MTC DataMart: Changes in Regional Travel Characteristics: 1960-1990

Changes in Regional Travel Characteristics and Travel Time Expenditures in the San Francisco Bay Area: 1960 - 1990

by:
Charles L. Purvis

Metropolitan Transportation Commission
101 Eighth Street
Oakland, California 94607

December 1993
Revised, May 1994

Paper submitted to the Transportation Research Board for Presentation at the 73rd Annual Meeting
January 9-13, 1994
Washington, D.C.

Paper Outline

This research is an update of a 1984 study by Kollo and Purvis. Results from the 1990 household travel survey conducted in the San Francisco Bay Area are compared to results from surveys conducted in 1965 and 1981, and to decennial census data. The study shows a decline in trip frequency per household and per person between 1981 and 1990, which is offset by an increase in average trip duration, yielding an apparent constant travel time expenditure per person and per household. Regularities in average travel time expended per household vehicle are also analyzed. Changes in Bay Area demographic characteristics, 1960 to 1990, are described to provide context to the changes in aggregate travel characteristics. Changes in household trip rates, by market segment, and by trip purpose and travel mode, are also summarized. Findings show a decline in home-based non-work and non-home-based trip rates per household, and increases in home-based work trips per household. Bay Area results are compared to household travel surveys from other metropolitan areas.

This research project is an update to the 1984 study by Kollo and Purvis (1). The authors' 1984 study compared results from the 1965 and the 1981 household travel surveys conducted in the nine county San Francisco Bay Area. Comparison was also made to journey to work characteristics from the 1960, 1970 and 1980 U.S. decennial censuses. This study updates the information provided in the 1984 paper to provide results from the 1990 Bay Area household travel survey, and includes new information related to changes and regularities in travel time expenditures of Bay Area households based on analysis of the 1965, 1981 and 1990 household travel surveys.

There is a basic need for a careful and comparative review of results from metropolitan area household travel surveys to detect survey strengths and weaknesses, devise strategies and methods for correcting problems and biases, and for planning strategies for estimation of new sets of regional travel demand forecasting models. A thoroughly structured travel survey analysis project related to cleaning, editing, weighting, expanding, linking trips, and flushing out survey idiosyncrasies and data outliers, is a critical precursor to travel demand model development activities. This study is but one element of the 1990 household travel survey analysis project.

This research also adds to a growing genre of literature related to comparative aggregate analyses of metropolitan travel characteristics. Most of this research is related to the temporal stability or regularities of travel characteristics, typically focusing on the basic presumption of constancy of trip generation, trip distribution and mode choice model coefficients. Selected studies of this genre include Kannel and Heathington's (2) study of Indianapolis travel characteristics based on household surveys conducted in 1964 and 1971 ; Yunker's (3) Milwaukee 1963 and 1972 survey analysis; Smith and Cleveland's (4) analysis of the 1953 and 1965 Detroit surveys; Cohen and Kocis' (5) Buffalo and Rochester study; the aforementioned study by Kollo and Purvis (1) for the San Francisco Bay Area; Norris and Shunk's (6) analysis of 1964 and 1984 travel characteristics in the Dallas region; and Walker and Peng's (7) study of Philadelphia region changes between 1960 and 1988. Other collections of results related to metropolitan area household travel surveys include an ITE Committee Report (8) from 1979; and the Characteristics of Urban Transportation Demand (CUTD) (9) manual published by the Urban Mass Transportation Administration.

Findings from this aggregate analysis of Bay Area household travel surveys generally supports theories related to travel time budget research conducted between 1961 and 1985. On the other hand, analyses of travel time expenditures by market segment (e.g., household size, vehicle availability) shows notable instability and irregularities in travel time expenditures. The vast research heritage related to travel time budgets includes early works by Tanner (10); and numerous efforts by Yacov Zahavi (11, 12, 13, 14); Zahavi and Ryan (15); Zahavi and Talvitie (16); and Zahavi, Beckmann and Golob (17).

Interest in travel time budget research apparently peaked in the late 1970s and early 1980s, culminating in a one-day conference on personal travel budgets held at the University of Leeds in the UK in May, 1979 (see the special issue of Transportation Research A,volume 15A, number 1, published January 1981). Precious few research related to transportation travel time or money budgets has appeared in the professional literature after 1985, perhaps due to the passing of a principal proponent of travel budget models, Yacov Zahavi, in the early 1980s, or perhaps due to lack of research material (or research budgets) for the continuing analysis of travel time budgets. Information included in this study may help to rekindle interest in travel budget and travel time expenditure research.

The Bay Area household travel surveys for 1981 and 1990 are apparently showing a real decline in trip frequency per household and per person. This is offset by a real increase in average trip duration (average trip time), yielding, on an aggregate basis, an apparent constant travel time expenditure of approximately 2.7 person hours of travel per household per weekday and 1.0 person hours of travel per person per weekday. This finding of an inverse relationship between trip frequency and trip duration is consistent with much of Zahavi's analyses conducted in the 1970s.

The remainder of this paper discusses comparability issues related to the 1965, 1981 and 1990 household travel surveys; changes in Bay Area demographic and economic characteristics between 1960 and 1990; changes in travel time expenditures and related characteristics between 1965 and 1990; changes in regional household trip rates, 1965 to 1990; and changes in Bay Area aggregate travel characteristics, 1965 to 1990.

