Traffic Forecasts

by on January 21, 2014 at 2:20 pm in Data Source, Travel, Uncategorized | Permalink

Official DOT forecasts of road traffic with actual road traffic.

Forecasts

Hat tip to Andrew Gelman, who compares it with some other famous forecasts.

Shane M January 21, 2014 at 2:31 pm

Kind of like projected vs. actual sales forecasts at the company I used to work for.

Shane M January 21, 2014 at 2:39 pm

The post indicates this data is simply a rollup of local level forecasts, so the faulty forecasts are distributed. We used to do local level forecasts, submit them, and then find that the rollups didn’t match what corporate wanted, so they’d send it back, we’d inflate local forecast, resubmit, ad infinitum until unrealistic result was achieved.

In this case, the inflated result seems to be organic at the local level.

Rahul January 21, 2014 at 3:46 pm

This is what happens when the forecaster’s wages (indirectly) are correlated to the forcasted series.

bdriscoll January 21, 2014 at 5:24 pm

…and what happens when the {government} forecaster’s wages are not correlated to their published work in any way ?

If their paychecks look the same regardless of the accuracy/inaccuracy of their forecasts — what incentives are they actually working under ?

Nobody at DOT will lose their job or get a pay cut over this outrageous traffic-volume forecast. More likely is a budget increase so DOT bureaucrats can “fix the problems” in their process.

Bigger issue is how much we should trust (or need) government statistics generally. The government spends vast resources on statistical collections/projections of every sort (central planners are insatiable); gross data errors like this DOT traffic projection are not the exception.

JWatts January 21, 2014 at 5:28 pm

It’s been my opinion, which might be entirely wrong, that Government agencies are pretty good at counting directly observed physical things (such as traffic on a road) but don’t seem to much care about accuracy in forecasting.

The exception would be NOAA, they get direct and negative feedback for notably poor forecasts.

Bob January 23, 2014 at 1:20 pm

No correlation with the quality of the forecast is no worse than correlation with the forecast matching something that has nothing to do with accuracy.

If my forecast on the cities with the most population growth was rated on whether my top guess was the city whose football team won the Superbowl that year, I bet it will not track population growth very well.

john personna January 21, 2014 at 6:09 pm

I thought it was pretty well known that prediction fails. I thought in fact that the best criticism of N Taleb was that “we knew all that.” But it seems we didn’t, if we just pretend it is the “government” part of “government prediction,” and not the “prediction” part.

Brian Donohue January 21, 2014 at 6:21 pm

But but but…

I think the point is that the misses are consistently in one direction.

john personna January 21, 2014 at 6:30 pm

Others have talked about regression to the “mean.”

I guess the funny thing is that “the best guess” often is “whatever happened last” even when “best guess” isn’t saying much. In 2008 the old trend line might have been the “best” guess, but with decreasing confidence. In 2010 what are you going to do? You can again use the old trend, and further downgrade confidence … or you have to give no answer at all. The alternative would be a trend in a random direction.

No answer is often better than “the best guess,” but everybody says “give us something!”

Indeed, people without a guess are often just called “clueless” and brushed aside in favor of bad prediction.

Sam January 21, 2014 at 6:33 pm

Well, yeah, because they’ve taken the growth rate from the 1990s and applied it to the current measurement. I will bet anything you like that the graph is plotted beginning at 1997 especially to show the suckiest possible predictions – I’d guess that the numbers would have been nearer the mark for the previous decade.

John Thacker January 21, 2014 at 6:41 pm

The growth was indeed near linear for 40 years, and tracked NGDP growth very very closely. The FHWA keeps expecting things to revert to the trend (as do, say, people talking about AD and NGDP). It keeps not returning to trend. When to update the model is an interesting question, but it does seem to look a bit ridiculous at this point.

John Thacker January 21, 2014 at 6:43 pm

Here’s the rolling average VMT trend back to 1988. You can see that it used to be linear, and track NGDP/AD, with small pauses in recessions. Now it doesn’t. Very interesting problem of when to update your priors.

John Thacker January 21, 2014 at 6:32 pm

I think everyone does know that prediction fails. One question is who pays when prediction fails. When the government fail to predict, or when something “too big to fail” fail to predict, or when so many people that they form a majority or a very significant voting bloc fail to predict, then everybody pays.

When a small private actor fails to predict, then we can all ignore it, or point and laugh if so inclined. But in those other cases, when we’re all left holding the bag as well, we care a little more.

