What do vehicle miles traveled and gas usage tell us about the economy?

 

 - by New Deal democrat

Today is the last day of our data drought before the onslaught of delayed reports that begins tomorrow with the November jobs report and CPI. 

In the meantime, there is a commenter on another economics site who generally believes that everything is OK as along as vehicle miles traveled YoY stay positive - and as of October, the rolling 12 month average was higher by 1.0%. Others robustly disagreed. So I thought I would take a look.

In the first place, here is the 12 month rolling average of total vehicle miles traveled for the past 50+ years in absolute terms:

[I am having a problem with Google giving me access to my photos, so instead here is a link to a FRED graph of the YoY% change in the 12 mile moving average of vehicle miles driven:

 https://fred.stlouisfed.org/graph/?g=1ONkm ]

It’s pretty obvious that total mileage traveled turned down, or at least decelerated markedly, at the onset of or shortly before all of the recessions in the past 50 years except for the pandemic.

But of course that just tells us that it is a *coincident* indicator, best for confirming in the rear view mirror what other data has already been suggesting. Notably, only in 1979 and 2000 did it turn down or decelerate in advance; in 1981, it did not turn down until months after the recession had already started.  Further, there was a significant deceleration beginning in the summer of 2005 that did not correlate to any recession in the next year; and for 4 entire years after the Great Recession miles traveled were completely flat, punctuated with several periods of small declines that did not coincide with recession, and so would have been very wrong signals had they been followed at the time.

Instead, I think that over the long term total vehicle miles traveled have told us that almost all US recessions in the past 50 years have at least in part been due to oil price shocks. And indeed the deceleration in miles traveled in 2005 coincided with gas prices hitting $3/gallon for the first time in the wake of Hurricane Katrina, and the four year period after the Great Recession was characterized by what I called at the time the “oil choke collar”; i.e., every time the economy would pick up, gas prices would shoot to $4, cooling the economy down, which would result in gas prices sinking back to $3, and the cycle would repeat.

In support of this, here is the YoY% change in vehicle miles traveled compared with the YoY change in the price of gas (red, /10 for scale) and for the period predating that statistic, the price of oil (orange, /20 for scale):

[To be supplied]

The oil price shocks stand out, as the YoY change in prices coincides with a sharp downturn in miles traveled. When the shock ends, vehicle miles recover. By contrast, in the 2001 recession and the pandemic there were no spike in gas prices, and the downturn or deceleration in usage was caused by other things.

Similarly, when we look at vehicle miles traveled compared with real GDP (red), we see a strong although not perfect (nothing ever is!) correlation:

[To be supplied]

 

In general, this suggests that we should expect at least a deceleration if not a downturn in vehicle miles traveled roughly coincident with the onset of recession. So here is the close-up of the past four years:

[To be supplied]

Through October, YoY vehicle miles traveled were generally steady, suggesting no recession had begun as of that time - but giving us no information with which to forecast.

But there is a very similar metric, which is gasoline product used, which is updated every week, and thus is current through the first week of December, and here is the story it tells:

[Liink to E.I.A. Data and grap of 4 week average of gasoline usage]

The four week average was significantly negative YoY through most of the summer before recovering in September into October. But for the past five weeks since the beginning of November, it has been very negative.

So, if the signs were positive into October, the decline in usage since then does not bode well for the November data were are about to be deluged with starting tomorrow.

Weekly Indicators for December 8 - 12 at Seeking Alpha

 

 - by New Deal democrat

My “Weekly Indicators” post is up at Seeking Alpha.

While there were no significant changes in the ratings, two big stories stand out: (1) the nearly total re-normalization of the yield curve in response to Fed interest rate cuts; and (2) evidence of further deterioration in the labor market.

The problem with Fed rate cuts, of course, is if they are in response to an economy that is about to roll over into recession. Yes, they lay the groundwork for a recovery, but you have to go through the recession first!

As usual, clicking through and reading will bring you up to the virtual moment as to the state of the economy, and reward me with a penny or two for my efforts in collecting and organizing the data for you.

Three important fundamentals-based indicators of the consumer economy: are they turning?

 

 - by New Deal democrat

Today is one day of quiet before a slew of updated statistics, including the November jobs and inflation reports, are to be reported next week (it is unclear whether others originally scheduled for next week, like building permits and starts, will also be updated). There was some housing and rental inventory and pricing data for Q3 released yesterday, but I will integrate reporting on that when the next housing data comes out.

Which means that today is a good day to highlight three fundamental datapoints that have in the past been very (although not perfectly!) reliable, all of which either may have just turned or may be on the verge of turning, down.

One fundamentals-based indicator which has a very long history (as in over 70 years) that has been a very good long leading indicator is per capita real retail sales, i.e., retail sales adjusted for both population and inflation (red in the graphs below). This on average turns down about a year before a recession begins. When people begin to cut back on spending for their households, unless things change it isn’t too long before it cascades into things like employment and hours of work, triggering a recession. Like many other indicators, it misfired in 2022-23 when a tsunami of supply-sided deflation created a positive “real” shock - but there is no such deus ex machina lurking now.

A second similar indicator is real personal consumption spending on goods, which is a broader measure of spending, and uses a different deflator (gold). In the graphs below below I have also normed it by population.

Finally, as I have pointed out many times, real aggregate payrolls of nonsupervisory workers has an almost perfect record of turning in the months before a recession has begun, going back over 60 years (blue). This is fundamentals-based as well; when workers in real terms are earning less money in the aggregate, they have less to spend, and this is usually an immediate trigger for a recession.

With all of that as background, here are all three indicators normed to 100 as of last December:

Neither real retail sales nor real spending on goods per capita have ever matched that level since, although their three month moving average has continued to improve in a decelerating fashion. Meanwhile, real aggregate payrolls have increased by 1.2% since, although only 0.3% of that has been during the last six months.

