Austin Contrarian

Intersection density is the wrong metric for connectivity

One last point about CNU's proposal to encourage street connectivity.  CNU proposes using a metric of 150 intersections per square mile.  This is the wrong metric.  If we want to mandate connectivity, we should use a metric that guarantees connectivity; requiring a certain number of intersections per square mile doesn't do the trick.

The street plan below illustrates how a developer could meet CNU's intersection-density standard without providing even minimal connectivity.  The box is one mile by one mile. The black lines are streets and the white spaces in between are for houses and green belts.

It's crude, I know, but it makes the point.  A developer could simply cut the interior of the subdivision into disjoint pods and cover each pod with a dense grid. I've drawn in the local streets for the pod in the lower left-hand corner. The blocks are supposed to be roughly 400-feet long, with one half-block. It has 26 intersections, which would scale up to 156 intersections for the entire subdivision.  Even this density doesn't guarantee connectivity within the pod. This pod is divided into four quarters, which are mutually accessible only along the pod's perimeter road.

In this example, there are neither east-west nor north-south routes through the subdivision interior. The only through routes lie one mile apart along the subdivision's perimeter. I don't think is what CNU has in mind when it talks about connectivity.

If we are after connectivity we should use a metric that guarantees connectivity. We could do this by requiring a minimal number of independent routes -- that is, routes without common streets -- between any two nonadjacent intersections.*  (For intersections interior to a street grid, that number is four.)   By also requiring a minimal number of subdivision entrances per mile (say every 400 or 600 feet), we would get the desired through connectivity and internal connectivity without having to worry about creative subdivision street layouts.  On the contrary, developers would get a bit more flexibility in laying out their subdivisions while providing the public benefit that is the point of this exercise.

*It isn't hard to determine the number of independent routes between two non-adjacent intersections.  It's equal to the minimal number of separating intersections -- i.e., the minimal number of intersections you'd have to eliminate to disconnect the original intersections.  This means it's enough to look for choke points.  For example, the intersections interior to the pod sketched above would be disconnected from the rest of the subdivision by eliminating the intersections at the pod entrance.  Hence they wouldn't satisfy a standard requiring more than two independent paths.

Flip flop

A couple of weeks ago, I suggested that cities give developers incentives to build dense, connected, local street networks rather than simply mandate them.  After thinking a bit harder about it, though, I'm not sure that's the best solution.

Cities should refuse to maintain subdivision streets that do not provide a minimal level of connectivity for other city residents (as Virginia will now do).  But should a city or state go one step further and mandate connectivity?  Should it pay developers to provide connectivity?  Or should it do nothing, other than leave the maintenance costs on subdivisions with functionally private streets?

Using incentive payments is sound in theory.  Connectivity provides a spillover benefit to residents outside the subdivision because it lowers congestion on major arteries, gives them more alternatives for getting around, and solves a collective action problem.  If bans on mixed-use development were also relaxed, connectivity could also shorten net trip length.  (Connectivity alone won't do the trick.)  The disconnected, cul-de-sac pattern is less expensive than a dense street grid.  But if a city offered developers incentives equal to the value of connectivity to the city, then developers would build dense, connected grids when it made economic sense to do so. 

There are a couple of practical problems with this, though.  One is that no city could afford to do it, at least not for enough subdivisions to make a difference.  The second is that it would be hard for the city to figure out what to pay.  A city could pay for all subdivision streets.  But see objection no. 1 (and cities shouldn't be subsidizing subdivision streets anyway).  A city, in theory, should pay only for the extra street length needed to get up to a minimal level of connectivity.  But that depends on what the developer would have built in the first place, and the city wouldn't be able to determine that. 

That leaves regulation (mandating connectivity) or doing nothing.

Simply mandating connectivity eliminates the drain on the city fisc.  It is more feasible, politically, than simply dumping money on developers, as Virginia has demonstrated.  Regulation is a tricky thing, though.  The cost-benefit analysis does not change merely because the city is no longer paying its share.  And, again, connectivity can be very expensive.    

