better and shakier and shakier terms with less and less cushion. I saw that going on even in the more stable heartland in a town when developers bought more and more land to build on. Now one sees corn and bean fields going back in among vacant homes and condos. Both sides, lenders and borrowers, include tricksters trying to get a leg up, not just the lender side. Enough said, sorry to go on, but a serious point that is being often skirted, the conspiracy of home buyers with lenders.

  44. AnttiK says:

    The problem is not if it’s correlated or saying that is must be causated.

    The problem in real-life is analyzing what/when is a causal link between A and B, if any?

    That’s the field of bayesian inference and it can only give you probabilities with margin of error in most cases.

    And that’s after huge numerical computation runs filtering out variables and simulating sensitivities.

    And for that you need lots of source data, to confirm the models.

    Basic frequential statistics give you very little tools to do this. They are mostly just more elaborate versions of correlation measures, not studies in statistical causation.

  45. DeDude says:

    To prove causation you must have correlation in the absence of any change in confounding variables. Since social and economic sciences are not able to control all other potentially influences (confounding variables) it is almost impossible to make solid scientific proof of anything. The next best thing is modeling. You make a model that takes into account all the possible influences of everything you could imagine influencing the parameter of interest; and then you use all available observations to fine tune the model so that it gives the best possible explanation of it all. The models are not perfect, but they generally are much better than any of the alternative (i.e., stupid single variable cherry picked examples, or pulling something out of your ass). Problem with models are that most people simply cannot understand them and, therefore, refuse to accept that they are better at approximating the truth than above mentioned alternative approaches (global warming being a poster boy example of that).

    A real good model will give you all the parameters that have causal effects and the ability to estimate how much causation each of these parameters can be assigned in a given scenario. In the absence of a good model your 3 questions are a probably a good way to evaluate causation. At least it should capture the worst cases of mistakes from the above mentioned “alternative” approaches.

  46. Moe says:

    I think what’s playing in the media can be a good reflection on ourselves:

    Godzilla has been acknowledged as a reaction (in Japan) to the nuclear age of destruction…perhaps in the U.S., our current fascination with “zombies” and “werewolves” is merely a reflection of an apathetic/ineffectual public…or maybe I need to switch to decaf…

  47. Rootless – All else equal is, of course, impossible but if we know the marjor factors that tend to influence sector pricing and then we can find instances in which one key factor (oil prices here) moves while the other factors remain neutral to the last session, then we can begin to isolate the effect. Obviously it’s a tricky route to navigate but where we find it most helpful is to help determin IF we should consider the isolated “cause” a major or minor factor over time. I tend to give weightings to various causal relationships so, for example, oil may be a 30% factor while business confidence is 20%, currency exchange rates 15% and utilization is 35% – that sort of thing.

  48. DMR says:

    Real world example of why correlation does not equal causation:

    http://tosh.comedycentral.com/blog/2011/11/16/jesus-on-a-dog-butt

  49. Jim67545 says:

    I don’t view the housing situation as a clear linear causality. There are, in my opinion, two types of factors: 1. Static complicit factors: To use an analogy, if you have a streambed which has rockfalls, boulders, etc. the flow is somewhat controlled. But, when you clean out all those impediments and the rainstorm comes, the result is more severe. Did the boulders cause the flood? No. But removing them reduced natural control and worsened the result. This is where low interest rates, repealing certain regulations, securitizing, implied government guarantee, CRA and the Alt-A and other programs come in.
    The second factor active factors: in the analogy is the rainstorm. We debate what stage the stock market is in. As the market rises investors notice, convert other assets to stocks, they pile in in greater numbers and to a greater extent. It’s a frenzy, feeding frenzy. That is what occurred in housing. Returning to the analogy, it’s like every molecule of water learned that this stream was no longer protected by boulders and the quickest way downhill is to go over to that stream. This is where the growth of private MBS, relaxed underwriting, homebuilder’s captive finance companies, subprime loans, etc. comes in (as originators flow toward that opportunity to expand volume and profit.)
    So, what am I saying? Causality is important. However, certain causes (especially human greed) led to a situation where effect reinforced cause, a feedback loop, and the situation when verticle. Much as the stock market can do.

