Blog - Investing Notes

March 29, 2007 - Personal Consumption Expenditures and the Stock Market

A reader, citing the book Ahead of the Curve by Joseph Ellis, inquired about the hypothesis that consumer spending drives economic cycles and is therefore a leading indicator for the stock market. For example, Mr. Ellis presents a chart relating quarterly annualized change in Personal Consumption Expenditures (PCE) to quarterly change in the S&P 500 index and states: "Most bear markets begin...when the year-over-year rate of growth in consumer spending is peaking... This suggests that finding an effective discipline for forecasting (downturns in the rate of growth of) consumer spending is essential to reducing stock market exposure..." To test the power of PCE to predict stock market behavior, we examine the relationship between monthly PCE data (the most granular available) and monthly S&P 500 index data for January 1959 through January 2007, a total of 578 months. Incorporating a one-month lag in the release of PCE (for example, aligning January PCE data with February stock market data) to avoid look-ahead bias, we find that...

The following chart compares the behaviors of seasonally adjusted PCE and the S&P 500 index over the entire sample period. As noted, PCE data lags S&P 500 data by one month based on its release schedule. Although both series generally rise over time, there is no obvious relationship between their variations. PCE shows very little volatility compared to stocks.

For a closer look, we compare monthly changes in the two variables.

The following scatter plot depicts the relationship between monthly changes in PCE and monthly changes in the S&P 500 index. As above, PCE data lags S&P 500 index data by one month based on its release schedule. If the two series move together on a monthly basis, the plot should cluster along a line passing diagonally from the upper right quadrant to the lower left quadrant. No such clustering is evident. In fact, the Pearson correlation between the two series is 0.00, indicating no relationship.

Might changes in PCE, as indications of future economic trend, lead changes in the stock market? Or, might changes in the stock market, as indications of fluctuations in personal wealth, lead changes in PCE?

The next chart shows the Pearson correlations for various lead-lag scenarios involving a maximum of 578 monthly changes in PCE and monthly changes in the S&P 500 index. As noted above, the coincident correlation (0 months lead-lag) is 0.00. Offsetting the two series such that changes in PCE lead changes in the S&P 500 index by 1-12 months produces very small mostly negative correlations. Offsetting such that changes in the S&P 500 index lead PCE changes by 1-12 months produces very small mostly positive correlations. Although there is a whiff of evidence that the stock market leads rather than lags PCE, the correlations are so small that assuming no relationship on a monthly basis is a safe bet for investors/traders. There is no trading signal.

What if the relationship is cumulative, visible not with monthly changes but perhaps with quarterly changes?

The final chart shows the Pearson correlations for various lead-lag scenarios involving a maximum of 192 quarterly changes in PCE and quarterly changes in the S&P 500 index. The coincident correlation (0 quarters lead-lag) is again 0.00. Results suggest weakly that the stock market may lead PCE by a few months (increases in stock wealth precede spending). There is also a hint that PCE is a contrary indicator for stock returns two quarters hence, but the correlation is too low to invite trading. The R-squared statistic for this latter relationship is just 0.03, indicating that the variation in PCE explains hardly any of the future variation in stock returns.

In summary, Personal Consumption Expenditures is not a leading indicator for the stock market. There is very weak (untradable) evidence that PCE is a contrary indicator for stock returns two quarters hence.

Within the web site companion to his book, Mr. Ellis presents updated versions of other Ahead of the Curve charts. We do not know how he has handled the reporting lag in economic data compared to stock market data, which can introduce look-ahead bias. More generally, we find visual inspection of such charts insufficient to infer reliable statistical relationships and, therefore, to define trading strategies. It seems that human visual processing is not calibrated finely enough to distinguish among lead, coincident, lag and randomness for pairs of complex time series. Such a subjective chart-reading approach seems to invite "see-what-you-want-to-see" bias. See our blog entry of 12/11/06 for a synopsis of the objective and rigorous analysis methods covered in Evidence-Based Technical Analysis: Applying the Scientific Method and Statistical Inference to Trading Signals by David Aronson.

See Blog Synthesis: The Economy and the Stock Market for other research on relationships between macroeconomic indicators and stock market behavior.