Thursday, September 22, 2011

Has Steve Williamson proved that our Universe is non-causal?


Steve Williamson has written a review of John Quiggin's Zombie Economics that is based neither on flippancy nor personal insults. Kudos to Steve! I encourage everyone to read the whole thing, which also serves as a pretty good gloss of the entire field of modern macro.

Which is not to say that I agree with everything Williamson writes. Far from it. For example, Williamson defends economists' failure to predict the financial crisis by appealing to the Efficient Markets idea:

By its nature, a financial crisis is an unpredictable event. We could have an excellent model of a financial crisis. The people living in the model world where the financial crisis can happen know it can happen, but they can’t predict it, otherwise they could profit in advance from that prediction. Similarly, an economist armed with the model will not be able to predict a crisis in the real world.

This idea can be generalized to the following principle:

"Assuming efficient financial markets, any event that has relevance for asset prices cannot be predicted in advance."

That is false. For example, consider large asteroid impacts. An impending asteroid impact is quite predictable with modern astronomy and good telescopes. You know when and where it will happen. And it definitely will have a large impact (heh) on asset prices. There is no efficient-markets paradox here. To see why, consider two possibilities:

Case 1: We cannot prevent asteroid impacts. If this is the case, the predictable effect of the asteroid will be incorporated into asset prices shortly after the astronomers confirm the impending event.

Case 2: We can and will prevent asteroid impacts. In this case, the asteroid impact won't happen (and will thus not have an effect on asset prices).

In Case 1, Williamson is wrong. Scientists can predict impending events, even though they can't profit from them. And society benefits from scientists' ability to predict the event (since we will have time to prepare and mitigate the effects of the disaster). The invention of modern astronomy turned what was previously unpredictable into something now quite predictable. John Quiggin is thus perfectly reasonable in calling for economists to improve their crisis-forecasting ability. The invention of better crisis forecasting techniques would not earn excess returns in financial markets, but would be of social benefit.

In Case 2, Williamson is semantically right but substantively wrong. If we knock the asteroid aside with a nuke, then technically, the impact never happened. But we know it would have happened if we hadn't acted. The fact that we stopped it from happening doesn't make the astronomers' prediction "wrong." In this case, "predicting" something really means making a contingent prediction: "IF we don't act, this thing will happen." John Quiggin is perfectly reasonable for asking economics to try to make predictions in this sense. If we were able to see a financial crisis coming and managed to head it off, I somehow doubt that Steve Williamson would fold his arms and snicker and say "See? Crisis didn't happen. Looks like you guys were wrong!"

Moral of the story: The EMH does not imply that we live in a non-causal Universe. That is a misuse of the idea. Predicting the future is often possible and socially beneficial, even when no one can use the predictions to make excess returns. And contingent predictions need not be reflected in prices at all. If you hear someone making the argument that the EMH absolves the econ profession from any responsibility to forecast future events...don't fall for it!

25 comments:

  1. ivansml7:24 PM

    As much as I like the creepy-wizard-with-crystal-ball picture, I don't think this is a valid criticism. EMH implies that news about asteroid impact will be immediately reflected in asset prices, which will thus fall well ahead of an actual impact. In which case, no, you couldn't have predicted the financial crisis in advance, as that would require to predict the news about the impact, not just the impact itself.

    ReplyDelete
  2. I think I'm going to side with Steve Williamson on this one.

    "Case 1: We cannot prevent asteroid impacts. If this is the case, the predictable effect of the asteroid will be incorporated into asset prices shortly after the astronomers confirm the impending event."

    The problem is that this analogy doesn't transfer from asteroid impacts to financial crises. Causation between asteroid impacts and asset prices goes only in one direction: asteroid impacts affect asset prices. But financial crises, by their nature, consist of changes in asset prices. If these changes in asset prices could be predicted in advance, the financial crisis would happen at the time when the prediction was made rather than at the time when the prediction was supposed to be fulfilled. But then, of course, the prediction is not really a prediction, it's simply part of the mechanism by which the financial crisis takes place.

