Calculated Risk reminds me that heavy truck sales is something that collapses during recessions. I wondered how the 12 month change in this variable compare against the corresponding change in vehicle miles traveled (suggested by Steven Kopits). The latter does pretty lousy.
Figure 1: Recession probabilities using contemporaneous 12 month changes in heavy truck sales (blue), and in vehicle miles traveled (tan). 50% threshold dashed red line. NBER defined peak-to-trough recession dates shaded gray.
The McFadden R2 from the heavy truck regression is 28%, vs. 7% for vehicle miles traveled (VMT). One can also see quite simply using the 50% threshold the truck regression captures all the recessions from 1970 onward, with no false positives, while the vehicle miles regression will miss every recession except the 2020, unless an absurdly low threshold of 25% is used (in which case we have a false positive for 1995). The results barely change is one uses Kopits’s preferred measure, the change in 12 month trailing moving average. Really, to be technical, VMT is a lousy coincident indicator of recession.
In neither case is there a recession indicated for May (13% for vehicle miles) or July (4% for heavy trucks) 2022 (latest data for vehicle miles and truck sales, respectively).
I think that correlation has held up well. One thing: the heavy truck sales line looks like the inverse of this:
Bruce Hall: Read…the…legend…to…the…figure.
I think Bruce has admitted on the blog before he’s very poor at analyzing things related to automotive, trucks or anything which traverses on America’s highways.
Did guys like Robert McNamara use graphs at auto companies like Ford before or during Bruce’s time??
Cut Bruce some slack, he’s a deep thinker. Sometimes he can even make out graph colors.
pgl, you should’t be posting your snide comments using Moses’ name
God you are a pathetic little boy. I do not pretend to speak for Moses – even when he has a very excellent point. Suck it up troll. No one believes a word you say as we all know you lie 24/7.
We need to help Brucie out as he is really SLOW:
Motor Vehicle Retail Sales: Heavy Weight Trucks
Menzie I… did… read… the… legend… to… the… figure. What… is… “RECESSIONF”?
Sorry if I didn’t understand your shorthand for something. My son who is an engineer looked at your post and legend and said, “I don’t think that’s a great legend.” I gather the gist of the post… economic activity is closely related to heavy truck sales. My observation was that the line look like the inverse of HT sales as reported by the St. Louis fed.
Esoteric writing is not always communication. So unless you are presuming that you are communicating within a closed loop system, you might want to be a bit clearer in your graphics.
And yes, heavy trucks sales are very sensitive to economic activity as I have been aware for at least 4 decades from my automotive background (no snark intended).
Bruce Hall: “F” denotes forecast, for variable being predicted, “RECESSION”.
I mistakenly thought that having the series bound between 0 and 1, given the legend and the context of the discussion, you would understand what was going on in the graph. Apparently I was wrong. Apologies.
Menzie, thanks. That wasn’t clear although if I looked closely at the X axis I might have possibly guessed that the data was beyond 2022 (although it certainly appears to end mid-2022). Regardless, I still think the legend could have been clearer, either with a heading “Forecast” or labels “Regression-FCST” since Fcst is commonly understood as an abbreviation for forecast.
That being said, what part is the “forecast” if the data indeed are actual. Are we to presume the 50% threshold not being crossed represents a “forecast”? In my experience in operations and business planning, a forecast chart would extend into the future. As I indicated in my original comment (which has been much maligned), “I think that correlation has held up well.” The chart, as it is shown, is a good indicator that our economy is not presently in a general recession, but it is in no way a “forecast” chart.
Bruce Hall: Unlike many bloggers, I try to be clear in the legends to the Figures, so that people understand what is plotted, and where the data sources have come from. I am sorry that this description: “Figure 1: Recession probabilities using contemporaneous 12 month changes in heavy truck sales (blue), and in vehicle miles traveled (tan). 50% threshold dashed red line. NBER defined peak-to-trough recession dates shaded gray.” was insufficient. It was, I thought, written in pretty plain English (I want to assure you English is my first language.)
Bruce Hall Are you familiar with or have you ever worked with probit/logit models? Based on this comment, it doesn’t sound like it:
That being said, what part is the “forecast” if the data indeed are actual. Are we to presume the 50% threshold not being crossed represents a “forecast”?
In this case the probit model forecasts (or, “predicts” if you prefer) the probability of a recession given a 12 month change in sales. The 50% threshold simply means that the model predicts a recession is more likely than not, which ought to seem pretty intuitive.
Menzie, my comment requesting clarification was relative to “forecast” for which there was none … unless you are attempting to imply that heavy truck sales are a leading indicator of economic activity which I don’t believe they are… more likely coincident or slightly lagging due to order times.
I did say that the correlation (historical relationship) was good. You may use “pretty plain English”, but in this case the key word in the legend, RECESSIONF, later clarified as “forecast” did not and does not align with the graph. It merely shows that the current situation relative to heavy truck sales does not indicate a general recession… at this time. I also agreed with that.
Plain English can be precise, but in this case it was not. Your overall point that there is a strong >b>historical relationship is not disputed.
