A Deeper Look at Those Skeptical Charts on Forward Earnings Estimates
by JEff Miller, A Dash of Insight
How useful are the forward earnings estimates of the sell-side analysts?
There is a widely-circulated chart that reinforces the popular viewpoint that analyst forecasts are hopelessly optimistic. It comes from a Morgan Stanley research note, first called to my attention by a reader who writes frequent and thoughtful comments, "CautiousInvestor". I knew immediately that there was something amiss, but I wanted to see the original context of the report. Now Josh Brown has joined in, providing a second chart from a different Morgan Stanley department. Josh is both open-minded and influential. He often uses forward earnings in his popular CNBC segments and also features reports from earnings expert Brian Gilmartin. This seems to provide the perfect occasion for a civilized, data-based discussion – difficult to get on a topic that seems to arouse passions!
Summarizing my position
I find the bottoms-up estimates to be very useful, with the most recent full update here. I have frequently noted that most pundits simultaneously claim two things:
- Analyst estimates are too optimistic.
- At the time of reports, the bar is "too low" so the "beat rate" is very high, but not meaningful.
If you think about these two propositions, it suggests that there is a crossover point where the estimates are pretty good. My research shows that estimates are pretty good if you limit the forecast to the next twelve months. The Morgan Stanley chart suggests that my conclusion is wrong. We would all like to have better forecasts of earnings, so this is a good topic for research. If we simply cannot forecast earnings effectively, so be it. Let us dig deeper.
The Attention Grabbing Chart
I will be working with the two charts from Josh's article, although the first has appeared in many places.
The message of the chart is pretty clear: Thirty-six years of data show that we should expect estimates to fall throughout the year. My own experience told me that something was wrong. Other observers – even savvy ones – have not worked as closely with the underlying data. Moreover, the conclusion fits neatly with the preconceptions everyone has.
Here are some of the questions that should immediately come to mind:
- What is depicted? Is the consensus for each month the arithmetic mean? If so, it is dramatically influenced by the occasional recession. Using a mean distorts the result when you have a few skewed cases. Suppose, for example, that Bill Gates and Warren Buffett are in a room with a group of other bridge players and we calculate the "average" income. Here is a quote from the report:
"Since 1976, the median year-over-year earnings growth forecast in January for the full year ahead is 14%, but expectations on average decline throughout the year to closer to the 5% average EPS growth we have seen over that period (Exhibit 1). In fact, in 29 of the 37 full years for which we have forward earnings data, the January sell-side analyst estimates proved to be too optimistic."
So the January figure is described as a median.
- What is a median value in this context? Did the researchers take the 37 years and choose the middle one? If so, why not tell us which year it was? Does that year really depict the typical behavior? Or do they look at the median change for each month? That would be strange, since the overall curve would not reflect any particular month. A median is best used when there is a cluster of casts that can be described as typical. Is that the case here?
- Why is there no actual earnings result for the year? This is a chart of changes in the forecasts. Normally when we want to evaluate a forecast we compare it with the actual result. That is not known for a few months after the end of the calendar year. Usually the actual beats the final forecast.