In any comparative analysis of household travel surveys it is advisable to provide the reader with information to help in understanding similarities and dissimilarities related to survey design, sample design, and survey analysis methods. Fortunately, the Bay Area household travel surveys of 1965, 1981 and 1990 are quite similar in design and analysis, and excellent documentation of all three surveys has been developed. This analysis required re-visiting the 1965 and 1981 survey files, especially in terms of calculating mobile persons, mobile households and travellers versus non-travellers.

The study area for the three Bay Area travel surveys has remained constant, including the same nine counties and the entire region of 6,900 square miles (17,900 square kilometers). Other regions, such as Atlanta, Buffalo, Rochester, and Philadelphia (5,7) have increased their study area size between survey years and require careful analysis to ensure appropriate comparisons between comparable areas.

The 1965 household travel survey, conducted by the Bay Area Transportation Study Commission (BATSC), was a home interview (face-to-face, in-person) survey of 20,486 households as to their weekday travel behavior. An additional 10,200 households were queried as to their average weekend daily travel behavior.

The 1981 household travel survey, conducted by the Metropolitan Transportation Commission (MTC), was a telephone survey of 6,209 households for weekday travel and an additional 882 households for weekend travel. Detailed survey methodology is included in the report by Crain and Associates (18) and in Reynolds, Flynn and Reinke (19).

The 1990 household travel survey, conducted by the MTC during the spring and fall of 1990, was also a telephone survey of over 10,800 households for weekday travel behavior. Of the 10,838 usable household samples collected by MTC, 9,438 households provided single weekday daily travel diaries, and 1,486 sample households provided either three-day or five-day (weekday) travel diaries. The survey results reported here represent the single day sample only, not the multiple weekday sample. Detailed survey methodology for the 1990 MTC travel survey is included in reference (20).

All three surveys were administered to all persons in households age five-and-older. All three surveys collected basic household information (household income, vehicle availability, length of residence, structure type, owner/renter tenure); data on each person (age, sex, race/ethnicity, relationship to head of household, employment status); and trip data (detailed means of transportation, origin and destination location and trip purposes, trip start time, trip finish time, vehicle occupancy). Certain "households" in all three surveys were actually group quarters units (boarding houses, fraternities, convents, prisons, etc.) and were excluded from all three sets of analyses. The analysis for all three surveys is of weekday, intraregional personal travel made by residents (age five-and-over) of Bay Area households. Therefore, the analysis excludes the following travel sub-markets: interregional travel made by Bay Area residents; travel made by non-residents (visitors and commuters); travel made by persons living in group quarters; and commercial travel.

Trips reported in this analysis are based on linked trip records. Trip linking procedures for the 1965 home interview survey (21), the 1981 telephone survey (22) and the 1990 telephone survey (23) are quite comparable. Trip linking is a technical necessity to remove incidental stops such as changing travel modes (e.g., walk to bus, drive to rail) and serving passengers (e.g., dropping off kids or spouse on way to work, picking up carpool passengers). Mode of access and mode of egress trip leg information is retained in special "extended" versions of the linked trip files for future work in estimating mode of access to transit sub-mode choice models. It is critical in a comparative survey analysis to identify if unlinked or linked trips are used. The Philadelphia and Dallas studies clearly indicated the use of linked trips in their analysis, and also provided a general description of trip linking procedures.

Weighting and expansion procedures were different for the three surveys. The 1965 survey was expanded to "backcast" estimates of households by single family/multiple family breakdown by 290 regional travel analysis zones of residence (24). The 1981 survey was expanded to the 1980 Census count of households by household size by 45 districts of residence (25). The 1981 survey weighting method reflected the fact that one-half of the 6,200 household samples were from the City of San Francisco. The 1990 survey was expanded to the 1990 Census count of households by household size, by owner/renter tenure, by auto ownership level, by 34 districts of residence (26). All of the results reported in this analysis reflect weighted survey results.

Reference (24) on the 1965 travel survey discusses screenline adjustment factors to account for trip underreporting in the travel diaries. These adjustment factors were calculated by county, by three general trip purposes (home-based work, home-based non-work, and non-home-based), and were applied only to the in-vehicle trips (not the transit trips). Screenline adjustment factors ranged from three to ten percent for home-based work in-vehicle trips, and from five to twenty-five percent for non-work in-vehicle trips. No screenline adjustment factors were required for the 1981 survey analysis, and the issue of screenline adjustment factors for the 1990 survey will be addressed in future MTC analyses. The results reported in this study are for reported travel characteristics before any screenline adjustment factors were applied.

Adjustment of household travel surveys to account for underreporting is discussed in Clarke, Dix and Jones (27) and Barnard (28). Clarke's summaries of U.S. and U.K. household travel surveys shows expanded survey data at 79 percent to 99 percent of screenline counts. Barnard's analysis of Australian household travel surveys shows expanded survey data at 55 to 95 percent of screenline counts. Both Clarke and Barnard note the differential underreporting by trip purpose and travel mode, with non-work vehicle trips the most likely trip market to be underreported, with transit trips the least likely to be underreported. The basic conclusion of Clarke and Barnard is that the standard household travel survey tends to underreport household travel on the order of 10 to 15 percent.