Willitts January 21, 2014 at 6:47 pm

Black swan.

john personna January 21, 2014 at 6:57 pm

As Taleb cynically notes, when prediction fails (in government or industry) the classic fix is a “do over.”

When you have endless predictions you can never really fail.

Erik January 21, 2014 at 2:48 pm

But where is 2012? In fairness, up through 2008 it seems perfectly reasonable to assume a reversion to the mean. After 2010, anyone looking at this should see that there is a new patter that has emerged and demanded a fundamental review. What happened to the 2012 forecast??

John Thacker January 21, 2014 at 5:21 pm

The reports are dated two years earlier from release. The line dated 2010 was in a report released on 2012 but based on 2008 data. (Weird, yes.) From the SSTI site that original made the graph:

The most recent C&P report, released in 2012, was dated 2010 and based on data from 2008. In Figure 1 the trend line from the 2010 report begins in 2008.

prior_approval January 21, 2014 at 2:50 pm

Almost as if the forecasters where in the pocket of the auto industrial complex.

Except for one little nagging question – what is that graph actually supposed to be measuring? Traffic volume as said in the article?

Because in terms of VMT, that graph is not particularly accurate –

‘For decades — through the rise of the two-car household, women entering the workforce, the growth of the exurbs — Americans reliably put more miles on their cars every year.

But no longer. Last year, for the eighth year in a row, vehicle miles traveled ticked down on a per-capita basis. The average American drove 37 fewer miles in 2012 than in 2011 — a 0.4 percent drop, according to new data from FHWA. It’s a small but significant decrease, continuing the downward slide of per-capita VMT that began in 2004, well before the economy faltered.

Experts attribute the reversal to a variety of factors including the gradual retirement of the baby boomer generation, volatile gas prices, decreased interest in driving by millennials, and the increasing popularity of walkable neighborhoods.

Meanwhile, population growth caused total mileage to tick up 0.3 percent in 2012. Total VMT, which has also seen a reversal of historical patterns, has declined three of the last eight years, for a net decrease of 0.9 percent over that time, reports the State Smart Transportation Initiative. Noting that total mileage has leveled off, SSTI advises state DOTs to rethink projects that add highway lanes — projects that are often justified based on faulty models assuming growth in VMT.’ http://dc.streetsblog.org/2013/02/27/for-eighth-year-in-a-row-the-average-american-drove-fewer-miles-in-2012/

The posted graph shows much larger declines than that of total VMT – but then, since that graph doesn’t actually both to mention what it is measuring, it isn’t as if one can criticize it based on actual information.

Yancey Ward January 21, 2014 at 3:21 pm

I am shocked the Auto Industrial Complex hasn’t had you whacked.

Willitts January 21, 2014 at 6:45 pm

Their black sedans were garaged with mechanical difficulties. Their black helicopters are grounded due to weather.

Marie January 21, 2014 at 6:47 pm

It has. That’s not the real prior approval. Shh.

Todd January 21, 2014 at 3:55 pm

retirement of the baby boomer generation, volatile gas prices, decreased interest in driving by millennials, and the increasing popularity of walkable neighborhoods.

I work with a number of large companies and in the past 10 years the number who allowed one or more work at home days has gone from 0% to 75%. That’s certainly a factor as well.

JWatts January 21, 2014 at 5:39 pm

I work for an industrial engineering firm. A substantial amount of our meetings are web ex and a lot of direct plant floor support is via VPN.

Indeed, it was positively awesome last Easter to be able to wake up to a phone call at 5am, fire up the laptop, VPN to a plant that was 620 miles away and have the problem fixed by 9:00am. In the past I would have been frantically booking the first available flight or, more likely for Easter morning, getting in the car and making a 13 hour drive. Fixing the problem, spending the night in a hotel room, and then spending the next day traveling back.

That level of support wasn’t possible when I started working in 1999. Granted, we had some remote support capabilities (PCAnywhere via 56K modem), but they just weren’t to the same level.

There have been some very large gains in avoided travel costs in the past 10-15 years.

Steven Kopits January 22, 2014 at 8:48 am

Mobility is virtually synonymous with GDP. Did vehicle miles driven fall with the rise of the internet, the spread of mobile phones? No, it fell when the oil supply stalled in 2005 and oil prices soared. The Chinese were (and have been) simply outbidding the US for oil, and that shows up in mobility. If in doubt, check the UK data: their VMT starts bending at the same time as ours. This is a big, big deal, really worthy of an article rather than a comment.