Perhaps the best solution is a hybrid approach.  Cities could mandate a certain minimal level of connectivity but let developers buy their way out of it.  That might provide enough of a nudge to solve the collective action problem without mandating street designs that cost more than they are worth.   

Federalizing local street networks

Streetsblog has an interesting interview with John Norquist, the president of the Congress for the New Urbanism.  He discusses boulevard design, highways' impact on cities, and the benefits of dense, local street networks.

He also discusses a CNU proposal to use federal funds to encourage the latter:

The first step is to allow federal funds to be used to bring street networks up to a standard of 150 intersections per square mile. So if you have a suburban sprawl kind of situation where the intersection density is like at 40 per square mile, if you have a project that’s going to bring that intersection density up to 150, then the state would be eligible for getting federal funding to go in and do that.

As much as I prefer grids to cul-de-sacs, I think this is a bad idea.

First, it’s important to distinguish between local street networks and major arterials like boulevards and highways.  States and the federal government (indirectly) pay for most highways, while cities pay for most boulevards and other arterials.  Norquist rightly complains that the feds have shaped highway design and location for decades.  Not to mention the mix of highway and transit spending.

Local streets are different.  Local street networks look like they do because cities don’t build them anymore.  They instead are platted, built and paid for by subdivision developers. They don’t become public streets until dedication, when cities take title and the responsibility for maintaining them.  Cities thus have privatized their grid design.  We should not be surprised, then, that developers ignore spillover benefits -- especially connectivity -- when laying out their streets.

Cities do plat and build the arterial roads that link subdivisions to the rest of the city. But we can’t get intersection density by redesigning arterials.  Arterials are one-dimensional; grids are two-dimensional.  CNU understands that, which is why it wants federal funding for local street networks with a minimum intersection density.

I’m sure developers would be perfectly happy with CNU's proposal. It would eliminate one of their two major infrastructure costs. (The other is running connections from the subdivision to the local water, sewer and electric lines.)  Developers would lose some developable lots because a dense street network needs more pavement than the minimal street networks developers build.  But I’m sure developers would gladly give up 10% of their buildable lots if cities would agree to build their streets.

Do we want to abandon developer-financed local streets?  Even if cities and the feds paid just part of the cost, this would be an immense expense.  Most cities don't have the money.  It would take billions per year -- every year -- to make a noticeable difference in city structure.

It is also a bad idea to subsidize developers and property owners.  Reducing the cost of building subdivisions would get us . . . more subdivisions.  If you dislike the relentless spread of single-family subdivisions, then you want developers to bear all the cost.  (Actually, you should want developers to bear all their costs regardless.)  

Such a subsidy could be efficient in theory if it were limited to the extra expense incurred by the developer to densify their street networks.  But developers would have no incentive to disclose their true costs, and local officials could not figure that out on their own.

Finally, it is a bad idea to get the feds involved in local street design.  If “incentives” meant “funding,” the feds would dictate street layout, just like they dictate highway width, turning radii, and drinking ages. Cities wouldn't  use their scarce dollars to build more than the feds require.  Developers certainly wouldn't. And the federal government might not do it the way you or I want it done.  Suppose the feds, thanks to lobbying by property owners and developers, decided to fund only 100 intersections per square mile? Or 75? Then 100 or 75 intersections per square mile is what would get built.

There would be enormous pressure to go with less dense networks since they are  cheaper.  Developers would continually push for less dense networks even if the cities were footing the bill -- sacrificing 20 houses for road funding might be a good deal, but sacrificing just 10 houses would be a better deal.

So whenever someone wanted to change the rules for local street design, they'd have to run to D.C.  That's worked real well in the past, right?

Downtown apartment owners should try this experiment

From City Beat:

Daimler will partner with the City of Austin to operate a car-share pilot program, Mayor Will Wynn announced this morning.