    ReplyDelete
  3. @Andy & ivansml:

    Well, let's think about this. Suppose we live in a Case 1 world: we can predict financial crises, but we cannot stop them from happening. First off, this seems pretty unlikely to me, since if a crisis must include a rapid fall in asset prices, it seems like information that would allow us to predict the crisis would accumulate slowly, probably causing asset prices to fall slowly as well as the probability of a crisis increased, and thus eliminating the crisis itself (i.e., this is a Case 2 kind of world, but without the need to nuke the asteroid, since simply making a contingent prediction of a crisis would prevent it from happening).

    BUT, for the sake of argument, suppose that a "financial crisis" is something other than a fall in asset prices; suppose it is a wave of defaults, for example. THEN we could be in a Case 1 world. At time t_0, our super-duper forecasts start upping the chances of default. Asset prices fall a bit. By time t_1, we've become 99% certain that there will be a default wave at time t_2, and asset prices have fallen a lot, to fully reflect the near-certainty of the default wave. But in the time t_1-t_0, firms, individuals, and the government have been able to prepare better for the default wave. This will mitigate the impact, making the default wave on the wider economy less severe, and making the total fall in asset prices less than it would have been in the absence of a default-forecasting technique.

    But note that I think this is unlikely. Far likelier, as I said before, is that the ability to forsee impending crises at long horizons would enable us to prevent those crises (a Case 2 world). If crisis prevention could be done by private actors, the prevention would be done through natural market processes, but if the crisis were a market failure that required government action, then we could actually see the "asteroid deflector" (i.e. crisis-prevention policy) in action.

    So yes, I DO think that it is completely reasonable to call for an improvement in our ability to warn of specific impending crises.

    ReplyDelete
  4. Hmm not sure that just b/c we can see a crisis coming, that we really have ALL the tools we need to head it off. E.g. - China is doing it's darndest to head off the impending real estate bubble, but sometimes government policy just doesn't have the right levers available to it.

    ReplyDelete
  5. I'm not sure, but I think we may be omitting this crucial issue: influential denialists who might respond to the predictions (asteroid hit, financial crisis, global warming) with denial, disregard for the evidence, ideological pronouncements, and so forth.

    Market actors responding to the impending crisis pull out of the market, or otherwise adjust. Market actors influenced by the denialists enter the void left by the others. They win big until the crisis hits.

    The market has not properly adjusted.

    ReplyDelete
  6. "By its nature, a financial crisis is an unpredictable event. We could have an excellent model of a financial crisis. The people living in the model world where the financial crisis can happen know it can happen, but they can’t predict it, otherwise they could profit in advance from that prediction. Similarly, an economist armed with the model will not be able to predict a crisis in the real world."

    IMHO, he's dishonest. The accusation is really that the macro guys boasted of the stability of the system, and it turned out to be highly unstable.

    He's posting a reply to a strawman accusation. And even with that, he's being unfair. People were pointing at the housing bubble before it burst, with number and such math-like stuff (not just some crazy ranters who turned out to be right by pure luck).

    It's important to remember that RBC/neo-liberal/Chicago/etc. macro is now shown to have really messed up, on a generational level. These people pursued false theories for their entire careers, and didn't let data get in their way (and still don't).

    Noah, it's part of the job of the current generation of econ grad students to intellectually bury these guys, which will be done against their will.

    ReplyDelete
  7. The funny thing is that a financial crises is still somewhat unpredictable when it's already happening. Can anyone tell the peak of a crises? In fact, that is the situation in Europe right now. A market report is a whole lot more complicated than a weather report. We might not even know what a market really is and how it works. Obviously a sentence like "The markets are now starting to attack italian bonds" is close to nonsense. It seems to be more something like a mixture of mass psychology and masses of little impacts.

    ReplyDelete
  8. "if the crisis were a market failure that required government action, then we could actually see the "asteroid deflector" (i.e. crisis-prevention policy) in action."

    Can you give any example of this, such that the private sector could not also have prevented the crisis if it had figured out what the problem was?

    When I look at the actual recent crisis, I see a lot of things that the private sector would have fixed if it had understood what was going on. First and foremost, there was the failure to consider whether the historical sample of mortgage default experience was from the same distribution as prospective experience given the actual behavior of the housing market in the 2000's. If economists had come, early on, to regard this as a problem, surely the private sector would have picked up on that, and the MBS market wouldn't have gotten all insane the way it did.

