Why we need projections
Thursday, August 2, 2012 | Charles Marohn America's post-World War II development pattern can not function without projections for future growth. Not only does the finance of our places depend on growth, but the sizing of key infrastructure components requires that we have some estimation of future demand. That estimation is impossible to do reliably, which only demonstrates that America needs a different model for local development.
We're building some momentum with our ongoing fundraiser for bringing the Strong Towns message to Pennsylvania. Our plan is to visit the state and hold 8 to 10 (or more) Curbside Chats. I anticipate some communities committing in the next few days, but we still have a long ways to go to make this happen. Please consider supporting this effort and pass it along to those you know in PA. We'd love to bring this message back to the Keystone State and change the conversation on growth statewide.
I've spent this week trying to explain why simply getting better traffic models won't help us make any better projections (Build a better model, July 30, 2012) and why, despite knowing that our models are flawed, we feel the need to rely on them (Better to be lucky than good, July 31, 2012). This entire conversation was prompted by The Projections Fallacy (July 23, 2012), a call for a different system of development, one that would be robust to modeling error. Before we discuss that system, it is important to understand why models -- as flawed as they are -- are so critical to the current system.
In the Suburban Experiment, the building increment for new development is the pod. For historical reasons that largely no longer exist (seen any smokestacks lately?), we separate different uses into pods via zoning. This obsession with separation has reached manic levels where we not only separate residential from industrial but we actually break down different types of residential (read: income class) and separate all of them from each other as well. I've seen zoning codes with over 50 different classifications with a dozen different residential zones.
For reference, this is described in eloquent terms in the fantastic book Suburban Nation. You can also get a good overview by watching this video series of a lecture given by Andres Duany.
The device that makes all of this pod-mania possible is the automobile. We separate everything into pods and then drive from one to the other. We start the day in a residential pod. We then drive to a business pod for work. At lunch we all stream out to a commercial pod and then return to our business pod. After work we may stop at a different commercial pod on our way home to our residential pod. On the weekends, we may even drive to a recreational pod. Sixty years into the Suburban Experiment, this approach is now the default setting. For us, it is normal.
To understand the role that projections play, it is critical to understand the hierarchical road network. Suburban development is organized in a hierarchy that is easily understood through street designations. Communities will have local streets, collector streets, arterials and major arterials. On the typical commute one would start on a local street (big), drive to a collector (bigger), the collector would empty into an arterial street (bigger yet) and the arterials would empty into the major arterials (biggest). Traffic is funneled into these major arterials to facilitate movement between pods. When the destination pod is reached, the cycle reverses until the destination is reached.
The arch-enemy of the suburban, auto-oriented development pattern is congestion, which brings the system to a standstill and destroys whatever "efficiency" this approach claims to provide. This means that, since traffic is constantly routed to the neck of the funnel (major arterial), that neck needs to be sized properly to handle all of the future growth on the periphery.
If you undersize the neck of the funnel -- the arterials -- congestion will ultimately overwhelm the system, backing up cars until, essentially, the system fails.
And here is the key financial insight necessary to understand our need for projections: It is vastly more expensive to fix congested arterials than it is to simply oversize them to begin with.
This is why a "conservative" approach is not one that reduces the size and budget of the project, opting for a more incremental solution. In the Suburban Experiment, "conservative" means massively over building everything under the guise of penny-wise, pound-foolish. To under design an arterial is a mortal sin for an engineer. It is also a budget buster.
Now imagine you are a local official charged with making decisions on transportation projects. It's 2012 so the only system of development you have is our Suburban Experiment, complete with a hierarchical road network. You may not like it, but what is the alternative when you are presented with a project that seems enormously expensive and appears to you to be completely out of scale? Are you going to demand a smaller project and risk levels of congestion that will render all your investments near worthless or are you going to fall back on that traffic projection, served up with all the credibility of a profession thousands of years old.
The hierarchical road network of the Suburban Experiment raises the stakes of every decision to critical levels. It leaves little room for error, at least for error of the fiscally conservative nature. The downside of over-estimating is modest -- a bigger bill (cheap insurance, as they say in the biz) and some excess capacity -- while the damage that can come from under-estimating is calamitous.
America's post-World War II development pattern can not function without projections for future growth. Not only does the finance of our places depend on growth, but the sizing of key infrastructure components requires that we have some estimation of future demand. As I've shown this week and last, that estimation is impossible to do reliably. This leads to only one rational conclusion:
America needs to abandon the failed Suburban Experiment and adopt a different model of local development.
There are a number of brilliant people that are working on what that model is, how it looks, how it will function, what its financing is and how to transition. In my next post on this topic, I'll do my best to give you my thoughts on what a different model would look like and, if I can, how we actually get there. And for those of you upset with me over the transit implications of this discussion, while you've never heard me say that Strong Towns realities should not apply to transit, I'll try and explain how we can still do large, significant projects with a Strong Towns approach. (Hint: Large projects are the culmination of many smaller, more incremental, successful projects.)
If you want to chat with Chuck and many others about implementing a Strong Towns approach in your community, consider joining the Strong Towns Network. The Strong Towns Network is a social platform for those working to make their community a strong town.
Engineering Profession, Public Policy and Planning 
Reader Comments (10)
No, actually it's very cheap to eliminate congestion by converting existing lanes to express toll lanes.
Man, you are just knocking them out of the park. The only problem is, I don't know what to do with my blown mind.
How does this analogy relate to the interstate highway system? Does this analogy scale up well? I would think it's cheaper to make just a few interstate highways than have one from every point to every other point on the map. Or does this analogy really only work in day-to-day, urban city dweller lifestyles?
