Why you should question analyst forecasts
Given how often ‘consensus earnings expectations’ are quoted in the financial press and feature in investor discussions there is surprisingly little awareness of what consensus really is. What does it really measure? Who contributes to it? What are its biases?
In the context of earnings ‘consensus expectations’ sounds like it should be the level of earnings or earnings growth that investors expect. But measuring what investors think about earnings is difficult, arguably impossible. However, measuring bottom-up analysts’ expectations is a pretty straightforward exercise and so this is typically used instead. But of course 'analysts' are not 'investors' and by using analyst consensus we need to be aware of analysts’ biases in order to appropriately interpret the data.
Bottom-up consensus tends to be overly optimistic at the outset. When the first consensus forecasts are calculated, two years before the start of the reporting period and more than three years before the actual figures are reported, analysts typically forecast earnings growth of around 10%. By the time the reporting period starts, growth expectations have on average been raised even higher, to 11.5% for the FTSE 100. But actual earnings growth has only turned out to be a more modest 6% per annum over the past few decades, in other words, we have a very consistent pattern of analysts cutting their earnings forecasts throughout the reporting period as reality has not measured up to lofty expectations.
We need to recalibrate how we interpret analyst consensus. We should assume that estimates ahead of the reporting period are likely to be too optimistic and overestimate growth, on average by around 5%.
This also means we should not assume that downward revisions to earnings estimates are necessarily bad news for share prices. Of course, downgrades for company-specific reasons can be the source of a genuine negative surprise for the market, but our base line assumption should be a gradual downward drift of estimates over the course of the year. Remember the 1990’s, one of the strongest bull markets in history, were characterised by consistent cuts to consensus earnings forecasts.
But perhaps we should cut bottom-up analysts some slack, after all, forecasting the future is a difficult business. Each individual company forecast can be achievable, but the persistent bias to over-estimate the growth of the companies they cover means that when you add up the forecasts to the market level the aggregate becomes difficult to reconcile with a likely macro-economic scenario. In every company’s case one could envisage a scenario how they could take market share, but not all companies can increase market share at the same time.
Another pattern in consensus estimates repeats with every quarterly reporting season. Consensus earnings estimates for the upcoming quarter generally fall ahead of the reporting season, the vast majority of companies then manage to beat those lowered expectations, and this in turn is followed by a wave of post-reporting upgrades.
There could be several explanations for this effect and it seems likely that all play a role. Firstly, pre-announcements are clustered in the weeks ahead of the reporting season and with pre-announcements naturally heavily skewed to the negative side they could be the catalyst for more company specific downgrades than upgrades, plus some spill-over effects to peers. It is also possible, that quarterly estimates are somewhat stale a few months out from the start of the reporting season and analysts only review their estimates as the reporting date approaches, which then reveals the gap between their bullish bias and reality.
Consensus earnings estimates for the upcoming quarter generally fall ahead of the reporting season, the vast majority of companies then manage to beat those lowered expectations...
Finally, it is possible that companies manage expectations lower, deliberately sounding more cautious ahead of the reporting season in order to lower the bar and making it easier to claim they beat expectations on results day.
Investor surveys arguably offer the most accurate glimpse of investor expectations.
Analyst consensus is used largely because the more useful investor consensus is difficult to measure. We can try to back out market participants’ earnings growth expectations from the price, assuming the current PE (based on analyst consensus earnings) reverts to the historic average, but this approach also has many limitations.
Investor surveys arguably offer the most accurate glimpse of investor expectations. There are always issues over how truthfully such surveys are filled out, sample size and the fact that surveys aren’t available on demand, but at least they offer an occasional snapshot of what the investment community thinks.