Travel Forecasting Assumptions '98 Summary

1998 Update of Regional Transportation Plan

Planning Section
Metropolitan Transportation Commission
101 Eighth Street
Oakland, California, 94607-4700
(510) 817.5700
/datamart/forecast/assume98.htm

August 1998

Return to MTC DataMart: Travel Models & Forecasts

Report Outline

Pricing Assumptions;
Travel Behavior Assumptions;
Demographic Assumptions; and
Network Assumptions.

List of Tables and Figures

Table 1: Peak Period Parking Costs
Table 2: Off-Peak Period Parking Costs
Table 3: Historical and Projected Auto Operating Costs
Table 4: Impact of Inflation on Bay Bridge Tolls, 1975-2020
Table 5: History of Transit Fares in the Bay Area, 1970-1996 (Not Available in HTML format)
Table 6.1: "Old-Style" Regional Highway Peaking Factors
Table 6.2: "New-Style" Regional Highway Peaking Factors
Table 7: AM Peak Period Calibration (Peak Spreading) Factors (Not Available in HTML format)
Table 8: Regional Work and Non-Work Trip Vehicle Occupancies: Historical and Projected
Table 9: Ratio of Gas Prices in San Francisco and Los Angeles

This report documents the travel forecasting assumptions for the 1998 update of the Regional Transportation Plan (RTP98). In preparing these travel forecasts, MTC uses four basic sets of assumptions:

Pricing Assumptions;
Travel Behavior Assumptions;
Demographic Assumptions; and
Network Assumptions.

Demographic and network definition assumptions are not included in this memo. The basic demographic assumption is that the RTP98 travel forecasts will be based on the socio-economic/land use forecast series Projections '98 developed by the Association of Bay Area Governments (ABAG).

Pricing assumptions include projected parking prices; gasoline and non-gasoline auto operating costs; fuel economy; bridge tolls; and transit fares.

Travel behavior assumptions include trip peaking factors; vehicle occupancy factors; and estimates of interregional commuters.

I. Pricing Assumptions

A. Parking Costs

The MTC demand models were estimated using nominal, or posted parking prices as opposed to actual parking prices. Actual parking prices would be the average parking price paid by a consumer, weighted by those who are subsidized by their employer and those who are not subsidized by their employer. For peak period parking cost, the monthly posted parking price is divided by 22 days per month to derive an average workday parking cost. The average workday parking cost is then divided by 8 hours to derive an average peak hour parking cost per hour in 1990 cents. In the home-based work mode choice model application, the per hour charge is multiplied by 8 hours, then divided by 2, to derive a per vehicle trip charge. Next, the per vehicle trip charge is divided by the vehicle occupancy so that parking costs are equally distributed between vehicle drivers and passengers.

Base year 1990 and 1998, and forecast years 2000, 2010 and 2020 peak hour parking costs, by the MTC 1099 zone system, are shown in Table 1. Off-peak per hour parking costs - 1990, 1998, 2000, 2010 and 2020 - are shown in Table 2.

The MTC assumption for parking costs is that they will increase, in real terms, between one and two percent per year between 1990 and 2020. The core of downtown Oakland, Berkeley and San Jose are assumed to grow by two percent per year between 1990 and 2020; in all other areas, by one percent. In addition, Palo Alto and Stanford, which were assumed to have free peak and off-peak parking in 1990, are assumed to have per hour parking charges roughly equivalent to downtown Berkeley in future year forecasts.

MTC staff periodically inventory parking garages throughout the Bay Area to monitor trends in parking prices. MTC staff is updating this parking cost inventory in summer 1998. Future parking cost forecast assumptions will be revisited based on this updated inventory.

B. Auto Operating Costs

The MTC travel demand models are based on non-linear auto operating costs which vary according to trip speed and distance. As speed increases, the fuel consumption rate (gallons per mile) decreases linearly. As distance increases, the share of "cold start" fuel consumption decreases. This internal model is used to derive trip-specific fuel economy (miles per gallon) which is multiplied by the per gallon gas price to derive per trip gasoline operating cost. A constant non-gasoline operating cost per mile is multiplied by trip distance to get per trip non-gas cost. Total auto operating cost per trip is the sum of the gasoline cost per trip plus the non-gasoline cost per trip plus any bridge tolls or parking charges. (Details on the auto operating cost model will be included in the BAYCAST Users Guide.)