Special imputation procedures were required in the 1990 travel survey analysis to account for persons who refused to answer the travel diary portion of the survey. MTC accepted, as complete, household samples where a majority of the adult (age 16 and over) members of the household provided trip diary information. Of the 24,300 persons interviewed in the 1990 survey, 1,100 refused to provide trip diary data. Of the 1,100 refusals, 500 were between the ages of five and fifteen. The higher share of diary refusals from children may have been due to the children's lack of interest in participating in the 1990 travel survey, or, as the MTC survey consultant reported, was due to parents refusing to divulge the whereabouts of their children because of fear of MTC (or the consultant) kidnapping their children.

The imputation process, basically a modified "hot deck" procedure, used the age, sex, employment status (student, full-time employee, etc.) and county of residence of the survey respondents to impute, i.e., "make up" trips for the trip diary refusers. This basically assumes that persons who refused the trip diary were actually mobile (made trips) on the assigned travel day, and that the trip frequency characteristics of a, say, eight year old boy is similar to the trip frequency characteristics of other eight year old boys. The imputation procedure only imputed trip frequency characteristics, not origin/destination data such as origin zone or destination zone.

Thirty percent of the households in the 1990 survey refused to provide information on household income. Hot deck imputation procedures using county of residence, owner/renter tenure, workers in household, and household size, were designed and applied to salvage these survey records.

A summary of key regional, aggregate demographic and economic indicators is provided in Table 1. These characteristics provide a context for later discussions on changes in travel time expenditures, trip frequency, and aggregate travel characteristics. As appropriate, weighted and expanded household travel surveys are compared to data from the respective census, or, as in the case of the 1965 survey, to independent demographic backcasts prepared by the Association of Bay Area Governments, the council of governments for the region.

The nine county San Francisco Bay Area is a large metropolitan region in Northern California with over six million persons residing in an area of over 6,900 square miles (17,900 square kilometers). Bay Area total population increased by 16.3 percent between 1980 and 1990. The number of households increased by 14.0 percent between 1980 and 1990, and total workers residing in the Bay Area increased by 23.4 percent in the 1980s. The recent upswing in regional average household size between 1980 and 1990 (2.56 to 2.61 persons per household) was the first census since 1960 where Bay Area household size has shown an increase, not a decrease with respect to the previous census year.

Growth in personal vehicle availability (+19.1 percent, 1980 to 1990) has outpaced growth in total population. Share of households owning zero vehicles has declined from 20 percent of all households in 1960, to 10 percent of all households by the year 1990. Communities with the highest shares of zero vehicle households are San Francisco (30.7 percent of households with zero vehicles in 1990), Oakland (23.3 percent) and Berkeley (19.0 percent). Vehicles per licensed driver is apparently approaching one vehicle available per driver, though state Department of Motor Vehicle records indicate that the actual number of drivers per Bay Area 1990 household is on the order of 1.87 drivers per household (contrasting to 1.76 vehicles per household).

Census data indicates a gradual decline in Bay Area total kindergarten through grade 12 school enrollment between 1970 and 1990. On the other hand, college enrollments have increased steadily between 1960 and 1990.

Regional mean household income has increased 11 percent in 1989 constant dollar terms between 1970 and 1980, and increased 17.9 percent between 1980 and 1990. Mean household income for households in the 1981 survey is lower than the mean income from the 1980 Census. In contrast, the 1990 survey reported incomes are slightly higher than the 1990 Census.

Key summary statistics that are reviewed as results are obtained from weighted, linked trip files are the total count of expanded trips, and trip rates per household and per person. Sample expansion and trip linking for the 1990 MTC household survey were completed in spring 1993. Soon thereafter, the unpleasant reality of a major (-13.3 percent) decline in total trips per household and per person revealed potentially embarrassing results, i.e., an absolute decline between 1980 and 1990 in the total number of trips made by Bay Area residents.

As shown in Table 2, total trips per household (all purposes and means of transportation) gradually declined from 8.78 trips per household in 1965 to 8.71 trips per household by 1981, and then dropped to 7.55 trips per household in 1990. Trips per capita (total persons in household) increased from 2.81 trips per person in 1965 to 3.39 by the year 1981, then dropped back to 2.93 trips per capita by 1990. All trip rates are expressed in trips per weekday.

One of the first reactions was: how does the Bay Area compare to other areas conducting travel surveys in the early 1990s? Results were compared to the Nationwide Personal Transportation Survey (29), Los Angeles (30), Sacramento (31), California (32), Dallas (6), Philadelphia (7), and other U.S. metropolitan areas (9). Other metropolitan areas, namely Los Angeles, Dallas and Denver, showed modest declines in trips per household comparing their 1960s and 1970s travel surveys to their 1980s and 1990s travel surveys. Only the San Francisco and Los Angeles regions appear to be showing declines in trip rates per person. If taken alone and not compared to the earlier 1965 or 1981 travel surveys, results from the 1990 Bay Area travel survey appear generally in line with other metropolitan areas.

While the evidence from Los Angeles and other metropolitan areas suggests that the Bay Area is not unique in terms of declining trip rates, this predicament of dropping trip rates has discomforting implications for the stability of trip generation model parameters. Survey-based, expanded "person" trips (mechanized modes only: vehicle driver, vehicle passenger, transit passenger) were compared to a recently completed year 1990 model simulation, using the 1981 survey-based travel demand models. Survey home-based work (HBW) person trips, shown in Table 3, were within one percent of model-simulated home-based work persons trips. This was encouraging. On the other hand, the survey-based home-based shop (other) (HBSHO) person trips were 20 percent less than the model-simulated trips; survey-based home-based social/recreation (HBSR) trips were 39 percent less than model-simulated trips; and non-home-based (NHB) person trips were 20 percent less than model-simulated person trips. This was discouraging. The non-work trip generation models basically responded to increasing household sizes, increasing auto ownership levels, and increasing real household income levels. Standard non-work trip generation models would only show an ever-increasing trip frequency based on these situations, and could not respond to the shifts in travel behavior that apparently occurred in the Bay Area between 1980 and 1990. Norris and Shunk's (6) comparative analysis also noted declines in home-based non-work household trip rates in San Francisco, Dallas, Denver and Atlanta.