  50. Bruman says:

    Rootless,

    There can be intervening effects that suppress observed correlations. If A is positively related to C, but it is correlated to something else (B) that has a negative correlation to C, then the correlation between A and C can be very low or even zero, because the effect of B on C “suppresses” the effect of A on C. Not surprisingly, this is called a “suppressor effect.” If you rerun the analysis and look for the partial correlation of A and C, controlling for B, you should get a positive correlation, so you are right that there should be something, somewhere, if there is causation. My point is that many times, what statisticians would call the “unconditional correlation” can be very low, even if there is causality, and so the correlations that you observe (without doing the controls) are not always as obvious as you think they would be. A can cause C, but you may have to control for the right factors in order to see a correlation.

    As for the relativity issue… the key idea is that the laws of physics (including laws of causality) are supposed to be the same in all inertial reference frames. If information can travel faster than light, then the consequence turns out to be that there is some slower-than-light inertial reference frame in which A causes B, even though B happens before A. Physicists basically had to choose between giving up on the idea of causality or giving up on the idea of traveling faster than light, and most chose the latter. This is why the suspicion that there were FTL neutrinos caused such a stir. My understanding is that they have since found a source of error that explains their results without requiring FTL neutrinos.

  51. [...] back and forth in last night’s discussion of what Causation actually [...]

  52. rootless says:

    Bruman,

    Thanks for clarifying the issue with the correlation. I can agree with this.

    As for the neutrinos. No, it’s not clear yet whether there is just an error with the measurements. There is a suggestion that there could have been a problem with the measurements, with the lengths of the proton bursts generating the neutrinos, which were much longer than the measured shift in the arrival time of the neutrinos. New experiment are being done with shorter bursts. I don’t have any information that the results are out already.

    See following blog by Matt Strassler where one can find some educated commentary:

    http://profmattstrassler.com/2011/10/26/a-few-tidbits-from-nagoya-including-opera-news/

  53. murrayv says:

    Haven’t read all of the comments, but have at least one other point to make re causation. You also have to distinguish between enablement and trigger. For instance, deregulation and slack regulatory enforcement enabled the credit bubble. Mark to market triggered the collapse of Bear Sterns and what followed.

    Barry, you wrote the other day of cognitive dissonance. You continue to believe in anthropogenic global warming, and yet it completely fails your tests of causation. Interesting.

  54. rootless says:

    murrayv, you wrote:

    Barry, you wrote the other day of cognitive dissonance. You continue to believe in anthropogenic global warming, and yet it completely fails your tests of causation. Interesting.

    How so? Please could you elaborate how you come to this bold assertion? If “believing” in anthropogenic global warming was an example of cognitive dissonance, as you apparently claim, what are the data and results from scientific research that allegedly invalidate anthropogenic global warming, which are being ignored and replaced by wishful thinking on the side of the “believers”? Please do tell.

  55. Taliesyn says:

    beaufou Says:
    November 16th, 2011 at 8:31 pm
    “I was actually thinking about this the other day; we could easily go back as far as Nixon in 1971 to find a cause.”

    Well my paying attention to current events didn’t get serious until the 2nd Oil Shock and all of the economic havoc that triggered. That’s why I feel a kind of empathy for the kids graduating university being dumped into an awful job market due to an economic mess they didn’ t cause , just inherited. The post Iranian Revo Oil shock / Runaway Inflation /dawn of the Reagan era wasn’t a pretty job market to graduate into either.
    I take the point about the Nixon era having to ultimately take us off the gold standard and thus allow the dollar to float ( I learned due in no small part due to the returning LBJ debt from Viet Nam & Great Society overspending ) alloyed with that 1st Oil shock thanks to Quadaffi’slybian standing down Big Oil except Armand Hammer’s Occidental blinked. I never met a Big Oil company that didn’t love a reason to raise prices and we had the hyper-inflation run rampant , bookended by that 2nd Oil Shock that really altered our economy.
    So , like Bush,jr having a major crisis on his economic & then international in his face , Reagan Era economic & financial gurus ( including Greenspan ) had an opportunity to try the “Let the Marketplace Go” true religion along with finding a reason for exponentially bloating *Deficit Spending*. David Stockman called it and was caused to leave his OMB station.
    So , in as tight a nutshell as I could compose , that’s where I see a *game-changing* causality of degree as well as implemented & executed ideological idolatry that inevitably led to policies& practices in both public & private sector that led to this , only the most recent economic debacle.
    BTW: This is just crying to be composed as an economic history parody based on Billy Joel’s “We Didn’t Start the fire”
    Anyway I just deleted my list of choices for pivotal causal moments in order to spare you all because I’ve already written my particulars of outrageous failures. It’s just not yet edited down to that level of pith-worthiness ripe for setting to Billy Joel’s music.