The city will set aside urban-core parking spaces for 200 of Daimler’s “Smart” cars. Rather than paying for the spaces, Daimler will let city employees use the cars for a number of hours that’s equal to the monetary value of the spaces. (That amount hasn’t been determined yet; it will be negotiated over the next few weeks, city officials said.)

The cars are small — two can fit in a typical parking space — and fuel-efficient.

The program will start in October and run for six months. Daimler will pay for fuel, maintenance and insurance during that time, said the city’s transportation director, Robert Spillar.

After six months, city officials will evaluate whether it saves money to lease the cars for the long term. Austin currently spends $6 million to $7 million on its fleet of 1,800 light-duty vehicles.

Employees would have to register to use the cars. They’d swipe a digital card across a car windshield, type in a pin code and drive.

Wynn said he hopes the program will reduce carbon emissions and encourage employees to use public transit or carpool into work and then use the Smart cars downtown.

Daimler started the program last year in Ulm, Germany.  It appears to be a succcess.

Downtown developers ought to be interested in this, too.  They pay bundles for structured parking -- $15,000, $20,000 or more per spot.  Building fewer spots would lower costs, which might make some marginal projects feasible.

Could a car-share program work for the apartment buildings downtown? Maybe.  The cost of purchasing a few cars every other year would be small relative to their annual operations budget.  They could recoup much (if not all) of the cost by renting out the cars to residents at an hourly rate.  The cost of storing the cars would be zero, since every apartment building has surplus parking spaces. And if fewer renters brought cars along with their futons, the apartments could be built with less parking.

Apartment managers could use the program to entice renters:  "Skip the car payment but not the car."  Getting to the rental car in the first place -- or the car to the renter -- is normally a big hassle, but it wouldn't be when the car was sitting in a garage ten floors down.

Developers will need to be persuaded.  They already take huge gambles by betting millions of dollars of their own money on markets that might not be there two years after they break ground.  This upfront gamble makes them (and their lenders) shun risky design.  If 900-sf apartments with granite countertops, stainless-steel fridges and 1.2 parking spots per unit have sold in the past, then that's what they'll build.

That's why it would it be neat to see one of the repeat players downtown experiment with the car-share program in an existing project.  Get some data.  If it works, then try it on a bigger scale in a new project.  Dispensing with lots of parking might even lower development costs enough to expand into lower-rent   markets.

O'Toole gets congestion pricing wrong

Randal O'Toole asks:

"The New York Times recognizes that, even in New York City, transit is perpetually hungry for subsidies. The state’s proposed solutions are to toll the remaining free bridges into Manhattan and new payroll taxes, all to subsidize so-called “sustainable” transportation. But why should drivers pay to subsidize transit and why should the state pay to subsidize Manhattan’s unsustainable densities?"

I don't think that anyone who really understands congestion pricing can ask that first question.  Properly pricing roads requires that transit be subsidized, at least until there is so much demand for it that it can pay for itself.

Let's look at the free bridges into Manhattan.  They are perpetually congested.  Pricing them properly would eliminate the congestion (most of the time).  As O'Toole very well knows, this is the economically efficient solution.

But congestion pricing does more than relieve congestion.  Congestion pricing tells us when a road needs more capacity.  Additional capacity costs money, and drivers are willing to pay only so much for it.  That "so much" is exactly equal to the price they are willing to pay to avoid congestion.  When the revenue collected by congestion pricing is low -- too low to finance new capacity -- we know we have enough capacity.  Drivers aren't willing to pay for more, so building it would be wasteful.

On the other hand, If a properly priced road starts generating "too much" revenue -- enough to cover the cost of expanding capacity -- then it's time to figure out how to add the capacity users are willing to pay for.

There are three ways to add new capacity: (1) add lanes; (2) build new, alternate routes; or (3) add rail. Sometimes (1) will make most sense. Sometimes (2) will be the right choice. And sometimes (3) will be the only sensible choice. A rail line will be the right choice when -- surprise! -- it is the most cost-effective way to add the capacity that travelers are willing to pay for.  

