click here for the WORD file tpbrept.doc

3.0 PROTOTYPICAL PROJECT EXAMPLES          1

4.0 INDUCED TRAVEL AND TRAVEL MODELS     4

The FOUR-STEP TRAVEL MODELING PROCESS                     4

5.0 RECENT FINDINGS FROM THE LITERATURE                      6

6.0 CONCLUSIONS ……………………………………………. 11

But, the story behind the headlines is not as simple. While interest focuses on the impacts of highway-related projects other modes can also cause induced travel. A complex blend of factors, meanwhile, influences travel growth and it is difficult to separate the effects of induced travel from other changes in travel behavior. The extent of induced travel associated with individual facilities depends on project-level characteristics. Sophisticated techniques to quantify the extent of induced travel are in their infancy and research has yet to provide definitive estimates of its impacts.

This briefing explains what induced travel is, what it is not, how it occurs across modes, how it is accounted for in the travel modeling process, and what recent research suggests about the potential magnitude of induced travel both at the project level and relative to other causes of travel growth.

Whenever a new transportation facility is constructed, changes in travel behavior occur in response to travel time savings created by that facility. Some of these changes in behavior induce new travel, but other changes merely redistribute existing travel.

Induced Travel. A new transportation facility that generates travel time savings is perceived by travelers to reduce the price of travel. In response, they may demand more travel in one of two ways:

· New trips. Completely new trips that did not occur previously.

· Longer trips. Additional mileage for trips that already occur.

The magnitude of these changes in travel behavior may be greater in the long-term as consumers respond to shorter travel times by changing from more central to less central

residential or job locations that increase trip lengths, and may be associated with more auto-intensive development patterns.

Redistributed Travel. A new transportation facility may also generate other types of travel behavior changes that shift existing travel to the new facility and away from other facilities. While these changes may generate additional travel on the new facility they do not create an area-wide increase in travel:

Mode Shifts. A change in mode to take advantage of improved travel times

Diverted Travel. A switch in routes to take advantage of a quicker route.

Time of Day Shifts. A switch in time of travel to take advantage of reduced congestion.

Most commonly, induced travel is considered to include new and longer trips but excludes diverted travel and time of day shifts. Mode shifts do not induce new travel, but they are sometimes considered to be part of induced travel because they generate additional vehicle miles of travel (VMT). The following consensus definition of induced travel reflects recent work by Cohen (1995), Heanue (1998), DeCorla-Souza & Cohen (1999), and Noland & Cowart (2000) among others.

An increase in total daily travel that results from a change in transportation system capacity within a defined geographic area.

New transportation facilities are frequently perceived by users to generate additional travel. Researchers agree that induced travel can occur when new or expanded transportation facilities are constructed, but not all travel growth that takes place on new facilities is necessarily new when considered on a regional level. In addition, the phenomenon of induced travel is not restricted to highway projects, as any type of transportation project can have an induced travel effect. This set of prototypical transportation project examples provides a qualitative illustration of the differences between induced travel and other factors that cause travel growth, and of how the extent of induced travel varies from project to project.

This project adds new lanes on a 20-mile stretch of multi-lane access controlled divided highway that is presently highly congested. The highway serves an area activity center outside the region's urban core with a mix of medium to low-density land use. Local bus transit service is available in corridor. The large scale of the project generates significant travel time savings for individual users:

Induced Travel. The expanded highway, which generates significant time savings, creates new trips. For example, a parent now opts to attend their child's soccer game, which previously took too long to drive to. Longer trips are also created, for example a worker moves to a new location further from his current job, but with similar travel times to previous location.

Redistributed Travel. The expanded highway causes a significant redistribution of travel. Trip timing changes occur, for example, an auto commuter sleeps an additional fifteen minutes because he can now get to work on time by leaving 15 n3inutes later. Diversion of trips occurs, for example, parcel delivery service trucks take advantage of new quicker route to cut travel time to and from the sorting depot by switching from arterial network. Mode shifts occur, for example, transit users switch to drive alone mode because travel time savings make driving more attractive.

