Unintended use for stock analysts

August 12, 2016

Despite much suspicion regarding the “quality” of analysts’ earnings forecasts, there is valuable information in these. An unintended consequence of this is that analysts’ earnings forecasts can be used to enhance year-ahead macroeconomic growth forecasts. In fact there is evidence that stock earnings forecasts are more useful for this than economists’ economic growth forecasts.


An unintended use for stock analysts

I have spent the vast bulk of my career physically surrounded by stock analysts. On many occasions I have listened to equity sales people, fund managers, etc make disparaging comments regarding the value of analysts, and I must admit I have thrown out the odd joke at their expense. In a former life, as Head of Quant for one of the big international investment banks, the Head of Strategy and I only half-jokingly suggested to the Head of Research that he fire the entire research team (except us of course!) and have strategy and quant replicate most of what the stock analysts did.

Do analysts actually add real measurable value? And if so, how? Through earnings forecasts? Through recommendations? By providing access to company management and deal flow?

The short answer is “yes”, there is statistically significant value in aspects of analysts’ output. The slightly longer answer is “it depends”, and beyond that it gets complicated fast. A few years ago I worked on a detailed review of the academic literature analysing analysts. It was immediately apparent that there is a vast body of research investigating analysts. One paper from 2008 identifies 250 related articles published in just 11 journals from 1993 to 2006. Expand the journal list and the time period covered and that number runs into the thousands – highlighting (a) what a complicated multifaceted question this is; and, (b) that there is plenty of research fodder for academics to cut their teeth on.

In this occasional research series I intend to periodically revisit various aspects of analysts’ output. For this current edition, I will focus on one particular point that I believe is extremely useful, but mostly unknown – that stock analysts’ earnings forecasts can provide fresh insight into the outlook for the macroeconomy.

Firstly, it is important to note that there is evidence of statistically significant information in analysts’ earnings forecasts. All jokes aside, and recognising forecasts are not without error, analysts do have some forecasting ability for earnings. Forecast quality is a very complicated question and is a function of myriad factors (analyst experience, firm investment banking activity, how analysts are remunerated, etc). I will address some of these issues in future papers. At this stage the key point is that, on average, there is some useful information in analysts’ earnings forecasts (many academics across a range of markets have concluded so, subject to sets of conditioning factors).

We also know that firm earnings and the macroeconomy will generally be positively correlated (with exceptions for certain firm types). In other words, we know there is a business cycle. So, if firm earnings are correlated with the macroeconomy and analysts’ earnings forecasts have some ability to predict future earnings, then can we use stock analysts’ earnings forecasts to help improve macroeconomic forecasts?

Yes we can. While for any given equity market there are usually significant differences between the industry composition of the market and the industry composition of the economy, the two are typically quite strongly linked. By aggregating across analysts so as to diversify away the stock-specific noise/error in their earnings forecasts we obtain the core common component, which is the market expectation. If aggregated earnings expectations have some predictive power for market earnings, and market earnings are positively correlated with the macroeconomy, then we have a stock-driven factor to assist with the formation of economic expectations. This is despite the fact that analysts’ forecasts show significant signs of herding (tight grouping, avoiding bold forecasts) and are typically anchored around management earnings guidance.

It should also be noted that analysts’ earnings forecast will, either implicitly or explicitly, incorporate an outlook for the macroeconomy. Texts on stock analysis commonly recommend analysis of the macroeconomy and relevant industries as the first step in developing a view on individual stocks. A relevant question is then whether analysts’ earnings forecasts are simply noisy estimates of the consensus outlook for the economy, or whether they provide additional information? My research suggests the latter. I have in the past produced academic research on this topic and found evidence that there is additional useful information in aggregated stock analysts’earnings forecasts for the macroeconomic outlook – beyond that available from economists’ forecasts and other typical factors employed in economic view formation (e.g. confidence surveys, inflation, interest rates, default spreads, etc).

However, it is not simply a case of aggregating the analysts’ forecasts and employing these in analysis. It is unfortunately somewhat more complicated than that, with a number of additional features of companies determining how useful their forecasts may be for macroeconomic expectations. Not surprisingly, the earnings forecasts of firms in cyclical industries (e.g. materials, industrials, consumer discretionary) are more helpful, given variation in their earnings is more strongly positively correlated with the macroeconomic backdrop. The same is true of smaller firms, and smaller cyclicals in particular.

One possible explanation for why smaller firms are more useful generally, is evidence of greater earnings manipulation by larger firms. In the first edition of this research series I discussed earnings manipulation in the form of loss avoidance. Other forms include smoothing and earnings aggressiveness (bringing forward earnings). Any manipulation of earnings can reduce the usefulness of forecasts in those companies. For example, I have found evidence of an inverse relationship between earnings smoothing by management and the predictive power of earnings forecasts for those companies; i.e. the more management smooths earnings, the less useful forecasts are for predicting the macroeconomy. Contributing to this phenomenon is herding of analysts’ forecasts around management earnings guidance. Hence, smoothing by management will be reflected in forecasts.

Complicating the application of this to practical asset management, there is time variation in the effect – not least because large cyclical firms tend to smooth their earnings more in economic downturns – so much so that the relationship reverses: more negative earnings expectations equate to stronger positive macroeconomic growth in the future. I should also note, for anyone contemplating the aggregation of analysts’ forecasts to model these effect, there is a range of critical (and often very messy) features of forecast datasets (e.g. I/B/E/S, Bloomberg, etc) that need to be accounted for to obtain useful information (e.g. rounding errors, incorrect announcement dates, historical actuals data errors, etc).

My academic research in this space focuses on the US (improving year-ahead macroeconomic growth predictions), given the depth of data available. Nonetheless, I have also found evidence of this phenomenon in many other markets and a number of the core concepts are incorporated in the systematic processes employed for the Firth Asia Systematic Equity strategy (FASE).

One of the tests I have performed on this phenomenon is the relative usefulness of aggregated analysts’ earnings forecasts versus economists’ forecasts for macroeconomic growth, i.e. which is more useful for forecasting the economy? Amusingly, despite economists explicitly forecasting factors like industrial production growth and stock analysts only implicitly doing so, I have found evidence of greater marginal value in the analysts’ earnings forecasts; e.g. aggregated analysts’ earnings forecasts are more useful for forecasting US industrial production growth (a year ahead) than economists explicit forecasts of industrial production growth. The analysts beat the economists.

In our FASE strategy we endeavour to incorporate a range of information sources that our research suggests are not fully incorporated into stock pricing. The usefulness of stock-level data for macroeconomic view formation is a key aspect of the process, and one which we believe is little-appreciated in the marketplace. While doing so can be somewhat complicated in practical terms, we believe there are significant benefits to be gained from assessing analyst earnings forecasts not only in terms of their direct implications for the stocks in question, but also their implications for wider (and longer term) market views. At the simplest level, practitioners should employ aggregated consensus earnings expectations as a core component of their macroeconomic forecasting.

In the next edition of GIGO I intend to return to the issue of earnings manipulation by management, with particular reference to earnings smoothing. Please do not hesitate to contact me regarding the hard numbers and research underlying the points made in this brief paper and/or to discuss the relationships between analysts’ earnings forecasts and macroeconomic prediction.

* Macalister, H.C. (2011), Aggregate Earnings, Forecasts and Revisions: Evaluation of the Information in, and Characteristics of, Aggregated Analysts’ Forecasts, University of Auckland

Written by: Dr. Hamish Macalister