The January 2011 Ceridian-UCLA Pulse of Commerce Index™ (PCI) fell 1.7% year-over-year, and 2.2% below year-ago levels. Econintersect uses the PCI raw data to help forecast Main Street economy – and our analysis is that the index is up 2.6% month-over-month and down 0.1% year-over-year.
The data has been in a long term “less good” trend.
Both the authors of Pulse of Commerce Index and Econintersect see diesel consumption as an important economic pulse point. Both use the data in different ways, apply different methodologies to analyze the data, and offer different conclusions
According to the economists at the UCLA Anderson School of Economics:
It seems difficult to square the behavior of the PCI with the evident improvement in a number of economic indicators, most notably the increase in payroll jobs and decrease in initial claims for unemployment. The PCI also appears out-of-sync with Industrial Production and Real Retail Sales which continue to grow in a healthy manner while the PCI is stalled out. We believe, however, that the year-over-year changes in the PCI are in fact in sync with the overall economic picture.
With respect to the monthly and quarterly results, we need to understand what is happening. A variety of hypotheses regarding measurement problems of the PCI do not find support in the data. It could be that the PCI is misleading because of a change in the transportation firms included in the PCI database, or a change in the growth within and outside the database. Neither of these ideas is supported by the evidence. Or, it could be that a shift in favor of intermodal rail is reducing the importance of trucks in the US economy. If true, the regional data should support this idea, but they do not, as the weakest regions are the least likely to be affected by a shift in favor of rail. It could also be that the recent spike in diesel prices is encouraging a more rapid deployment of fuel-saving technologies so that the amount of diesel fuel to ship a load is declining, but that would not account for such a substantial fall from month-to-month. Thus, as further explained in the pages that follow, we did our due diligence on these items and provided the full explanation.
That said, a look at the year-over-year changes in the PCI make it look very accurate. The year-over-year change of the three-month moving average peaked at 8 percent in July 2010 and has fallen steadily to essentially 0 percent in January. We analyzed the data of the three-month moving average as there is too much noise in the month-to-month data. That pattern we see is an early and amplified version of what has happened to Industrial Production, Business Inventories and Real Retail Sales. In other words, the PCI has correctly signaled slowing growth in all of these indicators, and is signaling still weaker growth ahead, unless trucking starts rebounding in February, March and April.
The transport network in the USA is almost exclusively fueled by diesel – and diesel consumption makes a good proxy for forward economic activity (as transport occurs, on average, one to two months before consumption). Caveat is that this index is based on road use of diesel.
The PCI is modeled using Ceridian’s diesel distribution network to forecast economic growth – primarily Industrial Production and GDP. Econintersect extracts the unadjusted (not modeled) diesel index for its economic model. Graphically, the unadjusted data has a slightly different feel.
The January three month moving average of the year-over-year data is negative – and the negative trend since mid-2010 continues. The size of the negative data in December was a concern – and the January data is also slightly negative making this the second month in a row of outright year-over-year contraction.
The question remains is this PCI properly reflecting the economy if all other major transport indices are reporting positive year-over-year improvement – pay special attention to the truck transport which is up 6% year-over-year:
- Pulse of Commerce Index: Down 2.2% year-over-year.
- Truck Transport (December 2011): Up 6% year-over-year
- Container Counts (December 2011): exports increased 1.3%, while imports increased 2.2% year-over-year
- Rail (January 2012): Up 0.8% year-over-year
Reading my thoughts, Ceridan-UCLA stated:
Actually the PCI is Doing a Good Job Showing a Weakening Economy
So, you may be surprised to learn now that the PCI is actually doing its job pretty well. How is that? The year-over-year change of three of the indicators that the PCI historically correlates strongly with, seems not to be following so well of recent, illustrated in the figure below, together with the PCI. The three indicators are Total Business Real Inventories, Industrial Production and Real Retail Sales. The PCI is the heavy line.
In the normal growth period from 2005 to 2007, these four were closely connected. In the 2001 recession, the PCI and Real Retail Sales did not decline much, while Industrial Production declined -5 percent year-over-year, and Inventories, with some delay, fell even more. All four series had similar substantial declines in the recession of 2008/09. The PCI and Real Retail Sales had year-over-year changes that bottomed out first. Then came IP and last Inventories.