JWatts January 22, 2014 at 10:54 am

“Mobility is virtually synonymous with GDP.”

Yes, but I believe that virtual mobility is also synonymous with GDP. ;)

“No, it fell when the oil supply stalled in 2005 and oil prices soared.”

I have no doubt that high oil/gas prices triggered a change in behavior. I just believe the access to high quality online shopping and remote work stopped a reversion to trend line and that the current indicators point to a permanent drop in the amount of miles traveled per capita. However, to your point, it will probably require a decade before we have any confidence in the trend one way or another.

Steven Kopits January 22, 2014 at 5:09 pm

The numbers disagree with you, JW. And the point is important.

If you are arguing that VMT is declining due to changing consumer preferences, then we would expect to see weak demand for oil, and low oil prices. Not so. Oil prices are at the limits of affordability for the US. US consumers are literally struggling to hold on to mobility. Oil consumption has collapsed across a number of sectors–VMT is the least of it. Jet fuel (through October) is down 16%, heating oil down 46% since 2005. Gasoline and diesel are down much less. Look at the stats, and they speak of virtual desperation on the part of the US consumer.

And we can see this is other studies as well. Here’s one from the indispensable Michael Sivak, of the University of Michigan’s Transportation Research Institute (UMTRI). His survey indicates that only 19% of people aged 18-39 without a driver’s license held a full time job. I was so stunned by this statistic that I really pushed Michael on the issue, but he said that’s what the numbers say. http://deepblue.lib.umich.edu/bitstream/handle/2027.42/99124/102951.pdf?sequence=1

Here’s another study, from the Highway Loss Data Institute, an insurers group. It concludes that unemployment represents 80% of the reason that young people are driving less. http://www.iihs.org/iihs/news/desktopnews/drop-in-teen-driving-tracks-with-teen-unemployment-hldi-study-finds These studies do not speak of consumers losing interest in physical mobility.

And further, if we thought the US was starved for oil, then we might expect demand to rebound sharply if prices fell. And it did. In Q4, the US added nearly 1 mbpd of consumption, 3-4%, in a matter of weeks, when oil fell to around $3 / gallon. This would not have happened if consumers no longer wanted to drive.

Let’s not kid ourselves. This country, as well as Europe, has been starved for oil. That’s what’s driving VMT (1 car in every 6 is missing from the road compared to trend), commercial aircraft departures (1 in 3 missing compared to trend), and heating oil, down by half. These are astounding numbers, and we have every reason to believe they are affecting the economy.

If you’re interested (and live in the NY area), I’ll be presenting on oil and the economy at Columbia’s SIPA, Room 1512, Tuesday, Feb. 11, 12:30. It’s open to the public. I’ll be going through supply-constrained models, how these relate to demand-constrained models, and their impact on the oil industry, consumers and the general economy. (And if Tyler would like to invite me, I’d certainly come present at GM as well.)

zbicyclist January 22, 2014 at 10:49 am

Indeed. No more rushing to the airport to visit a client site when you can dial in remotely (or webex/Skype to those who are right there). I feel no nostalgia.

This is a two edged sword, of course — since almost no matter where you are or what you are doing at the time, they can reach you.

Tom T. January 21, 2014 at 6:08 pm

The shift from brick-and-mortar to online shopping presumably contributes too.

RM January 21, 2014 at 8:13 pm

This may be temporary. With Amazon building warehouses around the country, it may well make sense to send your driverless, cheap to operate car to pick up your package. Or Amazon can send their driverless car with your package to your home. The driverless car is the alternative to drones. Walmart is in even a better position to deliver this way–whether your driverless car or theirs–because they already have warehouses everywhere.

Folks, we are coming upon the age of more, not less, driving

msgkings January 21, 2014 at 9:25 pm

Why wouldn’t they just make the UPS/Fedex trucks driverless? Still economies of scale delivering lots of packages to one area. So, no more driving (at least with that variable).

Anti-ummm January 21, 2014 at 10:42 pm

You have to pay for gas. Makes sense to retain the UPS model. They can be driverless but they will need a guy to deliver the boxes because the robots are too slow.