Viewed properly, then, transit is simply a means for adding capacity.  (Yes, I'm well aware of other arguments for transit, but I'm concentrating on the one that ought to appeal to the economics-inclined.)

The logic of congestion pricing, then, is that the old capacity is supposed to finance the construction and operation of the new capacity.  Or to put it in the charged language that dominates transit debates, the old capacity is supposed to subsidize the new capacity.  Indeed, if users had to pay "cost" for the new capacity, then too few would switch from the old road.  That makes no sense because it creates the wrong incentives.  

This does not mean the new capacity -- whether road or transit -- should be free.  It should be congestion-priced, too.  And as time goes by, the revenue raised thereby might cover the cost of operation and, in turn, finance even more capacity.  But until that transpires, we shouldn't insist that new capacity pay for itself.  (I'm distinguishing between "new" and "old" capacity for simplicity, but it doesn't matter which is built first.  As long as everything is priced right, the subsidies will run the right way.)

Now one can argue about whether New York City's mass transit provides more capacity than the city needs (I don't think that's true) or does so at too high a price (I don't know enough to say).  But either argument is irrelevant to the central point: the bridges into Manhattan should be priced, and any revenue generated above the cost of maintaining them should help finance transit.  

Cheap houses

Carpe Diem claims the average sales price of a Detroit home fell to $13,638 in January.

I had trouble believing that so I checked out a Detroit MLS site.  Go here to browse 88 pages of homes listed for $20,000 or less.  Included is this stately mansion for $19,900 and this 6-bedroom Victorian for $18,500.  This This quaint three-bedroom, one-bath is listed for $1 under $2 per square foot.  Imagine the disamenity of a city that cannot entice new residents even with a (practically) free house.  The market-clearing price for much of Detroit's housing stock is probably negative, but the price obviously won't go that low and the market won't clear. 

(Is there an arbitrage opportunity here?  How much would it cost to move these houses to, say, Chicago or Milwaukee?)

If you want a visceral sense of the loss, see this photo gallery, "The ruins of Detroit."

How highway spending reduces congestion (or doesn't)

The Texas Transportation Institute estimates that highway congestion cost Americans $78 billion in 2007.  The average person probably couldn't cite a specific number, but he knows it's a lot.

Quantifying the deadweight loss from congestion is one thing.  Figuring out how to fix it is another.  Proposed solutions fall into two categories:  "build more capacity"  and "do something else."  Depending on whom you ask, the "something else" means congestion pricing, adding light or commuter rail, or encouraging more walking, biking, carpooling and flextime.

The "more capacity" crowd relies on common sense:  if we add more lanes -- more capacity -- to a congested road, then the road won't be as crowded.  The "do something else" crowd argues that increasing the capacity of a congested road simply induces more congestion as the new capacity attracts users from transit, alternate routes, and off-peak travel times.  And the new residential developments that spring up along new roads inevitably pump even more cars into the system.  

This debate cannot be won or lost with ideological arguments.  It must be settled empirically.  Economists Clifford Winston and Ashley Langer have entered the fray to do just that.  Their empirical analysis shows that, although highway spending does tend to reduce congestion, it does so very inefficiently:

We estimate econometric models of the determinants of congestion costs to motorists, trucking operations, and firms and find that, on average, one dollar of highway spending in a given year reduces the congestion costs to road users only eleven cents in that year. 

They also found that if highway spending were targeted most efficiently -- i.e., to those states whose urbanized areas experience the greatest travel delays and to those stretches of road where spending is most effective at reducing congestion -- annual congestion costs would fall as much as 40 percent.

But even so, the congestion cost savings from one dollar of highway spending in a given year would amount to a modest twenty-five cents in that year, indicating that such spending, even if allocated more efficiently, is simply not a cost-effective way to reduce congestion. We conclude that the evidence strengthens the economic case for congestion pricing of roadways and provides some insight into policymakers’ preference for public spending over road pricing to mitigate congestion costs.