  56. AHodge says:

    A few thots
    1.Economics especially financial markets and policy is not a science like physics or medicine. its all highly tentative speculative. See any scientific agreement on what happened in 2008? No.

    2 econometrics has debated cause for a couple centuries, search econometrics/causation/correlation and read till your bored to death. Here is one summary finding
    QUOTE The basic idea of a causal model in economics traces back to the work of Mill (1854) and Marshall(1890) who verbally define causality in the familiar ceteris paribus fashion.1 X causes Y if, holding everything else constant, manipulations of X change the value assumed by Y. This intuitive idea was formalized by Haavelmo (1943) UNQUOTE this still looks some like correlation there it is

    3 causal modeling can be a partial success in macro if done right, (you can also do micro like what factors are driving your widget sales)
    ITS PRETTY MUCH A FLOP IN FINANCIAL MARKETS .

    4 Soros, frustrated by this 40 years ago, dug back into the what we know, studying Popper etc, and the behavior of markets
    And” proved”,( to me anyway) in the Alchemy of Finance, that you cannot predict markets
    meaning price or provable cause
    because among other things the actors are reacting to the price.
    Reflecting his strange sense of humor or lack of it is the title “alchemy”
    not science or chemistry as he explained it, but made up by the particiipant’s beliefs….

    At some level we cannot know causes and the complete clockwork of the universe. If there is one. Quantum mechanics and soros suggest not. so you could go that route and say WTF we dont know shit……
    but im willing to workd with a thighly tentative and humble lack of complete knowlege,
    so to my knowlege, is anyone else who has lasted trading
    Its all highly tentative assumptions. But as you say easy to disprove various obvious BS useless theories of which the Big Lie is one.

  57. murrayv says:

    Hi Rootless
    First the cognitive dissonance I was referring to was applying the causatioon tests to eg the housing bubble, but not applying the tests to AGW.
    The tests are:
    1) Is that factor proximate?

    2) Is it statistically valid? Asked another way, does any data eliminate that factor?

    3) Last, is that factor Necessary to the outcome? Is it Sufficient?

    There is no hard evidence that the factor is proximate in an open system like the atmosphere. even if the factor wrer proximate the response of temperature to increasin CO2 concentration is logarithmic and doubling CO2 now would only increase temperature 0.3 degrees C maximum. not the 2 to 7 degrees the catastrophists claim. The catastrophic increase comes only from models, and they require a fudge factor to get there. The fudge factor is positive feedback via water vapor. There is no scientific evidence for such feedback, but there is both theory and some (albeit not overwhelming) evidence for negative feedback via cloud formation. The feedback fudge factor clearly fails the proximate test.

    Temperature declined sharply from ca 1944 to 1953 and then bumped along the bottom to ca 1975 while CO2 concentration increased almost monotonically. Since Q4 1997 temperature has declined slightly although CO2 continues to rise, even faster than from 1944 to 1975. Clearly the correlation is not statistically valid for these periods.

    The models apply a fudge factor of aerosols to explain the 1944-1975 period, but there are no papers that quantify the aerosols, identify their source(s) or describe why they ceased to be effective after 1975. Solar variation (the clearly present ca 60 year cycle) explains both of the anomalous periods. Neither positive feedback nor aerosols are necessary nor are they sufficient.

