Get real about consensus estimates

Oversimplification of estimates implies equal chances of success and failure for the trader

Last week saw paradoxical price movements in three big stocks. HCL Tech and Reliance Industries beat the consensus estimates for Q4 and the full-year. Tata Consultancy Services (TCS) hit the upper end of consensus estimates. Yet, all three saw sell-offs. And, this happened in a market where the Nifty gained strongly on the days in question.

Does this puzzle conform to the adage of "buy on rumour, sell on news"? There is speculative action just before a result is announced (and when company-specific events like mergers, management changes, etc, are rumoured). Speculators enter before the event and book profits when things are "pucca". But this is rarely the Indian pattern, where trends often develop post-event.

Sometimes, results indicate that prior expectations have been met but the future going will be tough. In such circumstances, stock prices can drop. This, at least partially, is the case currently. Reliance has rising refining margins. But gas output from KG-D6 has declined drastically. More, the company is stuck in long-drawn negotiations about gas pricing.

HCL Tech will not see great improvements in margins. Projections suggest that earnings growth and revenue growth will slow to below the breakneck pace in 2012-13. In TCS, there are no storm clouds as such. But the market seems unimpressed by the IT giant just maintaining expected performance levels.

In a broader sense, the problem lies with using the so-called consensus estimate as a benchmark. It is too blunt and simplified a measure to be of much practical utility. The business channels on TV love it, since it makes it easy to measure performance.

But "consensus" uses one number to capture earnings expectations and that concept is flawed. The consensus estimate is an equally-weighted average of multiple estimates. In practice, different amounts of cash back each of those multiple estimates.

There is no way to weight the consensus with accuracy. One institution with deep pockets backing an outlying estimate can cause a big skew away from the consensus.

Consensus estimates are not meaningless for long-term investors but less so than people think. For any long-term investor, earnings are important but only one of the many factors under consideration. Quality of earnings, long-term growth expectations, likely peer performances, management changes, government policy and many other things will influence investor attitude.

For traders, it's easy to construct a simple strategy around consensus. This is implied in statements like "the stock rose after results beat consensus". A trader may for example, follow a rule like "buy if consensus exceeded; sell if results miss consensus". But such a simple system cannot really be traded for a significant edge, since prices will move the wrong way about as often as they move the right way.

Traders use many indicators to generate signals to exploit the extra volatility that occurs close to results. I like 55-day highs and lows when it comes to quarterly results, on the basis that a 3-month breakout (55-day high) or breakdown (55-day low) implies some change since the last quarterly results. Others look at seven day and 10-day moving averages. Some analyse volumes. All signals generated by these systems have high failure rates.

On the face of it, consensus estimates are logical indicators. They can be used but only with a complex trading system that builds in large error margins, strict time frames and disciplined stop-losses.

A more complex trading system could be something like "buy only if the consensus estimate is exceeded by 20 per cent, and if the 50-day moving average is also giving a buy signal, with a stop loss set at three per cent below the entry price". Even with such multiple filters, there will be failures. But the failure rate will be lower and the losses will be less when there is a failure.