Bob January 23, 2014 at 1:23 pm

The driverless car will be much cheaper to operate when it brings the packages of 50 people, so your mileage numbers wouldn’t be very different from those you get from the UPS truck

John Thacker January 21, 2014 at 5:18 pm

You do realize that that graph comes directly from the site of the State Smart Transportation Initiative mentioned in your linked news article, right? I agree it’s not the best graph (doesn’t start y axis at zero, and removing the caption at the bottom that gave the units didn’t help), but the underlying information obviously agrees since it’s from the same people.

Careless January 21, 2014 at 10:44 pm

You do realize that you’re addressing a lunatic, right?

John Thacker January 21, 2014 at 11:43 pm

Well, if we called everyone whose thinking was entirely driven by personal animus, a desire to agree or disagree based on the speaker and not the facts or arguments, and marked by an extreme selective reading of the facts and willingness to abandon positions and contradict oneself at the slightest hesitation, I think we’d have to call far too many of us lunatics, so I’m not sure it’s the right word.

Careless January 22, 2014 at 1:31 am

Yeah, but PA isn’t that rational. He can’t even reliably make an argument for or against GM, despite his intentions. He’s absolutely crazypants.

prior_approval January 22, 2014 at 12:28 am

Yep – and the written information does not support the graphs they use from SSTI.

For example, the decline in per capita VMT from 2011 to 2012 was 37 miles – which is essentially flat at a .4% decline.

Even more interestingly, though the text says that ‘Total VMT, which has also seen a reversal of historical patterns, has declined three of the last eight years, for a net decrease of 0.9 percent over that time.’ Yet the posted graph seems to have decline from something quite over 3 million at peak to a number a bit over 2.9 million – seems a bit more than of a decline than the actual one of 30,000.

Advocacy group posts colorful information that supports what it advocates, hoping that people will notice the pretty pictures and ignore the actual data.

I would have posted to the years long VMT coverage at calculated risk, but this is not exactly the sort of web site that cares about actual facts.

John Thacker January 21, 2014 at 5:24 pm

The posted graph shows much larger declines than that of total VMT

No, the graph is consistent, just the article (and graph) is confusing. One big problem is that the quoted article mentions “has declined three of the last eight years, for a net decrease of 0.9 percent over that time.” You’re noting that the decline is more than 0.9% from the peak, which I agree is probably how the article should have been written. Eight years ago (at the time of the quote) wasn’t the peak, so traffic both increased and decreased since that time. The decline is 0.9% from the start of that eight year period, and a greater percentage from the peak.

prior_approval January 22, 2014 at 1:15 am

So, since we are repeating the same positions (my fault for not noticing the bottom comment), let us use a bit more current data, and another creator of the resulting graphs from that data –

‘The Department of Transportation (DOT) reported:

◦ Travel on all roads and streets changed by 1.3% (3.4 billion vehicle miles) for August 2013 as compared with August 2012.

◦ Travel for the month is estimated to be 267.0 billion vehicle miles.

◦ Cumulative Travel for 2013 changed by 0.3% (6.1 billion vehicle miles).

The following graph shows the rolling 12 month total vehicle miles driven.

The rolling 12 month total is still mostly moving sideways.

In the early ’80s, miles driven (rolling 12 months) stayed below the previous peak for 39 months.

Currently miles driven has been below the previous peak for 69 months – almost 6 years – and still counting.

The second graph shows the year-over-year change from the same month in the previous year.’

http://www.calculatedriskblog.com/2013/10/dot-vehicle-miles-driven-increased-13.html

Which makes the peak around 2007, not 2005 – and the more that I look at those SSTI graphs, the more I’m convinced that calculated risk is exemplary in presenting information using graphs. Like having the recession periods marked on the graph, information per month, and clear text explaining what the graph actually details.

Further, the trend until 2007 is quite clear – Americans drove about three times as many total miles in 2007 as they did in 1971.

SG January 21, 2014 at 3:13 pm

The Fed is one of the worst offenders with this.

Evan Soltas had a great post back in 2012 on the Fed as little orphan annie

http://esoltas.blogspot.com/2012/08/the-fed-as-little-orphan-annie.html

Max Factor January 21, 2014 at 6:52 pm

The Fed really botched its LFPR projections. Here’s an IMF stumble:

http://www.zerohedge.com/news/2014-01-21/comedy-imf-forecasting-errors-global-trade-tumbles-more-50-imfs-2012-prediction

2013 world trade was expected to grow by 5.6% in April 2012. Now: it is more than 50% lower at just 2.7%!

