Tracking the Economy with CASS

Using freight loads as a measure of economic activity.
Source: Network Rail

Prophecy would be a wonderful thing - and I'm not talking about the biblical kind - just economic prophecy will do. Financial traders eagerly await numbers such as the US Non-farm payroll which has the power to move major markets or the Purchasing Managers Index which in the link is nicely mapped globally by Bloomberg. In a post last month I looked at yield curve inversions as an indicator of changes in real GDP and therefore of recessions. The conclusion was that they are indeed a statistically significant indicator of real GDP but only weakly so. A better indicator is simply the change in real GDP from the previous month. Using some statistical jargon GDP is autocorrelated. But what about other indicators of future economic growth (or decline)? There are a plethora of them. A good one might enable you to detect tipping points when an economy starts to move into recession.  In this post I'm going to examine a closely watched index of North American freight volumes and expenditures, the CASS Index

What is the CASS Index?

I first learnt about this index in an article Back on the rails by Russ Mould of AJ Bell, a UK investment platform. It is a measure of the amount of freight being carried in North America. The CASS Index has been collected since 1990 and is available for download from CASS Index or the St Louis Federal Reserve FRED database (just type CASS Index into the search box) The index aggregates data over all freight modes across North America and is updated monthly. This makes it as timely as other major government sources of data and the comprehensive nature of the coverage makes it potentially economically significant. Why should this data be useful? The transportation of goods around a continent is very likely a good indicator of manufacturing, retail activity and also construction. As the photo above shows (albeit form a UK construction site) freight can be closely linked with construction activity. Most retail goods are delivered on trucks and chemicals and industrial goods are often transported by rail. The length of the time series starting from 1990 makes the CASS Index a useful dataset to study as this amount of data should show whether it is statistically significant or not.

A Short Study using Python

In the chart below I've plotted the two forms of the main CASS Index:
Indices of North American freight.
Data source: St Louis Federal Reserve (FRED), CASS Information Systems Inc.
The CASS index comes in two forms: the expenditure index and the shipments index. The chart shows that although the shipments index, that measures physical volumes of freight, hasn't increased much in the last couple of decades the amount spent (Expenditure) on freight has. In this study I shall use the Expenditure Index as it most closely follows the US GDP. You can see this from the chart below. By comparing with the chart above you can see that the nominal US GDP has increased between1990 and 2020 by a similar amount to the CASS Expenditure Index. It therefore makes sense to try and model the nominal US GDP with the CASS Expenditure Index as a predictor. 

There is a subtlety though. Modelling the GDP itself would not be good idea. It is much better to model the quarterly change in the GDP. Both the CASS Index and GDP drift upwards in a slightly random fashion. By regressing one against the other the results would be badly affected by this longterm non-stationary relationship. It is better to model the much more stable numbers coming from changes in GDP and the CASS Index. I've done this using a Python program that you can see in the Appendix to this post. The results of the regression clearly shows that the CASS Expenditure Index is a statistically significant predictor of US GDP. It's a bit of a surprise as the GDP data is quarterly whilst the CASS Index is monthly and so to forecast the GDP you need to see 3 months ahead - that is impressive!
Regression coefficients and t-stats for model regression
Regression result using the Python Statsmodels module (ols)
The coefficient for the intercept reflects the average growth of the US economy of 4.4% and the t-statistic of 18.8 reflects that we can be very sure of this figure. The coefficient for the CASS Index is positive showing that growing freight expenditures are a sign that the economy is likely to grow. The t statistic of 2.6 shows that it is significant, in other words it's very unlikely that the coefficient of 6.68 is a statistical fluke so we can use it to make forecasts.

Prognosis and Discussion

Going back to Russ Mould's article on 21 November 2019 he noted a 5.9% decline in the October Shipments Index (The Expenditure Index also declined) prompting CASS to state that this is "signalling an economic contraction". The January 31st, 2020 release from the Bureau of Economic Analysis showed that US real GDP in fact held steady with the same rise of 2.1% as in the previous quarter. Current money GDP slowed though from a 3.8% rise in the 3rd quarter to a rise of 3.6% in the fourth quarter. Of course indicators are just that, indications and do not foretell the future. The t-statistic of my model is significant but not large like that for the intercept which gives the mean annual growth of US GDP. Also, the R-squared value, which is a measure of how much of the GDP data is explained by the CASS Index is only 0.05 so don't expect wonders. The CASS Index is yet again indicating a modest slow down in the US economy.

Appendix

In this appendix you can see the Python code I used to download the data and create the charts for this post. I've annotated it so hopefully you can develop it yourself if you wish. It's available in my GitHub repository SteveCBell/Quantifying. The program also generates the model residuals. You can see from these that the financial crisis of 2007/8 marks a regime shift in the model but I decided to keep the model as simple as possible.

Comments

Post a Comment

Popular posts from this blog

Introduction to Quantifying

Image
Introduction This blog is going to be about everything quantitative, focusing on applications of mathematical and statistical modelling. Not cool eh? Well, hopefully you'll find otherwise. Mathematics has been around for millennia and was first used for essentials such as counting things. More exciting applications soon developed in geometry and astronomy. This happened because people observed the regularity of the movements of the moon and the stars. Useful things came out of that very early such as the ability to navigate. But the ability to develop statistical models was a long time coming and wasn't to do with the lack of data. Conceptually statistics seems difficult and many ideas, even as simple as the mean, took a long time to gain acceptance. A great article on this is "An average understanding" by Simon Raper in the Royal Statistical Society's Significance magazine which you can find online. We take this concept for granted but it was only i...
Read more

Book Review: "Material World" by Ed Conway, published WH Allen 2023

Material World is an outstanding book. It takes six materials: sand, salt, iron, copper, oil and lithium and explores their role in the world economy. What is especially fascinating about this book is that the author takes both an historical approach and a forward look into the coming transition to a low carbon economy. The author is no scientist or engineer so whilst this can be frustrating at times it allows him to take a wider perspective connecting the past, present and future and considering many aspects of how these materials came to be so important. It is refreshing to share with him as he discovers more about his subject. This is a deeply researched book with an extensive bibliography and many interviews as the author travels the world to visit mines and quarries. A major theme of the book is that this 'material world' of mines, oil refineries and chemical factories is mostly hidden from us and so we take it for granted. We don't need to think about it too much any ...
Read more

Index Linked Gilts

Image
Short Guide to Inflation-indexed Bonds Summary In this post I introduce you to inflation linked bonds and how they can be used by investors in an investment portfolio. Introduction Most private investors focus their attention on the equity market. It's exciting and offers the possibility of big gains, albeit at the risk of big losses or worse every now and then. Others are nervous of the equity market and prefer government or corporate bonds with their predictable (mostly) flow of coupons and pay-out (hopefully) on the maturity date. If your nervousness has developed into paranoia you may invest in gold as an insurance against war and pestilence. And if all of these options and choosing between them seems like too much then, of course, cash is a respectable investment although in some countries you may now be 'rewarded' with a negative interest rate. But there's a good alternative to all of these options and that's inflation-linked bonds. What are I...
Read more