Heads must roll


SERIOUS MONEY:Investment managers have been using valuation models that are fundamentally flawed, writes Charlie Fell 

THE YEAR is almost over and it has been an extraordinary one, but not for the reasons that most believe. It is true that stock markets are on course for their worst year in decades, but it is our investment elite and their exceptionally poor performance that have shocked.

The performance record of our finest investment managers over one, three, five and 10 years can be described as nothing less than dreadful. They walked your money into a severe bear market not once but twice in just eight years. They ignored the lessons of history and the ferocity of a secular bear and they compounded the problem through the dismissal of cyclical indicators that warned of impending trouble a year ago.

The signals were flashing danger this time last year. They included a downward sloping yield curve, rising credit spreads and earnings that were at their fourth-highest level in more than a century versus their 10-year average.

The indicators went unnoticed and equity allocations remained high. Claims that the subsequent turmoil was an act of God simply do not stand up to scrutiny.

Perhaps some can forgive such obvious failings but, when added to the indictment that some managers appear to be incapable of valuing the assets in which they invest, clients must question the fees they pay for persistent value destruction. The flawed valuation models and the performance that has followed are nothing short of disgraceful.

The standard analysis calculates the price/earnings ratio on one-year forward operating earnings and compares the resulting multiple to a historical average based on trailing 12-month earnings.

The obvious bias created by the comparison of a future multiple with historic averages appears to be lost on practitioners.

The forward estimates of earnings are typically biased upwards and calculated before write-offs, while the historical earnings used for comparison purposes are net of exceptional and extraordinary charges. Not surprisingly, such faulty analysis almost always leads to the conclusion that stocks are cheap.

Furthermore, historical comparisons are also invalid as they typically span a period of no more than 20 years. This limited sample is not statistically significant and artificially inflates the average price multiples used for comparison as the data set includes the bubble-era valuations of the late 1990s. Stocks will look cheap when value is assessed relative to the most expensive valuations in US stock-market history.

Perhaps the most damning evidence against the use of single-point estimates is the fact that they have limited predictive ability. The price/earnings ratio has mean-reverting properties but, unfortunately, the eventual adjustment to a single-point multiple typically occurs through earnings changes and not through price. Thus, even if the analysis is properly constructed, it simply does not add value.

The highly paid investment experts are not done yet and out of their toolbox springs the so-called Fed model, which compares the earnings yield on stocks - the inverse of the price/earnings multiple - with the yield available on long-term Treasury bonds. Some routinely argue that stocks look extraordinarily cheap on this basis before the consideration of long-term growth prospects.

They are clearly unaware that the earnings yield, which approximates the long-term expected real returns for stocks, includes anticipated capital appreciation. Relatively simple algebraic equations demonstrate that the earnings yield equals the dividend yield plus expected long-term real growth.

Furthermore, use of the Fed model is theoretically invalid as investors are erroneously comparing a real variable with a nominal one. Stocks are a claim on real assets and the cash flows thereof should appreciate with inflation, while bonds are unambiguously a claim on nominal cash flows.

Inflation clearly affects bond yields but it should not affect earnings yields as a change in the discount rate should be offset by an equal change in the growth rate of dividends. Furthermore, investors fail to appreciate that inflation reduces the value of fixed-rate debt, which increases the value of equity.

Even if the comparison was theoretically valid, the model requires some implausible assumptions, but the most damning case against its use is the evidence that it simply does not work in practice.

A properly constructed price/earnings analysis should assess the stock market's long-term earnings power.

The use of relatively simple regression techniques generates an estimate of $66 per share for trend earnings. This easy-to-use technique is based on reported earnings through time and not those rosy operating numbers that exclude all the supposedly once-off bad stuff.

Indeed, these once-off charges have been positive each and every year in the past 20 years, ranging from 2 per cent of operating earnings in 1988 to more than 40 per cent in 2002. They have averaged more than 12 per cent over the entire period, with the obvious implication that what is extraordinary for one company is not so for the market in aggregate.

Valuations no longer look expensive in a historical contest on just 13 times trend earnings, but the damage of high equity allocations maintained at a time when prices were dangerously high has been felt by most pension funds. It should be clear that the valuation of common stocks is the cornerstone of successful investment, but without the capability of performing a valid valuation exercise, client returns inevitably suffer. Heads must roll.