GreenTRIP Connect Methodology

Understanding GreenTRIP Connect’s inputs and outputs

GreenTRIP Connect quantifies projected reductions in driving (called vehicle miles traveled - VMT) for residential developments. Connect recognizes that low-income households drive less, people living near great transit drive less, and households with easy access to GreenTRIP strategies (transit passes, carshare, bikeshare, unbundled parking) drive less. Buildings with limited parking also discourage driving. Vehicle miles traveled contribute to greenhouse gas emissions, so by reducing driving, these future residents are improving air quality, improving health outcomes for themselves and the region, and reducing traffic congestion.

The following methodology explains how Connect calculates VMT based on inputs that include location and user-determined characteristics of the building modeled in Connect. 

There are three research efforts that contribute to GreenTRIP Connect’s modeling

  1. A model that predicts vehicle miles of travel (VMT) for new multi-family developments in California

The VMT model is based on a working paper produced by CNT.  The California Strategic Growth Council commissioned an academic review of CNT’s initial paper in July 2015. The working paper was revised in response to these comments and on December 16, 2015 was posted hereThe paper was funded by the Ford Foundation and California Community Foundation, and prepared for the California Housing Partnership Corporation (CHPC). 

Like several models, GreenTRIP Connect uses indicators of location efficiency, such as level of transit service within ½ mile, as an important factor in determining VMT.  What differentiates this model most from those site-based models used is that it also is very sensitive to income variables -- low-income households drive less than moderate or high income households do, and this model captures that important distinction.  It thus clearly shows the benefits of reduced driving and associated greenhouse gas emissions as homes with more affordable rents are developed.  As explained below, this model uses detailed information from an extensive California household travel survey combined with transit service, land use, and census data. 

  1. Formulas for predicting further reduction based on GreenTRIP Strategies

The VMT prediction from the model can then be further reduced (known as “post-processing”) based on four strategies that have been shown to reduce driving.  In Connect these are called GreenTRIP Strategies and include unbundling parking costs from rent, the benefits of membership in carsharing and bikesharing, and the provision of transit passes.  More detail can be found in this paper.

  1. A parking use prediction model

The parking use prediction model is provided for the Bay Area only, as it is based on data gathered for the GreenTRIP parking database and with the methodology found in the Connect Parking Model paper. This model is similar to those produced by CNT for King County, Washington state ( and Washington, D.C. ( The work for Washington, D.C. culminated in a paper to be published in a future Transportation Research board publication, and received the best paper of the year by TRB’s Transportation and Land Development Committee (2016).  Parking data is now being gathered in the City of Los Angeles with the intention to add a parking module to cover the LA area once completed.

Note re: proposed for-sale/non-rental buildings:  The model is currently designed for inputs based on monthly rents, monthly costs of unbundled parking, and the monthly costs and benefits of GreenTRIP strategies.  To be used on “for-sale” condominium projects it is necessary to estimate what the total monthly cost would be for new residents, including the mortgage and predicting monthly building fees.  Similarly the costs to the condo-owner of purchasing an unbundled parking space along with the condo would have to be amortized into monthly payments.

Digging in further to the VMT Model

When a parcel is first selected a sample building with default number of units and rents is provided on the parcel[1].  This is done to make it easier for those people who do not have a building with specified values to propose.   This default building can be changed by adjusting the number of overall units.  You can also adjust the rents and the ratio of bedroom types by pressing the “customize units” button.  (Square footage can also be changed, but is not an input to the VMT model).

The VMT model then calculates how much residents of that building are predicted to drive two ways.

  1. On an average parcel  -- termed “average location” in the dashboard --  for that county.     This input includes all of the average location efficiency factors of the parcels in the county.  Since the amount of parking provided also has an impact on VMT (as explained below), it was necessary to make an assumption of average parking provisions.   All “average location” buildings are given 1.2 spaces per unit. (This ratio is the lower end of the Institute for Transportation Engineers guidance for parking in multiunit buildings.)
  2. On the selected parcel(s).  This input includes all of the location efficiency factors at the selected location. It uses the “total spaces” as the amount of parking provided. This number is automatically generated, and can be manually changed.  Lower parking provision results in lower VMT.

The VMT model uses four primary variables to predict household travel.

  1. Income

 Since the model is sensitive to income, we need a method to anticipate the income of a building’s future residents.  The proposed rents are used as a proxy, with rents equaling 33% of future household income.   Rents are divided into six categories, including the three HUD thresholds for affordable homes, extremely low income, very low income and low income.  In addition we used In addition we used the Moderate income as defined by California’s Department of Housing and Community Development (HCD) of 80-120% and then Middle Income of 120%-150%. Households earning more than 150% of Area Median Income are all considered the sixth category: High Income. Thus, the changes in the VMT prediction take place when rents pass one of the thresholds for these income categories.  This is true when putting in dedicated affordable homes, as well as when the market rate rents are changed (using the “customize units”) button.[2]

  1. Location Efficiency

After looking at a host of location variables, a combination of three was found to have the most powerful predictive ability.   These three variables can be seen by pressing on Connect’s map layers and include

  1. Employment Density -- The density of jobs within a half mile of the household was found to be a strong indicator for what is often referred to as a diversity of uses as well as for density.[3]
  2. Transit Availability -- This is the number of transit vehicles runs (in each direction) stopping within a half mile of the household in a typical week.
  3. Neighborhood Commute Distance -- The weighted average of the median commute distance by census block groups found  within a half mile of the household.
  1. Household Makeup

The members of the household naturally affect driving, so the model is responsive to the presence and number of the following list of household member types. These are assigned to the building by unit type and income using the county averages from the American Community Survey PUMS data.