This project expands the capacity of a moderately congested arterial along a few miles of existing roadway that serves a low-density residential area and provides access to retail mall. A mix of traffic engineering and ITS-related improvements are made, including new left turn lanes and real-time traffic signal synchronization across several intersections. The project generates a small travel time saving of a few minutes per user, as a result the potential for travel behavior impacts of any kind is anticipated to be small:

Induced Travel. Dowling & Colman (1998), in a survey of households in the San Francisco Bay area, found that respondents indicated a high degree of resistance to change in their travel behavior when offered travel time savings of between five and fifteen minutes per trip. This finding suggests that the arterial creates few new trips because the travel time saving for individual users does not provide a strong incentive to make additional trips. The improvement creates some longer trips. For example, some residents may choose to make a longer trip to a national-chain hardware store with lower prices that is now more accessible due to left turn lane.

Redistributed Travel. The arterial causes few trip diversions or mode shifts, again, because the travel time saving is not sufficient to induce significant changes in travel behavior. The arterial may cause some trip timing changes. For example, traffic back-up at intersections becomes less bad in peak hours, therefore a resident makes a trip to shop during rush hour.

This project extends Metrorail along a 10 to 15 mile suburban corridor, adding several new stations. The line links to a suburban activity center with low-density residential and commercial development and some existing bus transit. Congested conditions occur on parallel highway facilities:

Induced Travel. The line creates some completely new trips. For example, the new Metro access means a family chooses to attend a Wizards game at the metro accessible venue rather than watch it on television. The line also creates some longer trips. For example, an existing transit commuter moves to a new neighborhood now accessible by Metro, but further from their current job. Due to the auto-dependent design of their new neighborhood, they also make longer trips to stores, entertainment and other activities.

Redistributed Travel. The new line causes some mode shifts. For example, a resident who previously drove to a suburban job now utilizes Metro access to avoid traffic congestion by taking transit.

A new bike path is built that links an urban mixed use corridor, which includes homes,retail, office, and educational establishments. The new bike path parallels a congested local and arterial street network and it offers quicker and safer non-auto travel access to trip attractions. The bike path is time competitive with Metro for short trips and it supports a mix of recreational & utilitarian trips:

Induced Travel. The path creates some completely new trips. For example, the path's convenience prompts a resident to buy a bike and make new trips. The path also makes some existing trips longer. For example, a resident currently bikes to a neighborhood grocery store; the new path, however, allows the resident to extend his bike trip to a nearby shopping district with a better range of groceries and other items.

Redistributed Travel. The path diverts some bike travel from the road network. For example, an office worker commutes downtown by bicycle on a busy street network. The new path offers a quicker commute, so she switches routes.

Induced and redistributed travel are not the only factors that are responsible for growth in travel on each of the facilities described in the prototypical examples. A variety of regional trends also drive increases in travel. Heanue (1998) provides a review of the three main groups of factors that have driven the increase in total person travel and VMT in the United States and elsewhere over the post-war period:

Demographic factors, such as growth in population and households, auto ownership, and licensed drivers.

Economic factors, such as labor force participation, employment, transportation costs, and income.

Lifestyle choice factors, such as smaller household size and campus-style office parks.

Together, these exogenous factors may have a significant impact on travel growth in themselves.

Travel models are an invaluable tool for helping to estimate the impacts of transportation projects. The debate about induced travel, however, has generated discussion about how accurately travel models can determine the kinds of induced travel impacts described in the prototypical examples presented in this briefing. Assessing the effectiveness of travel models in this regard is difficult, since they do not disaggregate induced travel effects from overall travel growth estimates, which are the standard output of travel models. Review of the four-step of travel modeling process commonly used by MPOS, however, suggests that models capture most components of induced travel.

