More on corporate earnings and analyst error

As a follow-up to our earlier post on corporate earnings this season, we’ve put together some longitudinal data courtesy of Bloomberg’s wonderful little EA function, which those of you with a Bloomberg service can double check if you want to make sure we’ve got our maths right. Bloomberg data only goes back to Q405 earnings (and shows up as Q106 reporting), but still, that’s 17 periods including the current one, which is somewhat more than halfway finished. And, yes indeedy, something weird is going on this quarter—but it may just be a continuation of a trend that started in the first quarter of last year.

The handy little EA function, remember, lets us look at earnings surprises (and sales, too, if we’re interested in that). And it breaks the data down by geography—you can drill down to the individual country level, or keep it at regions, which is what we’re doing here. And if we compare EPS earnings surprises by region, and exclude the current quarter, we see a sharp difference between Western European companies and North American companies (US, Canada, and, um, Bermuda—don’t ask). For Western European companies, it’s pretty evenly split between positive and negative earnings surprises—in some quarters there are more positive surprises, and in some quarters there are more negative ones. Over the entire period (again, not including the current quarter), the positive/negative ratio is about as flat as flat can be—1.01 times. Here’s what the data look like (note that the Average does NOT include the data for the current quarter, since it’s not yet finished, and that the data is as of 1 March):

(If you click on the chart, it magically becomes legible!)

Now, this is really what you would more or less expect—all the percentages remain roughly the same over the entire period, except for the current one, where the total % of earnings surprises (both positive and negative) appears meaningfully higher than average. Now, it may be that this will regress to the mean as the rest of earnings come through—we still have lots of companies to report. So we’ll just have to wait and see. But if analysts are doing their jobs, then companies are doing what you would expect—some are a bit higher than expected, some are a bit lower, and it’s roughly comparable over time.

If we look at North America (which, let’s face it, is mostly US companies), a different pattern emerges:

In fact, several differences are apparent. First, there is no quarter when negative earnings surprises exceeded positive earnings surprises. In fact, the average positive/negative ratio is 1.33—there was only one quarter of Western European earnings that exceeded this ratio at all, and most quarters didn’t even approach this level. So North American data seem to be positively biased relative to Western Europe. Second, while negative surprises are a bit higher in North America than in Western Europe (20.73% versus 18.64%) this doesn’t look that material–it could be, I suppose, but I’m not sure there’s enough data here to say. However, positive surprises are a whole lot higher—27.06% versus 18.78%. Even for a data set this small, this looks pretty compelling. And this pattern has been pretty consistent over this entire period. Third, it’s this difference in positive surprises that accounts for the overall result of there being considerably more earnings surprises in North America than in Western Europe (47.43% versus 37.42%). Lastly, nothing in any of this so far in any way explains the sharp escalation in positive earnings surprises in 4Q09, where positive surprises account for 42% of ALL earnings monitored by Bloomberg so far this period, and total surprises are nearly two-thirds of all reported earnings.

Again, let’s be clear—we could still regress to the mean in the balance of this reporting season. But that would still leave unexplained the fact that North American companies routinely generate positive surprises at what appears to be a higher rate than Western European companies. Or, as we mentioned last time, North American analysts do a considerably poorer job of predicting earnings than do their European counterparts.

OK, now, this is all good fun for the statheads among us, but there are some needed caveats. First, this isn’t a real longitudinal survey—for that, you would need a decade or two of data, and we don’t have that through Bloomberg (maybe Thomson Reuters does, but I don’t subscribe to them). Did a similar pattern emerge in the 2001 recession, for example? We don’t know. Second, we have no way of knowing at this point whether the relative size of the data sets has any impact—the North American data set is about four times the size of the Western European one. It may be that there is a higher number of small cap names in the North American set that produces some excess variability—but that still doesn’t explain the directionality of the data. But it is a potential concern.

