Companies don’t like losses

Firm management can actively seek to avoid losses by manipulating earnings. However, if a loss is unavoidable, then there is a tendency to get as much bad news out of the way as possible in one hit. This behaviour means that extreme caution is required when assessing whether earnings surprise is really earnings surprise at all.

Companies don’t like losses, unless they are really big losses

The Firth Asian Systematic Equities Strategy is at its core a fundamental quant process. Effectively that means it consumes a substantial volume of accounting information for each stock analysed – dissecting income statements, balance sheets, cash flow statements, etc – in an effort to identify inefficient processing of that information by the market.

However, by relying heavily on such data an obvious question (and one that I have indeed been frequently asked) is how reliable such information is? This is clearly a critical point. If the data were to be completely unreliable then can any process using such data be robust? If the data is not completely unreliable, then how reliable is it? Is there value to be extracted from such data?

This brief paper is the first in an occasional series I will pen discussing various aspects of analyzing stocks with known (and unknown) errors in the data. From bias in analysts’ forecasts to managements cooking the books, each paper will introduce one quirk of stock analysis that investors should be on the lookout for.

Before focusing in on this first paper’s discussion point it is worth making the general observation that errors in data, manipulation of earnings components, distorted accounts, etc do not mean that analyzing such data is a waste of time. There are (usually) limits to the extent to which, for example, company management can skew results, so most sets of accounts have at least a kernel of truth in them. In addition, you may not want the data to be 100% reliable, given that added level of uncertainty assists in the generation of market mis-pricing of the fundamentals and consequently results in alpha opportunities.

It is also true that employing data with errors is quite acceptable (I would argue) if you understand the nature and potential magnitude of those errors. All investment analysis is subject to error and uncertainty. But if through analysis and experience you have a strong understanding of those errors you can adjust your strategy accordingly. If, say, a company announces earnings per share of $1, some will take that at face value. However, there are myriad assumptions, simplifications and potential manipulations that went into producing that number. The “true” number may be anchored around $1 (although it may not be) but there is a distribution of uncertainty around that. A key focus of the Firth Asian Systematic Equity strategy (FASE) is understanding the nature of that distribution and hence appreciating just how far data points can be relied upon.

The FASE strategy is an Asian strategy. In my too-many-to-admit-to years analyzing stocks in this part of the world I have encountered much skepticism from investors regarding the quality of Asian data. Some of that skepticism is warranted, and some is not. The Asia region encompasses a wide spectrum of regulatory regimes and accounting standards from international-quality developed markets (which may still have limitations!) to much weaker governance regimes in some emerging markets.

These differences matter. There is plenty of academic evidence highlighting a significant INVERSE relationship between earnings quality and the cost of equity, and, that lower earnings quality is associated with lower stock market volumes.

The first form of bias in stock data I will introduce in this series is loss avoidance. This is the phenomenon whereby companies will go to great lengths to avoid losses and/or disappointing relative to expectations.

That sounds rather obvious. We have all seen the damaging price effects of companies under-delivering relative to expectations. However, there is an interesting twist to this phenomenon that highlights manipulation of results by managements to avoid losses, and that manipulation therefore reduces the information (and thus the usefulness) of those accounts.

Prior to announcement of stock earnings, analysts will publish their forecasts. Putting aside issues regarding the forecasts themselves (which have more than enough quirks to keep this discussion series going for quite some time!), one might think that the actual announced results will follow a pretty standard symmetrical distribution around expectations. i.e. the company signaled their expectations to analysts and the market, and then announced something very similar to that, with actual results evenly spread between some a little less than expectations and some a little more than expectations.

However, looking over large numbers of announcements, that is not what happens. There is much academic evidence to show that companies will actively manipulate earnings to avoid disappointing. In fact some academics have gone so far as to find evidence for a hierarchy of management bias:

1. Avoid outright losses;
2. Report positive quarterly growth; and
3. Meet or beat analysts’ expectations.

One of the more damning pieces of evidence for management manipulation rather than simple random error is research suggesting that companies will not only try to avoid losses, but will, in particular, try to avoid small losses. It’s a little like a friend who breaks your trust. They only have to do it in a small way to ruin the relationship, and it takes a lot of good work to bring it back to where it was.

Similarly, companies know that a small loss will be very damaging to perception of the firm (and of course there will be all the flow-on implications for job security, bonuses, options, etc – all of which contribute to the bias in the first place). However, they also know that if you cannot avoid a loss then you may as well make it worthwhile…so pack in as much bad news as possible and make it a much larger loss. i.e. the marginal damage to firm reputation, the stock price, etc of a large loss may not be terribly large relative to the damage caused by a small loss. It’s a little like the loss itself is a binary event for the firm.

One of the early studies in this space found evidence of 30-44% of US firms with small losses manipulating earnings to generate small positive results (e.g. bringing forward future earnings or pushing out costs—we are not talking about illegal manipulation, but rather management via the flexibility of accounting treatment in GAAP). Many papers since have also found evidence of substantial manipulation through loss avoidance (including in Asian markets). Unfortunately there is also published academic research indicating that stock analysts are not particularly good at identifying which companies engage in such behavior.

This produces some very strange looking distributions of earnings results relative to expectations. Simple random error should result in a distribution of results relative to expectations (over many companies and market environments) akin to that in figure 1 – roughly symmetrical, but with a small bias towards positive surprise. In fact, there is academic evidence of the distribution looking more like that in figure 2 –

bias towards positive surprise plus an “avoid-at-all-costs” plunge in the distribution for small losses, and a more regular tail for large losses. So what does this mean for investors and how does it impact FASE?

In a perfect world we would want to identify error in the prediction of earnings before manipulation and error in predicting whether earnings are manipulated or not. However, that is exceptionally difficult. What we can do is simply take the position that small positive surprise may not be so positive. If many companies go out of their way to engineer positive surprise, then a small positive surprise may not be particularly good. Hence, we would generally view small positive surprise as really being no surprise at all.

Secondly, we seek to identify warning signs regarding firms’ ability to maintain earnings growth rates. Knowing that firms will try to avoid small losses until a loss is unavoidable, we attempt to identify where the quality of those earnings is deteriorating and thus signaling increased risk of disappointment. We then simply avoid such companies.

Thirdly, evidence of bias in losses/ disappointment towards trying to “clear the decks” and get as much bad news out of the way as possible, has interesting implications for market assessment of the risk of the firm. Our experience with the FASEF system indicates the market may not fully appreciate this behavior by company management. Specifically, the negative after-effects of loss announcement can linger far beyond the point in time where the loss is relevant for the stock price. Indeed on many occasions I have met with investors and discussed stocks that systematic strategies like the FASE system favour, only to have such stocks dismissed by the investor because of disappointment that may be years old. It is these kinds of inefficiencies in market perception (amongst others) that the FASE strategy is specifically designed to capture.

Loss avoidance therefore has a real impact on management behavior, market reaction and our systematic strategy.

This first paper, and those to follow in this series, are intended to be brief introductions to some odd (but very important) behavior relevant to fundamental stock analysis. Please do not hesitate to contact us for some of the hard data and the academic research behind these comments. Happy fundamental investing!

* Burgstahler, D., and I. Dichev (1997), Earnings Management to Avoid Earnings Decreases and Losses, Journal of Accounting and Economics 24(1), pp. 99-126.

Written by: Dr. Hamish Macalister