BS January 21, 2014 at 3:34 pm

Am I missing something, or do the years on the horizontal axis seem to be off by 2 (the 1999 forecast starts in 1997, 2002 forecast starts in 2000, and so on)?

John Thacker January 21, 2014 at 5:26 pm

You’re correct. As the SSTI says (and they made the graph):

The most recent C&P report, released in 2012, was dated 2010 and based on data from 2008. In Figure 1 the trend line from the 2010 report begins in 2008. Other trend lines reflect similar patterns.

Confusing dating by the FHWA, huh?

Semi-Senior Mitch Daniels Staffer January 21, 2014 at 4:08 pm

Looks like the City of Chicago & State of Indiana had great timing on their long-term leases of toll roads. Chicago Skyway for $1.8 billion in 2004 & Indiana Toll Road for $3.8 billion in 2006. http://www.in.gov/ifa/2328.htm

Ed January 21, 2014 at 4:16 pm

What exactly is the y access measuring?

In any case, sounds like the era of building new highways is over.
Now, we’ll just bury the ones we have underground so they’re out of sight.

Shane M January 21, 2014 at 6:46 pm

I think it’s trillions of vehicle miles traveled. (I’m assuming VMT is the acronym for vehicle miles travelled.)

http://www.ssti.us/2013/12/new-travel-demand-projections-are-due-from-u-s-dot-will-they-be-accurate-this-time/

Careless January 21, 2014 at 4:20 pm

In before “climate” makes it in the comments

dbg January 21, 2014 at 4:39 pm

linear extrapolation strikes again!

prior_approval January 22, 2014 at 1:17 am
RM January 21, 2014 at 5:14 pm

The big question is: how will driverless cars change this trend? With driverless cars we will see an increase in VMT. A few reasons why this might be the case: 1) sending the (electric and cheap to operate) car back home instead paying for parking; 2) getting to work at 8:30 and sending the back home to get the spouse to work at 10:00 for their part time job; 3) instead of chaining activities, returning home from shopping and sending the car to pick up the kids from practice; 4) the decreased opportunity cost of driving means many more pleasure trips.

(This may or may not be accompanied with less car ownership).

John Thacker January 21, 2014 at 6:30 pm

At that point, would you expect driverless taxis to finally make an appearance, as envisioned in many SF works?

Steven Kopits January 22, 2014 at 8:35 am

I expect driverless cars by 2020. Current offerings are pretty close already, for example, the Mercedes S500 could probably self-drive with little more than a throw of the switch. If can auto-brake, auto-accelerate, lane hold, auto-collision avoid, pretty much all the sensors you need for self-driving.

SDVs will materially increase traffic, I think, primarily through their effect on underserved segments, notably youth (how young) and the aged (a fast growing population).

John Thacker January 22, 2014 at 11:19 am

Right, but the question is, if cars are driverless, would you expect more people to shift to a taxi/Uber/Zipcar like model, rather than owning a car themselves?

Steven Kopits January 22, 2014 at 5:32 pm

You’d get a differentiated vehicle fleet. I would guess that self-driving electric vehicles would take a 25% share, primarily for local trips and used as a service, not as owned assets. Everything else stays pretty much the same, with shares adjusting pro rata. Average HH car ownership would decline (compared to trend) by about 0.5 vehicles or so. Total vehicles (including driverless taxis) might increase by 0.5 vehicles per household compared to trend. But I’m guessing. That’s not analysis.

Interestingly, the vehicle fleet would begin to look more like the electric power fleet, even as the power fleet might become more fueled solely by natural gas and renewables.

Aric January 21, 2014 at 6:12 pm

@RM: I have sometimes joked, “You think that environmentalists are mad about cars / SUVs driving around with one passenger in them? Wait until they have zero.”

RM January 21, 2014 at 6:50 pm

Totally true. Policy makers love the idea of the electric car because of the environmental benefits from less emissions. No account is taken of the fact that we will be exchanging one environmental problem for another: less emissions but more fragmented land uses in far exuburbs (loss of habitat, degraded water quality, etc.) as the costs of driving fall. Some serious analysis needs to be done on the environmental impacts of cheap to operate, driverless cars.

wrparks January 23, 2014 at 9:55 am

And apparently the environmental benefits may not materialize.