Interestingly, they also found that cities do not benefit equally from investments in new capacity.  Cities that begin with a heavy investment in capital stock -- i.e., a highly-developed road network with lots of alternate routes -- might see a reduction in congestion without inducing offsetting demand.  (Networks of any kind provide increasing returns as they grow denser.)  But . . .

we find that an increase in urban freeway capital spending in states (e.g., Maryland) that have merely above average investments in their capital stock increase a city’s congestion costs. In these states, the road network is sufficiently developed to serve a high volume of traffic. But when improvements to certain stretches of an urban freeway induce demand, the additional traffic may create congestion on other roads that may already be near congested conditions. Because these states have not invested in their road system to the highest extent, few alternative routings are likely to exist to diffuse the induced demand over an extensive road network. Thus, even if delays are reduced on part of an urban freeway that is improved, delays and congestion costs are likely to grow on unimproved roadways, resulting in a net increase in a city’s congestion costs.

That point deserves super-duper emphasis:  Spending money on highways to reduce congestion, paradoxically, increases congestion without a dense network of roads to diffuse the induced demand.  Adding capacity makes things worse.  Any solution therefore must come from the "something else" category, whether it be congestion pricing, transit, or more walking and bicycling.  

It ought to be clear that Austin is one of the "low investment" cities.  Austin has a primitive highway network for a city its size.  It's got I35, MoPac, the two northern cross highways, 290 and 183, and 290/71 along the southern mid-section.  The city has few practical alternate routes for getting from Point A to Point B.  Hence, an attempt to increase capacity on one of these roads will inevitably increase congestion on another road.

For example, TxDOT is pushing ahead with plans to increase the capacity of 290 from Manor west to 183 by making it a limited access toll road.  But the increased traffic flow on 290 will push thousands of more cars onto congested 183, where, once there, they really have no option.  What's worse, the "improvement" will feed thousands more cars into already hypercongested I35.  It's quite likely that the extra congestion on the much busier I35 and 183 will swamp any time savings on 290.

Given the primitive state of our highway system, we Austinites really don't have much choice.  Without a system like congestion pricing to manage demand, we're just pissing away our highway dollars.

Panoramic sprawl

Californian Matt Jalbert has taken a stunning set of panoramic photos of California sprawl.  My favorite is of this residential development in San Ramon.  I cropped a tiny patch for the photo to the right, but it does not begin to do justice to Jalbert's photo.

People use "sprawl" to mean lots of different things.  I suggest readers use the panoramic photo to self-administer a Rorschach test, a test to clarify your own definition:  What makes this sprawl?  (Or is it not?)

If you want to do it right, you have to assume you know nothing other than what's depicted within the four corners of the photo.  That is, you don't know how far the development is from the nearest major employment center, or whether it's two miles or ten miles outside a major city.  (Most Northern Californians know where San Ramon is, I imagine.)

So what makes it sprawl?  Note it's not vulnerable to the "sprawl is sparse" criticism -- i.e., that suburban large-lot homes chew up excessive amounts of open space.  This development is as dense, structurally, as an old Italian village.  The houses take up almost all of the space of their lots, leaving tiny yards at best.  

My take below the jump.

Baffled

By this:

Suppose the federal government gave you and your neighbor $500 each to buy a new bike, but what you really wanted was a $250 shopping spree for running gear instead. So you offered to sell your $500 federal check to your neighbor for $250 in cash so everyone’s dreams could be realized.

That is essentially what several cities in Los Angeles County planned to do with federal stimulus money, until the local transportation authority, its face slightly reddened, pulled the plug on the plans. A spokeswoman in Washington for the House Committee on Transportation and Infrastructure said Wednesday that the swaps would be illegal.

“We have already put governors, transit agencies and large metropolitan planning associations on notice that we intend to aggressively oversee that the funding is utilized as Congress intended,” the spokeswoman, Mary A. Kerr, said by e-mail.