N. Talib says that poor models are actually worse than having no models at all. The downsides of over estimating future traffic volumes are higher speeds, more crashes, more fatalities, higher societal costs, and more environmental degradation.
The downside of underestimating demand and minimizing road capacity? Maybe a few additional minutes spent going to grandma's house in the suburbs.
Modeling projections inherantly works better with a hierarchtical network in which all traffic funnels through one arterial. All you have to do is optimize the intersections along the way that lead to shopping mall A or freeway B. It's much harder for a model to predict what will happen in a network where there are numerous choices to get from A to B and lots of intersections to analyze.
I agree - Poor modeling is worse than no model. That's why you can't take human judgement, engineering experience, and local input out of the equation.
All roadway systems have a heirarchy, whether it's formalized or not. Even in downtown grid systems with lots of alternatives, certain streets are the workhorses as far as vehicular traffic. True, it's harder to model, which is why most regional models don't include all of the side streets. They are regional models. 10,000 foot level. Totally inappropriate for project specific, design level application. They just provide general guidance in terms of expected relative background traffic changes or for relative comparisons of two competing alternatives.
@Jeff Morrow
Sorry to have taken so long to get back to you - been out of town.
For small scale, human judgement is just fine. Large scale, I put as much faith in it than in modeling, maybe less.
I disagree that all roadways (assuming by your comments that you lump roads and streets together) have a hierarchy. I was just in Chicago for three days and much of the local street grid is not hierarchical. Multiple routes exist for any destination. As congestion increases, single occupancy cars are naturally priced out of the market and you can justify buses, taxis and other communal forms of (private sector) transportation. Even more congestion justifies subways and other rail investments. All of this happens on the same grid w/o the need for a hierarchy.
There is an important difference between roads and streets that is completely lost in our current transportation conversation. I would recommend you make a clear distinction between the two in your work.
What design decisions are made on large scale modeling? Large regional models are only appropriate for gross, 10,000 foot level planning or comparing two competing options of regional significance. I don't see that regional models have much if anything to do with New Urbanist concepts which are all focused on micro scale, place building efforts.
I was also recently in Chicago for a 3 day conference about a month ago. I found some streets to be more busy than others while I was walking from my hotel to the conference center. The following is a link to a downtown Chicago traffic count map.
https://data.cityofchicago.org/Transportation/Average-Daily-Traffic-Counts-Map/pf56-35rv
In downtown, core Chicago, near the "Magnificent Mile", It shows Ohio Street carries about 33,000 vehicles per day (vpd). Of course Ohio Street comes straight off the Kennedy Expwy into downtown so you would expect it to carry a lot of traffic.
South of the river, Congress Pkwy carries about 62,000 vpd onto the Dwight Eisenhower Expwy out of the downtown. Immediately north of Congress is Van Buren (which crosses the river) but only carries a little over 10,000 vpd. About 2 blocks north of that is Quincy, which is only about 3 blocks long. There is no count shown for Quincy most likely because it is so minor it's not worth bothering with, but I bet it's doesn't carry 10,000 vpd. Can't prove it, but common sense would say it doesn't.
So, within this very dense, downtown grid we have a major arterial (Ohio and Congress) a minor arterial or possibly a collector (Van Buren) and a Local Street (Quincy). Looks like a hierarchy to me.
You are probably right that I don't understand the difference between a street and a road. But that, in and of itself is a hierarchy. I guess you are saying there is no hierarchy between arterials, or no hierarchy between collectors, etc. Whats the point of that?
On hierarchy: it is important to distinguish between hierarchy in a dendritic (tree-shaped) road network and a grid of streets. The former has arterials that can range from 100-300 feet in width with 6 lanes and double left turn lanes. These are much wider (2-5x) than a typical residential street. The later has arterials that are much smaller (under 100 ft) and can be split into pairs. The hierarchy of an urban street network is both suble and complex, having to manage many demands of not only cars but also pedestrians, bikes, transit, and events (parades, fairs, farmers markets).
The decisions that are made at the macro level substantially influence the design outcomes at the micro level. If you create a road project of regional significance that is big enough to show up on a regional travel model, the easiest way to do that is to build a really big road to handle the current and projected capacity. Walkable places are not possible when a large road designed at the macro scale creates a pedestrian barrier. The market appropriately responds by building auto-oriented development.
What regional travel demand models do not take into account is change in priorities over time that add complexity. What happens if there is attempt to change the built environment when a transit project is proposed run by that big arterial? The model will say that the auto traffic needs to be accomodated and that slowing speeds and adding intersections will cause problems with the regional traffic needs. The engineer is not going to say, 'ignore the model since the traffic will go elsewhere and just build the place to want to see in the future.'
If a place is to evolve substantially over time, it's not going to get there by forecasting that the future will be a linear projection of the past.
@ Faith
On hierarchy - all true - I think we are quibbling over semantics.
On Macro/Micro decisions - disagree. Macro level only points to possible trouble spots. Micro analysis needed to determine if there actually is a problem and the amount of resources necessary to address it. You can't trust macro models at face value - handy to weed out some chaff, but not good enough to make dollar commitments.
Regional traffic demand models are a slave to priorities, not the other way around. True the model will say exactly what you suggest, and the engineer will say to the DECISION MAKERS if you choose A it will likely operate this way, if you choose B you can expect this other result, if you choose C etc. The engineer just recommends based on what the goals are for the project as set by the Decision Makers. If you want to create a "place" - great. Traffic will have to be handled elsewhere or in some other fashion. If your goal is to have smooth auto traffic operation - great. We can do that too. Just tell the engineer what you want and what level of congestion you're willing to live with and it will all happen.