The survey and model results for San Francisco and elsewhere indicate structural problems with non-work trip generation models. The results suggest the need for a better linkage between non-work trip generation models and work trip distribution models (i.e., total work trip duration). One hypothesis to be advanced and tested is that increases in work trip duration in a household are linked to lower non-work trip generation rates.

Another initial concern that warranted further analysis was the potential problem of survey respondents falsely claiming that they did not travel during the assigned travel day, basically to avoid the hassle of filling out trip diaries. The term "mobile" is used to denote persons or households who reported travel - by any means of transportation, including walk or bicycle - during their assigned travel day. The mobile share of population, by age of respondent, is reported for the 1981 and 1990 San Francisco Bay Area surveys, and compared to Wigan's (33) analysis of the 1981 Sydney, 1978 Melbourne and the 1977 Adelaide, Australia surveys (Table 4). The mobile share patterns of the Bay Area travel surveys are quite similar to the Australian metropolitan areas, averaging 82 percent mobile (18 percent immobile) for the two Bay Area surveys and 78 percent mobile (Sydney), 85 percent mobile (Melbourne) and 87 percent mobile (Adelaide) in Australia. Children ages five to eleven show the highest mobility share, ranging from 86 to 89 percent mobile in San Francisco, and from 86 percent in Sydney and 96 percent in Adelaide. Elderly persons age 65-and-over show the lowest mobility share, ranging from 60 to 65 percent in the Bay Area, and from 56 percent in Melbourne to 63 percent in Adelaide. These results are encouraging and suggest that the 1990 Bay Area travel survey is not biased due to excessive numbers of respondents falsely claiming no travel.

The analysis then turned to a review of average trip duration. It was felt that a real drop in household trip rates could make logical sense if the drop in trip frequency was offset by an increase in average reported trip duration. The 1990 survey indicated a modest, 10.4 percent increase in average trip duration between 1981 and 1990 (19.3 to 21.3 minutes, all trip purposes and modes) which offset a 13.3 percent decline in total trips per household. The 1984 study by Kollo and Purvis did not dwell too long on the changes in trip frequency or trip duration, basically due to an insignificant decline in the total trips per household between 1965 and 1981 (8.77 to 8.71) and a subtle increase in the total average trip duration between 1965 and 1981 (18.6 to 19.3). Given the lumpy, or spiky, distributions of reported travel times, differences of less than one minute in reported trip duration are probably not significant from a planning or statistical perspective.

The average reported trip duration for home-based work trips in the 1981 survey (26.6 minutes) is about 9 percent higher than the mean travel time for commuters as reported in the 1980 Census (24.3 minutes). Average reported trip duration for 1990 survey home-based work trips (29.2 minutes) is 14 percent higher than the mean travel time for commuters according to the 1990 Census (25.6 minutes). These increasing discrepancies between survey and census reported commute travel times bears further detailed analysis at a more precise geographic level to discern biases in either or both datasets.

Much of the confusion in the travel time budget literature is with respect to the definition of terms. Goodwin (34) makes a good case that travel time budgets should be based on all households, on all persons in households, and for all travel, including non-motorized travel. Much of Zahavi's research focused either on vehicle travel or travel by motorized means (vehicle driver, vehicle passenger or transit passenger). This study analyzes total travel time expenditures per total household and for total population, as well as the more restrictive definitions related to travel time per traveller or travel time per mobile person.

For reporting purposes here, the term "mobile" is used to denote persons or households who reported travel - by any means of transportation, including walk or bicycle - during their assigned travel day. The term "traveller" is used to denote persons or households who reported motorized travel (vehicle driver, vehicle passenger or transit passenger) during the assigned travel day. The term "total trip" refers to trips made by persons, age five-and-older and residing in households, by any and all means of transportation. The term "person trip" is a more restricted definition (similar to the person trips used in travel demand forecasting models) and refers to trips made only by motorized means of transportation.

The basic input data and resulting travel time expenditures and trip frequency rates are presented in Table 6. The share of population "travelling" (i.e., making motorized trips) increased from 67 percent of the population in 1965 to 76 percent of the population in 1990. The share of households "travelling" (i.e., one or more persons in the household making motorized trips) is rather stable at 88 to 90 percent of all households. The "mobile" household share (i.e., one or more persons in the household making trips by any means of transportation) is also stable at around 91 to 94 percent of all households.

All trips rates, per person (total, mobile and travelling) and per household (total, mobile and travelling) increased between 1965 and 1981, and decreased between 1981 and 1990.

Total travel time expenditure per mobile person increased 19 percent between 1965 and 1981, from 72 minutes per mobile person per day (1.2 hours) to 86 minutes per mobile per day (1.44 hours). Total travel time per mobile decreased slightly between 1981 and 1990, from 86.3 to 82.5 minutes. Total travel time expenditure per traveller is quite similar to average travel time per mobile person. Average travel time per traveller increased 15.8 percent between 1965 and 1981, and decreased by 5.5 percent between 1981 and 1990, dropping from 84.9 minutes to 80.2 minutes.

