On Wednesday, the Bureau of Economic Analysis releases its estimates of U.S. gross domestic product for the second quarter of the year. The BEA bureau cautions that its numbers are estimates and are subject to significant revision.
The reality is that the numbers BEA reports, which will be, as they always are, bandied about in the nation's news media, are not very good measures of what went on in the economy or is going on today, according to an analysis I've performed on U.S. economic performance from 1996 to 2012. They are even worse for forecasting the future. Instead, the stock market is a better leading economic indicator.
The government's initial estimate of the previous quarter's GDP, the "advance" estimate, is followed by revisions in the two subsequent months, called the "preliminary" and "final" estimate. The "final" estimate is often revised again in subsequent months and years to yield the actual GDP number. It is quite common for there to be significant differences between the actual GDP number and the initial estimate.
The average quarter between 1996 and 2012 experienced growth at an annualized rate of 2.4%. In 26 of the 67 quarters that I analyzed, the advance estimate the BEA missed the actual rate of growth (reported much later) by at least 1.2 percentage points—half the actual average growth rate. In nine of those quarters, the initial GDP estimate was off by more than the average growth rate itself, e.g., off by more than 2.4 percentage points.
Of more interest to the general public, as well as to economists, investors and policy makers, is the ability of current reported GDP numbers to shed light on the direction of the economy. Have we turned the corner? Are we headed for recession? Is the recovery gaining strength? Unfortunately, there is very little predictive power in the quarterly release of GDP figures—whether they be the advance, preliminary or final—for economic growth over the subsequent year.
In fact, there is no clear statistical relation of a given quarter's GDP growth to the growth rate during the following year. A good leading indicator should at least predict whether the future will be better or worse than average. But knowing where the current quarter's GDP growth rate is relative to its historic average predicts correctly where next year's GDP growth rate will be relative to its historic average only 61% of the time. This is not much better than a coin toss. The longer run view is no better. Last year's growth rate does not predict with any statistical reliability next year's growth rate.
Market indexes are better predictors of GDP growth. My analysis of 1996-2012 shows that the performance of the S&P 500 relative to its average predicts the performance of the economy in the next year relative to its average a full three-fourths of the time.
Indeed, knowing the current quarter's GDP growth rate adds almost nothing to the predictive power of the quarterly change in S&P 500. For the purposes of predicting next year's GDP growth, we are better off ignoring the current quarter's growth rate altogether.
For every additional 80-point quarterly change in the S&P 500—which is the average quarterly change (up or down) during the period I studied—the market predicts about one-half percentage point of additional GDP growth during the subsequent year with statistical reliability. (Here, statistical reliability means that the correlation between future GDP and the S&P 500 could have arisen from pure luck with a probability of less than 3-in-100.)
Given the performance of the S&P over the past three months (an 88-point gain) coupled with the poor growth of the past few years, we can expect that next year's growth rate will be around 1.9%. That's better than the 1.4% growth that would be expected had the market been flat, but hardly good news. In short, the market is predicting another year of just muddling along. There does not appear to be any strong reason to expect a return to pre-recession growth rates. Even less likely are the needed higher growth rates that would help us make up for lost ground over the last four years.
In sum, past GDP growth is not a good predictor of future GDP growth in part because there is a significant amount of measurement error in the numbers that are reported during the first few months. The revised numbers that come out much later are better at predicting future GDP, but by the time they're out, the future has already passed in most cases. The market is a better and timelier forecaster of future GDP, perhaps because the market has a financial stake in getting it right.
Mr. Lazear, who was chairman of the president's Council of Economic Advisers from 2006-2009, is a Hoover Institution fellow and a professor at Stanford University's Graduate School of Business.