So we’re back to the conundrum we mentioned last time—why do North American analysts consistently under-predict earnings in a way that their European counterparts don’t? Well, European managements could be feeding more interesting data and commentary to analysts than their North American counterparts. That’s something I would have entertained as a possibility fifteen or even ten years ago, when that sort of interaction was pretty routine, but it’s hard to see it occurring now, frankly. Second, maybe North American analysts have gotten more burned than their European counterparts, and are therefore more conservative in their predictions. The problem with this is that up to, say, the second quarter of 2009, there’s no evidence that North American analysts were burned at all—positive and negative surprises held pretty constant. But there is one interesting aspect of the North American data prior to this past quarter, and that’s the steady rise in the positive/negative surprise ratio during 2009—it’s higher each quarter. But that’s the only thing that really changes, and if anything it seems to derive from a decline in negative surprises during this period that mirrors an increase in positive surprises. But even if the current quarter is a continuation of a previous trend, what accounts for that trend?

Why do we care? Well, as is the case for anyone who works in the financial sector, I’m interested in a number of things, one of the key ones being sentiment. And we’ve had several quarters now of positive earnings surprises which get touted by the likes of CNBC as being meaningful of, well, something. We’re booming into a recovery, earnings show. Look how much better companies are doing. Well, leaving aside the fact that in general many companies are still doing worse than they were a year or two ago, and the fact that one fundamental reason for decent earnings at all is the pace of restructurings (ie, layoffs), what this might really show is that sentiment is possibly being distorted. North American (read Wall Street) analysts are doing an even poorer job of predicting the future than they usually do, but the spin is that companies are doing better than expected, so party on, dude. Given how much credence the business media in particular, and the investment community in general, places on these predictions, you would think that there might be some questions about whether this credence is properly placed. As the old saying goes, prediction is very difficult, especially of the future. If the current pattern holds, prediction is even more difficult that it was a year ago. Such sensitive souls. Perhaps we should pay less attention to these predictions—or, at the very least, less attention to the media spin-meisters and talking heads who still take them seriously. Maybe it’s time to remind people of that other old saying about Wall Street—“Often wrong, never in doubt.”

Categories: Business/Finance

11 replies »

  1. Let me begin with a caveat: I’m not a finance guy, and IR isn’t where my experience and expertise lie. So there may be lots of good reasons why I’m wrong here. If so, we have a reader or two who can point my errors out.

    That said, I DO know a little about the organizational psychology of American companies, and what I know suggests a theory. If XYZ Corp is a public company of the sort you’re talking about here, a lot of people are ultimately invilved in the end-to-end process of getting the company to its final predictions, which somebody at the company then feeds to the analystas. Those predictions encompass performance numbers and prognostications from all over the business, which means that there are plenty of points where something might go wrong.

    Now, let me refer you back to something I wrote last year on the dysfunctional role of goal-setting in companies. If you read that report, it becomes clear that people at all levels of an organization have every incentive to game the system, and this is going to be even more true in the kind of recession we’re dealing with right now. Whatever you do, you DO NOT miss goals. Now, the best way to make sure you clear the bar is to set that sucker as low as possible, and I have seen this very thing happen with my own eyes enough to assure you that this is a gospel fact of daily life in corporate America.

    If everybody in the company is gaming the goals downward in response to organizational goal programs, then it’s hard to imagine how the people close to the company/analyst touchpoint could effectively communicate accurate forecasts. There’s a systemic, cultural dynamic at work, and it assures that the company is going to lie to the analysts, leading them to lower projections which then result in the appearance of better performance.

    JS noted when we discussed this before that the market HATES missed projections, regardless of which direction they occur, and I believe that. However, a corporate director evaluating the Q2 performance of a group never fired anybody for coming in at 105%. The only problem that creates is that the people setting the goals say heck, we didn’t aim high enough, so let’s jack the Q3 goals up some more. No good deed goes unpunished, right?

    However, this only has to happen once for the rats to learn, and the logical response is to redouble their efforts to push goals down in the future – which means that the dynamic I’m theorizing here is of the “downward spiral” variety.

    If I’m wrong here, tell me why. At a glance, though, what you’re describing makes a good degree of intuitive sense to me.