    And there were a bunch of other problems that similarly could have been handled by the private sector if it had understood what was going on. Now this does mean that we have a lot to learn and that it would have been good to learn those things earlier rather than just learning them now that it is too late. But that isn't a general story about crisis prevention; it's a specific story about the particular factors that led to this particular crisis. In general, we can hope that economists will be as smart as they can be and thereby allow the private sector to be as efficient as it can be and thereby prevent, as often as possible, the conditions that develop into a crisis from developing. But crises that actually happen will continue to be unpredictable.

    ReplyDelete
  9. @Andy:

    Can you give any example of this, such that the private sector could not also have prevented the crisis if it had figured out what the problem was?

    No example that I know for sure! I can only guess and conjecture. But my guess is that the lack of OTC derivatives clearinghouses induced a market failure. There were a lot of smart guys shorting asset backs in the final year or two of the bubble, but a lot of them were doing it through the OTC derivatives market, which created big counterparty risks. But that is just a guess, like I said. I was speaking theoretically.

    In general, we can hope that economists will be as smart as they can be and thereby allow the private sector to be as efficient as it can be and thereby prevent, as often as possible, the conditions that develop into a crisis from developing. But crises that actually happen will continue to be unpredictable.

    Sure. But when people say "Hey economists, why didn't you spot the crisis?", they do NOT generally mean "Hey economists, why didn't you tip me off about the crisis so I could sell out before the peak of the bubble?" They mean "Hey economists, why didn't your models allow us to see that current conditions encouraged a crisis, thus allowing us to improve those conditions, whether through market actions or institutional design or policy, and hence avoid having a crisis occur at all?"

    That's what I mean by "contingent predictions" vs. "noncontingent predictions." If the Trojans had listened to Cassandra, her prophecies never would have come true, but she would have been a heck of a lot more useful than she ended up being. Economists should aspire to avert doom, not say "I told you so."

    ReplyDelete
  10. "I see a lot of things that the private sector would have fixed if it had understood what was going on. First and foremost, there was the failure to consider whether the historical sample of mortgage default experience was from the same distribution as prospective experience given the actual behavior of the housing market in the 2000's. If economists had come, early on, to regard this as a problem, surely the private sector would have picked up on that, and the MBS market wouldn't have gotten all insane the way it did."

    Shorter Andy Harless: the private sector didn't do what it demonstrably did.

    ReplyDelete
  11. Anonymous3:58 PM

    I've read the Williamson article.

    Well. Williamson seems to believe in 'utility'. I've read people like Samuelson, Gary becker, Lucas, Varian. None of these even tries to define utility in a way which enables direct measurement. Samuelson of course tries revealed preference, but that didn't work, either. The largest failure of the Lucas approach is that they haven't established any kind of independet statistical system which enables estimation of their key concept, 'utility'. The concept is basically as scientific as 'Phlogiston'.

    Stating that a sector (!) households tries to maximize utility is, therefore just semantics, not science. And don't start me on the utter lack of any idea how the sector households maximises whatever. Which political and individual and collective and sociological processes lead to this optimization of assumed 'utility' of the entire sector households? They have no idea. Nothing is directly estimated, nothing is explained.

    In my view: someting as abhorrent as astrology does a better job when it comes to a scientific investigation of the world.

    ReplyDelete
  12. Anonymous12:08 AM

    Longer Andy Harless: Market is not efficient. The private sector could not figure out anything because economists did not tell them..

    ReplyDelete
  13. Anonymous4:25 AM

    "Assuming efficient financial markets"

    In case you all missed it the finanacial markets suffered a cataclysimic FAILURE and had to be bailed out by the State and tax-payer.

    Eeesh, talk about fiddling while Rome burns.

    BP

    ReplyDelete
  14. Riffing on Aaron and Andy:

    Remember the old say 'the market can stay irrational longer than you can stay solvent'. Lewis' 'The Big Short' described a fundamental problem. People involved in finance who believed in a bubble, could not easily invest accordingly, because they didn't know *when* it would pop, and sustaining their position in the meantime could be impossible.