All urban transportation modeling systems used by MPOs are based on a "four-step" sequential process for organizing a series of computer models to simulate urban travel patterns. Figure One shows the typical process. Data inputs that describe land use and population patterns and the characteristics of the transportation system are determined exogenously. This information is fed into a sequence of computer programs. The results from one step feed into the computer models in the next step as shown by the solid arrows in Figure One and may feed back into previous steps (shown by the dashed arrows). The model produces a description of the flow of traffic on the system.

Step One - New Trips. This step of the model determines the number of daily trips that take place into and out of each modeling zone in the region. Several trip purposes are modeled including work trips, shopping, and other trips made to or from homes; trips based in a location other than a home (such as lunch or shopping trips made from a worksite); and commercial truck trips. To do this, the model generates estimates of the effects of household and business characteristics on trip generation and attraction. For example, using data from travel surveys, cross classification and linear-regression models estimate the average effect on household trips of factors like income, size of household, and auto ownership.

Person trip generation rates in the models are generally insensitive to changes in travel times. Hence, the impact of travel time savings on total trips is not reflected in forecasts. MWCOG is, however, currently developing a new model component to address linkages between changes in accessibility and trip generation.

Step Two - Trip Distribution. This step of the model takes trips generated in step one and allocates them into origin-destination pairs between zones based on relative accessibility. For example, the number of work trips produced by a district in Gaithersburg (Maryland) are matched with work trip attractions throughout the region to estimate the numbers commuting within Gaithersburg, to nearby suburbs, to downtown Washington, and elsewhere.

Trip distribution models generally account for travel time savings by sending person trips to further destinations due to accessibility improvements.

Step Three - Mode Choice. This step of the model determines the mode travelers are assumed to choose from the following transportation modes for each trip: (1) mass transit, (2) drive alone, and (3) carpooling. The model assumes that their choices are based on the relative availability and attractiveness of each mode. Mode-choice models estimate shifts in person trips among transit, HOVS, and low occupancy vehicles.

Step Four - Route Assignment. This step of the model determines the specific routes between zones that travelers choose to reach their destinations. To perform this step, the computer model selects the best path through the highway network for each type of trip, determining the shortest way both in ten-ns of time and distance to get from zone to zone. Traffic assignment models generally estimate the diversion of traffic from unimproved to improved facilities within a corridor.

Land Use Changes. Transportation infrastructure improvements may over the long run create induced demand as a result of increased development. Some travel models use forecasting processes to account for changes in total regional development that may occur as a result of improvements in accessibility. In the Washington metropolitan region, a cooperative land use forecasting process is used to provide detailed estimates of development that take into account transportation accessibility as well as other factors.

Figure One. Generic Urban Travel Model

Most researchers agree that when capacity is added on a congested facility or to a congested network, new travel is attracted to it. Recent empirical research has focused on determining "elasticity" values that quantify the relationship between the amount of new travel that occurs as a result of infrastructure improvements. Elasticity values are a useful tool for analyzing induced travel because they characterize the variation of travel demand in relative terms that can be used to predict changes in other corridors, regardless of differences in absolute traffic volumes or infrastructure improvement characteristics.

Figure Two demonstrates how an accessibility improvement shifts the travel supply curve from S to S₁, reducing the price of travel and generating a move along the travel demand curve that results in an increase in travel equal to Q₁-Q. The induced travel impact of traffic might be measured as a ratio of the change in quantity of travel demanded (∆Q) over the change in travel supplied (∆S). By using ∆Q(log) and ∆S(log), this value may be expressed as an elasticity. A low elasticity of close to 0 (characterized by a steep demand curve) indicates that consumers do not demand additional travel when travel times decrease. A high elasticity of close to I (characterized by a shallow demand curve) indicates that consumers are very responsive to changes in travel time.

The literature on travel demand elasticities focuses on highway-related travel, with little or no examination of induced travel on other modes. Most commonly, elasticities reported in the literature express changes in travel demand relative to changes in lane miles of capacity or travel time, both of which are considered to be closely related to changes in induced

travel. Elasticity fmdings from selected studies are summarized in Tables One and Two. In aggregate, these results suggest that regardless of the variable used to measure induced travel (lane miles of capacity or travel time), demand for travel is responsive to new capacity across a broad range of US and European study areas. The results also indicate that short-run elasticities are lower than long run elasticities. This suggests that travelers initially respond to new capacity mostly by diverting routes or switching modes, but in the long run they choose to modify job and housing locations, creating longer trips and new trips.

