ANATOMY OF INDUCED TRAVEL

USING AN INTEGRATED LAND USE AND TRANSPORTATION MODEL

IN THE SACRAMENTO REGION

by

Caroline J. Rodier, John E. Abraham, Robert A. Johnston, and  John Douglas Hunt

 

November 28, 2000

 

Abstract- Recent research on induced travel has placed renewed attention on the ability of currently available analytical tools to capture the induced travel effects of proposed new highway projects.  In this study, an integrated land use and transportation model, MEPLAN, is used to evaluate the potential importance of land use and trip distribution induced travel effects in the Sacramento, California, region.  The model is used to simulate a future base case scenario (low-build) and a beltway scenario for 25- and 50-year time horizons.  First, the scenarios are simulated with the full Sacramento MEPLAN model set, and its implied elasticities of vehicle miles traveled with respect to lane miles are compared to the empirical literature.  The findings indicate that these elasticities are similar.  Second, three sensitivity tests are performed in an attempt to isolate the contribution of different induced travel effects in the model.  The scenarios are simulated holding constant the following effects from the future base case scenario to the beltway scenario: (1) quantities of developed land, (2) land development and household and employment location (1), land development, household and employment location, and trip distribution.  Each of these scenarios represents methods of operating travel demand models to capture induced travel. Third, the California vehicle emissions model is used to estimate the air quality effects of induced travel in the scenarios.  Significant increases in VMT and emissions were found for the beltway scenarios run with the full MEPLAN model, and large errors were found when land use effects only were not represented and when land use and trip distribution effects were not represented.

 

Introduction

 

            The induced travel hypothesis is grounded in economic theory and predicts that an increase in roadway supply reduces the time cost of travel, and thus (to the extent that demand is elastic) increases the quantity of travel demanded (or vehicle travel).  This seemingly basic principle of induced travel has been the center of some debate between transportation planners and environmental advocates.  Historically, many transportation planners have asserted that since the demand for travel is derived primarily from economic activities it is largely fixed.  They have justified transportation infrastructure based on estimates of quantities of congestion or on simple calculations of time savings assuming fixed amounts of travel.  Environmental advocates have identified this weakness in typical transportation planning, and have used induced travel arguments to halt or slow proposed new highway projects because of the effect of increased travel on the environment. 

            Recent research has provided persuasive evidence for induced travel, and the principle has been acknowledged by leading transportation researchers (Transportation Research Board, 1995; Transportation Research Circular, 1998) and by the Environmental Protection Agency (EPA, 2000).

            One of the difficulties of testing the induced travel hypothesis is controlling for confounding economic activity variables such as population, income, and other demographic trends (e.g., women in the workforce).  Much of the recent induced travel research has attempted to control for these variables and has not been able to reject the hypothesis of induced travel (Goodwin, 1996; Hansen and Huang, 1997; Noland and Cowart, 2000; Chu, 2000; Fulton et al., 2000; Noland, 2000).  The results of this research have yielded fairly consistent long-term elasticities of vehicle miles traveled (VMT) with respect to roadway lane miles.  See Table 1 below.

 

Table 1.  Long-term elasticities of VMT with respect to lane miles.

Source	Geographic region	Elasticity range
Hansen and Huang, 1997	County and Metropolitan area	0.3 to 0.7 (county) 0.5 to 0.9 (metropolitan)
Noland and Cowart, 2000	Metropolitan area	0.8 to 1.0
Fulton et al., 2000	County	0.5 to 0.8
Noland, 2000	State	0.7 to 1.0

 

 

            The recent evidence for the induced demand hypothesis has brought renewed attention to the inability of most regional travel demand models to represent the effects of induced travel (Transportation Research Board, 1995; Transportation Research Circular, 1998).  This limitation may have important implications with respect to compliance with the Clear Air Act Amendments (CAAA) and the National Environmental Policy Act (NEPA). 

The CAAA mandate the conformity of state air quality plans and transportation plans to meet national ambient air quality standards.  Non-attainment regions use travel demand models to demonstrate that aggregate emission levels in their transportation improvement plans are not greater than the motor vehicle emissions budget in the approved state implementation plans.  If regional travel demand models do not account for the effect of induced travel, VMT and emissions may be underestimated in transportation plans that include highway capacity expansions.  If the requirements of the CAAA are not met, penalties can be imposed, including the loss of federal funds for transportation projects, the imposition of stricter requirements, and possibly litigation.

NEPA requires Environmental Impact Statements (EIS) for federal projects to provide information about the environmental effects of the project and alternatives to decision-makers and the public.  Many highway projects are still justified primarily by estimates of congestion reduction.  However, if a regional travel demand model does not account for the effects of induced travel, then congestion reduction from the highway project may be overestimated, and congestion reduction from alternatives (e.g., auto pricing and transit) may be underestimated.  In addition, analysis of the secondary impacts of highway projects (e.g., changes in land use) is also required (Council on Environmental Quality, 1987).  If a regional travel demand model does not capture induced effects, then it cannot assess secondary effects.   