    Finally the surface instrument data is not adequately corrected for all of the warming biases that are known to exist even for HADCRUT, and the NASA GISS data has been severely manipulated to get the last decade to be warmer than 1934-1944. I can send you several pages of urls on biases and manipulation. You can go to http://www.agwnot.blogspot.com for some treatments of solar cycles, and review the folloawing for GISS manipulation.
    http://wattsupwiththat.com/2011/11/05/ncdc-data-shows-that-the-contiguous-usa-has-not-warmed-in-the-past-decade-summers-are-cooler-winters-are-getting-colder/#more-50527 . there is a lot of data that eliminate4s the statistical validity of AGW.

    All tests failed. Murray

  58. rootless says:

    No one will read it anymore now, probably. But I’m going to reply anyway for the record.

    murrayv, you wrote:

    First the cognitive dissonance I was referring to was applying the causatioon tests to eg the housing bubble, but not applying the tests to AGW.

    How would you know where Barry Ritholtz has applied the causation test, and where he hasn’t?

    There is no hard evidence that the factor is proximate in an open system like the atmosphere.

    Can’t proximity exist between a factor A that is thought to be a causing one and an effect B, if the system is an open system? Then it would be true for any relationship between any causing process and effect. What about “solar cycles” then? Why would that be? Sounds to me like an absurd claim. Proximity means that A and B are close enough in space and time so that cause and effect can be mediated. Greenhouse gases have increased since pre-industrial times, the globally averaged surface temperature has increased since pre-industrial times. So, proximity in time is given. Both processes occur in the same physical system, the atmosphere. Proximity in space is given. What evidence is lacking for the proximity? That the atmosphere was an open system is just not a relevant fact for the argument here.

    even if the factor wrer proximate the response of temperature to increasin CO2 concentration is logarithmic and doubling CO2 now would only increase temperature 0.3 degrees C maximum. not the 2 to 7 degrees the catastrophists claim.

    Where does this denialist claim about the 0.3 degrees come from? From any scientific research? Or where from? And how would you know this number was right, in contrast to what the results from in peer-reviewed journals published studies of mainstream science are?

    The catastrophic increase comes only from models,

    No, it doesn’t. The scientific statements about the expected temperature increase for an increased CO2-concentration, all other climate drivers staying the same, are also based on observations. On measurements and paleo-data. Your claim is not based on facts.

    and they require a fudge factor to get there. The fudge factor is positive feedback via water vapor.

    It is news to me that the positive water vapor feedback simulated by the climate models was introduced into these models by a “fudge factor”. Please could you reveal the source of your claim? Or did you make this up yourself? Perhaps you can show me where I find this alleged “fudge factor” in the NASA GISS Earth System model ModelE? I would say this claim by you is also not based on facts.

    There is no scientific evidence for such feedback,

    There isn’t? Are you saying there isn’t any scientific evidence for the validity of the Clausius-Clapeyron equation, from which directly follows that a volume can hold more water vapor with increasing temperature before saturation occurs? Are you saying there isn’t any scientific evidence that water vapor is radiatively active in the thermal range of the radiation spectrum? Both establishes important aspects of the theoretical physics basis for the water vapor feedback. If you answer these two questions with “Yes”, you would have the burden to explain on what grounds you deny the validity of what is generally accepted textbook knowledge in thermodynamics and radiative transfer theory.

    Are you saying there isn’t any evidence from real world data for a positive trend of the water vapor content in the atmosphere? This claim wouldn’t be based on facts again, since there is also evidence for such a trend from reanalysis:

    Dessler, A. E., and S. M. Davis (2010), Trends in tropospheric humidity from reanalysis systems, J. Geophys. Res., 115, D19127, doi:10.1029/2010JD014192.

    but there is both theory and some (albeit not overwhelming) evidence for negative feedback via cloud formation.

    The mere existence of negative feedbacks doesn’t invalidate AGW, or any of the criteria above. No one says there weren’t any negative feedbacks, involving clouds or others variables. So just stating the fact that there were such feedbacks isn’t sufficient to make this a relevant argument for the question at hand.

