Musings on Scenario Development for Travel Forecasting

Truth be told, I volunteered to be a Foresight Ambassador for NCHRP. My and the several other ambassadors’ mission was to spread the word about NCHRP Foresight Reports 750. This series of six reports, as you have probably heard, develops future scenarios for transportation development and planning. All rookie ambassadors were supplied with a one-day briefing, a really snazzy canned keynote-type presentation and PDFs of the six reports. This canned presentation was oriented to state DOT executives and others of that ilk, so it was quite long on concept and quite short on details.

When the Innovations in Travel Modeling Conference rolled around, Lori Sundstrom of NCHRP suggested that I stop by and give a lightning talk about those reports. The expertise of my audience meant the canned presentation wouldn’t do. Instead, I needed to quickly tell modelers the particulars of how those reports could benefit them.  To prepare for this presentation, I did something I am loathe to do — I read those reports. Well, not every page, but quite a lot of it. And I found out that making sense of those reports, holistically, was not as simple as I thought it would be before diving in.

So there were six separate reports looking at six different theme areas written by six distinct sets of consultants.*  As far as I could tell, there was little attempt to coordinate content or style across those reports. So any given modeler wishing to benefit from those reports must both distill and synthesize the massive amounts of information contained therein.

Let me step back momentarily and explain why modelers need scenarios, preferably canned scenarios. Scenario planning, by now, should be SOP for LRTPs. Alternatives need to be correctly evaluated within varying contexts, not just those that are particularly favorable. The most robust plans should evaluate well across a broad set of reasonable futures. Ideally, every alternative should be tested against every scenario.  More specifically for modelers, those same scenarios become inputs of the modeling process.  But a scenario is not particularly helpful if it cannot be modeled. Thus, modelers need to advise the scenario creation process to assure that the technical limitations of the models are not violated.

A model also needs to be validated; and one recommended way of validating is to stress the model through a sensitivity analysis. There are too many inputs to find the sensitivity to each one, so scenarios are needed to package variations to base-case inputs in a consistent fashion. If the model produces a reasonable answer for difficult scenario, then there is a good chance it won’t produce a difficult answer for a reasonable scenario.

The creation of scenarios can be hazardous. A model can be otherwise perfect, but a wrong-headed scenario can ruin a lot of hard work.  In my experience the worst mistakes made during plan evaluation have been caused by poor assumptions about future conditions. Given the clear possibility of a planning disaster, planners (and modelers) could use a lot of help in how scenarios are formulated. If someone else, more capable, is tasked with the scenario creation chore, then the modeler is offered some degree of professional protection from unforeseen events.

Thus, while far from a god-send, Reports 750 are about the best resource humanly possible for creating planning scenarios. But none of the six consultant teams saw the big picture. So the scenario creation job, as far as we are concerned, remains unfinished. If an LRTP or a model sensitivity analysis needs at most eight to twelve scenarios, then the information found in the various Reports 750 will need substantial post-processing.

The horizon year for Reports 750 is 2050, which is in rough alignment with ongoing LRTP efforts. Keep in mind that we do not actually expect to realize one of the future scenarios when writing an LRTP. The scenarios help establish a direction for plans and policies that will be redirected and revised, incrementally, at 5 year intervals or so.

The six theme areas covered by the reports are: freight, climate change, technology, sustainability, energy and socio-demographics. Modelers can pretty much skip over climate change and technology, since these reports deal with issues that, for the most part, do not directly affect model inputs. There may be indirect effects, such as regulations to limit green-house gas emissions, but some of these are covered under sustainability.

Scenarios are entirely contained within their theme; for example, a climate change scenario only deals with climate, but not with the interaction between global warming and population growth in coastal regions.

Most scenarios, with the notable exception of those in the energy report, are highly compressed. Such scenarios are best explained as “latent dimensions” or packages derived from a host of elementary trends. Those elementary trends are generally explained well, so they are there for the taking should custom scenarios be desired.

Scenarios within a theme are distinct and contradictory. They don’t represent points on a continuum of possibilities. Rather, they sprawl across a multidimensional space defined by some of the elementary trends. A reader must dig for those elementary trends (but they are there).

The socio-demographics report also provides a systems-dynamics (SD) model for testing scenarios, but this SD model strikes me as being too simple to be useful for transportation plan evaluation in metropolitan areas.

The freight report is particularly interesting in the rigor used to assemble its scenarios. The report demonstrates how workshops of stakeholders can result in a small set of scenarios from a much larger set of elementary trends. It also demonstrates how difficult scenario building can be when there is a desire to be inclusive of all opinions and facts.

The consultants got to borderline cuteness when naming the scenarios, but they clearly had an objective of keeping the scenario names as memorable as possible. I am listing the scenarios, without explanation, for the three of four relevant reports that had them.

The five scenarios in the socio-demographics report have the proper names of Momentum, Technology Triumphs, Global Chaos and Gentle Footprint. There are even distinctive logos of each of these.

Sustainability also had five named scenarios, Crisis World, Mega World, Suburban World, Wonder World and Green World. These bear little relationship to any of the scenarios in the socio-demographic report.

Lastly, the four scenarios of the freight report are named Global Marketplace, One World Order, Millions of Markets and Naftástique!. And these bear little relationship to any of the scenarios in the socio-demographics and sustainability reports.

You can pretty much guess what each scenario contains by interpreting its name. Each of these fourteen scenarios are composed of several elementary trends within its theme area.

The consultants for the energy report did not actually create composite, named scenarios. Instead they presented the elementary trends, likely in the hope that the reader can create his or her own small set of scenarios. While less entertaining to read, the energy report may actually be structured more usefully for travel modeling purposes. To get an idea of the elementary trends, some of them in the energy report are price of oil, fuel economy, fuel mix, vehicle cost and amount of travel.

A quick inspection of the fourteen named scenarios of those three themes shows that they do not necessarily mix and match well. It would be difficult to take any one scenario from socio-demographics, take any second scenario from sustainability, take any third scenario from freight, and have them add up to a satisfying whole. Somebody needs to determine which of the 100 combinations (not even including energy) makes any sense at all.

So here are my takeaways from NCHRP Foresight Reports 750.

You can save a heck of a lot of time by zeroing in on the descriptions of the elementary trends and the composite scenarios. The sections on background and process can be left for those people who are particularly interested in the topic of scenario building itself.

Rigorous scenario building, as illustrated by the freight report, can be difficult but rewarding. Something similar, but maybe briefer, might serve well for anybody in a planning agency who wants to build their own scenarios from scratch. I don’t recommend that you start entirely from scratch, so steps can be accelerated or skipped.

Scenarios (and their elementary trends) are created with reference to currently perceived conditions and pivot around what is already known. Thus, the results are well grounded. There are no flights of fancy (or flights of cars).

It is still necessary to build scenarios that span the four pertinent theme areas (freight, sustainability, energy and socio-demographics), as all these themes influence travel forecasts. Here is where planners and modelers must exercise their own logic and creativity.

Leaning on the Foresight Reports 750 reports can make the job of scenario creation easier and far less professionally hazardous.

Alan Horowitz, July 4, 2016, Whitefish Bay

*A variation on an old riddle: If six consultants can write six reports in sixth months, how long does it take six thousand consultants to write six thousand reports? (Original units were soldiers, cigarettes and seconds.)