Most travel demand models account for mode and route shifts associated with induced travel, but many do not account for other induced travel effects such as changes in land use, trip generation (or number of trips), trip distribution (or destination choice), and departure time choice.  All of the behaviors except departure time choice can change the travel models’ estimates of VMT.  Representation of departure time choice can change estimates of congestion if peak-spreading occurs; for example, less severe congestion would be projected during the peak period for the future base scenario, and by comparison, a highway alternative would appear less effective in reducing congestion.  It is generally acknowledged that changes in mode choice, route choice, and departure time choice are effects of induced demand; however, the importance of land use, trip generation, and destination choice effects has been a source of controversy (DeCorla-Souza, 1998). 

The empirical and the modeling literature provide scant evidence on the subject (DeCorla-Souza, 1998; Dowling and Colman, 1998; Noland and Cowart, 2000).  Dowling and Colman (1998) use a travel behavior survey and find that travel demand models may underpredict trips induced by a major new highway project by 3% to 5%.  Coombe (1996) reviews the results of several modeling studies in the U.K. and finds that the estimates of induced travel, which include analyses of the effects of trip generation, trip distribution, mode share, and land use, in these models is not large overall.  However, there is evidence that elasticities implied by transportation models calibrated against cross-sectional data in the U.K. are lower than those found in the empirical literature (Halcrow Fox and Associates, 1993).  In the U.S., travel modeling studies in the Salt Lake City, Nashville, and Sacramento regions suggest that changes in trip distribution may be a significant effect of induced travel (COMSIS, 1996; Johnston and Ceerla, 1996).

            In this study, an integrated land use and transportation models, MEPLAN, is used to evaluate the potential importance of land use (land development and location of population and employment) and trip distribution induced travel effects in the Sacramento, California, region.   The model is used to simulate a base case scenario (low-build) and a beltway scenario for 25- and 50-year time horizons (from 1990 to 2015 and 2040).  First, the scenarios are simulated with the full Sacramento MEPLAN model set, and its implied elasticities of VMT with respect to lane miles are compared to the empirical literature. 

Second, three sensitivity tests are performed in an attempt to isolate the contribution of different induced travel effects.  Calibrated relationships in a model may provide some guidance about the relative magnitude of separate effects of induced travel (Coombe, 1996).  The scenarios are simulated holding constant the following effects from the future base case scenario to the beltway scenario: (1) the quantities of developed land in each zone, (2) land development and household and employment location, and (3) land development, household and employment location, and trip distribution.  Each of these scenarios represents various methods of operating travel demand models to capture induced travel.  Scenario (3) is equivalent to a travel demand model without feedback of assigned travel times to trip distribution; that is, only the mode choice and traffic assignments of induced travel are represented.  This is still a common method of operating travel demand models in the U.S.  Scenario (2) is equivalent to a travel demand model with feedback to trip distribution; that is, the trip distribution induced travel effects are added to scenario (3).  This scenario is analogous to a state-of-the-practice travel demand model.  Scenario (1) is equivalent to a travel demand model with feedback that is integrated with an activity allocation model; that is, the locations of different types of employment and population can vary with the scenario, but not acres of land developed.  Very few travel demand models are applied in this way in the U.S.  Elasticity is calculated for each sensitivity test, and the results provide some insight into the relative contribution of land use and trip distribution effects of induced travel in the Sacramento region.

Third, the California vehicle emissions model (DTIM2 with EMFAC7F1.1 emissions factors) is used to estimate the air quality effects of induced travel in the simulated scenarios.

 

The Sacramento Region

 

            The Sacramento region is located in the central valley of Northern California.  In 1995, the region was estimated to have a total population of 1.8 million and total employment of about 700,000.  Population is expected to grow annually at a rate of 1.9% to 2015, and employment is expected to grow annually at a rate of 2.2% to 2015 (Sacramento Area Council of Governments, 1996).  Average household income in 1995 was about $63,000 dollars.  In the past, the employment base of the Sacramento region has been largely government and agriculture; however, more recently there has been a rapid expansion of high technology manufacturing.  The residential and employment densities of the region can be characterized as medium to low.   Current mode shares for home based work trips are approximately 76% drive alone, 17% carpool, 3% transit, 2% walk, and 2% bike.

 

 

Methods

The Sacramento MEPLAN Model

The MEPLAN modeling framework is described in Hunt and Echenique (1993).  The basis of the framework is the interaction between two parallel markets, the land market and the transportation market.  This interaction is illustrated in Figure 1.  Behavior in these two markets is a response to price signals that arise from market mechanisms.  In the land markets, price and generalized cost (disutility) affect production, consumption, and location decisions by activities.  In the transportation markets, money and time costs of travel affect both mode and route selection decisions.

 

Figure 1. The interaction of land use and transportation markets in MEPLAN.