Typically, the immediate recession signal is when these indicators turn negative YoY. Here’s what that metric looks like now:

All of these have decelerated, several sharply, since the March-April timeframe when consumers and producers alike were front-running tariffs. None have turned negative yet, although if their present rate of deceleration continues, that is likely to occur within 4-6 months of their last datapoint for September. Next week all of them are scheduled to be updated through November, so we will have a much more current view of whether fundamental economic conditions are stormy for the average American household.

Is Thanksgiving seasonality masking a possible longer term positive regime change in jobless claims?

 

 - by New Deal democrat

This week’s update of initial and continuing jobless claims is a demonstration of two frames of seasonality: one in the immediate term, and one longer term stretching back several years. When we parse them out together, they suggest there may have been somewhat of a regime change that began in July and is still ongoing. 

Let’s start as usual with the raw numbers. Initial claims rebounded from last week’s near 50 year low by 44,000 to 236,000. The four week moving average, which irons out most of this seasonality, rose 2,000 to 216,250. Continuing claims, which it is especially important this week to note lag one week, declined dramatically, by -99,000 to 1.838 million, the lowest since early April:

Of course, the prior week was Thanksgiving, and as I wrote last week, the seasonal adjustment “expects” a big decline, but this year’s was even bigger. I expected a rebound this week, and we got it. Next week I expect a similar rebound in continuing claims.

That’s the immediate term seasonality issue.

But this week the graph above covers not just my usual frame of two years, but three years, to show that a regime change may be afoot. That’s because in the immediate post-pandemic years of 2023 and 2024, there were apparent pandemic related unresolved seasonality issues: claims rose from January until mid-year, and then declined during the second half of the year until the next January. This year the unresolved seasonality has been much more muted, especially in the second half of this year. Claims did rise into June, but then sharply declined in July, and have generally remained in that range since.

Of course, seasonality issues should be negated in the YoY% comparisons, which as I always point out, is more important for forecasting purposes. There, initial claims were lower by -1.3% this week, the four week average by -3.2%, and continuing claims by -1.9%:

As per usual, I score this as a positive, as this is what happens during expansions. In other words, it forecasts no recession in the very near term, which is good news compared with some other data we have recently received, including the JOLTS report for September I wrote about on Tuesday.

Additionally, because jobless claims lead the unemployment rate, our usual look suggests that there is no upward pressure on that rate, and if anything there is an increased likelihood of a small decline in the unemployment rate in the next several months:

That’s good news. And that plays into the possibility of a regime change in the trend in claims since the middle of this year.

Below is a graph of the YoY change in the actual number of initial claims filed plotted both weekly (thinner, gray) and monthly (thicker, blue) in 2025:

In the first half of this year, jobless claims typically were in the +10,000 range YoY. That all changed since the end of June. In the 23 weeks since, jobless claims have averaged just under -4,000 lower YoY. While I speculated during the summer that the change was school year related, and indeed there was payback in the form of sharply higher numbers early in September, the trend of lower YoY numbers has continued, even against the comparisons of very low numbers at this time last year (remember the residual seasonality of 2023-24 meant low numbers in December into January).

This suggests to me that the post-pandemic residual seasonality has been evaporating in large part, and further it is evidence, contrary to most of the monthly data we have gotten this year - including most recently from both the regional Feds and the ISM as I have documented in the last few weeks - that have suggested that the labor market may have tipped over. Instead, initial claims may point to at least a slight recovery.

Weekly data is noisy, but it always captures trend changes first. While the official monthly government data is still stale, I will pay especially attention to the regional Feds reports on employment trends in their districts, the first of which is scheduled to be released on Monday.

Q3 employment costs: probably the “least positive” since the pandemic

 

 - by New Deal democrat

The employment cost index, which was updated this morning through Q3, typically gets much less attention than the monthly payrolls report. But in this circumstance it is entitled to more notice, since the monthly data has only been posted through September as well.

Additionally, one important advantange of the Employment Cost Index is that it is adjusted for the type of job performed, while the monthly average statistics are not. Thus, for example, since many low-paid service workers were laid off during the COVID lockdowns, the latter metric was distorted by the job mix, whereas the former measure was not.

The news from this morning’s report was mixed. On the one hand, quarterly compensation increased just under 0.8%, whether measured by wages and salaries alone (blue) or by total compensation including benefits (red). On the other hand, the quarterly increase in wages was among the lowest in four years, and for total compensation it was the lowest (note: graph subtracts this quarter’s changes from both datapoints so that they norm to 0 for easier comparison):

On a YoY% basis, median wage compensation increased 3.6%, on par with the previous two quarters, while total compensation was the lowest since the pandemic, although higher than at any point between the Great Recession and the pandemic:

Although it is somewhat noisy, the employment cost index tends to in tandem with, but inversely to, the unemployment rate:

There really is not any leading/lagging relationship here, and my reading of this graph is that both median compensation and the unemployment rate were relatively stable this year though September, as the uptick in each is within the range of noise.

Finally, since the employment cost index measures wages and other compensation normed by occupation, an interesting comparison is with the Atlanta Fed’s wage tracker, which measures wage increases between those who switch jobs (presumably for better pay and/or benefits) and job stayers (updated through August):

During most of the post-pandemic era, when the unemployment rate was especially low, employees could get substantially better wage increases by switching to a new job. This year that has ended. Job switchers are doing no better than job stayers.

In sum, this morning’s data tells us that when it comes to wage growth, workers are still doing better than they were at any point during the last long expansion before the pandemic, but although the news was still positive, it was the least positive, relatively speaking, since then. This is especially true given the increase in inflation since early this year.