Under the federal stimulus package intended to improve the nation’s infrastructure, the Los Angeles County Metropolitan Transit Authority was set to dole out roughly $215 million in sums of at least $500,000 each to the county’s 88 cities to get their projects moving.

Many cities on the list, however, did not have qualifying projects because they are too small or cannot move as quickly as the stimulus law stipulated. So the transit agency encouraged the cities to do with the stimulus money what they often do with other money — swap it with other cities at a discounted rate.

These swaps sounded innocuous enough:

Irwindale, which has roughly 1,500 residents, agreed to sell its $500,000 allocation to the city of Westlake Village — 9,000 residents — for $310,000 in cash, which would go into Irwindale’s general fund.

“We have a general fund deficit this year,” said Robert Griego, the city manager of Irwindale, which got offers from a half-dozen cities for its money. “So we probably would have used it to avoid people getting laid off.”
Torrance, a city in southern Los Angeles County, moved to snap up transportation dollars to spruce up major arteries, and planned to buy a $500,000 share from Bradbury, another small city, for about $315,000.

“We thought it was kosher,” said Eric Tsao, the finance director of Torrance. “We could have leveraged more for our projects, and the other cities would have had some money to do what they would like. You can’t do much with $500,000 in infrastructure anyway.”

But when the exchanges were brought to light on Monday by The Pasadena Star-News, the transit agency quickly intervened. It sent letters to cities clarifying that the only swapping allowed would be federal money for equal amounts of state transportation money.

I understand why the federal government wants its transportation dollars spent on transportation projects.  But unless the Times left out part of the story, the cities purchasing the "transportation dollars" intended to use the dollars for transportation projects.  If they hadn't, they presumably wouldn't have been able to buy the transportation dollars at such steep discounts.

The feds' intervention was not necessary to protect the stimulus.  Stimulus dollars do not have to be spent as efficiently as private investments, but that does not mean they have to be spent as inefficiently as possible.  For example, paying 100 guys $10 an hour to dig holes in a desert might be stimulative but it wouldn't provide as much stimulus as paying 50 bus drivers $20 per hours to keep the buses running.  The  bus drivers provide a social benefit, which leverages the stimulus dollars; the hole diggers do not.

The cities' swaps obviously made everyone better off.  The cities that sold the dollars got cash to offset budget deficits and cuts in city services and employment.  The cities that bought the dollars got more transportation than the sellers would have gotten for much more money.  The feds, almost by definition, got more stimulus and more transportation.  Thus, I'm baffled.

Simpson's paradox

Counterintuitive fact o' the day (and one I'm a little embarrassed I just learned).

Let's say you run an experiment on a sample population.  You do this by dividing the population into two subgroups and running the experiment on each separately.  The experiment succeeds for each subgroup.

If the experiment succeeded for each sub-sample, then it necessarily succeeded for the combined sample, right?

Nope.  Sometimes an experiment's apparent success on two samples disappears  when the samples are combined.  This is Simpson's paradox.

Example:  A pharmaceutical company wants to determine whether a new drug performs better than an old drug.  The manufacturer tests New Drug on 300 patients in Chicago and 300 patients in Cleveland.  In Chicago, 240 of the patients get New Drug and the other 60 get Old Drug.  Ninety of the former (37.5%) and 20 of the latter (33.3%) recover.  Thus, New Drug performs better than Old Drug in the Chicago trial.

In Cleveland, 60 of the patients get New Drug and 240 get Old Drug.  Thirty of the former (50%) and 110 of the latter (45.8%) recover.  Hence New Drug performs better than Old Drug in the Cleveland trial, too.

But New Drug's apparent superiority disappears once the samples are combined -- 120 of the 300 that took New Drug recovered (40%) while 130 of the 300 that took Old Drug recovered (43.3%).  Thus Old Drug did better than New Drug in the aggregate sample even though New Drug did better than Old Drug in each of the two sub-samples.*

Weird.

(*Example lifted from Simon & Blume, Mathematics for Economists)