Average travel time expenditure per mobile household increased by 8 percent between 1965 and 1981, from 173 minutes (2.88 hours) per mobile household in 1965 to 187 minutes (3.12 hours) per mobile household in 1981. Average travel time expenditure per travelling household showed a 10.7 percent increase between 1965 and 1981, and a 5.5 percent decrease between 1981 and 1990.

Average travel time expenditures per total household and per total household population are shown in Table 7. Travel time expenditure per total household was 2.7 hours per household in 1965 and 1990, and 2.8 hours per household in 1981 (Figure 1). Travel time expenditure per total persons in households increased from .86 hours per person in 1965 to 1.07 hours per person by the year 1981. Travel time expenditure per total persons in households apparently declined to 1.03 hours per person in 1990. This represents a 2.1 percent decrease in average travel time expended per household, 1965 to 1990; and a 19.8 percent increase in average travel time expended per person in household, 1965 to 1990.

The three sets of travel surveys were further stratified by auto ownership level and by household size to detect any other regularities in travel time expenditures by market segment. Average travel time expenditure per person, by auto ownership level, increased between 1965 and 1981, and generally remained constant between 1981 and 1990. The most significant changes, 1981 to 1990, is an 11 percent drop in travel time per person in zero-vehicle households (1.10 to .98 hours/person).

Average travel time per person by household size is also shown in Table 7. Average travel time per person decreases with increasing household size. This is because single person households must perform all household travel chores, whereas multi-person households can share household travel chores between household members. A single person household in 1990 spent 1.30 hours per day travelling (any means of transportation). A five-or-more person household in 1990 spent 5.06 hours per day per household, or .86 hours per person in household. The travel time expenditures per person show moderate increases between 1965 and 1981, and a general stability between 1981 and 1990. The travel time expenditure for two-person and three-person households increased between 1981 and 1990; travel time expenditures decreased for one-person, four-person and five-or-more person households over this same period of time.

Careful examination of the coefficients of variation by market segment is required to understand the statistical significance of these minor to moderate changes in mean travel time expenditures per household and per person. Errors in the reporting (and coding) of trip start and trip finish time are prone to occur in household travel surveys and can significantly affect average travel times in the aggregate and by market segment.

Changes in aggregate regional vehicles available, vehicle trips and vehicle hours of travel per household is summarized in Table 8. The surveys show a more than doubling in the number of vehicles available and vehicle hours of travel in the Bay Area between 1965 and 1990. Average vehicle trip duration decreased slightly between 1965 and 1981, from 18.4 minutes per average vehicle trip in 1965 to 18.0 minutes by the year 1981. Average vehicle trip duration increased 14 percent between 1981 and 1990, from 18.0 minutes to 20.5 minutes.

Vehicle trips per vehicle has shown a steady decrease over the three survey time periods, declining from 3.24 trips per vehicle according to the 1965 survey, to 3.08 trips per vehicle in the 1981 survey, to 2.71 trips per vehicle in the 1990 household travel survey. Vehicle hours of travel per vehicle available is rather stable at around 0.92 to 0.99 daily hours expended per vehicle. The 1990 travel survey indicated that the average vehicle was on the road approximately 0.93 hours (56 minutes) per day.

Further analysis of vehicle travel time expenditures in the Bay Area should investigate changes in average trip length, in miles, per vehicle for the three household travel surveys. A network-based evaluation of vehicle miles of travel and average trip speeds using travel survey records is needed for a careful outlier analysis to edit and correct, or delete trip records, and as a precursor step for trip distribution model development.

Changes in regional household trip rates, comparing the 1965, 1981 and 1990 Bay Area household travel surveys are summarized in Tables 9 through 13. Changes in household trip rates by trip purpose and travel mode are shown in Table 9. Total trips per household by market segment are in Table 10 (household size), Table 11 (vehicle ownership level), Table 12 (housing structure type), and Table 13 (income level).

The only trip purpose showing increasing trips per household between the 1981 and 1990 surveys is home-based work trips. The 5.3 percent increase in home-based work trips per household between 1981 and 1990 represents a 20.1 percent increase in regional, aggregate home-based work trips. This compares to a 20.0 percent increase in regional, aggregate employed residents. This simply means that work trips per worker did not change between 1981 and 1990.

Home-based non-work trips are broken down into three trip purposes: home-based shop (other), home-based social/recreation, and home-based school. Home-based shop (other) is a catchall trip purpose and includes shopping, personal business, medical/dental, unlinkable serve passenger and change travel mode purposes, etc. Home-based social/recreation trips include indoor and outdoor recreation trips, visiting, and eating meals. Home-based school includes student trips from home-to-school and school-to-home, regardless of grade level.

All three home-based non-work trip purposes show steady declines in trips per household over the three Bay Area household travel surveys. Home-based social/recreation trips per household increased slightly between 1965 to 1981, only to show a precipitous drop of 34.5 percent (1.26 to 0.83 trips/ household) between 1981 and 1990. This might mean that Bay Area residents aren't having fun any more, or that household members are trading off out-of-home social/recreation activities for in-home (or weekend) social/recreation activities. The Bay Area travel surveys do not indicate what people are doing at home - whether they are asleep, working or telecommuting, playing, eating, socializing or watching television. Thus it is impossible with current survey data to understand the true nature of the tradeoff between in-home activities and out-of-home activities.