The MTC auto operating cost model is based on work conducted by Cambridge Systematics, Inc., as part of the Urban Transportation Energy Conservation study, published in 1978 (known as "UTEC"). (The UTEC models were also used to derive auto operating costs for the Southern California Association of Governments new set of travel demand models.)

The basic inputs to the BAYCAST model system, in terms of auto operating cost, are gasoline price (in 1990 constant dollars); the fuel correction factor (to represent fleet turnover and more fuel efficient vehicles); and the non-gasoline operating cost (in 1990 cents per mile.) Data on historical, 1990 to 1998, and assumed future year auto operating costs are detailed in Table 3 and Figures 1 and 2.

The notes to Table 3 indicate some of the major assumptions going into these auto operating cost forecasts. For gasoline prices, MTC uses future gas price estimates provided by the California Energy Commission (CEC). The projected year 2020 average gas price, as developed by the California Energy Commission, is $1.384 per gallon (retail price) in 1998 current dollars.

MTC is assuming no change in fuel economy relative to 1990. This respects the overall fuel economy trend as established by the US Energy Information Agency (EIA) in their "Household Vehicles Energy Consumption Report" (September 1997.) The EIA found no significant increase in overall passenger vehicle fuel economy between their national surveys conducted in 1988 and 1994. Overall this means that we are projecting that overall auto operating cost per mile will decrease from 8.81 cents per mile in 1998 to 8.40 cents per mile in the year 2020 (all in 1990 constant dollars).

A question was raised about the differential gas prices in the San Francisco versus Los Angeles regions. Table 9 shows the ratio of San Francisco to Los Angeles gas prices between January 1995 and October 1997. Over this time period, San Francisco gas prices have been, on average, two percent higher than Los Angeles gas prices. This is not a significant difference, so the recommendation is to use the CEC statewide gas price forecast unadjusted for Bay Area price differential.

The other key assumption is that non-gasoline operating cost (maintenance and repair, motor oil, parts, accessories) is 40 percent of total auto operating costs. This 40 percent figure is based on US Bureau of Labor Statistics data on consumer expenditures (see Table 4 of the MTC report: Consumer Price Indices: Bay Area & U.S. Cities: 1950-1997.) In a typical household, between five and six percent of a household's expenditures are related to auto operating costs. Gasoline cost has fluctuated from 55.6 percent to 73.5 percent of total auto operating costs over the past twenty years.

(Auto ownership costs, which comprise around 10.2 percent of the average household's budget, are not used in determining trip running, or variable costs. These fixed costs of auto ownership are more important in determining the number and quality of vehicles to own or lease. Given the difficulty in projecting automobile quality and costs, household income is used as a surrogate in predicting auto ownership levels.)

C. Bridge Tolls

The Bay Area bridge tolls are expected to drop from $2.00 to $1.00 in the year 2008 (Table 4, Figure 3). Given an inflation assumption of 4 percent per year, a year 2020 toll of $1.00 is equivalent to approximately 33.7 cents in 1990 constant dollars. This MTC bridge toll assumption is consistent with the financial forecasting assumptions used in projecting bridge toll revenues.

All Bay Area bridges had a standard automobile toll of $1.00 per crossing in 1990. Commute ticket booklets offer 15 to 32 percent discounts off of the $1.00 toll, as follows:

1990 Base Year Bridge Tolls

Bay Area Bridges	Auto Toll	Commute Tickets	Commuter Toll ($/ticket)	Free Toll for SR3+ During Peak Period?
Antioch	$1.00	$27 / 40 tickets	$0.68	No
Benicia/Martinez	$1.00	$27 / 40 tickets	$0.68	No
Carquinez	$1.00	$27 / 40 tickets	$0.68	No
Richmond/San Rafael	$1.00	$34 / 40 tickets	$0.85	Yes (since 10/89)
Golden Gate	$1.00	$20 / 23 tickets	$0.87	Yes
SF/Oakland Bay	$1.00	$34 / 40 tickets	$0.85	Yes
San Mateo/Hayward	$1.00	$34 / 40 tickets	$0.85	Yes
Dumbarton	$1.00	$34 / 40 tickets	$0.85	Yes