    Another example is that it was clear by early 1999 that the dot-com bubble was just that, but the overall NASDAQ didn't decline until (IIRC) spring of 2000. One would either have bet on individual companies, or would have had to maintain a losing position for well over a year, when that position was deeply in the red.

    In addition, in both cases some Big Money people were making a lot of money on this, and the easiest way to keep making the money was to ride the bubble. That influenced anybody who didn't want to get stepped on by the Big Money guys.

    ReplyDelete
  15. 'By its nature, a financial crisis is an unpredictable event. We could have an excellent model of a financial crisis. The people living in the model world where the financial crisis can happen know it can happen, but they can’t predict it, otherwise they could profit in advance from that prediction. Similarly, an economist armed with the model will not be able to predict a crisis in the real world.'

    He has obviously never read Minsky ??

    The crisis was not a black swan, or an external shock. It was an endogenous event that built over time, and was inevitable given the mechanics of financial institutions.

    Together, Williamson and Sumner epitomise the under disregard for the scientific method that economists demonstrate. Their theories will never be falsified in their heads, no matter what happens.

    If a scientific model had failed so catastrophically and repeatedly, it would be thrown out, no question asked, no ifs no buts. Just abandoned.

    ReplyDelete
  16. Anonymous4:45 PM

    Two events stand out that caused the crisis: the housing bubble and the creation of the shadow banking system.

    Some economists were calling out the housing bubble and pointing to past bubbles, like the recent Internet bubble. They were dismissed with "this time it's different."

    A shadow banking system arose that was clearly vulnerable to a Diamond-Dybvig-type crisis, i.e. a bank run and yet no one looked at the new shadow banking system and thought is was vulnerable. The Internet bubble deflated without much of a problem. The housing bubble nearly brought down the economy.

    ReplyDelete
  17. Noah: "But when people say "Hey economists, why didn't you spot the crisis?", they do NOT generally mean "Hey economists, why didn't you tip me off about the crisis so I could sell out before the peak of the bubble?" They mean "Hey economists, why didn't your models allow us to see that current conditions encouraged a crisis, thus allowing us to improve those conditions, whether through market actions or institutional design or policy, and hence avoid having a crisis occur at all?" "


    I not only second this, but state that I've never seen the first example at all, just the second.

    Williamson is using the first, to dodge his failure, and the failure of his colleagues.

    ReplyDelete
  18. The funny thing is how we look back at the great depression and think, well at least we won't make those mistakes again - and then proceed to do our best to copy some of the bad ideas from that time.

    In the face of greater understanding, the interests of some people still prevent the best actions.

    ReplyDelete
  19. To claim that markets are not efficient assumes that there is such a thing as objective value. Some try to define it using "utility," others avoid the issue in other ways, but in the end there is no such thing.

    All value is determined by the opinions of individuals, and market prices are simply the result of myriad opinions.

    If two men are alone in the world and one of them says "I think this piece of art is worth ten chickens," and the other thinks it is worth 20 chickens, which of them is correct? What is the "objective" value of the art?

    ReplyDelete
  20. Stone,

    I might agree with you but:

    a) People suffer from various cognitive biases that prevent them from making coherent decisions all the time. (Please don't respond with 'the government has biases too' because that is poor reasoning. I can design a safety net system for a tree climbing complex whilst myself being vulnerable to falling).

    b) Markets, in the real world, are *not* the result of countless individual decisions. They are a product of complex social and legal interactions and many players have far more power than others. Many are also corporations or unions rather than individuals, which changes the dynamic.

    ReplyDelete
  21. Cahal,

    A "coherent" or "rational" decision is the same as an "objective" value. Everyone is always making what they consider to be a rational, coherent, reasonable decision. Anyone claiming that another person or group made an irrational decision is claiming to have a better grasp of that person's (or group's) needs and wants. How can anyone make that claim with a straight face?

    An objective value is as impossible as an irrational decision. If any decision is deemed irrational, we must always ask "irrational according to whom?"

    ReplyDelete
  22. 'Everyone is always making what they consider to be a rational, coherent, reasonable decision.'

