Illargi [Co-editor of The Automatic Earth]: As you probably know, I’ve written extensively on the Consumer Metrics Institute and its graphs and data. The CMI is, in my eyes, very useful, and different from others, in that it tracks the behaviour and spending habits of consumers, who make up 70% of US GDP, today, and updates its data on a daily basis.
By comparison, the US Bureau of Economic Analysis, which reports on “official” GDP numbers, publishes its findings once a quarter, and then only one full month after a quarter has already expired. In other words: we won’t know what GDP did in the first two weeks of July until early November. This makes CMI a leading indicator vs the lagging BEA.
In late August, I showed what this implies, in the following graph, based on Doug Short’s great graphs, which are in turn based on CMI data. The idea behind it is that if you let the GDP data where they are in a graph, you can shift the CMI ones forward by roughly a quarter. Similarly, the S&P 500 lags the BEA data by about an additional quarter, so it can be shifted backwards about that much.
I use terms like “about” and “roughly” to indicate that what interests me here is not scientific rigour, but exploring and lining up trends. Looking at the data, it was clear that peaks and troughs line up quite strongly, once you allow for those “time-shifts”. Not all that crazy a notion, as the CMI itself confirms. More on that later. First, that graph I made in late August:
Inside that yellow ellipse on the right hand side was my “prediction” of US GDP, based on the 91-day trailing CMI Growth Index. And that kept irking me a bit; it seemed a little steep. It took a while to realise what was happening.
I knew the CMI also publishes 183-day and 635-day trailing Growth Indices, but I’d never seen those in a graph. So I contacted the Institute, hoping they could help. Turns out, they could. Actually, they were happy to. The first lines from Rick Davis at CMI were:
Ilargi: Thank you for your coverage of the Consumer Metrics Institute! I have become a devoted fan of your writing — I aspire to skewering the economic establishment with similar panache.
It’s always easier to talk on that sort of basis, mutual respect. About the graph above, in which I shifted the timing of the data around, Rick says:
I really like the chart you made, complete with the implications that our data has for the GDP. The wild-cards in all of the GDP data are inventory builds, exports and industrial stimuli — all of which should reverse or soften in the 3rd and 4th quarters. It will be interesting.
Rick graciously agreed to do a graph just for me with all three Growth Indices combined, about which he says:
Attached is the chart I promised which includes all three of the growth indexes that you requested (91-day, 183-day and 365-day). We had never looked at all three of them superimposed upon each other like this before, and I agree with you that the crossing points are interesting.
Before getting to that graph, though, I think it’s good to show you another one, made again by Doug Short, because it shows to what extent the 91-day Index already smooth es out the day-to-day data:
And that smoothing was what I expected, and was hoping to find in the combination of all three indices. Found it! Here’s the graph Rick Davis at CMI made for me:
The implications are plain to see: the more time an Index covers, the more the extremes are dulled. Also, there’s a time shift happening here as well, in that the more time an Index covers, the later the peaks and troughs appear. Where in the first graph above, GDP data, extrapolated from the 91-day Index, seemed to fall to -5% (averaged out) in the fourth quarter, an average of all 3 Indices would show a less hefty picture.
To bring this into the scenario painted by Doug Short’s graphs, and the one at the top that I based on that, here’s two more graphs. Please note that since CMI and Doug Short don’t use the same colours for every index, I had to play with those as well in order to combine them. So, in the CMI graph, 91-day is blue, but below it is red. The 183-day is green for CMI, and yellow for me. The 365-day is red at CMI, and light-blue for me. And then there’s an adjustment for horizontal and vertical proportions too.
We can do either one of two things with this. We can allow for the time-shift inherent in the CMI data when positioning them inside Doug Short’s GDP graph, like this:
Or we can leave them as CMI itself has allowed for:
This last graph has the advantage of providing more of a glimpse into the future. Granted, this means we’re cheating a little here and there, but not nearly enough to dismiss the data offhand. The upshot is that, as mentioned before, the CMI Indices follow consumer behaviour on a daily and constantly updated basis. The 91-day Index, which has the least data to rely on, will always show the most volatility, and point forward more than the other two. You choose which one you think is more accurate.
Please note the GDP Q3 and Q4 data (dark green bars), which I this time around based on a guesstimate of the averages of the 3 CMI Indices, not just the 91-day. They are less negative than in the first graph, but still solidly less than zero (-1% in Q3, -3% in Q4). Note also that in the case where we’ve provided for the time-shift inherent in the graphs, GDP would likely be far more negative than in the case where we don’t. For now, I based the projection on the mid-case, the yellow line 183-day Index. Since things have worsened substantially towards the more recent data, my projections are more likely to underestimate the fall than to exaggerate it.
Finally, lest we forget, Doug Short also incorporates the S&P 500 into his CMI graphs. Now, obviously the S&P is updated daily. Somewhat curiously, though, it proves itself to be a quite severely lagging indicator. If you align the peaks and troughs we’ve been looking at, which are quite pronounced and far too similar in shape to ignore, the S&P runs about 3 months behind the BEA’s official US GDP data, which means that it’s as much as 6 months behind the 91-day CMI numbers.
And I know people may claim again that The Automatic Earth consistently says that the next big drop in the markets will forecast the next big one in the real economy. How can that be if the data show that the S&P 500 lags American consumer behaviour by half a year? Well, it’s not as strange as it may seem. Apart from flagrant manipulation of the stock markets, of which there is more proof than we would ever even care to see, there’s another factor in play.
People who invest in stocks have no access to real consumer data. Caveat: they would if they paid attention to the Consumer Metrics Institute numbers, which is precisely why I pay so much attention to them. But because nobody does this, the S&P 500 is not an indicator of now, but of quite a few yesterdays ago. The time lag it has allows for a rising stock market at the same time that unemployment rises (forget the equally lagging U3 9.6% unemployment), and foreclosures surge. It’s all just a matter of time, or timing if you will.
If stocks keep on trading at the very low levels they have for months now, it’s certainly possible for the Fed or the Treasury or the Plunge Protection Team, or anyone else (HFT?!), to manipulate the data upward. And from what we’ve seen lately, there’s little doubt they’ll try. Still, this doesn’t really change anything solid, other than the time lags in the graphs we just looked at. The consumer part of the GDP, which is some 70% no matter what, has been showing negative growth for a long time according to the CMI data. And there’s doesn’t seem to be any way, other than divine intervention, that this will not eventually reflect in the GDP and S&P 500 numbers. Again, it’s all just a matter of time.
Here we go, adding in the S&P. First “bare”:
And then with GDP projections for Q3 and Q4:
Now where do you think, looking at the correlations between the various data, that the S&P is most likely to go in Q4 2010? How about Q1 2011? The Automatic Earth is not here to dole out investment advice, but all the same, does this look to you like a good time to buy stocks? Sure, there will always remain questions about the above until we see the actual numbers. But by the same token, we’re way beyond crystal balls and tea leaves here; the Consumer Metrics Institute are not exactly a bunch of empty cone heads.
And besides, you won’t know what really happened until 3-4 months after it did happen. And that, certainly in the case of such relatively powerful swings as we’ve been contemplating, will probably be too late for you to change course.
Simple as that, really. Feel lucky?