  1. Household has a disabled member
  2. Number of adult (18-64) students in household
  3. Number of employed workers in household
  4. Number of children (0-5)
  5. Number of school-age children (6-17)
  6. Number of adults (18-64) in household
  7. Number of senior citizens (65+) in household
  1. Regional Context

Regional patterns influence auto travel, and recent California research revealed that the four major metropolitan centers demonstrated similar VMT.  This finding led to a tripartite classification in the model, with every area being classified into either:

  1. Metro region
  2. Small city
  3. Rural area

These 3 geographies can be seen in the map of California on Figure 1 of the working paper.

Additional VMT Reductions using GreenTRIP Strategies

Once Connect calculates the baseline VMT, it is possible to get additional reductions in VMT by using GreenTRIP strategies.  As in the GreenTRIP certification program, only those strategies that were considered to have sufficient data to estimate benefits were included: travel demand management strategies (TDM) like transit passes, unbundled parking, carshare and bikeshare.

TDM post-model VMT reductions

  1. Car share memberships provide a VMT reduction per participating household of 7 miles per day for the first membership, with a significantly smaller reduction for  a second membership in the household.  It is assumed that 25% of households in a given building will participate in the car share program and get those reductions).[4] 
  2. Transit passes provide a 4.5% reduction in VMT for each transit pass given to eligible residents
  3. Bike share memberships provide a VMT reduction per membership of 0.02 miles per day.
  4. Parking Policies

There are two ways to reduce projected VMT based on parking in Connect. The first is unbundling the cost of parking from the rent (i.e. charging for parking separately).  The second is reducing parking provision under 1.2 spaces per household. The calculations used in Connect follow the general principles outlined in the report CAPCOA Quantifying Greenhouse Gas Mitigation Measures (2010).   This is essentially the same report that forms the basis of CalEEMod and some other models.  As in CalEEMod, parking policies combined provide up to a 20% reduction in VMT.  

For more information on the TDM reductions on VMT, please read this report.

The Parking Prediction Model only applies in the San Francisco Bay Area, and it is based on the GreenTRIP Parking Database sites. The calculations include the following building variables:

  1. Parking supply
  2. Average rent
  3. Parking price
  4. Average bedrooms per unit
  5. Transit passes
  6. Extremely-Low Income affordable units
  7. Very Low Income affordable units
  8. Carshare memberships
  9. Neighborhood variables include: Block size (walkability), Job density within a 30 minutes transit ride, and Transit Connectivity Index - TCI (frequency of transit).

More information and details on the Parking Model are available here. TransForm is working with the city of Los Angeles collecting data to support an LA Parking Database in the future, with an enhanced GreenTRIP Connect parking model for LA once the Parking Database is in place.

[1] In order to get started the GreenTRIP Connect Tool populates the building inputs with a typical multi-unit building in the region using the 2014 American Community Survey Microdata from the US Census (or Public Use Microdata Sample -- PUMS-- data) The first thing determined is the number of units in the building by using the average size of all multi-unit buildings in the county from the PUMS data. Then using this building size, the PUMS data is examined at the regional level to determine the mix of units by number of bedrooms.

 For each unit type (Studio, 1 Bedroom, 2 Bedroom, and 3+ Bedroom) the average number of people living in each is calculated from the county PUMS data. Using the household size the income for such a household is calculated using the HUD definition for 120% of area median income (AMI) household. This income is then used to estimate a starting value for rent at 33% of monthly income. Thus each unit type’s rent is estimated so that the likely tenants would be the average household size for the given number of bedrooms and whose income is slightly more than 120% of AMI (assuming the usual 30% of income going to rent).

[2] The impact of income can be seen in Figure 2 (page 34) in the CNT Paper.  It shows how in Metro Regions and small cities a larger share of VMT change (compared to moderate income) is attributable to income. 

[3]Employment density is calculated by buffering a household with a half-mile radius and including jobs from the underlying US Census Block Groups based on the proportion of each block group covered by that buffer.  The underlying job data come from the Longitudinal Employer-Household Dynamics (LEHD) data provided by the US Census Bureau. 

[4] This participation level is based on information from developers that have worked with GreenTRIP and may vary significantly based on the proximity of pods, the availability of parking and nearby transit, and how much marketing is done.