Table 2. Selected VMT Elasticities Relative to Lane Miles of Capacity

As Tables One and Two illustrate, results of the research to date vary widely. Some of this variation may be caused by differences in the characteristics of different study regions or time periods examined. For example, in regions where parking or other travel costs are high the impact of new capacity on travel demand may be depressed. The study of travel demand elasticities, however, presents complex methodological challenges and some of the variations observed in results to date may be the result of weaknesses in the research methodologies used.

Selection of Variables. Researchers frequently choose to measure travel demand with respect to additional lane miles of capacity because data on lane miles are readily available on an area-wide basis over time. Cohen (1995), however, observes that capacity addition is a poor independent variable because induced travel is a function of existing congestion levels. Hence, adding new lane miles to uncongested freeways will produce no induced travel, while removing a bottleneck may generate significant new travel without additional lane miles. For this reason, travel time may be a more appropriate variable for measurement. Unfortunately, accurate data on travel-time changes associated with facility improvements are difficult to obtain.

Geographic Extent of Study. Some studies of induced travel are based on data from a narrow geographic region, or a certain class of highways. As a result, trips diverted onto the new facility from other facilities outside the study area may mistakenly be reported as induced travel, thus over-estimating the induced travel effect. Early studies of induced travel frequently feature this flaw. More recent research has focused on region-wide studies that examine all categories of highway travel (for example Noland & Cowart, 1999 or Fulton et al, 2000).

Multicolinearity of Variables. Travel growth is also driven by a range of demographic and economic factors that may vary with increases in lane miles or improvements in travel speed. The impact of these variables can be accounted for using complex statistical techniques, however, uncorrected multicolinearity effects may result in over-estimates of induced travel.

Simultaneity Bias. The direction of cause and effect of induced travel remain uncertain. Added capacity may in fact be a planned response to anticipated travel growth rather than vice versa, resulting in an inaccurate estimate of induced travel effects.

Short-run Versus Long-run Effects. In the short-term, elasticity tends to be low, since homes and businesses cannot immediately relocate to take advantage of new capacity, however, in the long-term, such shifts are more likely to occur and as a result elasticity is higher. If studies do not examine long term trends in travel demand, for example measuring travel on a new facility for the first two years after it is opened, induced travel may be underestimated.

Interpreting Elasticity in a Project Context. Elasticity values are often thought of as a physical constant. For example, an elasticity of -0.5 is interpreted to mean that a 10 percent decrease in travel time will generate a 5 percent increase in traffic. For a given elasticity and given capacity increase, however, the quantity of traffic induced varies directly with congestion levels. Induced travel is also a dynamic phenomenon, thus as induced travel occurs, travel speeds decrease and demand for new travel correspondingly diminishes.

DeCorla-Souza and Cohen (1999) use a sketch planning tool to illustrate the dynamic nature of induced travel on a hypothetical 8 mile long corridor with a freeway facility that is widened from 4 to 6 lanes, i.e. a 50 percent increase in highway capacity. They apply low and high elasticity values of -0.5 and -1.0 under scenarios of low and high congestion to determine the impact of the new facility on travel demand.

As Figure 3 shows, assuming an extreme elasticity of -1.0 and high congestion, the 50 percent capacity increase generates an corridor-wide VMT increase of about I I percent. Under a more moderate elasticity of -0.5, and high congestion, corridor-wide VMT increases by about 8 percent. In contrast as Figure 4 shows, under low congestion conditions, these VMT increases are 9 and 5 percent respectively.