Non-home-based trips per household increased substantially between 1965 and 1981 (1.91 to 2.34 trips/household) only to fall back to a level moderately higher than the 1965 trip rate (2.10 trips/household).

Total transit trips per household decreased slightly between 1981 and 1990, from 0.56 to 0.48 trips per household. School bus trips per households also showed a slight decrease between 1981 and 1990. In-vehicle person trips showed the most significant absolute decline between 1981 and 1990, dropping from 6.81 to 6.12 trips per household. Total vehicle trips per household (not shown in Table 9) decreased from 5.23 trips per household in 1981 to 4.86 vehicle trips per household by 1990. A steady decline in walk trips per household can be shown between the three household travel surveys, dropping from 1.35 walk trips per household in 1965, to 1.00 trips per household by 1981, then levelling off at 0.75 walk trips per household by 1990.

Bay Area average household size declined from 3.12 persons per household in 1965 to 2.61 persons per household by the year 1990. This major change in household size is reflected in a major shift between household size cohorts between survey years. It is interesting to note that the 1990 trips per household by household size are, for the most part, higher than the 1965 trips per household by household size (Table 10). One person households made an average 3.34 trips in 1965; in 1990, 3.64 trips per household. Two person households made an average of 6.17 trips in 1965; in 1990, 6.34 trips per household. On the other hand, five-or-more person households made 15.22 trips on average in 1965; the average five-or-more person household in 1990 made 14.30 trips, a six percent decline from the 1965 trip rate. There was no change in total trip rate per household, 1965 versus 1990, for either the three-person or four-person household group. Even though the total trips per household by household size cohort is typically higher in 1990 compared to 1965, total trips per household declined from 8.78 trips per household in 1965 to 7.55 trips per household in 1990.

Trips per household by vehicle ownership level (0, 1, 2, 3-or-more vehicles) is shown in Table 11. Though the term "ownership" is used, the three household travel surveys, as well as the decennial censuses, collect information on household vehicle availability, not ownership.

The most interesting finding is the minor decline in total trips per household for zero-vehicle households, comparing the 1981 survey (4.00 trips per household) to the 1990 survey (3.89 trips per household). All other vehicle availability cohorts show a steady decline in total trips per household comparing results from the three surveys. All of the home-based non-work trips per household show a decline by all vehicle availability levels between 1981 and 1990. The no-vehicle group shows a slight increase in non-home-based trips per household between 1981 and 1990. Home-based work trips per household is typically higher in 1990 than 1981, except for the three-or-more vehicle group.

Total trips per household by trip purpose by housing structure type (single family, condominium or townhouse, duplex, apartments) is summarized in Table 12. Though there are significant differences in total trips per household by housing structure type, there is no apparent pattern in the declining trip rates by structure type. Single family dwelling unit total trips per household declined from 10.13 trips per household in 1965 to 8.64 trips per household by 1990. Apartment dweller trips per household declined from 6.13 (1965) to 5.50 (1990) trips per household. Duplex householders showed the slowest decline in average household trip rates, dropping just 3 percent from 6.87 trips per household in 1965 to 6.66 trips per household by the year 1990.

Total trips per household by trip purpose by three household income levels is shown in Table 13. The income tertiles used in this analysis are roughly one-third of all households for the three surveys, though due to differences in the income group definitions for the travel surveys, the results are not strictly comparable. The most significant finding is the subtle decline in total trips per household for low income households, similar to the finding of a minor decrease in the trip rate for zero-vehicle households. All income tertiles are showing increases in home-based work total trips per household; and all income cohorts are showing minor to moderate decreases in home-based non-work trips per household.

This section analyzes changes in trip purpose shares, modal shares and vehicle occupancy comparing the three Bay Area household travel surveys. For work trip modal share analyses, comparisons are made to decennial census journey-to-work modal shares.

Share of trips by trip purpose, by means of transportation, is highlighted in Table 14, and Figure 2 and Figure 3. The home-based work trip share of total trips remained quite stable at 21 to 22 percent of all trips between 1965 and 1981. By 1990, home-based work trips accounted for 26.4 percent of all trips. The increase in home-based work share, 1981 to 1990, was offset by a decrease in home-based non-work share. Non-home-based trip purpose share for trips by all means of transportation increased from 26.8 percent in 1981 to 27.8 percent of all trips by 1990.

The most significant change between 1965 and 1981 for transit trip purposes was the increase in non-work transit trips and a decrease in the share of home-based work transit. The 1990 survey shows a reversal in this trend with 41.8 percent of all transit trips being home-based work trips. The three surveys also indicate a steady increase in the non-home-based share of transit trips.

Modal share for home-based work trips, or for commuters in the case of the census journey-to-work data, is reported in Table 15. The column heading "HWC" denotes home-to-work census comparable survey data, and excludes home-to-work related and work related-to-home trips to provide a better comparability to census journey-to-work statistics. The column heading "HBW" refers to the standard home-based work person trips, and includes the home-work and home-work related trips. The modal share comparisons between the 1981 survey and the 1980 Census, and between the 1990 survey and 1990 Census, provide quite encouraging results. If the 1990 Census data can be considered "observed" data, then the 1990 survey is showing slightly too high transit shares (10.5 percent survey versus 9.9 percent 1990 Census), slightly too low walk shares (3.0 percent survey versus 3.8 percent Census) and basically the correct distribution of vehicle driver versus vehicle passenger work trips.