Hot off the presses.

http://pubs.acs.org/doi/abs/10.1021/es4045677

“While the scenario parameters influence EDV deployment, the EDV deployment does not in turn produce a discernible effect on total system-wide emissions.”

prior_approval January 22, 2014 at 1:23 am

Or wait until one car is shared by many people. Like with Stadtmobil, which has forty thousand people sharing 1800 vehicles – German only http://de.wikipedia.org/wiki/Stadtmobil

With a driverless car, the sharing becomes even more practical – after the car has made its journey, it simply makes the one next one, instead of waiting around in a parking lot for a typical 8 hour work day.

The idea that everyone will own an expensive driverless vehicle is pretty silly, to be honest.

Instead, think personalized mass transit, making use of a vehicle much more efficiently than today.

John Thacker January 22, 2014 at 11:20 am

Indeed, it would be like a driverless taxi.

Which is why it’s bizarre that cities (generally Left-leaning) decide to regulate companies like Uber and taxis so much more than people owning a private vehicle.

Willitts January 21, 2014 at 6:41 pm

If these lines are a compilation of state forecasts, it is even scarier.

Have they never heard of an ARIMA model? Structural breaks?

Anyone? Bueller?

DK January 21, 2014 at 9:32 pm

I am sure DOT taxes have absolutely nothing to do with any of it!

Steko January 21, 2014 at 10:18 pm
John Thacker January 21, 2014 at 10:24 pm

Yes, long term oil prices combined with the long term VMT graph make it clear that it’s not just oil prices causing this, but something else as well.

John Thacker January 21, 2014 at 10:27 pm

After all, VMT increased linearly on a 12 month rolling average throughout the 70s gas crisis. The linear increase in VMT was not affected by crude oil prices before. It’s clear that something has changed now. (Oil prices being high surely should play a role, but why has the apparent elasticity increased so much compared to the previous 40 years?)

Steko January 21, 2014 at 10:34 pm

Sorry I hit the reply in the wrong box below and before I was finished getting my quotes:

“To explore the possibility that traffic might grow more slowly than assumed, an alternative HERS analysis was conducted assuming for illustration that VMT will grow at the average annual rate of 1.23 percent, the historical average from 1998 to 2008. Modifying the input forecasts to match this VMT growth rate would reduce the benefits associated with pavement and capacity improvements, so that an annual spending increase of only 3.52 percent (translating into an average annual investment level of $80.2 billion) would be sufficient to fund all potentially cost-beneficial projects by 2028. If spending were instead sustained at 2008 levels, HERS projects that average speeds would improve by 2.1 percent under this alternative compared with a decline of 0.7 percent under the baseline assumptions.” – See more at: http://marginalrevolution.com/marginalrevolution/2014/01/traffic-forecasts.html#comment-158027003

Another sensitivity test concerns the growth rate between 2008 and 2028 in motor fuel prices relative to general rate of inflation. The baseline HERS assumption is of no difference between these rates. An alternative assumption was based on the High Oil Price case from the Energy Information Administration, Annual Energy Outlook 2010. In this case, the ratio of gasoline prices to the consumer price index nearly regains its 2008 level by 2012 and increases thereafter through 2028 at the equivalent of 3.4 percent annually. The change in assumption from the baseline case causes HERS to reduce its projection of future travel growth and reduces the model’s estimate of the average annual investment level needed to fund all projects with a benefit-cost ratio of 1.0 or higher by 2028 to $96.9 billion.”

Steko January 21, 2014 at 10:39 pm

I don’t expect a third generation reblog of an item to due much diligence esp when an item plays right into their blinders but you’d think a specialist outfit like SSTI would have more rounded coverage.

John Thacker January 22, 2014 at 12:02 am

I don’t expect a third generation reblog of an item to due much diligence esp when an item plays right into their blinders but you’d think a specialist outfit like SSTI would have more rounded coverage.

I don’t possibly see what due diligence you would want Alex to do here (as opposed to Andrew Gelman or SSTI). Reality has not lived up to the projections. One can certainly blame FHWA, or one can note that reality and trends have shifted suddenly from what held true for many years and even completely excuse them for not updating the model after 10-15 years of divergence. I think that people with all sorts of blinders and preconceptions would be able to project onto this graph, and Alex’s sparse commentary does not push the reader in any direction.

