Monday, 20 July 2009

Andrew Lo's adaptive markets and the Slow EMH

Andrew Lo, writing in the FT, says:
...human behaviour is hardly rational, but is driven by "animal spirits" that generate market bubbles and busts, and regulation is essential for reining in misbehaviour.
Regular readers won't be surprised to see me agreeing with this, and indeed I have a proposal for how that regulation could work.

However I am suspicious about any theory which does not make testable predictions, and I fear that Lo's "adaptive markets hypothesis" may fall into this category.
This "Adaptive Markets Hypothesis" (AMH) - essentially an evolutionary biologist's view of market dynamics - is at odds with economic orthodoxy, which has been heavily influenced by mathematics and physics...The formality of mathematics and physics, in which mainstream economics is routinely dressed, can give outsiders a false sense of precision.
...fixed rules that ignore changing environments will almost always have unintended consequences...The only way to break this vicious cycle is to recognise its origin - adaptive behaviour - and design equally adaptive regulations to counterbalance human nature.
Just like Shiller and Akerlof's Animal Spirits, which has influenced Lo's proposal, and like Nicholas Taleb's Black Swan - the AMH appears to be a narrative proposal requiring creative interpretation, instead of a model with specific outcomes that can be checked. That is the domain of literature, not science.

In essence, I agree with the project that Lo (and Shiller and Akerlof) have outlined. But instead of sidelining the mathematical tools of economics, we need to fiercely apply them to make this concept work.

I suggest we start with a smaller step: not to overthrow the EMH with a set of generic, unpredictable exceptions, but to modify it to take into account certain imperfections of behaviour in markets. Specifically a "slow EMH" which would work like this:
  1. In the long run, market prices do reflect all available information.
  2. However, some of the facts they reflect are not exogenous (externally given facts); they are the opinions of market participants. (Others are exogenous, but hard to observe - I will come back to those)
  3. These opinions take time to form and test; the key mechanism for testing them is to attempt to make or take a price, and observe whether people accept it. For example, when selling a house, a seller may set a price 5% higher than that achieved by a neighbour, in order to find out whether buyers are willing to pay.
  4. These tests, in turn, influence the opinions of other observers - those opinions then also become facts which are relevant to the price.
  5. The nature of such price-setting tests are that they rarely jump straight to the "correct" level; buyers and sellers are rarely willing to pay a price wildly out of line with the last price paid, and so the increments are more gradual than the EMH would imply.
  6. Over time, therefore, prices will gradually move towards a stable level which reflects external information; but in the meantime there may be a free lunch, if you are better or braver at interpreting the external information than the market on average.
This mechanism may explain why there are relatively long-term bull and bear markets in both housing and equities. The standard EMH implies that, say, house prices should jump directly to their new level as soon as a demographic change becomes apparent. However, instead we saw a continuously rising housing market from around 1992 to 2007, and a steady fall since then.

It may also explain markets which never reach stability: if the adjustment process is not complete by the time new external facts arise, a new process will start, overlapping the previous one, and stability may not be achieved.

About that free lunch: it won't be easy to earn, but consider that there are hard-to-measure exogenous changes and you may be able to predict the market's reaction to them. For instance, demographic changes and restrictive planning laws, which were largely predictable throughout the 1990s and 2000s, could be considered to have led to the house price boom in the UK. You could make a good case that the correct level of prices, based on these facts, was probably around the 2005 level (perhaps 2006 and 2007 were an overshoot). But prices in 1999 did not immediately jump to the level they would be at in 2005. Instead, many people had a general belief that prices should be higher, and tested that belief with gradual price increases. As buyers continued to bite, confidence in this belief grew, and that in turn justified a higher price.

If you can identify a market rationale for higher prices, combined with market mechanisms that slow down price adjustments, you may be able to make money. Even if you can't, you can understand markets better and know better when to rely on their signals.

This I call the Slowly Efficient Market Hypothesis, or Slow EMH. It is less universal than the AMH, but has the strength that it, seemingly unlike AMH, can be mathematically modelled in a testable way. Look out for a future article which will do just that.

p.s. I am not at all convinced about the 'free lunch' idea in the above. I am certainly not ready to put any money behind it right now. But I do believe the EMH cannot operate instantaeously and some form of Slow EMH is definitely at play.

p.p.s. The Wikipedia description of the AMH indicates it may actually make some more testable predictions than I assumed from reading the FT article. I have not yet read the 2004 paper but will do so and report back.

1 comment:

Anonymous said...

Dear Sir:
any new update on this?