Modal shares by trip purpose are summarized in Table 16. The regional transit trip share of all trips increased from 5.5 percent in 1965 to 6.4 percent in 1981, levelling off to a 6.3 percent transit share by 1990. The transit share for home-based school trips (11.3 percent in 1990) is slightly higher than the transit share for home-based work trips (10.0 percent in 1990). Transit share for the other non-work trip purposes is hovering around 3 to 4 percent transit for all three household travel surveys.

Walk and other mode modal shares have shown a steady decline over the three survey periods. Only non-home-based walk/other shares have stabilized at about 14.8 to 15.0 percent. Home-based school trips have shown a major decline in walk/other mode share, declining from 52 percent share in 1965 to a 26 percent share by 1990.

Average vehicle occupancy by trip purpose, calculated by dividing in-vehicle person trips by vehicle driver trips, is shown in Table 17. All trip purposes have shown a steady decline in vehicle occupancy except for home-based school trips. Vehicle occupancy for home-based school trips showed an unexpected increase from 2.23 persons per vehicle according to the 1981 survey to 2.52 persons per vehicle in the 1990 survey. Home-based work trip vehicle occupancy from the 1990 survey (1.10) is basically the same as commute vehicle occupancy from the 1990 Census (1.097 persons per commute vehicle).

Results from the 1990 San Francisco Bay Area household travel survey provides major challenges to Bay Area transportation planners. An apparent decline in trip frequency per household and per household is offset by an increase in average trip duration, yielding an apparent stability in average travel time expended per household and per person. Findings from this study are generally consistent with the travel time budget studies of the 1970s and early 1980s. Findings may also rekindle interest in travel time budget analyses and alternative travel demand forecasting models based on activity analysis, time use studies, and travel time budgets.

Comparison of model-simulated trips to 1990 survey trips shows an excellent match for home-based work trips, yet a moderate overprediction of non-work trips with respect to 1990 survey person trips. The non-work trip generation and trip distribution models in use in the Bay Area are not structured to account for this inverse relationship between trip frequency and trip duration. Trip generation models are typically built to provide ever-increasing non-work trips per household based on assumptions of real income growth and growth in auto ownership per household. New and improved non-work trip generation models may need to incorporate some direct linkage with work trip distribution models (e.g., total work trip duration).

There is a significant potential for underreported trips in the 1990 survey, especially for non-work trips, and for vehicle driver, vehicle passenger, walk and bicycle trips. There is probably a minor (5 percent) underreporting of transit trips in the 1990 household travel survey. As a part of a "cross-validation" project, MTC will assign the raw, expanded survey trip records to regional highway and transit networks for analyzing screenline loadings and for analysis of vehicle miles and person miles of travel. This project would also allow for the editing and correction of survey outliers (e.g., trips with absurdly low or absurdly high travel speeds).

There is a relatively stable share of mobile persons and households for the three Bay Area household travel surveys. This means that the share of persons falsely claiming no travel is not a major problem with the three household surveys.

The 1990 survey was conducted in less than ideal situations. Survey response rates declined between 1981 and 1990. In 1981, 69 percent of eligible households contacted completed the survey; in 1990, 49 percent completed the survey. The survey consultant reported problems due to interviewer fatigue as well as interviewee fatigue, and a higher degree of interviewer turnover than expected. It is also unclear what impact the October, 1989 Loma Prieta earthquake had on moderating travel patterns; the impact of the economic recession in the U.S. on moderating out-of-home travel; the nature and extent of the in-home substitution for out-of-home activities; and the nature and extent of weekend travel substitution for weekday travel.

The most challenging aspect of this future research could be the integration of travel time expenditure concepts into a disaggregate travel demand model system for use by regional transportation planners. The findings of this aggregate analysis should be used to inform a more detailed and rigorous disaggregate travel behavior analysis.

1. Hanna P.H. Kollo and Charles L. Purvis. Changes in Regional Travel Characteristics in the San Francisco Bay Area: 1960-1981. In Transportation Research Record 987, TRB, National Research Council, Washington, D.C., 1984, pp. 57-66.

2. Edward J. Kannel and Kenneth W. Heathington. Temporal Stability of Trip Generation Relations. In Highway Research Record 472, HRB, National Research Council, Washington, D.C., 1973, pp. 17-27.

3. Kenneth R. Yunker. Tests of the Temporal Stability of Travel Simulation Models in Southeastern Wisconsin. In Transportation Research Record 610, TRB, National Research Council, Washington, D.C., 1976, pp. 1-5.

4. Robert L. Smith, Jr., and Donald E. Cleveland. Time-Stability Analysis of Trip-Generation and Predistribution Modal-Choice Models. In Transportation Research Record 569, TRB, National Research Council, Washington, D.C., 1976, pp. 76-86.

5. Gerald S. Cohen and Michael A. Kocis. Components of Change in Urban Travel. In Transportation Research Record 775, TRB, National Research Council, Washington, D.C., 1980, pp. 42-47.

6. Bahar B. Norris and Gordon A. Shunk. The 1984 Home Interview Survey in the Dallas-Fort Worth Area: Changes in Travel Patterns, 1964-1984. In Transportation Research Record 1134, TRB, National Research Council, Washington, D.C., 1987, pp. 1-9.