 

 

The cornerstone of the land market model is a spatially-disaggregated social accounting matrix (SAM) (Pyatt and Thorbecke, 1976) or input-output table (Leontieff, 1941) that is expanded to include variable technical coefficients and uses different categories of space (e.g., different types of building and/or land).   Logit models (McFadden, 1974) of location choice are used to allocate volumes of activities in the different sectors of the SAM to geographic zones.  The attractiveness or utility of zones is based on the cost of inputs (which include transportation costs) to the producing activity, location-specific disutilities, and the costs of transporting the resulting production to consumption activities.  The resulting patterns of economic interactions among activities in different zones are used to generate origin-destination matrices of different types of trips.  These matrices are loaded to a multi-modal network representation that includes nested logit forms (Williams, 1977) for the mode choice models and stochastic user equilibrium for the traffic assignment model (with capacity restraint).  The resulting network times and costs affect transportation costs, which then affect the attractiveness of zones and the location of activities, and thus the feedback from transportation to land use is accomplished. 

The framework is moved through time in steps from one time period to the next, making it “quasi-dynamic” (Meyer and Miller, 1984).  In a given time period, the land market model is run first, followed by the transportation market model, and then an incremental model simulates changes in the next time period.  The transportation costs arising in one period are fed into the land market model in the next time period, thereby introducing lags in the location response to transport conditions.  See Hunt (1994) or Hunt and Echenique (1993) for descriptions of the mathematical forms used in MEPLAN.

The specific structure of the Sacramento MEPLAN model is shown in the diagram in Figure 2, and Table 2 defines the categories in the diagram.  The large matrix in the middle of the diagram lists the factors in the land use submodel and describes the nature of the interaction between factors.  A given row in this matrix describes the consumption needed to produce one unit of the factor, indicating which factors are consumed and whether the rate of consumption is fixed (f) or price elastic (e).

 

Table 2.  Description of categories in figure 2.

 

Type of Category	Category Name	Category Description
Industry and Service	AGMIN	Agriculture and mining
	MANUF	Manufacturing
	OFSRV-RES	Services and office employment consumed by households
	OFSRV-IND	Services and office employment consumed by other industry
	RETAIL	Retail
	HEALTH	Health
	EDUCATION	Primary and secondary education
	GOVT	Government
	PRIV EDU	Private education
	TRANSPORT	Commercial transportation
	WHOLESALE	Wholesale
Households	HH LOW	Households with annual income less than $20,000
	HHMID	Households with annual income between $20,000 and $50,000
	HH HIGH	Households with annual income greater than $50,000
Land Use	AGMIN LU	Land used for agriculture
	MANUF LU	Land used for manufacturing
	OFSRV LU	Land used for services and office employment
	RETAIL LU	Land used for retail
	HEALTH LU	Land used for health
	EDUCATION LU	Land used for education
	GOVT LU	Land used for government
	RES LU	Land used by residences

 

 Figure 2.  Hunt diagram of the Sacramento MEPLAN model.

The Sacramento MEPLAN model uses eleven industry and service factors that are based on the SAM and aggregated to match employment and location data.  Households are divided into three income categories (high, medium, and low) based on the SAM and residential location data.  The consumption of households by businesses represents the purchase and supply of labor.  The consumption of business activities by households represents the purchase of goods and services by consumers.  Industry and households consume space at different rates and have different price elasticities, and thus there are seven land use factors in the model.  Constraints are placed on the amount of manufacturing land use to represent zoning regulations that restrict the location of heavy industry.  Each of these land uses (except agricultural land use) locates on developed land represented by the factor URBAN LAND.  Two factors are used to keep track of the amount of vacant land available for different purposes in future time periods (MANUF VAC LAND and TOTAL VAC LAND), and the development process converts these two factors to URBAN LAND.  The MONEY factor is a calibration parameter that allows differential rents to be paid by different users of the same category of land.

The single-row matrix just above the large matrix in Figure 2 shows activity that is demanded exogenously, which includes exporting industry, retired households, and unemployed households.  This corresponds to the “basic” economy in a Lowry model.  

The matrix directly above at the top of the diagram shows the structure of the incremental model that operates between time periods.  The r’s for the industry and household factors indicate the economic growth in the region, and the r’s for the land use factors show how vacant land is converted to urban land.

            The matrix on the left below the large matrix indicates the structure of the interface between the land use and transportation submodels.  Each row represents one of the matrices of transportation demand and indicates the producing factors (in the corresponding columns in the matrix above) whose matrices of trades are related to that flow. 

            The remaining three matrices at the bottom show the structure of the transportation model.  Five modes are available, and each mode can consist of several different types of activity on different types of links.  The matrix directly to the right shows that all modes are available to all flows (m).  The matrix below this, on the right, indicates the travel states (s) that make up each mode.  The matrix on the left shows which travel states are allowed on each transportation network link and whether capacity restraint is in effect (a) or not (w).  The design of the mode choice and assignment models is based on the Sacramento Regional Travel Demand model (DKS Associates, 1994).  A more detailed description of the Sacramento MEPLAN model design can be found in Abraham (2000).

            The parameters in the Sacramento MEPLAN model were estimated with a sequential approach in which parameters of individual submodels are estimated, and then the overall model is considered.  The submodels in MEPLAN and other local models used to inform the calibration of the MEPLAN model are shown in Figure 3.  The local models are on the left and right side of Figure 3.  Parameters (shown as l) were taken from the input/output economic model of Sacramento from the California Department of Water Resources and the Sacramento regional travel demand model (which uses some outside parameters in its mode choice model) for use in the Sacramento MEPLAN model.  The parameters in LUSB, TASB, and FREDA submodels were estimated separately, but the LUSA and the TASA submodels could not be estimated separately.  The “spatial interaction” data at the center of the top of Figure 3 consists of detailed tables describing how much interaction occurs between different amounts of economic activities by type by zone.  Observed data at the required level of detail were not available, and thus TASA could not be run independently of LUSA.  The accessibility numbers at the center of Figure 3 were not available either, and thus LUSA could not be run independently of TASA.  As a result, most of the parameters in both LUSA and TASA were estimated in the overall estimation process.    A more detailed discussion of parameter estimation and calibration can be found in Abraham (2000).

 

Figure 3.  The submodels of the Sacramento MEPLAN model and other models used to inform parameters.

 

Emissions Model

The California Department of Transportation’s Direct Travel Impact Model 2 (DTIM2) emissions model and the California Air Resources Board’s EMFAC7F emissions factors are used in the emission analysis.  The outputs from the MEPLAN model used in the emissions analysis include the results of assignment for each trip purpose by each time period (AM peak, PM peak, and off-peak).  The Sacramento Area Council of Governments (SACOG) provides regional cold-start and hot-start coefficients for each hour in a twenty-four hour summer period.  The 2015 emissions factors are used for the 2015 scenarios, and the 2020 emissions factors are used for the 2040 scenarios.  The 2020 factors were the latest available from EMFAC7F.

 

Scenarios

The major transportation network improvements are made in the year 2005, and thus land use is affected in the years 2010 to 2040 (in five-year increments).  See Figure 4 for a map of the scenario network.  Regional population and employment totals are approximately the same across scenarios (i.e., the percentage change from the future base case is less than 1%) and income is consistent across scenarios.

 

Figure 4.  Map of the Sacramento scenario network.

 

Base Case.  The base case scenario represents a financially conservative expansion of the Sacramento region’s transportation system and serves as a point of comparison for the other scenarios examined in this study.  This scenario includes a relatively modest number of road-widening projects, new major roads, one highway HOV lane segment, and a limited extension of light rail.

            Beltway.  The beltway scenario adds two regional beltways (in the north, south, and east areas of the region) and an extensive expansion of the region’s HOV lane system.   This scenario includes 591 new lane-miles of highways, six new interchanges for the beltways, 65 lane-miles of new arterial roads to serve the beltways, and 153 lane miles of new HOV lanes.  This scenario represents a 54 percent increase in new freeway lane miles and a 588 percent increase in HOV lane-miles over the base case scenario.

            Sensitivity tests of the model components that capture the induced travel effects were applied to the beltway scenario.  See Table 3 below.  The scenario was first simulated with the full MEPLAN model to represent all the induced travel effects, including land use, trip distribution, mode choice, and traffic assignment, captured by the model (Scenario A).  Next, the scenario was simulated holding only acres of land developed constant from the future base scenario (Scenario B).  Then, the scenario was simulated holding land development and population and employment in each zone (Scenario C).  This scenario is analogous to a regional travel demand model system with feedback of assigned travel times to the trip distribution step (until the model converges).  In other words, the trip distribution steps are elastic with respect to changes in generalized travel costs.  State-of-the-practice regional travel demand models would include these model processes.  Finally, the scenario was simulated holding land development, population and employment location and trip distribution constant (Scenario D).  This scenario is analogous to a regional travel demand model system without feedback of assigned travel times to trip distribution that is sensitive to changes in travel time and cost.  Such a model would use fixed trip distribution matrices.  Many regional travel demand models in the U.S. are still currently operated in this manner.  

 

Table 3.  Summary of scenarios simulated in the sensitivity analysis with the Sacramento MEPLAN model.

 

Induced Travel Effects	Beltway A	Beltway B	Beltway C	Beltway D
(1) Quantity (acres) of land developed	X
(2) Population & employment location & redevelopment	X	X
(3) Trip distribution	X	X	X
(4) Mode Choice	X	X	X	X
(5) Traffic Assignment	X	X	X	X

RESULTS

 

In this section, the land use results from the full model simulation of the base case and the beltway scenarios are described, and then the travel and emissions results for the beltway sensitivity tests are compared.  

Land Use

Table 4 presents the household and employment land use results by superzone for the year 2015 and 2040.   In the Base Case scenario, land development from 1990 to 2015 and 2040 occurs north, east, and south of the City of Sacramento.  There is limited land development in Yolo County because of exclusive agricultural zoning in the county.  Over time for both the 2015 and 2040 time horizons, households and employment tend to locate primarily in existing, built-up areas northeast, east, and immediately south of the central business district (CBD).  In 2040, however, households are more likely to locate in relatively more remote sections of these areas (e.g., South Sutter, Southeast Sacramento County, and El Dorado Hills).  In general, household and employment location tends to follow land development; however, density increases in some zones.  The land use results for the other scenarios are discussed in comparison to the future base case scenarios. 

Table 4.  Percentage change from the base case scenario to the beltway scenario by superzone.

Households	2015	2040
Sacramento CBD (13, 15,50)	1%	1.6%
Citrus Hgts/Roseville (70,71,4)	1%	1.7%
Rancho Cordova/Folsom (6,12)	0%	1.1%
Inner Suburbs (1-3,7-11,14,16,25)	2%	-9.2%
Outer Ring (remainder)	-1%	6.7%
EMPLOYMENT	2015	2040
Sacramento CBD (13, 15,50)	4%	3.0%
Citrus Hgts/Roseville (70,71,4)	1%	0.0%
Rancho Cordova/Folsom (6,12)	12%	18.2%
Inner Suburbs (1-3,7-11,14,16,25)	3%	-1.1%
Outer Ring (remainder)	-12%	-3.6%

            Roadway expansion in the beltway scenario allows industry to locate further away from the households that it serves and employs.  Employment location is more intense in the existing, built-up areas northeast, east, and immediately south of the CBD, and in the CBD for both the 2015 and 2040 time horizons.  Differences in employment location, however, are more dramatic in 2015 than in 2040, and the opposite is true for households.  In 2015 there is a movement of households further away from employment compared to the base case; however, this shift is more intense by 2040, as more households locate in the most remote eastern sections of the region.

Businesses are moved around more easily than households in the Sacramento MEPLAN model in the shorter term.  The constraints of existing commercial building stock are not well represented in the model. The model does not have a specific representation of floorspace development, and thus important differences among types of buildings cannot be distinguished and there is no representation of the cost to redevelop a building space.  It is relatively easy, for example, for the model to show retail operations to moving into a former warehouse or an office moving into a former retail space.  A floorspace model would better simulate the difficulty of such moves by distinguishing among building types and representing the time and money needed to redevelop buildings for new use.

In the beltway scenarios for both the 2015 and 2040 time horizons, the distant eastern zones that include the cities of Auburn and Folsom lose commercial employment and become more like “bedroom communities” compared to the base case scenario.  As a result of increased roadway capacity, retail activity can shift from local commercial to more remote zones where “big-box” retailing is likely to occur (although the model has no representation of establishment size).  In both scenarios and time horizons, Rancho Cordova becomes increasingly important as a commercial node east of the City of Sacramento and west of Folsom.

 

Travel

 

The daily VMT results for the sensitivity analysis of the beltway scenario are provided in Table 5 and in Figure 5.  The beltway scenario simulated with the full MEPLAN model (Scenario A) generates a relatively large increase in VMT compared to the base case, and this increase grows over time (13% in 2015 and 18% in 2040).  Greater distances between the home and the workplace and faster auto travel times that result from roadway construction in the beltway scenario increase VMT.  The error resulting from the failure to simulate the various induced travel effects in MEPLAN (see figures in parentheses in Table 5) is, in most cases, relatively large and this error increases over time.  In Scenario D, when only the mode choice and traffic assignment effects of induced travel are represented, MEPLAN predicts a small reduction in VMT because of the HOV lanes in the beltway network.  In Scenario C, when the trip distribution effects of induced travel are added, the model captures approximately half of the increase in VMT found in Scenario A.  Comparing Scenario C to Scenario B indicates that shifts in categories and amounts of population and employment in zones also makes a significant contribution to induced travel in MEPLAN.  Comparing Scenario B to Scenario A indicates that, when only quantities of land developed is held constant from the future base case scenario, the error is small compared to other beltway scenarios (Scenarios C to D).  Thus, changes in acres developed make a relatively smaller contribution to induced travel than do changes in employment and population location.

 

 

Table 5.  Daily VMT results for the Sacramento Region.

Scenarios: model component(s) represented  from the future base case scenario	2015 percentage change VMT	2040 percentage change VMT
Scenario A: (1)   Quantity of developed land (2)   Activity location & redevelopment (3)   Trip distribution (4)   Mode choice (5)   Traffic assignment	13%	18%
Scenario B: (2) Activity location & redevelopment (1)   Trip distribution (2)   Mode choice (5) Traffic assignment	11% (-2%) a	17% (-1%)
Scenario C: (3)   Trip distribution (4)   Mode choice (5)  Traffic assignment	6% (-7%)	10% (-8%)
Scenario D: (4)  Mode choice (5)  Traffic assignment	0% (-13%)	-1% (-19%)

^a  Figures in parentheses are the percentage point change in VMT from Scenario A.

 

Figure 5. Percentage point contribution of induced travel effects

to daily change in VMT from Scenario A in 2015 and 2040.

 

            The results presented in Table and Figure 5 raise some issues.  First, the contribution of land development and population and employment location to the total change in VMT seems to fall over time. Land use changes occur more quickly because of the absence of specific representation of different types of floorspace development (which was discussed in the previous section), and thus land use changes are probably overestimated in 2015.  This result may also be specific to this scenario in this region.  In this scenario, beltway freeways, which are fairly centrally located, are in place by 2005 and after that there are no more additions to the transportation network.  During the 2005 to 2015 periods the beltways open up substantial quantities of previously undeveloped land.  By 2040 the position of the beltways is such that they would not open up much more land for development; people would likely have already located in the area around the beltway because of population growth.  By 2040 they affect travel destinations more than location and development. 

            Second, the contribution of land use changes to the total change in VMT seemed large to some reviewers.  In the Sacramento region there are relatively large amounts of undeveloped land.  This would increase the development response in this region compared to older and more built-up regions.  In addition, the dominance of the automobile in the Sacramento region would tend to reduce the mode choice response compared to cities (for example, in Europe) that have a higher rates of transit ridership and cycling.

            Third, it is important not to generalize the results of this study to other scenarios and other regions.  The results presented will vary based on the location and timing of new highway projects in the region (e.g., congestion levels and types of geographic regions connected) and the type of new highway capacity (e.g., HOV lanes included in the network).  In addition, the calculated results are based on a model that was calibrated on cross-sectional data and not longitudinal data that included induced travel effects.  This is, however, typical of regional travel demand models.

            The results of the elasticity of VMT with respect to lane miles for the sensitivity tests are presented in Table 6.  The arc elasticity of VMT with respect to lane miles is calculated as the percentage change in VMT from the base case scenario to an alternative Beltway scenario (i.e., Scenarios A to D), divided by the percentage change in total lane miles from the base case scenario to an alternative Beltway scenario (i.e., Scenarios A to D).  Note that the log arc elasticity and the mid-point arc elasticity were also calculated and the results were the same as the as the arc elasticity calculated with the formula just described.  The arc elasticity is not exactly comparable to the point elasticity in the empirical literature; their comparability depends on the shape of the demand curve and the relative size of the change in the cost or supply variable.   

            The elasticity results for Scenario A, in which the full model was run, are similar to the empirical elasticity results from aggregate studies at the metropolitan level described above (0.8 for 2015 and 1.1 for 2040).  The very long-term elasticity for the year 2040 is somewhat higher than that found in the empirical literature.  Elasticity tends to increase over time as expected.  The elasticity is zero when the MEPLAN model simulates only the mode choice and traffic assignment effects of induced demand (Scenario D).  Again, this is because of the HOV lanes in the beltway network.  When the trip distribution effects are added (Scenario C), approximately half of the induced travel effects are captured.  Comparing Scenario C to Scenario B indicates that changes in the locations of population and employment account, approximately, for the other half of the induced travel effects.  Comparing Scenario B to Scenario A indicates that the failure to represent changes in acres of land development accounts for a relatively smaller portion of the elasticity compared to the location of employment and households. 

 

Table 6.  Elasticity of VMT with respect to lane miles results for the Sacramento Region.

 

Scenarios: model component(s) represented from the future base case scenario	2015 Elasticities	2040 Elasticities
Scenario A: (1)   Quantity of developed land (2)   Activity location & redevelopment (3)   Trip distribution (4)   Mode choice (5)   Traffic assignment	0.8	1.1
Scenario B: (2) Activity location & redevelopment (1)   Trip distribution (2)   Mode choice (5) Traffic assignment	0.6 (-16%) a	1.0 (-1%)
Scenario C: (6)   Trip distribution (7)   Mode choice (5)  Traffic assignment	0.4 (-54%)	0.6 (-43%)
Scenario D: (4)  Mode choice (5)  Traffic assignment	0.0 (-100%)	0.0 (-100%)

^a  Figures in parentheses are the percentage changes in the Scenarios B, C, and D elasticities from Scenario A.  

The similarities of the elasticity results in this behavioral model with the elasticity results from the aggregate studies (described in Table 1) increase the confidence that the results in this model and the aggregate statistical studies are reasonable.  One of the critiques of the empirical induced travel studies has been that they use aggregate statistical data as opposed to disaggregate behavioral data.  This study begins to address this concern because the model is more behavioral than statistical, but only certain parameters of the model were established using disaggregate data.

Vehicle Emissions

The daily vehicle emissions results are presented in Table 7.  When the full MEPLAN model is used to simulate the beltway scenario (Scenario A), there is a relatively large increase in emissions.  However, when the induced travel effects of only mode choice and traffic assignment are represented in the MEPLAN model (Scenario D), emissions decrease because of the reduction in VMT resulting from the HOV lanes in the beltway scenario.  When the induced travel effects of trip distribution are added (Scenario C), emissions are largely predicted to increase, but the increase is generally less than half that obtained from Scenario A.  Some pollutants are reduced in Scenario C because of increased speeds, and thus reduced vehicle hours of travel.  The errors due to the failure to represent the induced travel effects of land use are relatively large in scenarios C and D.  Again, when acres of land developed are held constant, the errors are comparatively smaller than errors from changes in the locations of types of employment and population.  In general, emissions increase, but the error due to the failure to represent the induced travel effects is relatively stable over time.     

 

 

Summary and Conclusions

            In this study, an integrated land use and transportation models, MEPLAN, was used to evaluate the potential importance of land use and trip distribution effects of induced travel in the Sacramento, California, region.   The model was used to simulate a base case scenario (low-build) and a beltway scenario for 25- and 50-year time horizons (from 1990 to 2015 and 2040). 

First, the scenarios were simulated with the full Sacramento MEPLAN model set and its implied elasticities of VMT with respect to lane miles were compared to the empirical literature.  This scenario includes changes in land use (acres of land developed and employment and population location), trip distribution, mode choice, and traffic assignment.   Very few regions in the U.S. analyze all these induced travel effects of proposed highway projects.  The calculated elasticity for the beltway scenario was 0.8 in 2015 and 1.1 in 2040.  These elasticities are similar to elasticities reported in the empirical literature, which range from 0.5 to 1.0 for metropolitan regions.

Second, three sensitivity tests were simulated in an attempt to isolate the contribution of different induced travel effects.  The future base case and beltway scenarios were simulated holding constant the following effects from the future base case scenario to the beltway scenario: (1) quantity of developed land, (2) land development and household and employment location, and (3) land development, household and employment location, and trip distribution. 

When only the mode choice and traffic assignment effects of induced travel were represented in the model (3 above), no induced travel was captured, and the elasticity was zero.  In part, this was because the beltway network included HOV lanes, but it is still fair to conclude that very little induced travel was captured by changes in mode choice and traffic assignment in the Sacramento MEPLAN model.  This scenario is analogous to a regional travel demand model that uses fixed trip distribution matrices.  Such travel demand models are still commonly used in the U.S.

When the trip distribution effects were added to the mode choice and traffic assignment effects of induced travel (2 above), approximately half of the induced travel effects were captured.  This scenario is analogous to a regional travel demand model that includes trip distribution steps that are elastic with respect to generalized travel costs.  State-of-the practice regional travel demand models in the U.S. include such processes.  

When land development was held constant from the future base scenario, the results suggest that changes in acres of land developed make a relatively smaller contribution to induced travel than changes in the location of various types of employment and households.  However, the two effects together account for approximately half of the induced travel.  In general, we found that the contribution of land use changes became somewhat less important over time.  This, however, is partially caused by the absence of a floorspace model in the current model.  The model tends to somewhat overestimate the mobility of employment in shorter time horizons.

This study has shown that the lack of representation of floorspace types in the current MEPLAN model leads to a suspiciously fast response in employment location.  The Sacramento MEPLAN model is currently being changed to include a more dynamic and behavioral submodel of development and redevelopment.

Finally, the California vehicle emissions model (DTIM2 with EMFAC7F emissions factors) was used to estimate the air quality effects of induced travel in the simulated scenarios.  When the full MEPLAN model was used to simulate the beltway scenario, it was found to significantly increase VMT  (13% in 2015 and 18% in 2040) and emissions (approximately 11% in both time horizons).  When the land use effects only were not represented and the land use and trip distribution effects were not represented, large errors were found for the estimates of VMT and emissions and, in the latter case, the rank ordering of the scenarios was altered.  When origins and destinations are held constant, emissions are projected to decrease for all pollutants compared to the base case scenario because of travel time and distance saved resulting from more direct available routes to destinations (provided by the new highway capacity). 

The results of the study indicate that the induced travel effects represented by the Sacramento MEPLAN model (and not typically represented by regional travel demand models) for the scenario evaluated make a relatively significant contribution to projections of VMT and emissions. The magnitude of change between the scenario and the base case is significantly altered.

Sometimes merely spatially rearranging a given amount of population and employment is discounted as a serious induced demand effect.  The argument has been made that the growth would have occurred anyway but just somewhere else and so it can be ignored.  The results of this study suggest that it can count for quite a bit.  The effect on VMT of spatially rearranging a given level of population and employment can outweigh the effects of attracting new development that wouldn’t have occurred otherwise.

Induced demand for vehicle travel occurs because people take advantage of the mobility provided by new infrastructure — people's travel and location patterns change in response to transportation infrastructure.   This is a substantial mobility benefit of transportation infrastructure.  This study has shown how the different mechanisms in MEPLAN contribute to induced demand, and has shown how that induced demand has costs related to the environment.  Future studies should use land use and transport interaction models to establish the mobility benefits of induced demand for comparison with the environmental costs.  New generations of land use and transportation models designed to enable benefit calculations may be well suited to this task.

