The great data blank


By Michael R. Sesit

The Straits Times

DO YOU find yourself spending days examining countless bits of data before deciding where to invest? The bad news is you might be wasting your time. The good news, if you can call it that, is you have plenty of company – particularly in the professional investment community.

'Our industry is obsessed with the minutiae of detail,' says Mr James Montier, chief global equity strategist at SG Securities in London and a specialist in applying psychology to finance.

One reason is that analysts seek to impress with quantity over quality, often producing voluminous reports that amount to the ostentatious presentation of the obvious. Another is that 'analysts are often petrified of saying, 'I don't know',' Mr Montier says.

Academic research shows more information does not necessarily imply better information – except perhaps when computers are used to process the data.

Expert handicappers were just as accurate at predicting winners of horse races when they had five bits of information as when they had 10, 20 or even 40 bits, according to a 1973 unpublished study by Mr Paul Slovic, a psychologist at the Oregon Research Institute.

More recently, three researchers at the University of Chicago's Graduate Business School – doctoral candidate Claire Tsai and behavioural scientists Joshua Klayman and Reid Hastie – came up with similar results after testing the ability of American football fans to accurately forecast the winners and point spreads of collegiate football games.

The participants – University of Chicago students who had to pass a test showing their understanding of football – were given six random bits of data, or cues, in each of five successive stages. The teams weren't revealed. On average, the participants correctly predicted the winners about 62 per cent of the time, regardless of whether they had six cues or 30.

Interestingly, the students' confidence increased when they were fed additional data, even if their predictive accuracy didn't. That was also true of Mr Slovic's handicappers.

When the Chicago experiment was replicated using a computer model, the model was correct 56 per cent of the time with six cues, gradually increasing in accuracy to 71 per cent with 30 cues. But people aren't computers, and the human brain can absorb and process only so much information.

Psychologist George Miller, in the 1950s, observed that the average human working memory – which Mr Montier calls 'the brain's scratch pad' – was able to handle seven items, plus or minus two. 'The span of absolute judgment and the span of immediate memory impose severe limitations on the amount of information that we are able to receive, process and remember,' he wrote in the 1956 edition of the Psychological Review.

Ever hear the old quip: The more you learn, the more you know; the more you know, the more you forget; the more you forget, the less you know? This scientific research unfurls it in spades.

'Too much time is spent trying to find out more and more about less and less, until we know everything about nothing,' Mr Montier says.

'Rarely, if ever, do we stop and ask what do we actually need to know.'

Instead of assembling gobs and gobs of data, investors should focus on the information that is really important.

That may vary, depending on whom you ask. There is no magic formula, of course. If there were, everybody would be using it. Better yet, there wouldn't be a stock market, because there would be nobody to take the other side of a trade.

Mr Montier says he measures a stock's appeal according to some valuation criteria: its momentum, which gauges the strength of its price movement, and balance-sheet data indicating a company's financial soundness, such as its so- called F score, which takes account of profitability, debt and operating efficiency.

Another item would be a corporation's capital discipline, such as how much money it is paying out in dividends and stock buybacks or keeping as retained earnings.

Again, another discipline may hold lessons for investors. A study of two Michigan hospitals found that emergency department doctors were sending about 90 per cent of patients with severe chest pain to their hospital's cardiac care units (CCUs). It also turns out that, by chance, they were admitting 90 per cent of those who belonged in the CCUs and about 90 per cent of those who didn't. The physicians might as well have tossed a coin.

The problem was the doctors were looking at a wide range of risk elements such as age, gender, smoking, cholesterol levels and family history. While good predictors of heart attack potential, they are poor diagnostic tools for determining whether someone is actually having an attack.

In another experiment, researchers Lee Green and David Mehr designed a 'decision- support tool', cards that matched heart attack probabilities against diagnostic data.

Doctors' decision-making improved. Surprisingly, though, their accuracy rate rose even when they didn't use the charts, reflecting the physicians' enhanced focus on key diagnostic elements.

'Teaching simple decision-making strategies might effectively reduce unnecessary CCU utilisation,' the authors wrote.

In response, researchers Green and Mehr came up with an even simpler decision tool of stepped yes/no questions relating diagnostic cues to heart attack probabilities. It worked even better.

Mr Montier recommends building similar information-filtering tools for investing. If you answer 'yes' to the initial question, go to the next query. If the answer is again affirmative, move to the third question, and so forth. If at any time you answer 'no', find another stock to buy.

Need a New Year's resolution? One word: Simplify.

The writer is a Bloomberg News columnist.