Articles

How do stale estimates distort consensus EPS?

A consensus average includes estimates nobody has touched since the news. Clean it by date before you say the company beat or missed.

Ten analysts cover a company that cut its full-year guidance 30 days ago. The raw consensus for the year is $3.203, but four of the ten estimates are older than the cut. Using only the six current ones, consensus is $3.058, so the raw figure is 4.7% too high. Reported EPS of $3.07 is a 4.2% miss against the raw number and a 0.4% beat against the cleaned one.

The assumptions

Illustrative full-year EPS estimates, invented figures
AnalystFY EPS, $Days since updateStatus
Analyst A3.4295stale
Analyst B3.3888stale
Analyst C3.0512current
Analyst D3.109current
Analyst E3.48120stale
Analyst F3.0220current
Analyst G3.086current
Analyst H3.4070stale
Analyst I3.1215current
Analyst J2.984current

The company cut guidance 30 days ago, so any estimate last updated more than 30 days ago was written before the news. Reported EPS is $3.07. The sample is ten estimates, which is small; with a larger sample the same method holds but the gap may differ.

Step 1: the raw consensus

Sum of the ten estimates = 32.03

Raw mean = 32.03 ÷ 10 = $3.203; median = $3.110

Step 2: the cleaned consensus

Keep estimates updated within 30 days: C, D, F, G, I and J

Sum = 18.35; cleaned mean = 18.35 ÷ 6 = $3.058

Gap = 3.203 − 3.058 = $0.145, or 4.7% of the cleaned figure

Excel: =AVERAGEIFS(B2:B11,C2:C11,"<=30")

The four stale estimates average $3.420 against $3.058 for the current ones. Because they sit above the rest, they pull the mean up by $0.145, while the median, $3.110, is less affected.

Step 3: what the difference does to the story

Reported EPS of $3.07 against each consensus
ConsensusEPS, $Surprise
Raw mean of ten3.203−4.2%
Raw median of ten3.110−1.3%
Cleaned mean of six3.058+0.4%

At a 15x multiple, a target based on the raw mean is $48.05 and one based on the cleaned mean is $45.88, a difference of $2.17 a share from the same ten analysts.

What if: a looser definition of stale

Consensus by how recent an estimate must be
Keep estimates updated withinEstimates keptConsensus, $Surprise at $3.07
30 days63.058+0.4%
75 days73.107−1.2%
90 days83.141−2.3%
No filter103.203−4.2%

The date filter matters because the news date is known. Widening the window to 75 days keeps analyst H, whose $3.40 was written before the warning, and the surprise shifts to −1.2%.

The common mistakes

Takeaway

Questions readers ask

What is a stale estimate in consensus EPS?

An estimate that an analyst has not updated since an event such as a profit warning. Here four of ten were last updated more than 30 days ago, before the guidance cut.

How much can stale estimates move consensus?

In this example the raw mean is $3.203 against $3.058 once the four are removed, a gap of 4.7%.

Should you use the median instead of cleaning the mean?

The median is less affected, $3.110 here, but it still contains stale figures. Cleaning by date addresses the cause and gives $3.058.

The workbooks behind this article

The free companion files for The Equity Research Analyst rebuild a fictional coverage model with live formulas: segment earnings, reported and adjusted EPS, a consensus cleaned of stale estimates, the price target and the rating grid. The estimates in this article are invented and are not in them. No account and no email address.

Open the companion files →

Not on Amazon: 19 paperbacks by Julian R. Sterling on markets, funds and compliance, printed on demand and shipped worldwide. Browse the bookshop →

Also on this site

Reading guide: corporate finance, valuation and markets → · All 1908 articles →