    • Further thinking. It seems like there are two pivot points of interest. First would have been SarbOx, which went into effect in 2004. That would theoretically have driven more conservatism with corps, which would in turn have led to more positive surprises. But we don’t have that data here – like you say, we’d need to go back further, and in this case, did the numbers change in 2004. The second pivot we see – 2009, when the recession really began hitting hard. I think my org theory above may have played a bigger role in this. The underestimate dynamic was already in place, but something made it worse.

      I passed this onto the CFO here, and he wonders if the differences between US and international accounting principles might have something to do with it?

  2. You could pretty easily design an experiment to test the difference between US and accounting principles. The hard part is getting (ie paying for) the data you’d need to test it. Unlike climate science, there appears to be precious little “open source” financial data of the type that Wuf uses above.

  3. Sam–what you’re suggesting in your first comment does make some intuitive sense. I’m not sure it’s the complete explanation, though–while there are lots of companies that provide guidance to the street and will “help” analsyts (to the extent they are legally allowed to) construct thei models, there are lots of companies that absolutely won’t. And, of course, analysts aren’t stupid–if they think that a company management is deliberately lowballing them, they’ll say so. Keep in mind that there are a couple of things at work here–and one of them is reputational. Analsyts like making money, and one of the things that helps them make money is by being voted top analysts. And consistently missing earnings is a pretty good way to have that NOT happen. While investors probably would prefer that companies beat estimates rather than miss when they do miss, they would rather not see the misses in the first place. And investors are just as likely to blame analysts for this as compay managements. Analysts who consistently get this sort of stuff wrong don’t last all that long as analysts.

    Which desn’t explain the US pattern of the past four years, though. But keep in mind this is all companies that are followed by analysts that Bloomberg covers, and that’s quite a lot, including small caps. It would be interesting to see an analysis of the relative performance of cpompanies in the S&P 500, which is the market benchmark, to determine whither this shows a similar pattern.

    I’m not sure what impact SarbOx has had. An interesting point, but I have nothing on which to base a conjecture. OTOH, I’m pretty sure that differences in accounting standards don’t mean a whole lot–what we’re looking at here is EPS, which is pretty manipulable under both US GAAP and IAS (by manipulating tax rates form one period, for example), which is what most European companies use.

  4. Very interesting.

    I can only think about these things by doing logic trees, and I was able to come up with seven possible reasons analysts might consistently miss low, but none of them are really compelling because as Wufink says, you don’t keep that house in the Hamptons by being wrong a lot. You’d think there would be some sort of natural corrective process, even if it were a crude one, e.g., “We’ve undered Mega-Corp ten times in a row, this time I’m adding 3 cents to the estimate, darn it.”

    One thing I do wonder though is whether analysts who are independent, e.g., Sanford Bernstein, have a different error rate (and bias) than analysts who are associated with companies that make a living through investment banking, selling stocks, or holding inventories of stocks, e.g., on account or through funds. I’d conjecture the latter group might find themselves under more pressure (likely subtle) to err on the side of positive surprises, especially if it’s true as I’ve been told, that the market “punishes” negative surprises more thant it “rewards” positive ones. Any thought, Wufink?

  5. John Harvin–that is an interesting thought, and one could probably investigate it, although it wouldn’t be easy. These are Bloomberg data, which means the surprises are in terms of Bloomberg estimates, and that data set probably fluctuates wildly in terms of market cap–companies with larger market caps, I suspect, probably have a higher number of forecasts going into the consensus for each company. So you have a potential size effect as well as a potential affiliation effect that you could look at as well–are predictions better for larger companies because of a higher number of forecasts gong into the consensus estimate, or worse? Bloomberg usually does give you the names and affiliations of the analysts it uses for each company. Actually digging this out for a large data set such as this one would be extraordinarily tedious, of course, unless Bloomberg itself had a handy little function that let you dig into it sensibly (which they might–I’m constantly surprised at the range of tasks you can do with these things). But then you could actually parse by affiliation, if you actually know them all, and knew whether they were independent or not. Sounds like a good project for someone looking for an advanced finance degree. The notion is certainly plausible. Again, I know Thomson Reuters has a collection of data like this as well, which may go farther back than the Bloomberg data does. Maybe Nelsons does too.

