There is Little Value in Broker Economics
Originally published in the Financial Times, 4th January 2006.
The FT reports regularly on the views of economists and stockbrokers. Readers might be intrigued by the conflicts between them but they should not be surprised. The purposes of the two groups are completely different. Economists are in pursuit of the truth and stockbrokers of commissions.
The first principle of stockbroker economics is that all news is good. In a weak economy, interest rates tend to fall and in a strong one profits usually rise. It has therefore become an item of stockbroker faith that falling interest rates and rising profits are both good for stock markets. Neither theory, however, seems robust under testing. For example, the US stock market has moved in the same direction as profits only 54 per cent of the time. As both shares and profits have a long-term tendency to rise, this suggests a purely random relationship.
The second principle is that the stock market is always cheap. This requires more flexibility than the first principle and is achieved by inventing meaningless measures of value, such as the bond yield ratio, and using whichever one of these absurdities happens to give the most bullish answer at the time.
Data mining is the key technique for nearly all stockbroker economics. There is no claim that cannot be supported by statistics, provided that these are carefully selected. For this purpose, data are usually restricted to a limited period, rather than using the full series available. Statistics, it has been observed, will always confess if tortured sufficiently.
The greatest single triumph yet achieved by data mining is the invention of the bond yield ratio. This claims that equities can be valued by comparing bond yields and earnings yields. These ratios showed a strong correlation in the US from 1977 to 1997. But the exact opposite relationship ruled from 1948 to 1968. It is, of course, possible to use all the available data, thereby flattering the prejudices of economists but offending the key principle of data mining. If this is done, it shows that there is no relationship at all between bond yields and earnings yields.
Readers can, nonetheless, be confident that the use of the bond yield ratio will not disappear simply because it cannot be supported by either theory or experience. Claims based on data mining are not discarded simply because they do not work. They are put into the pending tray with the standard excuse that "the relationship has broken down". While this cannot be logically distinguished from "there never was a relationship", it has two great advantages. First, it sounds a great deal better and, second, it demands less effort to reuse old nonsense than to invent new follies.
Nonsense about value comes in many other forms. The most common is the belief that the markets can be valued in relation to the current price-earnings ratio. This survives in the face of its obvious failure in the past. For example, PEs were relatively high in 1932 when the market was extremely cheap. But current PEs can be relatively high, when the market is overpriced, as indeed they are today. There has therefore been a demand for ways to avoid the obvious conclusion that markets are expensive. Supply has risen nobly to meet this demand. In addition to restricting the years in the sample, the average is miscalculated by using the arithmetic rather than the geometric mean. (By using the arithmetic mean of the PE, the market appears cheap for well over 50 per cent of the time. If the arithmetic average of the earnings yield is used, the market appears to be expensive more than 50 per cent of the time. The preference for expressing the relationship between price and profits by PEs should not, therefore, cause surprise.) Further help can be obtained by using forecast rather than historic earnings. As these are unknown they can be raised to whatever level is needed to fit the argument being used.
Value is a problem for fund managers as well. Stephen Wright and I published a book on the overvaluation of the US stock market which, by great good luck, was published at its peak in March 2000. We showed strong evidence that the stock market was more overvalued than ever before, including at its previous peak in 1929. We calculated that the chances of the market falling over the next year were a bit more than 70 per cent, though this rose to more than 90 per cent over five years. This raised the interesting question of what fund managers should do. For their own money, the answer was clear - sell. But if they went liquid for clients they had a 30 per cent chance of ruining their businesses and careers.
Since dog should not eat dog, I hasten to point out that, while a lot of stockbroker economics finds its way into the financial press, this is at least partly reasonable. What is widely believed may be nonsense but it can still be news. The other problem is that the columns have to be filled. It is not clear that we need to be reminded, at least as often as we are, that "stock markets abhor uncertainty, but the devil is in the detail". It would nonetheless be a truly unkind editor who red-pencilled this. Happily for journalists, uncertainty and detail are always with us.