DeCorla-Souza and Cohen's results show that while an corridor-wide VMT increase of between 5 and 1 1 percent occurs as a result of the new facility, there may be more significant VMT increase at the facility-level. Under an elasticity of -0.5 and high congestion, VMT increases on the facility by 40 percent and under an elasticity of - 1.0 and Assessing Induced Travel in the Washington Metropolitan Region 8 high congestion, the percent increase in VMT on the facility is 48 percent. In both cases, however, a majority of this increase is caused by travel that is diverted from arterial facilities where VMT is reduced as a result, and travel time improves for all travelers in the corridor.

Figure Three. Corridor-wide Induced Travel Impact under High Congestion

Induced Travel Relative to Other Causes of Travel Growth. DeCorla-Souza and Cohen's examine the impact of induced travel on base VNff, however, their analysis does not include an assessment of induced travel relative to other causes of travel growth. Heanue (1998) applies a range of induced travel elasticities from the literature to actual data about VMT growth over time for a typical American city to determine the relative impact of capacity expansion on travel growth. Using an empirical value of -1.0 for elasticity of travel demand with regard to travel time (from Goodwin, 1996), he estimates induced travel to be responsible for 22.1 percent of total VMT growth over time. Using Hansen's (1997) elasticity of O..9, he finds induced travel accounting for only a six percent share of total VNff growth. Heanue's analysis suggests that even using the most conservative estimate, approximately 78 percent of VMT growth can be attributed to nonsystem supply factors such as population and economic growth.

Induced Travel is Multimodal Phenomenon -- The literature (& policy debate) focus on highway-related induced travel, but any type of transportation system project can have an induced travel effect.

Travel growth is influenced by many factors -- Induced travel is only one factor that influences travel growth. Other factors include travel behavior changes at the facility level and regional growth pressures. Causes of travel growth are difficult to disaggregate Assessing Induced Travel in the Washington Metropolitan Region 9

Research to date has focused on major highway improvements - Little is known about induced travel impacts of other types of transportation improvements such as ITS, arterial widenings, and transit Assessing Induced Travel in the Washington Metropolitan Region 10

References

Cohen HS (I 995) Review of empirical studies of induced traffic. Appendix B of expanding Metropolitan Highways: Implicationsfor Air Quality and Energy Use. TRB Committee for Study of Impacts of Highway Capacity Improvements on Air Quality and Energy Consumption. TRB Special Report 245. Transportation Research Board.

Domencich TA et al (1968) Estimation of urban passenger travel behavior: An economic demand model. Highway Research Record 238, Highway Research Board

DeCorla-Souza P & Cohen HS (I 999) Estimating induced travel for evaluation of metropolitan highway expansion. Transportation 26

Dowling RG & Colman SB (1998) Effects of increased highway capacity: Results of a household travel behavior survey. Transportation Research Circular. No. 48 1.

Fulton LM et al (2000) A statistical analysis of induced travel effects in the US Mid-Atlantic region. Presented at the 79th Annual Meeting of the Transportation Research Board, Paper no. 00-1289.

Goodwin PB (1996). Empirical evidence on induced traffic: A review and synthesis. Transportation 23 (1).

Hansen M & Huang Y (1997). Road supply and traffic in California urban areas. Transportation Research A 31(3)

Heanue K (1998) Highway capacity expansion and induced travel: evidence and implications. Transportation Research Circular. No. 48 1.

Noland, RB & Cowart W (2000) Analysis of metropolitan highway capacity and the growth in vehicle miles of travel. Presented at the 79h Annual Meeting of the Transportation Research Board, Jan. 2000

Standing Advisory Committee on Trunk Road Assessment (SACTRA) (1994). Trunk Roads and the Generation of Traffic. Conducted for the Department of Transport, London,UK

Induced Travel: I-270 Case Study Figure 1: Comparison of Round 6.2 and Round 2 Cooperative Forecasts - Households

Induced Travel: I-270 Case Study Figure 2: Comparison of Round 6.2 and Round 2 Cooperative Forecasts - Employment

References

Send mail to Info@TrafficSTINKS.org with questions or comments about this web site.
Last modified: October 04, 2001