7. W. Thomas Walker and Haiou Peng. Long-Range Temporal Stability of Trip Generation Rates Based on Selected Cross-Classification Models in the Delaware Valley Region. In Transportation Research Record 1305, TRB, National Research Council, Washington, D.C., 1991, pp. 61-71.

8. ITE Technical Council Committee, 6F-12. Consistency of Origin-Destination O-D Characteristics Through Time. In ITE Journal, Vol. 49, No. 10, 1979: 32-39.

9. Charles River Associates, Inc. Characteristics of Urban Transportation Demand: An Update. Report DOT-T-88-18. UMTA, U.S. Department of Transportation, Washington, D.C., Revised Edition, July 1988.

10. J.C. Tanner. Factors Affecting the Amount of Travel. Road Research Technical Paper No. 51, TRRL, Crowthorne, Berkshire, UK, 1961.

11. Yacov Zahavi. The TT-relationship: a Unified Approach to Transportation Planning. In Traffic Engineering & Control. August/September 1973, pp. 205-212.

12. Yacov Zahavi. Traveltime Budgets and Mobility in Urban Areas. Final Report. Report FHWA PL-8183. FHWA, U.S. Department of Transportation, Washington, D.C., May 1974.

13. Yacov Zahavi. The UMOT Project. Report No. DOT-RSPA-DPB-20-79-3. Research and Special Programs Administration, U.S. Department of Transportation, August 1979.

14. Yacov Zahavi. Travel Regularities in Baltimore, Washington, London and Reading. Technical Memorandum. Mobility Systems, Inc., Bethesda, Maryland, December 1981.

15. Yacov Zahavi and James M. Ryan. Stability of Travel Components over Time. In Transportation Research Record 750, TRB, National Research Council, Washington, D.C., 1980, pp. 19-26.

16. Yacov Zahavi and Antti Talvitie. Regularities in Travel Time and Money Expenditures. In Transportation Research Record 750, TRB, National Research Council, Washington, D.C., 1980, pp. 13-19.

17. Yacov Zahavi, Martin J. Beckmann and Thomas F. Golob. The UMOT/Urban Interactions. Research and Special Programs Administration, U.S. Department of Transportation, January 1981.

18. Crain and Associates. 1981 Bay Area Travel Survey. Metropolitan Transportation Commission, Berkeley, California, August 1981.

19. Marilyn M. Reynolds, Sydwell M. Flynn, and David B. Reinke. 1981 San Francisco Bay Area Travel Survey. In Transportation Research Record 877, TRB, National Research Council, Washington, D.C., 1982, pp. 51-58.

20. E.H. White and Company, Inc. 1990 Bay Area Travel Survey: Final Report. Metropolitan Transportation Commission, Oakland, California, March 1991.

21. Hanna P.H. Kollo and Pat Hackett. Documentation of the BATSC Linking Procedure 3. Technical Memorandum. Metropolitan Transportation Commission, Berkeley, California, April 1972.

22. Design of Trip Linking Procedures 1981 MTC Travel Survey. Working Paper 1. Metropolitan Transportation Commission, Berkeley, California, December 1983.

23. Trip Linking Procedures: 1990 Bay Area Household Travel Survey. Working Paper 3. Metropolitan Transportation Commission, Oakland, California, revised June 1993.

24. COMSIS Corp., Barton Aschman, Inc., and Cambridge Systematics, Inc. Travel Model Development Project: Task 6 - Data Base. Technical Paper. Metropolitan Transportation Commission, Berkeley, California, August 1976.

25. Sample Weighting and Trip Expansion: 1981 MTC Travel Survey. Working Paper 4A. Metropolitan Transportation Commission, Berkeley, California, April 1983.

26. Charles L. Purvis. Sample Weighting and Expansion: 1990 MTC Travel Survey. Working Paper 2. Metropolitan Transportation Commission, Oakland, California, revised June 1993.

27. Mike Clarke, Martin Dix and Peter Jones. Error and Uncertainty in Travel Surveys. In Transportation, Vol. 10, 1981, pp. 105-126.

28. Peter Barnard. Evidence of Trip Under-Reporting in Australian Transportation Study Home Interview Survey and Its Implications for Data Utilisation. In New Survey Methods in Transport, Elizabeth S. Ampt, Anthony J. Richardson, and Werner Brog, editors, VNU Science Press, Utrecht, The Netherlands, 1985, pp. 311-331.

29. Patricia S. Hu and Jennifer Young. Summary of Travel Trends: 1990 Nationwide Personal Transportation Survey. Report FHWA-PL-92-027. FHWA, U.S. Department of Transportation, Washington, D.C., March 1992.

30. Applied Management & Planning Group. 1991 Southern California Origin-Destination Study: Summary Findings. Southern California Association of Governments, Los Angeles, California, February 1993.

31. Household Travel Survey: Report #1. Sacramento Area Council of Governments, Sacramento, California, December 1992.

32. Don L. Ochoa and Les Jones. 1991 Statewide Travel Survey: Summary of Findings. California Department of Transportation, Sacramento, California, November, 1992.

33. Marcus R. Wigan. Australian Personal Travel Characteristics. Special Report 38. Australian Road Research Board, Vermont South, Victoria, 1987.

34. P. B. Goodwin. The Usefulness of Travel Budgets. In Transportation Research-A. Volume 15a, 1981, pp. 97-106.