Friday, January 3, 2014

Fighting "Data" with Data

Two can play this game.  I'm a bit fired up.  So let's delve straight into it shall we?  If you recall, in this morning's post I took a brief look at the Reason Foundation's latest and greatest.  Well, I have all the same data they use from (some of their own reports) as well as the US Census.  What I won't use is their Travel Time Index because it is abstract, arbitrary, and meaningless.

Over the past year or two I've been compiling all of this data into a few very large spreadsheets that list the top 250 metros or so.  I'll stick to the top 20 or so for these purposes since they're the cities with at least some modicum of transit investment.  Plus, we need to compare apples to apples.  For much broader data dump that breaks cities down by size, see this link.

Keep a close eye on the r-squared values because as I shift from the type of data they're focusing on to the type of data that I consider more critical to what their stated end is, better use of public resources towards better public decisions and behavior patterns, you'll catch a noticeable change.  (higher the r-squared, the greater the correlation, the more significant the data set).

First, we have Lane Miles per Capita with Average Commute Times

 photo Untitled-6_zps9d2d5b9b.jpg

The Reason Report is suggesting we build more roads.  If you deliver more highway capacity, you are going to increase the number of highway lane miles per capita.  You'll notice STL and Hou way up at the top for highway capacity per capita.  Then there is a giant mass in the middle with little correlation between anything.  And then two off to the right with higher commute times.

The first point to be made is how statistically little significance there is here.  There is a very VERY slight relationship between more highway capacity and reduced commute times.  The primary issue is because our transportation infrastructure shapes our real estate markets in a way that produces a statistical equilibrium in commute times.  We like being about 25 minutes from work.  Every city revolves around 25-30 minutes.  The kind of infrastructure is fairly irrelevant.

Now about those two off to the right with higher commute times.

Here we have Congestion vs Average Commute Times.

You'll see those same two cities off to the left in terms of longer commute times.  As you see it's DC and NYC (and this is for metropolitan statistical areas or MSAs).  Yes, DC and NYC have the highest commute times of the 20 biggest metros.  They also have the highest transit ridership.  

In Reason's world, this is bad.  High transit, high commute time, baaaaaad.  What it also means is those commuters can be more productive with their time while spending less to commute to and from work.  In the world of transit vs car-dependence, the issue is high public spending/reduced private spending vs high public spending/high private spending.  Car dependence costs a lot and doesn't save much time.  

As you see, there is virtually zero relationship between congestion, measured as traffic per travel lane to commute time.  So if congestion has nothing to do with commute times, why are we bothering?

Next, we have Density vs Average Commute Times.

 photo Untitled-8_zps725d98b3.jpg

No relationship.  Because once again, on average, we all look for some distance between home and work.  Some separation between the disparate tasks of our day and the mindsets between them.  Cities adapt to equilibrium.


Vehicles Miles Traveled vs. Average Commute Times

 photo Untitled-9_zpsb7096079.jpg

Not a terribly strong relationship here, but there is some decrease in commute time for driving more.  Is that a good trade-off?


Car Dependence vs Lane Miles per Capita

 photo Untitled-10_zps78dffd41.jpg

The relationship is beginning to strengthen ever so slightly and based on the data points available, the trendline doesn't show the relationship as strong as it actually exists when you start factoring in non-US cities.  When you do, you start seeing an even stronger and sharper relationship between highway capacity per capita and car dependence, which I measure as commuting by individual car or carpooling.

The nearly free good (since highway users only cover about 43% of costs through user fees (tolls and gas taxes) of free, fast highway travel skews behavior patterns.  This is basic economics and behavioral science at work.  People like free stuff (and hate to let go of their entitlements) but there are few things more expensive than free.

Travel Time vs Lane Miles per capita

 photo Untitled-11_zps31fbac96.jpg

This is a similar graphic to the one at the top.  Same categories.

However, I want to show it in a way here that illustrates how little effect new highway capacity has on reducing travel times.  The slope is virtually flat.  But think about it this way. A doubling of highway capacity per capita, from the lower end to the higher end of this spectrum (from 5 to 10) comes with a cost.  For a metro the size of DFW to double highway capacity (which would have little effect on commute times), would costs $304 billion at $80 million per mile.

So, yay?


Lane Miles per Capita vs Vehicle Miles Traveled per Capita

Now we're seeing a big jump in r-squared, thus correlation between data sets.  More lane miles, in all likelihood you're going to be driving more.  Highways disconnect as much as they connect, but when they connect, they connect long distances.  Thus, you're driving more.


Travel Time vs Car Dependence

They're right!  The higher transit use the longer the commute time, which echoes the DC/NY information above.  Also, by increasing car dependence, which in my estimation is their goal, they can suggest gains in commute times.  But by increasing dependence, which at these levels suggests choice is undermined, is that really the ideal?  Again, what's more efficient in terms of energy, infrastructure, and use of time.

Furthermore, why would people live in DC or NYC in the first place by this logic?  Why isn't Saint Louis the ideal?  It is Reason's ideal.