    Really? Even when they are drunk or are suffering from mental disbailities? When they do not have appropriate knowledge? Children? Someone who walks infront of a bus without looking?

    'An objective value is as impossible as an irrational decision. If any decision is deemed irrational, we must always ask "irrational according to whom?"'

    Well behavioural studies have discovered things like peak end evaluation and preference reversal, showing people often find it very difficult to evaluate the decisions they have made based on how they are framed etc.

    You are subscribing to rationalism, which was abandoned centuries ago because of its blatant unfalsifiability. Instead, we use the scientific approach. It really does make more sense.

    ReplyDelete
  23. Anonymous4:22 PM

    Williamson really likes his arguments about how economic models can never predict crises. But he is wrong about what the optimal response to such a prediction is. If macroeconomist develop a model that says that stock prices will peak and then crash at precisely 2pm on october 16th, then Williamson says he would just sell his stocks on the 15th and the crash would happen on the 15th not the 16th. But Williamson is wrong. The optimal thing to do is to sell at precisely 2pm on the 16th, because the model says that the price will continue to rise between the 15th and the 16th. So if people have total faith in the model, the crash would happen exactly as predicted, regardless of whether there was a real basis for the prediction in the first place.

    Of course, people don't actually pay attention to economic models, and these never have an impact on the markets. This only reaffirms the notion that market crashes can be perfectly predicted in principle.

    ReplyDelete
  24. There are a couple of problems I see here: first, we can never know how many past financial crises were accurately predicted and successfully avoided because they never occurred so we don't have models for them. Using the asteroid comparison again, a model assumes that things will follow a predictable path but cannot account for all contingencies so it smooths them out. An asteroid hurtling toward earth will hit assuming certain low probability contingent events do not occur. But even a slight deflection occurring far enough away from earth would alter its trajectory enough to create a huge variance in its actual vs. predicted path. The asteroid that misses earth appears to prove the model wrong, so we toss out the model.

    Second just a map is not the terrain, so too a model is not the actual event. Things are more likely not to unfold according to the model. Remember Descartes? We can start with a ray of length n, and with some fairly basic math develop a model of a perfect chiliagon (a 1000-sided polygon). But it doesn't follow that we can construct one, or, if we could, that it would conform to the model due to limitations of the materials, measurements, construction techniques, time, and other unpredictable events not considered by the model.

    I think there's a huge difference between predicting a specific economic meltdown and predicting the effects of a hypothetical meltdown given specific economic conditions. Planning for the latter allows for a wide range of models based on various contingencies.

    ReplyDelete
  25. I think Brian's model on this issue is spot-on. Efficient Markets doesn't mean markets always know what's going down - the issue is whether any individual can systematically know it better. The EMH doesn't really say markets are always spot on and perfectly efficient, rather that they are "efficient-as-you-get"
    http://synthenomics.blogspot.com/2012/04/efficient-as-you-get-market-hypothesis.html

    Claims of endogenous credit cycles seems pretty self-serving; now that the financial crisis has happened, it's fairly obvious. Why, if they were so self serving, it's seems curious why you would still want to come comment on a random (but very entertaining) blog. You could have just short the securities and moved on with your merry life.

    This hints at the issue with timing; we just don't know WHEN the crisis will hit. So Noah, while the concept of contingent forecasting makes sense, I'm not too sure if there's enough evidence that there's a strong correlation between certain economic indicators and the presence of a crisis. However, with the concept of debt cycles and fragility, there seems to be room to look at the market as it's evolving and take precautionary measures to avoid the fat tail, negative Black Swan events.

    In a sense,
    (1) The EMH predicts an individual cannot beat the market as a whole. Individual stocks cannot consistently outperform the market as a whole Even if there are cognitive errors, they can disappear through arbitrage

    (2) The EMH also predicts the market, as a whole, can exhibit certain fragilities. Unknown unknowns can spread through the market, and catch people unawares, as the EMH is fundamentally an informational proposition.

    With these two perspectives, it seems like the correct role for regulation is not to go after specific sectors (you can't predict individuals), but rather to worry about common aggregate dangers, such as debt.

    ReplyDelete