<b<Bruce Hall I did say that the correlation (historical relationship) was good.
Normally when we say that two things are correlated it’s the case that both variables can take a wide range of possible values. But what happens if one of the variables is binary; e.g., “0” if no recession and “1” if recession? You really can’t use the normal approach. That’s why we use probit or logit models. The dependent variable is binary; i.e., it’s either a “0” or a “1”. In this particular case the issue is whether the change in VMT is good at telling us whether or not we’re in a recession. The answer is no; however, heavy truck sales are good indicators at predicting “0” or “1” at the 50% threshold level. The model isn’t predicting a future recession; it’s telling us whether or not we’re currently in a recession.
Years ago a former colleague and I were asked to analyze whether or not two bridge players were cheating at the national bridge championships. We presented several pieces of evidence including a pretty clever application of a logit model that predicted binary signals (actually, the physical positioning of the cards) between the two players as captured by overhead cameras that monitored the games. Anyway, we proved that they were cheating and they were permanently banned from ever playing competitive bridge.
“(I want to assure you English is my first language.)”
Important clarification for any white nationalist readers out there.
My son is an engineer too. And unlike you and your son – he gets basic finance and economics.
You’re stretching to make an invalid point.
Did you bother to read Bill McBride’s post on this? Dr. Chinn’s graph was more clearly marked than was Bill’s. But if you actually READ and Understood the context of their posts, the message would have been clear to even a 5 year old.
Damn Bruce – you are SLOW. I guess it was your son’s mom that got him through school.
pgl… the Fed and McBride both define “heavy trucks” as over 14K GVW, but that is not standard industry classification. Heavy Trucks are Class 7&8 (over 26K); Classes 3-6 are Medium Trucks; under 10K are Light Trucks. By the way, that’s the official definition, not some finance take on it.
Your version of plain financial English may not hack it. Stick to day trading.
Forecast means: predicting a future event or trend.
I know what Menzie meant to communicate. It just wasn’t quite up to his usual standards. Your comments are erroneous and irrelevant.
August 4, 2022 at 5:38 pm
You’re stretching to make an invalid point.”
Brucie must take yoga as he stretches a lot to make totally bogus “points”!
“the Fed and McBride both define “heavy trucks” as over 14K GVW, but that is not standard industry classification.”
This is how you excuse all the stupidity you have exhibited here? Noting that there are different kinds of trucks. Like Damn! Brucie notes the obvious! BRAVO!
“Forecast means: predicting a future event or trend.”
Really? 2slug and I should buy you FINANCE FOR DUMMIES for Christmas. BTW – are you working with Princeton Steve on a new version of economic terms no one should ever use?
August 2, 2022
China’s weekly road logistics price index rises
BEIJING — China’s road logistics price index rose last week, driven by increasing demand and a stable supply of transport, industry data shows.
The index came in at 1,039.40 points in the week ending July 29, up 1.36 percent over a week earlier, according to a survey jointly conducted by the China Federation of Logistics and Purchasing and the Guangdong Lin’an Logistics Group.
The sub-indices of all types of vehicles registered slight week-on-week growth. The figure for full truckload logistics prices, which mainly measures bulk-commodity and regional transportation, stood at 1,041.09 points, up 1.41 percent from the previous week.
[ A detailed price index that has become increasingly important. ]
Heck, Henry Weingarten has a gimmick. Why shouldn’t Stevie? Who needs serious analysis?
(Cue 60s pop music…)
When Henry Hub is in the second house
And WTI aligns with Mars
Then miles driven will guide the economy
And gasohol will power the cars!
This is the dawning of the Age of the Low R Squared
The Age of the Low R Squared…
Macroduck: Excellent. And so accurate.
I live to entertain.
If Nancy Reagan wasn’t a subscriber before her death, I’m not buying any of it.
Off topic –
I’m not sure what to do with this headline: “Scientists say it’s time to prepare for human extinction”
By “prepare” they mean sell duration, buy volatility? Go ahead and eat fatty foods? Swipe right every time?
“Click here” (but you knew that already)
This is where a hefty silver-haired former Marines pilot says “Yes Sir!!!! And you nearly always succeed” and lets out a very guttural laugh.
MD, it is the blind belief in these crazy models-based predictions that shown ignorance in the subject to climate change. Have the models been accurate for any prediction? If so list them. BTW, did you know the previous warm periods during this interglacial were equal and even higher than that predicted? Did you know those periods are considered the sweet spots for human civilization progress?
We need proxies to estimate these periodic temperature peaks. Here’s just one: http://jonova.s3.amazonaws.com/graphs/lappi/gisp-last-10000-new.png
Gee a graph that goes back to 8000BC! I bet the intelligence of the person who drew this goes back to the cave man days.
Heavens! What brought this on? CoVid has a twist in his knickers again.
Once again, an incomprehensible graph. But look on the bright side, at least no log scale!
Here’s the vmt graph I use. (Feel free to publish it.)
