The Efficient Markets Hypothesis
by George Hatjoullis
During my brief tenure as a finance academic I gave careful consideration to the Efficient Markets Hypothesis (EMH). It asserts that markets instantaneously and accurately reflect all publicly available and relevant information in asset pricing. This sounds appealing and quite innocuous. A great deal of money can be made if one has an informational advantage and there are armies of analysts continuously digesting every announcement trying to get an edge. Surely, this is self-evident?
In science, even social science, the self-evident must be explored. A whole academic industry emerged based on ‘rigorously’ testing the EMH. The test design involved testing a joint hypothesis on what constitutes ‘value’ and informational efficiency. It is impossible to do it any other way. For the design to work one must hold constant all other influences on price and one must have a view on how the piece of information being considered should impact price.The tests involved taking single items of information and seeing if they could be used to gain a financial advantage.The hypothesis is that one can gain an advantage and failure to do so is consistent with the EMH.
The joint hypothesis designed bothered me. However advocates of the EMH saw it as one of its virtues. There are four possible situations.
a) The model is correct and the market is efficient. One would expect the item of information to confer no advantage and to reject the joint hypothesis of inefficiency.
b) The model is incorrect but the market is efficient. One would expect evidence of inefficiency even though it is generated by the false model and not to reject the hypothesis.
c) The model is correct and the market is inefficient. One would expect evidence of inefficiency and not to reject the hypothesis.
d) The model is incorrect and the market is inefficient. One would expect evidence of inefficiency and not to reject the hypothesis.
This seems like a very robust test design. There is a prior expectation of finding evidence of inefficiency in three out of four possible situations so if no such evidence is found it is powerful evidence in support of the EMH.
The problem I had with the design was the irrelevance of the model of value used. Research papers used a whole range of models of value in their tests and yet the published papers invariably rejected the joint hypothesis of inefficiency. Were all these models of value correct? Given the obvious lack of knowledge of the ‘true’ value why was the hypothesis not rejected more often (or at all as far as published papers were concerned).
I considered the possibility that the papers that failed to reject the hypothesis were simply not published. Hard to prove but the suspicion remained. Evidence in a number of disciplines since this time (1977-80) suggests this was far more likely than you might imagine. I also considered the possibility that the research designs biased in favour of rejection. After reading every paper that came up in my searches (this was part of my doctoral work) only one thing of significance emerged.
There were a great many papers that failed to reject the joint hypothesis initially but the inefficiency was rejected as economically irrelevant as transactions costs would wipe out any advantage.The conclusion was to reject the joint hypothesis of economically relevant inefficiency. Many papers were thus amending the hypothesis ex post (anathema) and admitting poor research design (transactions costs should have been included in the design). However, they were still published I noted!
My supervisor (a theoretician) seemed less animated by this than I was but he did suggest I write a working paper highlighting my observations for discussion (this it turns out is the academic equivalent of a select committee). This I did (1980) and whilst I found it cited quite often over the years it was never published.
My main observation was that marginal transaction costs are effectively zero. Trillions of dollars of assets are bought and sold daily and there is no additional costs to using information in deciding when and what to buy and/or sell. Moreover, no one use single items of information. Analysts merge together many types of information in coming up with recommendations. The EMH test design was seriously flawed and rejected evidence of inefficiency on specious grounds.
Shortly after putting out this working paper I abandoned the academic world. In the (real) world of finance I found a whole industry of people using publicly available information to predict price movements and indeed a whole society dedicated to doing this . There is even a qualification . It is of course technical analysis. I know a great many more wealthy technical analysts than I do finance academics though this proves little (maybe they are taking more risk). Also virtually every fund manager and proprietary trader will look at a simple price chart when making decisions. I thought I would share…