Do we really get top 5% jumps in the ISM during recessions?
Strong jump this month in ISM - The ISM headline rise is so strong this month that it is in the top 5% of all mo/mo jumps among all readings since 1951 (about 700 observations). The jobs jump month-to month is in the top 5.3% of all monthly jumps over the same period. This is obviously a big deal. The improvement in the month is big by historic standards. But that is not the end of understanding this report.
The venerated ISM - The ISM is a venerated manufacturing (MFG) survey most of whose readings date back to 1951. Its export and import gauges are newer and so are order backlogs. The ISM expresses readings on the MFG sector overall and in terms of some specific components by looking at monthly gains. Data are presented as a 'diffusion index' which tells us how widespread the changes are for better or worse compared to the month before. Survey responses are limited to 'no change,' 'up' or 'down'. We aggregate them across respondents and express each category as a 'diffusion index' which consists of all the 'up' responses plus half of the 'unchanged' responses with the sum taken as a percentage of the total. In this framework the value of '50' represents neutrality. This a simple scalar diffsion index for each category condenses the information in the three responses into a single gauge. The higher the reading, the more widespread are increases; the lower the reading, the more widespread are declines. Fifty is the line of demarkation between absolute net declines and absolute net increases. People talk of this breadth as though it is strength, largely because there is a statitistical association between strenght and breadth. But the survey is really one of the breadth of the month's trend, not of the strength of the underlying components.
Why diffusion gauges are imperfect - The main way folks look at the ISM readings, is to observe the diffusion values. But there is more to understanding the ISM than that. The various ISM categories are all sampled the same way; component indices are constructed the same way, but they all respond differently and with different degrees of variability over the business cycle. That fact makes direct comparison across categories a bit confusing. Each componet diffsuion index has a different habitat over which it ranges. While each component can theoretically range from zero to one hundred, none have taken on both (or even either) of the extreme values in the past 50-plus year period of monthly observations. Some hold to much narrower ranges while others use much of their theoretical range. Inventories, the PMI index itself and order backlogs vary over just little more than a 40 point range within their 100 points of possibility. Production varies over a near-60 point range while new orders span a 50 point range. Supplier deliveries and prices range over a span of 80 points each.
Putting ISM readings on a more equal footing - So to put these in context we like to look at ISM components (1) as a percent of their means or (2) to see where the current value stands as a percentage of the range overwhich the component has varied or (3) to rank the components within their own ranges. Until some sort of adjustment like that is made we can't really compare readings across categories. For example the highest lifetime average reading among ISM components is 62.8 for prices and the lowest average is 49.2 for employment. These differences remind us that prices almost always rise - even if by small amounts. So, of course, the price gauge will tend to be high. In MFG, in contrast, due to rising productivity and competiveness constraints, jobs in MFG have actually been falling over the years - in absolute as well as in relative terms. This is clear as the diffusion index averages less than 50. Clearly a reading of 55 for employment would mean something very different than a reading of 55 for prices. For prices it would represent a low reading, below its mean - a good inflation result even though prices are still indicated to be rising at a reading of 55. But for employment it would be a relatively high reading (Indeed the employment reading was last as high or higher than 55 in November of 2005, nearly four years ago). Prices which are at a reading of 55 in July have risen from abnormal lows in the recession and were last this strong or stronger in August of 2008.
This month's readings - The levels of the variables this month range from new orders and order backlogs in the low 60th percentile of their respective ranges to the 21st percentile for inventories and the 42nd percentile for employment. Most reading are around the 45th percentile of their respective ranges. That is below the midpoint but not by much. Indeed, the low standing of inventories points to how lean they are and that is a lurking positive, since when demand picks up lean inventories will be rebuilt and will stimulate output.
Understanding the differences - Three components are above their mean values: new orders, order back logs and production. That means they are in really quite ordinary shape, and do not lie at recession levels or even at levels seen in a slump. At 71% inventories are the lowest relative to their mean; at 87%, prices are the next lowest; at 92% of mean we find employment and then the PMI index itself.
Sorting out the meaning - The more we look at the data across the components the more it is clear that the weakness is now much less pronounced than it once was, while the upward momentum is strong. Many readings are still below par. But they do not lie below normalcy by much and some of the more important leading components are among the strongest readings, like orders and production. The change in the ISM tells us about momentum.
The ISM: It's a 'good news' report that is being regarded as demonstrating that the economy is either in recovery or close to it.