For the state-owned bridges for FY 1989/90, MTC staff calculated an average auto toll weighted on commuter ticket usage and full toll usage, as follows:

Computation of Average Auto Toll

Bay Area Bridges	Commuter Tickets	Total Autos & Trailers	Tickets as % of Total	Average Auto Toll
Antioch	225,569	1,605,516	14%	$0.96
Benicia/Martinez	3,696,160	13,643,902	27%	$0.91
Carquinez	4,724,623	17,585,673	27%	$0.91
Richmond/San Rafael	1,257,179	8,428,199	15%	$0.95
SF/Oakland Bay	4,227,393	36,521,920	12%	$0.96
San Mateo/Hayward	1,845,246	12,131,171	15%	$0.95
Dumbarton	2,085,757	8,381,841	25%	$0.92

The average toll for the Golden Gate Bridge was 94 cents per revenue vehicle between July and December 1990 (source: Golden Gate Bridge District. Comparative Record of Traffic for the Month of December 1990).

For purposes of travel forecasting, the one-way toll is halved so that both directions on every bridge are allocated one-half of the total average toll. This is a technical necessity to counter the toll collection direction bias.

Note that free tolls for three-or-more person carpools were instituted on the Carquinez Strait bridges (Carquinez, Benicia/Martinez and Antioch) in October 1995. This is the only change in toll assumptions from the 1990 base year. The final tolls used in the 1990 model simulation are as follows:

Bridge Tolls for Travel Forecasting: Base / Future Years

Bay Area Bridges	Drive Alone & Carpool-2	3+ Carpool	Off-Peak Tolls
Antioch	$0.48	$0.48 / $0.00	$0.48
Benicia/Martinez	$0.46	$0.46 / $0.00	$0.46
Carquinez	$0.48	$0.48 / $0.00	$0.48
Richmond/San Rafael	$0.48	$0.00	$0.48
Golden Gate	$0.47	$0.00	$0.47
SF/Oakland Bay	$0.48	$0.00	$0.48
San Mateo/Hayward	$0.48	$0.00	$0.48
Dumbarton	$0.46	$0.00	$0.46

D. Transit Fares

The tradition with transit fare forecasts is to assume that they keep pace with inflation.

Base and top end transit fares by Bay Area transit operator, 1970 to 1998, are shown in Table 5. Historical and projected base fares are charted in Figure 4.1 (Muni), Figure 4.2 (AC Transit), and Figure 4.3 (BART). These charts show base transit fares in current and 1990 constant dollars. The current dollar fares are based on a four percent per year increase in consumer price indices.

Transit operator fares were revised to incorporate recent (1990-1998) fare changes. The most notable increase is in BART fares. The current BART base fare of $1.10 is equivalent to 86 cents in 1990 dollars, or a 7.5 percent real increase in the BART base fare (80 cents in 1990 to 86 cents in 1998). The current BART top fare of $4.70 is equivalent to $3.67 cents in 1990 dollars, or a 22.3 percent real increase in the top fare ($3.00 in 1990 increasing to $3.67 in 1998.)

MTC will use all transit operators fares as of 1990, except for BART, where we will use 1998 real fares deflated to 1990 dollars.

II. Travel Behavior Assumptions

A. Vehicle Peaking Factors

The new MTC BAYCAST model system is oriented to the production of daily and AM peak period traffic assignments. PM peak period traffic assignments may also be produced from the BAYCAST model system since the basic output of the demand models are daily trips by trip purpose and travel mode. In addition, the user can factor the two-hour peak period vehicle trip tables to peak hour tables using peak hour-to-peak period factors by trip purpose.

In contrast to the old MTCFCAST model system, the new BAYCAST system directly simulates the number of AM peak period home-to-work vehicle trips, derived from the home-to-work departure time choice model. This is basically a "peak spreading" model which will predict fewer trips in the peak period when congestion levels increase. The standard approach of using fixed shares for all other trip purposes is still needed to augment this new departure time choice model.

