On the last Tuesday of every month two pieces of data are released. We get the monthly update on the Case-Shiller Home Price Index, which is based on thousands of real estate transactions, in which real families spend what is often their life savings. And we get the monthly number from the Conference Board’s Consumer Confidence Index, which is based on a short questionnaire sent to 5000 households asking them how optimistic they feel.
Yesterday the C-S 20 City Composite showed a nice little uptick, +0.8% for April, +3.8% year on year. That reversed a few months of downticks and was reassuring to those of us rooting for stability in the housing market.
The CCI, on the other hand, was down to 52.9 from 62.7. That undoes two months of gains and returns us to a point just above the March level.
How did the stock market react? Was it cheered by what households actually did with their money in April or depressed by what they said to pollsters in June? The S&P was down –3.1%.
As I said just yesterday, I take all survey results with a lump of salt. And I am particularly impatient with surveys that ask about what somebody intends to do when you could just as well look at what they actually do.
Of course, there is the question of time lag. We like to hear exit polls on election day because the real numbers won’t be available for several hours yet. Similarly, although the Bureau of Economic Analysis publishes a monthly number for consumer expenditure, they take nearly a whole month to put it together, meaning that the May number just came out two days ago.
The Conference Board, on the other hand, puts out its June number at the end of June. So you don’t have to wait to find out how consumers actually voted with their dollars, you can just read what are essentially economic exit polls.
Ah, but just how accurate are those polls? The answer is modestly. The CB charges some crazy amount of money for its historical data, but I managed to find a bootleg series on the web. The monthly CCI correlates with the change in personal consumption expenditure for the same month at 0.29. That’s higher than I was expecting to find, but I think that most people who treat the monthly CCI as meaningful news would have assumed much higher still.
Moreover, the change in consumer confidence and spending has a lot to do with change in consumer income. The correlation between change in income over the past six months and the CCI is 0.68 and the correlation between six month change in income, lagged a month, and monthly change in expenditure is 0.19.
So just looking at the six month change in income as of May gives you about two thirds of the CCI’s accuracy in predicting June’s expenditure number. Six months change in income was the first thing I tried. I am sure that given a day to fiddle I could build a model out of BEA numbers that did just as well, if not better than the CCI.
And there is another problem with that modest 0.29 correlation. The numbers I have are revised. What the CB releases on the last Tuesday are “preliminary” numbers. The final, revised, numbers do not come out until a full month later. So yesterday, when the CB told us that the preliminary June number was 52.7, it also informed us all that the final May number was 62.7 rather than the 63.3 they previously reported. April had originally been 57.9 before being changed to 57.7.
If I had the patience, I could go through years of monthly news articles and compile my own data series of preliminary CCI numbers. I really don’t.
Those revisions do not appear to be large, but cynic that I am I will assume that the correlation between preliminary CCI numbers and consumer expenditures will not be quite as high as when the final numbers are used.
But more to the point, revising a number that is primarily useful as a short-term indicator a month after it is first released, long after anybody is paying attention, is a red flag. I am truly scratching my head trying to imagine what these revisions could possibly be based on. The data is from a survey of 5000 households conducted by TNS, “the world’s largest custom research company.” What do they need an extra month to work out?
So if the CCI is not so hugely useful in predicting consumer expenditures, which are in turn only a portion the US economy and a tiny slice of the world economy, why does the stock market, and everybody else, take it so seriously? Because it does.
Keynes once brilliantly explained the stock market as being similar to a form of beauty contest then popular in British newspapers. Readers were shown pictures of dozens of young ladies. They were asked to pick their five favorites, with a prize going to the reader who picked the five women most often picked by others. To win the prize, the smart reader would not pick the five he thought most attractive so much as the five he thought others would think most attractive. (Unless he was really smart, in which case he would pick the five that others would think others would think were most attractive.)
That’s how the stock market works. I can eloquently argue that the CCI is not really that incrementally useful and the the C-S Home Price Index is way more important, but I am not likely to get anywhere. Even if I were to convince an investor that he should ignore the CCI, unless I was able to convince him that I was likely to convince most others of the same thing he will continue to react to the monthly CCI update because everybody else does.
In fact, what would really get him interested is if I could build a model to predict what the CCI would say a few days before it is released. That it may ultimately have no power to predict the future of the real economy is irrelevant. Come to think of it, I’ve got some free time….