    Since Q4 1997 temperature has declined slightly

    Show me the data. Your statement is un-scientific nonsense. I equally could claim that the temperature has increased quite a lot since Q4 2000. It wouldn’t be right either, though. Scientifically valid statements about a trend can’t be based on two cherry picked data points, and even if a trend analysis is done with all the data, the results needs to be statistically robust, which is not the case for such a short time scale, to make a valid statement.

    although CO2 continues to rise, even faster than from 1944 to 1975. Clearly the correlation is not statistically valid for these periods.

    Since no one says that CO2 was the only climate driver that had an effect on the global temperature so that there would be a linear correlation, and no one says that a correlation needs to be found on every time scale, even on short ones, not finding a linear correlation only disproves a straw man argument that is your own invention.

    The models apply a fudge factor of aerosols to explain the 1944-1975 period

    The same as for the water vapor feedback. What “fudge factor” where in the models, e.g., in ModelE, and what is the source for your claim?

    but there are no papers that quantify the aerosols, identify their source(s)

    Again, a baseless, non-factual claim as following abstract should be sufficient to show:

    http://www.giss.nasa.gov/meetings/pollution2002/d2_koch.html

    One also could look in the IPCC-report of 2007 to find a compilation of scientific references for the effect of aerosols on climate.

    describe why they ceased to be effective after 1975.

    Of course there aren’t any papers for this, since no one claims that aerosols “ceased to be effective after 1975″. You just have made up your own reality, again.

    Solar variation (the clearly present ca 60 year cycle) explains both of the anomalous periods.

    How would you know this is the case? Just because you assert this? This assertion clearly contradicts the findings of mainstream science, doesn’t it?

    Finally the surface instrument data is not adequately corrected for all of the warming biases that are known to exist even for HADCRUT,

    Again, a claim for which no evidence is provided. What is the source for this assertion? Any scientific publications? Or just the assertions made by other pseudo-skeptics, which they tell each other in their opinion blogs?

    and the NASA GISS data has been severely manipulated to get the last decade to be warmer than 1934-1944.

    I call such an accusation libelous against the GISS scientists involved in the analyses. There is no basis in reality for such an accusation. Although it is a known strategy applied by the denialists to defame the scientists on personal grounds, because the scientific findings aren’t liked, but can’t be refuted on scientific grounds. It’s a giveaway where the phenomenon of cognitive dissonance can really be found. And no, citing the same assertions by other denialists won’t qualify as proof.

    You can go to http://www.agwnot.blogspot.com for some treatments of solar cycles,

    What could an opinion blog written by a pseudo-skeptic layman offer me to learn about solar cycles? I have higher standards for my sources of knowledge than what can be found there.

    and review the folloawing for GISS manipulation.

    Yet another pseudo-skeptic opinion blog.

    http://wattsupwiththat.com/2011/11/05/ncdc-data-shows-that-the-contiguous-usa-has-not-warmed-in-the-past-decade-summers-are-cooler-winters-are-getting-colder/#more-50527

    What is the evidence there for the alleged manipulation? I don’t see any. The fact that data used and methodology applied have been revised in the course of time, which led to slight changes in the results of the analysis? More data become available in time, sometimes data are corrected, methodologies are improved, sometimes scientists also make mistakes, which are being corrected. All this is a normal part of the scientific process. You don’t seem to think so.

    And that the temperature in the contiguous US hasn’t increased over the last decade doesn’t invalidate the global long-term warming trend due to anthropogenic greenhouse gases either. It’s just another pseudo-skeptic straw man argument. The contiguous US cover only about 1.5% of Earth’s surface. To conclude from the trend in US about the global trend is a logically fallacious generalization. So is basing a conclusion about the global warming trend on about only a decade of data.

    there is a lot of data that eliminate4s the statistical validity of AGW.

    I asked you for the data and the findings from scientific research on which you base this assertion. You haven’t provided any. You only have delivered the usual mix of straw man arguments, arguments that lack relevance, assertions that aren’t based on facts, and libelous accusations of forgery against scientists. But no data that contradict the causal relationship between CO2 and global temperature, and no scientific publications.

    All tests failed. Murray

    Only on your side.