Identifying the Freight Industry Business Cycle
To identify freight industry expansions and recessions, our goal was to replicate as closely as possible common approaches used to identify the business cycle for the broader economy.
There is no universally accepted definition of a recession. The National Bureau of Economic Research (NBER) — a nonprofit that is the most authoritative arbiter of U.S. economic expansions and contractions — considers a wide range of metrics and allows some subjectivity in making the call. In practice, recessions can be approximated as two consecutive quarters of declining real output (Gross Domestic Product).
The challenge for us, then, was to create a metric that approximates real GDP, but for the freight industry.
The Federal Reserve Board (FRB) publishes a monthly index of industrial (manufacturing, mining and utilities) output, including sub-indexes at the detailed industry level — known as Industrial Production (IP). Much of this output — and many of the inputs used to create it — touches the country’s freight network at some point, often multiple times.
To account for differences in the distribution of the goods that come out of the nation’s factories and mines, and the goods that travel on the country’s freight network,  we aggregate each detailed industry’s seasonally-adjusted production index weighted by the industry’s share of for-hire truck freight reported by the U.S. Census Bureau’s 2012 Commodity Flow Survey (CFS).  We then smooth the result using LOESS (locally estimated scatterplot smoothing).
To test whether or not this Freight-Weighted Industrial Production Index successfully captures demand dynamics for the freight industry, we compare it to the Cass Shipments Index — an industry benchmark for freight demand.
- Annual changes in the Cass Shipments Index and our Freight-Weighted Industrial Production Index show a correlation of 0.77 from the start of the Cass series (January 1999) to the most recent data (June 2019).
- The Cass Shipments Index tends to be more volatile (standard deviation of 0.091) than the Freight-Weighted Industrial Production index (standard deviation of 0.036).
One advantage of the Freight-Weighted Industrial Production index, is that monthly data are available starting in January 1972, much earlier than most other industry metrics, allowing us to explore freight demand over a relatively long history.
To identify turning points in the Freight-Weighted Industrial Production index, we apply the Bry-Boschan algorithm to the smoothed index. The Bry-Boschan algorithm is a common approach for translating the somewhat subjective official definition of macroeconomic expansions and contractions into a programmable set of rules.  Following convention, we specify a minimum phase of six months — meaning any period of contracting or expanding economic activity must last at least half a year (two quarters) in order to qualify as a complete phase — and a minimum cycle of 15 months — meaning that each adjacent expansion and contraction must last a combination of at least five quarters.
To identify the common supply trend, we find the maximum autocorrelation factor for heavy duty truck production (NAICS Code 33612) and truck trailer production (NAICS Code 336212) as reported by the Federal Reserve Board in their Industrial Production data. Heavy truck and trailer production tend to lag Class 8 Truck Orders — a common industry benchmark — by about six months, but also may more closely reflect real changes to capacity since truck orders can be canceled (which tends to be especially common at turning points in the freight cycle). Although they are limited by a longer lag, a much longer time series is available for heavy truck and trailer production than for Class 8 Truck Orders.
For the price trend in this analysis we use the Bureau of Labor Statistics’ (BLS) Producer Price Index for General Freight Trucking, Long-Distance Truckload Primary Services. Real-time truck freight price indexes such as those produced by Freightwaves and DAT Solutions are more commonly used as the most reliable industry benchmarks, but the BLS Producer Price Index series is closely correlated with those metrics and offers a longer time series dating to 1992. The BLS index includes both contract and spot rates and, as a result, is less volatile than other industry price indices.
The results of these metrics show that the freight industry has its own rhythm independent of the broader business cycle. Freight industry booms and busts tend to be shorter than macroeconomic expansions and contractions, and while they do tend to lead the overall business cycle, using freight recessions alone as a leading indicator of economic recessions yields many false positives.
View our economic commentary disclaimer here.
 The CFS does include coverage of businesses in construction, some retail and services, farms/fisheries and government-owned enterprises/agencies. It is not obvious that accounting for these omissions would materially shift the trend.
 In some cases where there is no IP index available for a category reported in the CFS, we associate it with a related index — for instance, for the CFS category “electrical and electronic goods merchant wholesalers”, we associate it with the IP category for “electrical equipment, appliance, and component manufacturing.” For truck freight associated with retail sales — a category for which there is no obviously related industrial production index — we create a real retail sales index based on the U.S. Census Bureau’s Advance Retail Sales (excluding Food Service) report, inflation adjusted using the Consumer Price Index (CPI), indexed to 2012, and backcasted to 1972 (the Census retail sales data begin in 1992) to mirror the IP data.
 Another industry benchmark of freight demand is the Bureau of Transportation Statistics’ Transportation Services Index, which begins in 2000.
 For a more complete discussion of the Bry-Boschan algorithm, see European Commission, Business Cycles Analysis and Related Software Applications, 2003.