ISM NMI and Stock Returns

Each month, the Institute for Supply Management (ISM) compiles results of a survey “sent to more than 375 purchasing executives working in the non-manufacturing industries across the country.” Based on this survey, ISM computes the Non-Manufacturing Index (NMI), “a composite index based on the diffusion indexes for four…indicators with equal weights: Business Activity (seasonally adjusted), New Orders (seasonally adjusted), Employment (seasonally adjusted) and Supplier Deliveries.” ISM releases NMI for a month on the third business day of the following month. Does the monthly level of NMI or the monthly change in NMI predict U.S. stock returns? Using monthly NMI data and monthly closes of the S&P 500 Index from January 2008 through January 2012 (only 48 months), we find that:

The following chart compares monthly behavior of NMI and the S&P 500 Index over the available sample period. The average level of NMI during this period is 50.7. Visual inspection suggests that the two series mostly move together, but such inspection is generally imprecise in assessing whether one series leads another.

For a closer look, we relate next-month S&P 500 Index returns to monthly levels of NMI and monthly changes in NMI.

The following scatter plot relates next-month S&P 500 Index returns to monthly levels of NMI and to monthly changes in NMI over the available sample period. The respective Pearson correlations are 0.02 and -0.04, and both R-squared statistics are 0.00, indicating no relationships.

Might there be material non-linearities in the relationships with next-month stock market returns? Over the entire sample period:

• The average next-month return for the S&P 500 Index when NMI is below (above) 50 is -0.4% (+0.4%).
• The average next-month return for the S&P 500 Index when NMI is falling (rising) is 1.1% (-0.3%).

Especially since NMI likely varies with the business cycle, the available sample is very small for this kind of analysis.

Are there any startling lead-lag relationships between stock market returns and NMI?

The final chart relates monthly S&P 500 Index returns to monthly levels of NMI and to changes in NMI for various lead-lag relationships, ranging from stocks lead NMI/change in NMI by six months (-6) to NMI/change in NMI leads stocks by six months (6). Results suggest that:

• Strong past stock market returns indicate strong future NMI readings.
• NMI readings tell little about future stock market returns.
• The erratic lead-lag relationships between monthly changes in NMI and stock market returns suggest noise (and too small a sample for inference).

Again, the available sample is very small for these analyses.

In summary, evidence from simple tests on a very small sample offers little support a belief that ISM’s NMI (measuring U.S. non-manufacturing activity) usefully predicts next-month U.S. stock market returns.

Cautions regarding findings include:

• The sample is very small in terms of number of economic cycles and any subsample analysis.
• Retroactive adjustments to NMI data may confound measurement of its predictive power for stocks.
• This analysis does not rule out the possibility that surprises in NMI, relative to some measurable expectation, more usefully forecast stock market returns.