John Thacker January 21, 2014 at 11:58 pm

Is your point here that they did issue an “alternative analysis?” That is certainly welcome, but note that the actual performance since 2008 ended up being considerably slower than their alternative analysis, since things have not even returned to the 1998-2008 trend, much less the preceding 1970-1998 trend that all the standard analyses have continued to use.

And while certainly very interesting, I see nothing in the quote to suggest that gas prices (as in your FRED graph) are sufficient to explain this change in the growth of VMT. It certainly didn’t have this effect during the gas crisis in the 70s. FHWA really has no idea why it’s happening either, so they stick to linear interpolation. Their preferred model remains the 30 trend line which held good until 1998 or so, but their alternative model used a recent 10 year trend line. Neither of them are accurate. The FHWA (and associated agencies) does not do sophisticated forecasting that tries to determine why, they simply extrapolate– which for many years served them quite well. Now it does not, which is a mystery.

Steven Kopits January 22, 2014 at 6:10 pm
Careless January 21, 2014 at 11:04 pm

Doesn’t explain the beginning of the trend, and the trend follows the beginning, so it really doesn’t explain any of it. And even if it did, it wouldn’t explain the pathetic failure of the later lines.

Steko January 21, 2014 at 10:31 pm

“To explore the possibility that traffic might grow more slowly than assumed, an alternative HERS analysis was conducted assuming for illustration that VMT will grow at the average annual rate of 1.23 percent, the historical average from 1998 to 2008. Modifying the input forecasts to match this VMT growth rate would reduce the benefits associated with pavement and capacity improvements, so that an annual spending increase of only 3.52 percent (translating into an average annual investment level of $80.2 billion) would be sufficient to fund all potentially cost-beneficial projects by 2028. If spending were instead sustained at 2008 levels, HERS projects that average speeds would improve by 2.1 percent under this alternative compared with a decline of 0.7 percent under the baseline assumptions.”

Bill January 21, 2014 at 10:47 pm

All of the projections were made in the Bush years, when you consider the latest projection, that for 2010, was made in 2008..

Anything more recent?

John Thacker January 21, 2014 at 11:51 pm

That’s not quite correct. As I’ve said twice else in this thread, if you follow the link to SSTI’s site, “The most recent C&P report, released in 2012, was dated 2010 and based on data from 2008. In Figure 1 the trend line from the 2010 report begins in 2008.”

The latest projection was dated for 2010, made in 2012, and was based on 2008 data. If you mean that the latest projection was based on Bush year data, that’s correct, but it was certainly made in 2012. Are you arguing that being based on 2008 data means that it was “made in 2008?” I can sort of see an argument for that, but looking at the graphs, it seems unlikely that if they repeat their previous model anything would change. They certainly decided in 2012 to continue to apply the same model to the 2008 data, even though they had some real life experience of 2009-2012 to caution them against doing so.

The first projection based on 2010 data will be released this year, but the report will be dated 2012 for some confusing reason.

Bill January 22, 2014 at 9:14 am

john. since this is an aggregation of estimates of others, I dont see your point. It is still 2008 data, and still an aggregation of state projections. it would seem states have an incentive to skew projections.

John Thacker January 22, 2014 at 11:22 am

The actual data used to seed the projections is from 2008. The projections were made in 2012, when preliminary data from 2009-2011 (and monthly VMT estimates from Traffic Volume Trends for 2012) were available as well. It should have been obvious in 2012 that running the same model of the previous 30 years on the 2008 data was a bad idea.

John Boyce January 22, 2014 at 9:36 am

The most priceless thing at the link was the PA Samuelson graph of the USSR overtaking the USA, along with PA’s moving the intersection date to a later year in each edition! If libs could only control the weather…

JWatts January 22, 2014 at 11:00 am

“Samuelson’s influential textbook has been criticized for including comparative growth rates between the United States and the Soviet Union that were inconsistent with historical GNP differences. The 1967 edition extrapolates the possibility of Soviet/U.S. real GNP parity between 1977 and 1995. Each subsequent edition extrapolated a date range further in the future until those graphs were dropped from the 1985 edition. Samuelson concluded the economic description of the Soviet Union and marxism in 1989: “Contrary to what many skeptics had earlier believed, the Soviet economy is proof that … a socialist command economy can function and even thrive.” The Collapse of Communism happened during the same year and the Soviet Union broke up two years later.”

Ah yes, another Nobel Prize winning Economist.

Comments on this entry are closed.

Previous post:

Next post: