Understanding the Transit Multiplier

The transit multiplier is the ratio of the total VMT reduction associated with transit, including both direct and indirect effects on VMT, divided by the VMT reduction associated with transit use alone. The mechanism by which transit leverages larger reductions in VMT, and produces a multiplier effect, is straightforward. Transit creates opportunities for TOD around transit stations, ``compact, mixed-use development near transit facilities with high-quality walking environments" (Cervero, Murphy, Ferrell, Goguts, & Tsai, 2004, p. 11). TOD may reduce per capita VMT independent of transit usage, as residents walk and bike more and make shorter automobile trips in a mixed-use environment.

However, researchers have yet to reach a consensus on the magnitude of the multiplier effect. Table 1 summarizes these studies, which draw on data from different cities and use different methods, and have produced estimates for the multiplier ranging from 1.29 to 9. Estimates of the multiplier can even vary widely within a given study.

Newman and Kenworthy (1999) calculate that for every one mile traveled on public transit, there is a displacement of 3 to 4.5 VMT in the United States and 5 to 7 VMT in other countries. They propose four reasons for VMT reductions due to the transit multiplier: 1) transit concentrates population and employment, which reduces trip lengths; 2) transit increases ``trip chaining," as transit users combine multiple errands into a single public transit trip; 3) transit results in car shedding, as households with good access to transit service own fewer cars than their counterparts in unserved areas; and 4) transit, and the compact land use patterns it produces, increase walking and biking.

Holtzclaw (2000) estimates that, in San Francisco, every one passenger mile traveled on public transit displaced 9 VMT by car, while in two other California cities, Walnut Creek and San Ramon, transit displaced only 1.42 VMT. He attributes the difference in the multiplier to differences in population density; the population density in San Francisco is significantly higher than the other two cities.

Bailey et al. (2008) conducted a national study based on data from the 2001 National Household Travel Survey. Using structural equation modeling, they find that each mile traveled on U.S. public transportation reduced VMT by 1.9 miles. They also find a significant correlation between public transit availability and reduced automobile travel, independent of transit use.

While the work to date on the multiplier effect has made important progress in exploring a complex phenomenon, there are several outstanding issues that make it difficult for transportation planners to apply the research summarized in Table 1:

• Widely varying estimates undermine confidence in the results of these studies and make it difficult to select a value from within the range of estimates. Usually in these cases, researchers can compare values or explain differences between studies due to samples, methods, and definitions, but in the case of the multiplier effect, the range of estimates is so wide that this has proven difficult.

• Results are difficult to generalize and apply to smaller metropolitan areas or areas that do not currently have extensive transit coverage. Most of the studies summarized in Table 1 draw data from the most populous metropolitan areas in the United States and, in one case, Europe; moreover, several of these studies are more than a decade old.

• It is difficult to separate cause and effect in many studies to date. Station areas may be compact because of transit or transit lines may be planned through such areas to maximize ridership.

• Assumptions are counterintuitive in some studies. For example, Bailey et al. (2008) estimate a model that assumes that birth rates and death rates are the ultimate drivers of VMT, as opposed to variables that are much more directly related to travel demand, such as total population, lane miles, transit guideway miles, and fuel prices.

• Studies do not produce reliable results. Some of the multiplier studies, such as New York MTA (2009), produce highly variable multiplier estimates depending on the assumptions that researchers make.

Fundamentally, the very concept of a fixed multiplier that does not vary over space and time is questionable. A transit project in an already compact metropolitan area may produce greater indirect benefits than a project in a sprawling area because the compact area encourages walk trips generally, or the reverse could be true because there is so much room for densification in a sprawling area. Land use patterns take a long time to develop, and thereafter are slow to change, so the multiplier effect will likely be greater 10 or 20 years after a transit project is built than it will be either 1 or 5 years after completion. Any measure of transit's indirect effects on travel patterns should take this temporal factor into account.

Transit Impact Area

A transit multiplier applies specifically to an area around a transit station in which the conditions of the built environment and travel choices change after service is initiated. The area may densify, become more diverse in terms of land uses, or become more pedestrian-friendly in terms of street design. The change in the built environment affects walking for purposes other than transit access, and may even affect automobile driving distances, both of which affect VMT independent of transit use. Most planners talk of a half-mile impact or catchment area around rail stations. However, several recent studies have challenged this standard and argued that rail, at least, has effects on land uses and values and travel behavior at a greater distance than previously estimated (Canepa, 2007; Iacono, Krizek, & El-Geneidy, 2008; Ker & Ginn, 2003; O’Sullivan & Morrall, 1996; Petheram, Nelson, Miller, & Ewing, 2013). If people come to a transit station by modes other than walking, the catchment area will expand with the speed of the mode. The catchment area will be larger for bicycling than walking, and larger for feeder bus service than bicycling. Regardless of the mode of rail access, land use patterns in proximity to transit stations and travel choices of residents may change.

One way to know how far transit's influence extends around a transit station is to assess how the market responds to station proximity according to distance. In Salt Lake County (UT), Petheram et al. (2013) find a positive relationship between LRT station proximity and rental apartment building values out to 1.25 miles. Since development density naturally increases with land prices, land uses might be affected by transit out to 1.25 miles. In another study, Canepa (2007) argues that housing and employment density and urban design can push catchment areas to a mile or more. The transit catchment area can be increased by removing physical barriers between land uses and transit stations, especially by providing direct pathways to transit stations.

Our Research Approach

To assess the impact of Portland's Westside LRT line on household VMT, we have gathered and pooled data from two travel surveys in Portland: one for 1994 before the line and then for 2011 well after the line, thereby permitting longitudinal comparisons. These two dates, separated by 17 years, allow us to study the effect of transit investments on VMT in and around transit stations. By studying changes in an otherwise comparable corridor without LRT, we can control for general trends in development and travel on the west side of Portland. With a bit of manipulation, we can also separate direct effects from indirect effects.

Our work is in contrast to the vast majority of studies of travel and the built environment that are cross sectional in nature. Longitudinal studies that use data for the same places over time to predict behavior are rare because longitudinal data are rare. Yet the approach we take responds to a 2005 National Research Council report calling for studies of so-called natural experiments, changes that occur naturally when some public or private action alters the built environment. ``When changes are made to the built environment—whether retrofitting existing environments or constructing new developments or communities—-researchers should view such natural experiments as ‘demonstration’ projects and analyze their impacts on physical activity" (National Research Council Committee on Physical Activity, Land Use, & Institute of Medicine, 2005, p. 12).

We take advantage of the fact that a natural experiment occurs every time a new transit line is built. We expect well-located transit lines will attract new development, changing the built environment of the station areas and thus creating a corresponding change in household travel behavior. Ours is a classic quasi-experimental study design referred to as a pretest–posttest (pre-intervention and post-intervention) design with a comparison group (Shadish, Cook, & Campbell, 2002). By comparing the difference in outcomes between pretest and posttest for the treated group and the comparison group, we can determine whether the treated group has experienced greater changes from pretest to posttest than has the comparison group. If the differences are statistically significant, this is relatively strong evidence of a treatment effect.

We focus on the Westside LRT line (western portion of the Blue Line), the portion that starts almost two miles west of downtown Portland and extends through Beaverton out to Hillsboro. This 15-mile section, with 17 stations, opened in 1998, after the first Household Travel Survey and well before the second. Much of the alignment is through land that was ripe for development or redevelopment. Station areas have had many years to densify and thereby affect travel behavior. This represents our best opportunity for a pre-intervention and post–intervention comparison.

Only the Westside LRT line would be expected to substantially affect land use and travel patterns between the two survey years, 1994 and 2011, and to have effects that are readily isolated methodologically. The Eastside LRT line was completed in 1986, eight years before the 1994 Household Travel Survey, and thus had already had much of its ultimate impact on development patterns by the time of the survey. The Airport LRT Red Line, opened in 2001, mostly travels through land that is industrial (surrounding the airport). Only one station serves a residential neighborhood, which is bounded by highways. The Downtown Streetcar also began service in 2001. Any reasonable buffer around its stations would encompass LRT stations as well, making it difficult to isolate the streetcar's effect on land use. The Interstate LRT Yellow Line, which opened in 2004, may not exemplify the potential of rail to affect development patterns due to its alignment along the interstate. Portland's fifth LRT line, the Green Line connecting downtown Portland to Clackamas County, was opened in 2009, too recently to have had much effect on development patterns in the corridor.

Control Selection

To estimate the impact of the Westside LRT we have to select an appropriate control, what is often referred to as a ``counterfactual," in this case comparing a group of people who did not receive the treatment—new light rail services—with those who did (Shadish et al., 2002). Since we cannot actually observe a counterfactual, we select a control group that measures what most likely would have happened in the Westside transit corridor in the absence of LRT. Since the region is constantly growing and suburbanizing, and transportation investments are constantly being made, it would not be reasonable to assume that nothing would have happened even in the absence of the light rail.

In this particular case, we have a good idea what would have happened in the absence of LRT. In the early 1990s, the Portland regional transportation plan featured heavy highway investments on the west side, including a Western Bypass (beltway) and supportive arterial improvements. The now famous land use–transportation–air quality (LUTRAQ) study was undertaken to decide between the plan and an alternative scenario, which featured an LRT line out to Hillsboro and supportive TOD development in the corridor. This was referred to as the LUTRAQ scenario. The study showed that the LUTRAQ scenario would outperform the highway/sprawl scenario with respect to VMT and most other performance indicators. Therefore, the former was selected as the preferred alternative, LRT was built, and the Western Bypass was scrapped. Thus, a reasonable counterfactual for LRT expansion is a highway-oriented alternative.

We considered three highway corridors as potential controls for the Westside LRT corridor. All three lie on the west side of Portland, and extend out from the Portland core. These were selected over east-side corridors both because of their location within the region, and the fact that none of them received an LRT intervention during the period under study. In contrast, any potential east-side corridors lie along or intersect with other LRT lines.

The three highway corridors are shown in Figure 2. One heads to the northwest along the NW Sunset Highway (lightest shading). The second heads to the southwest along State Highway 210 (medium shading). The third heads further south along Interstate 5 and the SW Pacific Highway to Tigard and beyond (darkest shading). The LRT corridor is shown with a crosshatched pattern. As a match with the LRT corridor, we selected the corridor running along the SW Pacific Highway. The corridor is 12.5 miles and has 14 major intersections along it. Of the three potential comparison corridors, this corridor had the most similar sociodemographics to the LRT corridor in the two survey years. In particular, it had comparable household income before the intervention (in 1994) and, importantly, comparable household income after the intervention (in 2011). Hence, we do not have to worry about income confounding the relationship between LRT construction and travel characteristics. With the NW Sunset corridor, household income was significantly higher before and after the intervention. With the State Highway 210 corridor, household income was significantly higher after the intervention.

Longitudinal Analysis of Transit's Land Use Multiplier in Portland (OR)

All authors

Reid Ewing & Shima Hamidi

https://doi.org/10.1080/01944363.2014.949506

Published online:

10 October 2014

Figure 2. Transit corridor versus three potential highway corridors.

Display full size

In a quasi-experimental study, the comparison group should be as similar as possible to the treated group. If the two groups were equivalent, this would be a true experiment, but because they can never be truly equivalent in planning practice, a quasi-experiment is the best we can do. In the next section, we test for rough equivalence of travel and other statistics before the intervention. The existence of big differences before the intervention would create statistical problems, most notably the likelihood of regression to the mean.

Data

This study uses geocoded household travel data to explain household VMT in terms of sociodemographic, built environment, and travel variables. The sample of households varies with the specific test being conducted or model being estimated. The data come from surveys conducted in 1994, four years before the opening of the Westside LRT line, and 2011, 13 years after the opening. The 1994 survey was a two-day travel survey. We selected the first travel day for each household. There is a known tendency of trip reporting to drop off with time in a multi-day survey. The 2011 survey was a one-day survey that covered a larger sample of households.

This study also uses socioeconomic data for surveyed households, built environmental data for buffers around household locations, and transit service data for households and buffer areas. Both travel diary surveys provide X–Y coordinates for households and their trips, which allows travel to be modeled in terms of the precise built environment in which households reside and travel occurs. Purpose, travel mode, travel time, and other variables are available from the survey data set for individual trips. Distance traveled on each trip was either supplied or computed with GIS from the X–Y coordinates. Household size, household income, and other variables were available from the survey data set, which allowed us to control for sociodemographic influences on travel at the household level.

Other data sets have been collected for the same years as the travel surveys to estimate values of many built environmental variables for .25-mile, .5-mile, and 1-mile radius buffers around each household. These include a geocoded parcel land use layer, geocoded street and transit layers, and travel time skims, population, and employment by traffic analysis zone as supplied by the regional metropolitan planning organization, Portland Metro.

We acquired parcel-level assessor data in the survey area from each individual county to estimate the amount and type of each land use within the buffers. We converted parcel features to centroid points, allowing us to join parcel attributes to the buffer polygons. We used roadway centerlines to collect intersection points, counting and summarizing centerline intersections. We collected transit stop geographic locations from all operators serving the travel survey area, merging stops according to bus or rail categories. We joined bus and rail stop locations to buffers for stop counts. We determined population density by weighting census block group population estimates with residential parcel square footage, and then applied population density per square foot to residential parcels intersecting each buffer. We obtained employment data at the traffic zone (TAZ) level from Portland Metro, which we used, along with interzonal travel times, to compute employment accessible within 10-, 20-, and 30-minute auto travel times and within 30-minute transit travel times. We generated employment for individual household buffers by weighting the size of the TAZ in proportion to the buffer and then multiplying by the number of jobs in each intersecting TAZ.

Table 2 shows the variables used in the study. Measures of household size, employment, VMT, and trip frequency refer only to household members who completed travel diaries; we have no data for other household members. We defined and measured all variables consistently for the two survey years and adjusted household income for consumer price inflation. Even adjusting for inflation, incomes rose substantially between 1994 and 2011 across the prosperous Portland region.

Importantly, we have used a 1.25-mile network distance to define the study area around the transit stations on the Westside LRT line and around the intersections on the SW Pacific Highway. This buffer produces an adequate sample of households for statistical purposes, and also squares with the ``transit impact area" literature cited here. Surveyed households are represented by dots in Figure 3, with light dots representing households surveyed in 1994 and dark dots representing households surveyed in 2011. By using a larger buffer than the conventional half-mile, we are not suggesting that the effects of LRT on travel and density are uniform within the buffer; we are suggesting, however, that average effects over a larger area can be used to define transit's impacts on VMT.

Longitudinal Analysis of Transit's Land Use Multiplier in Portland (OR)

All authors

Reid Ewing & Shima Hamidi

https://doi.org/10.1080/01944363.2014.949506

Published online:

10 October 2014

Figure 3. Surveyed households in transit and highway corridors in 1994 and 2011.

Display full size

Impact of Rail Investments on VMT

The best-fit models are shown in the Technical Appendix. The models include most of the so-called D variables from the household travel literature (Ewing & Cervero, 2010; Ewing et al., 2011). The measure of density is activity density. The measure of diversity is land use entropy. The measure of design is the percentage of four-way intersections. The measure of destination accessibility is the percentage of regional employment within 10 minutes by auto, which controls for regional location (as opposed to local conditions). Previous studies have found that regional accessibility is the most important determinant of VMT, more important than the local D variables. The remaining D variables capture household sociodemographics.

The likelihood of a household having any VMT increases with household size and household income and decreases with destination accessibility, entropy, and percentage of four-way intersections. The three variables of ultimate interest are also significant. Households that use transit are less likely to drive. Households that walk are also less likely to drive. Households living at higher densities are less likely to drive, independent of the numbers of transit and walk trips.

For households that drive, household VMT increases with household size, number of employed members, and household income, and decreases with destination accessibility and percentage of four-way intersections. The three variables of ultimate interest are also significant. Households that use transit drive less. Households that walk drive less. Households living at higher densities drive less, independent of the numbers of transit and walk trips.

We next apply these relationships to the ``difference in difference" in average transit trip making, walking, and density between the two corridors to estimate how much VMT per capita changed based on direct and indirect effects of LRT. The assumption in this quasi-experimental study is that transit trip making, walking, and density would, in the absence of LRT in the transit corridor, have changed between 1994 and 2011 to the same degree as they changed in the control corridor. This is a standard assumption in pretest–posttest quasi-experimental studies with control groups (Shadish et al., 2002). The cross-sectional VMT model is simply used as a vehicle to implement the quasi-experimental design.

Assuming transit use would have increased by this same amount in the transit corridor as in the highway corridor in the absence of LRT, the net effect of LRT on transit trip making in the LRT corridor (the difference in difference) is 0.60 transit trips per household. Assuming walking would have increased by this same amount in the transit corridor as the highway corridor in the absence of LRT, the net effect of LRT on walk trips in the LRT corridor (the difference in difference) is 0.60 walk trips per household. Finally, assuming activity density in the transit corridor would have increased by this same amount as in the highway corridor in the absence of LRT, the net increase in the activity density (the difference in difference) attributable to LRT is 0.14 thousand persons per square mile, a marginal increase. Both corridors densified substantially over the 17-year period, and much more than expected for the highway corridor.

To determine the direct effect of the LRT line on household VMT, we substitute the net change in the average transit trip rate into the VMT models. A 0.60 increase in the number of transit trips reduces the expected average VMT from 15.44 to 14.97 vehicle miles per household per day, a reduction of 0.47 VMT. A 0.60 increase in the number of walk trips reduces the expected average VMT from 15.44 to 14.72 vehicle miles per household per day, a reduction of 0.72 VMT. A 0.14 thousand persons per square mile increase in activity density reduces the predicted average VMT from 15.44 to 15.18 vehicle miles per household per day, a reduction of 0.26 VMT. When all three effects occur simultaneously, the models tell us that household VMT would drop from 15.44 to 14.01.

The transit multiplier is just the total reduction in household VMT when all three effects are considered (1.43) divided by the reduction in household VMT due to the direct effect of transit trips alone (0.47), or 3.04. That is, the total effect of transit on VMT due to compact land uses near stations and increased walking is three times that of the direct effect due to ridership increases.

Policy Implications

A growing body of research suggests that transit investments have both direct and indirect effects on VMT, that the indirect effect through TOD around stations may be greater than the direct effect through transit ridership alone, and that the so-called transit multiplier (total effect divided by direct effect) ranges from 1.29 to 9.0. In this study, we assess the direct and indirect effects on car use (as measured by household VMT) of the Portland Westside Max LRT extension over a 17-year period. Unlike other studies, we use longitudinal data, which are far superior to cross-sectional data, to estimate effects of discrete changes in the built environment, and we compare a treated transit corridor with a highway corridor that serves as a control.

We find that the corridor in which the Westside LRT line was built became denser, generated more household walk and transit trips, and experienced a slower rise in VMT per household than did the SW Pacific Highway Corridor, which became our control corridor for purposes of comparison. Local governments, Portland Metro (the regional government), TriMet (the transit operator), and the even the state government have all taken steps to encourage TOD around transit stops; apparently, it has worked. The cities of Hillsboro and Beaverton have made major modifications to zoning around stops to encourage TOD, and have also designated urban renewal areas and made public investments in station areas using tax increment financing. Finally, they have added amenities close to stations such as a main street theater, performing arts center, and civic center, with a large plaza for civic events. Portland Metro has launched a major educational campaign to promote TOD called ``Get Centered," channeled a disproportionate share of transportation dollars into centers and corridors, and purchased easements in TODs to encourage denser development. TriMet, the transit operator, has developed its parking lots and other excess lands in active uses at stations themselves. Also, the State of Oregon has enacted tax abatements, subject to local approval, for development in centers.

Naturally, this analysis has its limitations, suggesting several possible future investigations. First, our household sample sizes are small, particularly for the transit corridor in 1994, which had only 65 surveyed households. The difference-of-means tests are affected by small sample sizes, and our confidence in their results is thereby limited. Second, we have no ability to account for micro-design characteristics of the built environment that may render one environment more pedestrian-friendly than another, and account for differences in walk and transit usage. That is to say, much of what makes an area pleasant for pedestrians (Ewing & Clemente, 2013) is too fine-grained to be captured by our metrics. Third, we have no ability to account for residential self-selection; that is, the tendency of those who want to walk or use transit selecting to live in TODs near stations. While residents of the two corridors are matched for household size and income, they may have entirely different attitudes toward walking and transit use. Those who have selected to live near transit stations may be more inclined to use transit and walk, regardless of where they live; we just cannot measure such differences without attitudinal data. It may be this phenomenon rather than the environment of transit station areas that accounts for different rates of walking and transit use between the two corridors.

As debates about air quality, energy, and climate policy have increased, more attention has been paid to the roles of urban form and transit infrastructure in addressing these policy challenges. The vigor that has accompanied research in the area, however, has sometimes given rise to warnings against overexuberance. A special Transportation Research Board panel recently acknowledged that land development patterns likely have an influence on travel, but signaled that it did not have as much ``verifiable scientific evidence" as it would like to have to support its conclusions (American Public Transportation Association, 2009, p. 131). Our study contributes verifiable scientific evidence, based on a strong quasi-experimental design, on the size of transit's land use multiplier when complemented by supportive public policies and investments. Future studies can build on these findings.

Notes