You are into moving sums. CoRev the barking dog is into moving averages. And of course Bruce Hall and his engineering son cannot read a simple legend.
Yes – we have the Three Stooges of basic economics!
The method of presentation depends on the data. If you work with the VMT data, you know there’s lots of chop, and the seasonally adjusted figures, well, may not accurately represent seasonal variation. That’s why Menzie’s graph looks like such spaghetti.
So how to present? You could do 3 mma over prior year 3 mma. Still too choppy. You could present annual totals or averages by calendar year, but you’ll miss turning points. You can do monthly changes, a la Menzie, and get garbage.
But one option is a rolling 12 month average. That eliminates the whole seasonality question and provides a nice, smooth graph. The downside is that it may be late to show an inflection point and, in the current case, may be distorted the data by the rise in VMT in 2021.
So you have to make decisions, and each of those involves some sort of compromise.
I think you’ll agree, Menzie’s graph is hardly legible, much less possible to understand. I have no idea why he felt a need to do it, because he already had a visual the last time we discussed the topic.
In any event, my graph, by contrast, tells a pretty simple and clear story with nice aesthetics and easier to understand legend.
Steven Kopits: The probit regression in this post is estimated using a 12 month change in VMT.
Wow – you are THE EXPERT in data manipulation designed to deceive!
“This graph shows heavy truck sales since 1967 using data from the BEA. The dashed line is the July 2022 seasonally adjusted annual sales rate (SAAR).”
Bill McBride’s graph. Hey – he did not use moving sums either. And you like Bill McBride. Huh. Did you tell Bill his graph was awful? I dare you to do so as I would love to see how he snaps back at your BS.
If you have time you may want to update the RV shipment model as related to recession.
The monthly data are a bit tedious to record from the RV website, but it looks like 2019 shipments were 405,596, 2020 shipments were about 470,454. I had to estimate March 2020 based upon March 2021 comment about % difference between two months. 2021 shipments were 600,240. 2022 YTD through June were 324,031. For 2021 June YTD shipments were about 50% of total shipments for the year. So, 2022 total shipments could be about 628,000.
It looks like an annual data probit model forecasts about zero chance of recession in 2022.
Expert enlightenment welcomed.
I gotta soft spot for the RV data. Maybe because my Dad liked RVs and would sometimes take us to the RV shows. Wish he could have had some of his retirement years in an RV as I think he semi-dreamed of. But it wasn’t in the cards. He travelled a lot when he was younger, so it’s not a sob story, just would have been a nice “cherry on top” if he could have gotten a couple years of RV travel in there near the last.
Now that the discussion about truck sales and our present economic situation (recession or not) has been amply tossed about, how about an actual forecast?
Note section 5 (especially 5.2). That might be more indicative of future economic activity as Class 8 (long-haulers) are quite sensitive to economic conditions. Medium trucks (classes 4-6) are more local haulers so it’s easier to get mixed messages.
Moses, you were saying…?
What are you saying? Did you not understand what your link was writing? Let me make this simple even for you. The prices charged by truckers are falling. That should be good news since you are obsessed with inflation. Dr. Chinn was talking about activity not prices.
Bruce – you have done a world class job in one respect. You have demonstrated over and over again you have no clue what this discussion was ever about.
Now we have all said you are really dumb. Way to prove the point!!!!
Read my comment slowly. Quit trying to be snide; you only come across as stupid.
Read your own link SLOWLY. We have asked you to do so but I guess that is over your little head.
Section 5′ title:
5. Contract vs. Spot Rates
Prices not activity. Bruce Hall is too dumb to even understand the title of section 5!
5 Truckload Market Trends to Watch in Q3
2. Class 8 Truck Orders
There, spelled it out for you. Duh!
Yea it had a lot of data on a lot of specific trucks. One graph showed orders for one type of truck? Gee willerkeys! The main point of this weird link of yours was about the prices for freight. Oh wait you are into reading 725 page documents you do not even remotely understand. Got it!
I thought the threshold probability was a demarcation which “optimizes” the probability forecast of recession when in fact recession is the state and “optimizes” the probability of no recession when the forecast probability is below the threshold. For example, at the 50% threshold, the probability of the recession forecast being correct when the state is recession may be 82% and the probability of no recession may be 80% when the forecast indicates no recession. Maybe you or someone else can do a better job explaining this.
AS Strictly speaking the threshold value is the value used to score a predicted probability as a “hit” or a “miss.” For example, if the threshold value is 0.5 for y = 1 and the probit model predicts a 0.3 probability, then it would be scored as a miss. If the predicted probability is 0.6, then it would be scored as a hit. A 2.x2 matrix is commonly used to summarize the hits and misses. So if there are 100 observations and the model correctly predicts 30 zeros and 40 ones, then one measure of the goodness of fit would be (30 + 40) / 100 = 70%. That would correspond with the 82% in your example. But I doubt that Bruce Hall is particularly familiar with probit & logit models, so in his case it’s best to understand Menzie’s 50% threshold line as a visual aid. It shows how many individual observations predicted a recession as more likely than not.