Now take a look at what has happened since 2010. All four series have experienced peak year-over-year values, followed by significant declines. The PCI year-over-year peaked first in June 2010 and fell the most, the early and amplified characteristics of an effective leading indicator. The year-over-year Industrial Production peaked a month later, July 2010, and experienced a significant decline to +4 percent in January 2012 compared with -1 percent for the three-month moving average PCI. Real Retail Sales year-over-year peaked in November 2010 at 6 percent and have subsequently floated down to around 4 percent. And Inventories year-over-year peaked at 7.5 percent in May 2011, and have fallen to 5 percent in November 2012, the most recent data available.
In other words, the PCI year-over-year peak in 2010 and the deterioration throughout 2011 has correctly anticipated the same movement of Industrial Production, Total Business Real Inventories, and Real Retail Sales. The weakness in the PCI is suggesting either further weakness in these indicators or a big gain in trucking February, March and April.
Although I remain unconvinced by many of the Ceridan-UCLA arguments – this remains one of the more interesting indexes, with several interpretations. Diesel usage should be an excellent economic pulse point, but in practice is creating more questions than answers.
Caveats on the Use of this Index
This is a post Great Recession index which has little real time history on foretelling economic activity. This model works in hindsight. A positive point for this index is that there is usually little backward revision.
Diesel consumption per ton mile is improving at rate which Econintersect has no means to quantifying in the U.S. But on a global basis, this improvement is likely well over 1% and could be as high as 5% per year such as:
- There have been significant inroads for fuel conservation by placing trailers on higher efficiency railroads;
- There has been some conversion of diesel trucks to using LPG;
- Not only has current environmental standards forced conversion to more efficient diesel technology – the rising price of diesel alone has forced truckers to upgrade to the higher efficiency trucks / engines / trailers/ use management;
- Tractor design continues towards more aerodynamic design.
Although it is true that diesel moves the goods necessary for the economy, using diesel data without an efficiency adjustment likely will provide incorrect conclusions. Therefore, it is trend lines, not specific values, which are important. It is very likely this index is UNDERSTATING the economy by an amount equal to the indeterminate efficiency improvement rate.
Monthly diesel use can vary with the weather or other natural causes making is index noisy. For this reason, Econintersect uses the three-month moving average for modeling economic activity.
The PCI diesel consumption is based on roadway diesel sales – not railroads, sea or air transport. The Achilles heel of this index might be its inability to adjust for alternative non-truck transport. Ceridan-UCLA dispute there is any evidence that rail is making inroads into road transport – but did not use a tonnage comparison.
Shifting to Rail? The Regional Data Doesn’t Say So. The railroads are reporting a significant increase in intra-modal activity with containers shipped first by rail for long-hauls and then transferred to truck for the shorter routes. If this shift to rail were a substantial part of the story in 2011, it ought to be evident in the regional data, perhaps in the Mountain region which was crossed by longhaul truckers and now by trains, or in the Pacific where rail might be displacing trucks for cargo destined for the East Coast. The chart below illustrates growth of the U.S. PCI since December 2010 until January 2012. The U.S. PCI three-month moving average peaked in May 2011 and has been on the decline ever since. The regions are sorted according to where they end up at the right, from the West South Central and to West North Central. The regions that seem to have contributed to the shape of the U.S. overall have wide lines and the others have thin lines. The Mountain region has a shape that amplifies the ups and down of the U.S. overall which might the effect of railroads. But the Pacific region was actually improving in the second part of 2011, not what might have been predicted by the shift-to-rail hypothesis. There are also evident shifts around March in the behavior of several other regions too: WSC, MAT and WNC. Thus, there is no clear evidence of the shift to rail in the regional data. It could be critical, but there is no smoking gun in the regional data.
Econintersect determines the month-over-month change by subtracting the current month’s year-over-year change from the previous month’s year-over-year change. This is the best of the bad options available to determine month-over-month trends – as the preferred methodology would be to use multi-year data (but the New Normal effects and the Great Recession distort historical data).