Technical Synthesis Topic
Fuel Price Synthesis: Travel Model Uncertainties

(PDF version)

September, 2007

Travel models are developed to support informed decision-making regarding future transportation alternatives. Whether that is feasible in an environment of uncertainty when observed data and trends are not available was the focus of various discussions regarding the potential impact of prolonged higher fuel prices.

Six key questions posted to the email list during different periods capture the broad level of uncertainties associated with capturing the influence of higher fuel prices on overall travel:

• Lacking observed data to support behavioral responses it is unknown whether significant shifts in travel patterns, trip making characteristics and land use patterns may occur as a result of higher fuel prices so how does one forecast those effects?
• If the price of fuel becomes overly burdensome to all drivers, can we expect corresponding advances in vehicle technology to improve overall fuel economy (as seen in the 1970's) and how do you forecast vehicle fleet mix if this comes to fruition? Along with the uncertainty associated with the evolution of fleet technology, the switch to alternative fuels should also be considered.
• Could vehicle miles of travel (VMT) actually increase as a result of individuals reducing their driving costs by migrating to hybrid or fuel efficient vehicles?
• Is the discussion regarding fuel prices and out-of-pocket expenses even relevant in mode choice discussions except for the lowest income stratifications?
• What kind of economy will the United States have if the price of energy doubles or triples and the cost of transportation represents a greater share of economic input and what would be the corresponding effect on overall travel?
• Is it even a proper role for travel models to forecast gas prices when conformity determination and other planning applications require consistent planning assumptions to measure the scope of the changes?

Travel Model Uncertainties

As noted in previous syntheses on the topic, quantifying the potential impacts to travel behavior as a result of fuel price changes is not as straightforward as it initially appears since responses to these changes have evolved over time. A number of contributors were cautious about drawing direct relationships between gas price shifts and driver behavioral responses. Several contributors observed that while there is ample evidence that people do alter their travel behavior during previous (i.e. early 1970's) price spikes; nevertheless, more recent data is less conclusive. Moreover, consistent responses from which to draw conclusions to use as defensible travel model inputs do not yet exist. Consequently, the lack of consistent behavioral responses limits the certainty of how to address these underlying issues in current travel demand model practice.

For example, recent observations indicate that higher gas prices have had minimal impact on driver behavior. Economists have noted several probable explanations for these recent findings:

• Increase of dual-income households in America has created more disposable income and a greater ability to absorb increases to specific portions of the household budget (e.g. auto operating expenses);
• Increasingly longer commute trips make households more dependent on vehicles to get to and from work. This is especially true in regions where the automobile played a substantial role in determining land use patterns. Exacerbating the problem is the lack of viable alternative modes of transportation to support the longer commute trips;
• America's love-affair with vehicles has made people more willing to absorb out-of-pocket costs associated with higher gasoline prices by modifying expenses elsewhere in the household budget (e.g. eat out less, go to movie theaters less often or avoid unnecessary purchases). Another contributing factor is people simply switching to more fuel-efficient vehicles rather than choosing an alternative mode of travel.

This level of uncertainty led one respondent to hypothesize that a potential bias may exist to overstate probable shifts to alternative modes of travel if fuel price changes are incorporated in the model structure. Furthermore, the respondent cautioned that, "the challenge to the modeling community is to not overstate the changes in vehicle miles of travel (VMT) and mode share as a strict relationship between gas prices and travel behavior." In response it was recommended that larger macro-economic conditions need to be considered when interpreting past data regarding changes to VMT and mode share during periods of fuel price adjustments.

Further highlighting the difficulty in forecasting probable outcomes is the interdependence of the strength of the overall economy and transportation. As one contributor noted, transportation (the value added by goods and people delivered, and the mechanisms thereof) constitutes a quarter of our economy. Changes to that relationship could also impact overall travel.

Conclusions

Though ostensibly this synthesis has addressed the intricacy of properly forecasting behavioral responses to fuel price changes, the underlying issue has been one of adequately accounting for uncertainty in the modeling process. Imbedded also within the discussion is an understanding that addressing model uncertainty can be accomplished by acknowledging existing data limitations. Moreover, it appears that there is consensus in the modeling community to develop a process that provides modelers with the ability to communicate potential impacts while recognizing the difficulties of accurately quantifying the impacts.

Despite the level of uncertainty associated with a lack of observed data, it is also evident from the initial questions that considerable thought has nevertheless been given to potential outcomes. Indeed, the six questions themselves offer a framework for addressing the inherent uncertainty. For example, suggestions were offered for developing a forecast scenario given a set of known values (e.g. auto operating costs, tolls, transit fares, in constant dollars) and then determining a range of potential outcomes given different forecast variables (e.g. fuel price increases). Forecast scenarios could likewise be developed for other questions posed throughout the discussion. Thus acknowledging limitations (e.g. unknown behavioral response) while incorporating them into a range of potential outcomes (e.g. shifts in overall travel, trip lengths, and mode splits) appears to be a viable approach until there are greater levels of historical data to draw more specific conclusions.