    The other way you could test it, I suppose, would be to see what happened at Citi (my old firm, although I was on the credit side) when Sally Krawcheck was hired from Bernstein to oversee equity research. Did analyst predictions get any better, or worse, or what? I have some thoughts on the subject, but I’ll keep them to myself.

    Of course, even if you came up with a plausible explanation for why US analysts consistently missed low, you’d still have to explain why European analysts don’t.

  6. Actually, I’ll check at work tomorrow and see if there’s any further possible granulation to the Bloomberg data. If you could limit it to particular indices (S&P 500, for example), you could probably look at the potential size effect that way. And I’ll check on the affiliation issue as well. You never know.

  7. Well, I did some further digging, and I have to say it doesn’t really clarfy anyting–if anything, the reverse. Bloomberg doesn’t let you disaggregate the forecasters by institutional type–bank, broker or independent research house. However, it does let you look at the data for a particular index–the S&P 500, for example. This is perfect, because this lets me deal with the large cap/small cap issue. So if you do the same anaysis on the S&P 500–500 of the largest, most representative and most analyzed companies in the US, what happens? It gets worse. I won’t bother reproducng the table, but the gist is that the trend to positive suprises, which in the overall universe gave us a 1.3X positive/negative surprise ratio over the past four years, just simply explodes. That 1.3X ratilo doubles, to 2.6X. And the total number of surprises, both positive and negative, also increases. We again see the pattern of a dramatic increase in this ratio in 2009, though, which I suppose is vaguely comforting, although this leads to a 5.2X multiple in 3Q09. The current quarter is running about 3.35X. So the larger and more widely followed the company, the greater the likelihood of (1) not getting it right in the first place, and (2) an increased likelihood of missing low. So the vast majority of smaller cap companies in the overall universe actually moderates the impact of the 500 largest companies. Not what I would have expected.

    For Europe, we can use the Bloomberg 500 as a proxy, even though it only consists of about 300 names or so. But it’s the only cross-Europe Bloomberg has that one can use. And It starts looking more like the overall US universe–what had been a flat as a pancake positive/negative ratio rises to 1.2X (it’s running well below that this quarter, though, at 1.06X so far). There are still positive and negative quarters, just fewer negative ones, and the positive ones are a bit more positive. There is no noticable 2009 effect as we saw in the North American or S&P 500 analyses, though. And as is the case for the North American and S&P universes, the overall number of surprises increases.

    So what does this tell us? Darned if I know, other than the fact that a universe of large, important and widely followed companies produces more analyst errors than does the overall universe in both North America and Europe. Great.

  8. I would be very interested in your thoughts on whether Krawcheck had any effect at all, especially since she was so outspoken for so long on the need for independence.

    I look forward to the size effect analysis, if it comes off.

    This is cool stuff. Very thoughtful.

  9. Well, I have to qualify my comment by noting that (a) I was in credit research, and (2) I was in London. So two degrees of separation. That being said, we in London got lots of feedback from New York equity and credit analysts. She was liked, I think, but salad dressing, Sandy going over the top as usual. No one could believe what she was being paid, so that probably got a whole lot of noses out of joint, whch certainly didn’t make her job any easier. The main thing is that the heavy lifting in response to Grubman was already being implemented by legal and compliance. And the main–indeed, only–impact, aside from endless compliance meetings and money laundering internet tests, was to cut everyone off from Capital Markets–but not from sales and trading. Hah. So much for independence. In fact, it reached a point when we couldn’t even be called analysts any more–we became Credit Sales Specialists. It was an interesting irony that the main impact of Grubman/Spitzer was to more or less leave equity research intact, but to decimate credit research in both New York and London, going from one of the top ranked teams to no team at all.