Old-style (MTCFCAST) AM and PM peak hour vehicle peaking factors are shown in Table 6.1. New-style (BAYCAST) AM and PM peak period vehicle peaking factors are shown in Table 6.2. The AM peak period is defined as 700-900 AM. The PM peak period is defined as 400-600 PM.

As a part of the peak period traffic assignment calibration and validation process, a set of peak period calibration factors were developed. These calibration factors, documented in Table 7, reflect the subregional variation from the regional peaking factors shown in Table 6.2.

Data from the 1990 household travel survey show that the AM peak hour (0730-0830) is 58 percent of total vehicle trips occurring in the AM peak period (0700-0900) (930,038 vehicle trips / 1,610,546 vehicle trips, from Survey Working Paper #4, page 160, Table 2.3.7A.) So, a rough rule of thumb is to multiply any AM peak (two-hour) period traffic assignment by 0.58 to get a rough estimate of peak hour predicted traffic volumes.

B. Vehicle Occupancy Factors

In the old MTC model system, vehicle occupancy assumptions were important input assumptions to the home-based shop, home-based social/recreation and the non-home-based mode choice model system. These vehicle occupancy assumptions were used, and are still used, for dividing the vehicle trip cost between vehicle drivers and passengers.

All of the new mode choice models either split the number of person trips by vehicle occupancy level (i.e., drive alone, shared ride 2, shared ride 3+), or they split the in-vehicle person trips by vehicle driver and vehicle passenger modes. So, the forecasting system is in a sort of paradox: vehicle occupancy is both an input assumption as well as a forecasting output! The issue in auto occupancy forecasting is to ensure that the input occupancy assumption is reasonably consistent with the forecasting output vehicle occupancy rate.

Historical vehicle occupancy rates, from MTC household travel surveys, and BAYCAST predicted vehicle rates for 1990 and 2015, are shown in Table 8. The 1990, 2015 and 2020 model-simulated vehicle occupancy rates are based on MTC model validation and forecasting work completed between 1997 and 1998.

For the home-based work, home-based shop and home-based social/recreation mode choice models, trips are split by occupancy level (DA, SR2, SR3+). For the three home-based school mode choice models and non-home-based trips, person trips are split into vehicle driver and vehicle passenger. For home-based grade school trips, vehicle driver is not an available mode. This means that the vehicle driver trip for escorting children to school is typically included as a home-based shop/other shared ride 2 or shared ride 3+ trip; the vehicle passenger (the child) is classified as a home-based grade school vehicle passenger trip.

This is clumsy and confusing, but reflects the nature of travel: where persons in a particular vehicle may be traveling to different activities. The parent's trip purpose is to escort the child to school (home-based shop/other); the child's trip purpose is to attend school (home-based school).

Historical and projected vehicle occupancy factors are shown in Table 8. Note that these are not assumptions per se but model simulations.

C. Interregional Commuters

Assumptions about the number of interregional commuters is key in two respects: first, intraregional home-based work productions and attractions need to be adjusted to reflect in-commuting and out-commuting from and to Bay Area jobs and households; second, interregional vehicle trips are needed to augment the intraregional trips included in the standard BAYCAST travel demand models.

Interregional commuters are estimated by factoring the 1990 Census journey-to-work data file (STP214) using a 46-by-46 matrix that comprises the 34 Bay Area superdistricts and the 12 Bay Area neighbor counties. The factored year 2020 interregional commuter matrix is used as the basis for estimating background interregional daily and peak period vehicle trips. This is basically a "sketch planning" effort to complement the formal models used to predict intraregional personal and intraregional commercial travel.

III. Demographic Assumptions

MTC uses ABAG's Projections '98 census-tract level forecasts for the year 2020 as the horizon year in the 1998 update of the Regional Transportation Plan. MTC combines and allocates these tract level forecasts to MTC's 1099 regional travel analysis zone system.

IV. Network Assumptions

A major part of the RTP update is the definition, coding and simulation of a variety of network alternatives. Alternative definitions are needed for each study alternative, for each of the three types of networks being created: highway networks, transit networks, and pedestrian/bicycle networks. Definition of network alternatives will be described in the Regional Transportation Plan.