 

ACKNOWLEDGEMENTS

We would like to thank the Federal Highway Administration for funding this study through their Dwight David Eisenhower Transportation Fellowship.  We would also like to thank the University of California Transportation Center and the Mineta institute for their support of this work.  We would like to give a special thanks to Patricia L. Mokhtarian for her important comments and advice. Any errors are those of the authors.

 

REFRENCES

Abraham, J. E. (2000) Parameter Estimation in Urban Models: Theory and Application to a Land Use Transport Interaction Model of the Sacramento, California Region.  Diss., University of Calgary, Canada.

Chu, X. (2000)  Highway Capacity and Areawide Congestion.  Preprint for the 79^th Annual Meeting of the Transportation Research Board.  National Research Council, Washington, D.C. 

 

COMSIS Corporation (1996)  Incorporating Feedback in Travel Forecasting:  Methods, Pitfalls and Common Concerns.  Office of Environmental Planning, Federal Highway Administration, DOT-T-96-14, Washington, D.C.

 

Coombe, D. (1996)  Induced traffic: what do transportation models tell us?  Transportation, 23, 83-101.

 

Council on Environmental Quality (1987)  Regulations for Implementing the Procedural Provisions of the National Environmental Policy Act, 40 CFR parts 1500-1508.

 

DeCorla-Souza, P. and Cohen, H. (1998)  Accounting for Induced Travel in Evaluation of Metropolitan Highway Expansion.  Preprint for the 77^th Annual Meeting of the Transportation Research Board.  National Research Council, Washington, D.C. 

 

DKS Associates (1994)  Model Development and User Reference Report.  Sacramento

Area Council of Governments, Sacramento, CA, October.

 

Dowling R.C. and Colman, S.B. (1998)  In Transportation Research Circular.  Highway Capacity Expansion and Induced Travel: Evidence and Implications.  Transportation Research Board. National Research Council. 481 (February), 21-32. 

 

Halcrow Fox and Associates (1993)  Review and Specifications of Model Elasticities.  Report for the Department of Transport by Halcrow Fox and Associates, Accent Marketing & Research, and the University of Leeds, as part of the London Congestion Charging Research Programme.

 

Hunt J. D., and Echenique, M. H. (1993)  Experiences in the application of the MEPLAN framework for land use and transportation interaction modeling.  Proceedings of the 4^th National Conference on the Application of Transportation Planning Methods, Daytona Beach, Florida, USA, (May), 723-754.

 

Hunt, J. D. (1994)  Calibrating the Naples land use and transport model.  Environment and Planning 21B, 569-590.

 

Fulton, L. M., Meszler, D. J., Noland, R. B. and Thomas, J. V. (2000)  A Statistical Analysis of Induced Effects in the U.S. Mid-Atlantic Region. Preprint for the 79^th Annual Meeting of the Transportation Research Board.  National Research Council, Washington, D.C. 

 

Goodwin, P. B. (1996) Empirical evidence of induced traffic, a review and synthesis, Transportation, 23, 35-54.

 

Hansen, M. and Huang, Y. (1997) Road supply and traffic in California urban areas, Transportation Research A, 31, 205-218.

 

Leontief, W. (1941) The Structure of the American Economy 1919-1939, Oxford University Press, New York, NY, USA.

 

McFadden, D. (1974) Conditional logit analysis of qualitative choice behavior.  In Frontiers in Econometrics, Editor: Zaraembka P. Academic Press, New York, USA, 105-142.

 

Meyer, M. D., and Miller, E. J. (1984) Urban Transportation Planning:  A Decision-Oriented Approach.  McGraw-Hill, New York, USA.

 

Noland, R. B. (2000) Relationships between highway capacity and induced vehicle travel.  Transportation Research A (forthcoming). 

 

Noland, R. B. and Cowart, W. A. (2000) Analysis of Metropolitan Highway Capacity and the Growth in Vehicle Miles of Travel.  Preprint for the 79^th Annual Meeting of the Transportation Research Board.  National Research Council, Washington, D.C. 

 

Pyatt, G., and Thorbeck, E. (1976) Planning Techniques for a Better Future.  International Labour Office, Geneva, Switzerland.

 

Sacramento Area Council of Governments (1996) 1995 Regional Housing, Population, and Employment Projections.  Sacramento, CA.

 

Transportation Research Board (1995) Expanding metropolitan highways: implications for air quality and energy use, Special report 245, National Research Council, National Academy Press, Washington, D.C. 

 

Transportation Research Circular (1998) Highway Capacity Expansion and Induced Travel: Evidence and Implications.  Transportation Research Board. National Research Council. 481 (February).

 

U.S. Environmental Protection Agency (2000) Induced Travel: A Review of Recent Literature with a Discussion of Policy Issues.  Office of Policy. Energy and Transportation Sectors Division, Washington, D.C.

 

Williams, H. C. W. L. (1977) On the formation of travel demand models and economic evaluation measures of user benefit.  Environment and Planning 9A: 3, 285-344.