Guest contribution by Ilargi
October 28, 2010
This is a guest post by Ilargi, editor-in-chief for The Automatic Earth.
In anticipation of tomorrow’s announcement of Q3 GDP by the BEA, here’s a comparison of the BEA numbers with a series of other data.
Let’s start with a bit of “history”. I first published on the Consumer Metrics Institute‘s CMI growth Indices in August ’10, after seeing Doug Short’s graphs in which he combined the CMI 91-day index with BEA’s GDP numbers and the S&P 500.
While the comparison is inviting, as is obvious to anyone who sees it, it’s not perfect. So I began to tinker with Doug’s work. The first step was to move the CMI 91-day data forward vs GDP by roughly one quarter. Not only does this line up peaks and troughs with remarkable accuracy, it also addresses the fact that the 91-day index is a leading indicator, something that’s easy to see when you realise that it’s updated daily, while this week’s Q3 GDP numbers are published a full 4 months after Q3 began. It looks like this:
I then also shifted the S&P data, something that’s a bit harder to explain as both an act in itself and as a phenomenon. What I can say is that the S&P is highly volatile and follows the GDP numbers with, again, remarkable accuracy, only it does so with about a one quarter delay. Now, most of us will feel that the markets have a leading role in the economy, and they do, but it’s not the only role they play. They are very much followers too, something I’m convinced we’ll see confirmed when in Q4, our present point in time, economic numbers will deteriorate. The result is:
The problem of course remains the same: we won’t be allowed to know what happened in the past 4 weeks until 3 months from now. At least not by the BEA. However, if we can build a sufficiently solid case for CMI data as a leading indicator for “official” GDP numbers, perhaps we won’t have to wait that long. We won’t be able to offer 100% certainty, since we have no crystal ball, but we might yet be able to make you give this some serious thought.
Friday’s Q3 GDP numbers are not all that important. As Doug has previously noted, the “real” data won’t be known until well after the November 3 mid-term elections, when the BEA publishes its first revision. Note that the Q2 numbers only recently went through a second revision, which was down 33% from the first one, lowering GDP growth from 2.4% to 1.6%. That is, more than 5 months after the start of Q2, the BEA still needed to revise its data by a third.
Back to CMI. The 91-day growth index is not the only index they publish, and in fact I find it gets a bit too much attention compared to the others. Which are the 183-day and the 365-day index. Since I had never seen these in a graph, I contacted RIck Davis at CMI in September, and he provided one, saying that they’d never done this graph before, and that it does indeed look interesting. Here’s the graph that Doug made which adds these two indices, both with the same one-quarter shift as the 91-day:
I first contacted Doug last week, and I think that was the first time he saw my versions of his own graphs. He called them “fascinating” and asked me for the post you’re reading. The reason I wrote to him, as well as to Rick Davis at CMI, was a question Doug fielded from a dhort.com reader, who wondered how it was possible that the US Census Bureau reported a +7.3% y-o-y rise in retail sales in September ’10, while the CMI 91-day index is tumbling.
Earlier, I had seen a Gallup report on consumer spending that showed a drop of -10.6% over that same period.
These are the kinds of things that make me itchy. All parties involved, Gallup, Census Bureau, BEA and CMI, may have different survey methods, and they may measure somewhat different areas of spending. But an 18% difference in findings is simply too much, it lacks credibility.
So who’s right? At a certain point it comes down to one’s own trust in the various methods. I did ask Gallup for their thoughts on the gap between their data and the government’s, but they’ve unfortunately failed to respond to date. Which is a shame, since even with a bucketful of caveats, the Census Bureau seems to make Gallup look grossly incompetent. And not only Gallup, CMI too.
It would be interesting if Doug can, for a future post, obtain the complete Gallup and Census Bureau files, and incorporate them in a graph. For now, though, we still have to interpret the possible CMI indices. This is no easy matter. As I said, we have no crystal ball. However, we can’t deny that the CMI 91-day index, when shifted, has that eery compliance with the BEA GDP numbers. Still, when we look at the 183 and 365-day graphs, we see a much smoother “motion”. They take away much of the volatility that’s inherent in the 91-day.
We need to recognise that this also means that the 365-day can shift us away, whether up or down, from more recent events, in that it averages out an entire year. And there lies a big clue as to why it is much smoother than the BEA GDP numbers, which measure only 91-day increments. This means that the 365-day then becomes a very solid and reliable index, potentially much more so than the GDP, even if, as I said, it can miss out on very real, more extreme moves.
The expectation among “experts” is that the initial report on Q3 GDP will come in around +2%. Remember, this can undergo strong revisions later on. If we look at the 365-day index in the graph above, we can make a preliminary “prediction” (remember, no crystal ball) for GDP going forward.
As the graph is right now, that is with the 3-month shift forward, Q3 GDP according to the CMI 365-day index comes at around +1.25%. And if I can be bold for a moment, I would say chances are that no matter what number is published tomorrow, the BEA Q3 GDP will be very close to that after all revisions have been done. Sadly, we won’t know until say, Christmas. If we look at what the 365-day would appear to say about Q4, I would say we come to about -0.5% (average the graph over the 3-month window).
While it’s still all too close to crystal ball territory for (my) complete comfort, I am convinced that this is a far more solid way of interpretation of the data then the “bare” comparison of BEA and the CMI 91-day index, because both, on account of their limited timeframes, are subject to high volatility.
NOTE: it is for instance entirely possible that we will decide at some point that the 365-day needs to be shifted forward a little more, which would move the fall in GDP also further into the future. But that is not the most important thing here: we’re not aiming for details, but for trendlines. And the CMI 365-day index, which as of this morning stands at -2.86% (even as the 91-day shows signs of “recovery”), looks like a solid indicator for such a trendline, far too solid to ignore. Also, seeing that in the most pronounced part of the last downturn, Q4 2008, the 365-day never went below -2%, while BEA GDP scraped -7%, we may well have something worse than -2.86% ahead of us.
Lastly, I remain curious about the discrepancy found in the Gallup and CMI vs the “official” data. And I’m not ashamed to state that I, like many others, have strong doubts about just about any number the US government publishes these days, whether it’s GDP or unemployment.
In the latter case, we have a “second opinion” available in U6 vs U3 data, which indicate a US unemployment rate of 17.1% instead of 9.6%. What we try to do with the CMI data is seeking to find a similar “second opinion”, this time for GDP and consumer spending.
To be continued, no doubt.