250 likes | 384 Views
Household Travel Surveys Lessons / Issues / Plans . Presentation to the AMPO Travel Modeling Working Group October 24, 2006 Ron Milone MWCOG/NCRTPB Washington, DC. Dynamics shaping the planning process in the Washington, DC region. Sprawling development continues
E N D
Household Travel Surveys Lessons / Issues / Plans Presentation to the AMPO Travel Modeling Working Group October 24, 2006 Ron Milone MWCOG/NCRTPB Washington, DC
Dynamics shaping the planning process in the Washington, DC region • Sprawling development continues • The Washington region is second to New York for the percentage of workers with "extreme commutes” • Home buyers trade off lower housing prices with longer commutes • Public money for new construction limited • Local share of funding for transportation costs is increasing • Virginia is considering public-private partnership for building HOT lanes • Managed highway pricing is planned in Maryland • The number of immigrant residents/workers has increased, helping to counter the number of baby boomers who are retiring • Extensions to the completed 103-mile Metrorail system now planned. Increasing transit service and ‘Smart Growth’ are cited by many as congestion remedies
Why are HH surveys conducted? • Sample measurements: Household travel surveys are intended to identify localized relationships between travel ‘desires’ and land use, system, policy factors that can be forecasted • Household travel surveys are not designed to count demographic and travel quantities with a high level of geographic precision.
The HH survey is one component of a regionalinventorythat informs models
Needs of conventional modelers • Trip Rates, by purpose • Production-end rates • Attraction-end rates • Trip Length Frequencies, by purpose, by O-D pattern • Modal Share, by purpose, by O-D pattern • Time-of-Day profile by purpose, by mode, by directionality
How low can the surveys go, with confidence? • Regional level - Yes • Regional Level, by socio-economic stata -Yes • State Level - Probably • County Level – Maybe • By Sector – Maybe/No • By TAZ or finer- No
Organizational Issues • A HH survey is a substantial, yet infrequent undertaking; it can be a ‘shock-wave’ to the work program • Identifying funding sources is a challenge • The knowledge/skills requirements are unique • Administration of survey is increasingly being out-sourced – How well do surveyors know the region? • Interagency cooperation and coordination required • Interfacing with the general public is always a delicate matter
Past HH Travel Surveys in Washington, DC • 1968 Home Interview Survey • Face-to- Face Interviews with an ‘army’ of interviewers • 26,000 Households sampled (1 in 20, 1 in 33) • 6 Jurisdictions • 1987/88 Home Interview Survey • Mail / CATI combination (conducted by MPO) • 8,000 Households sampled (1 in 166) • 8 Jurisdictions • 1994 Spring/Fall Home Interview Survey • Mail / CATI combination (conducted by consultant) • 4,800 Households sampled (1 in 300) • 13 jurisdictions
Trends Impacting Surveys • Study area is steadily expanding • Cost of data collection is increasing • Sample sizes are steadily decreasing • Ability to collect data by telephone- increasingly difficult • Cell phone market share increasing • Telephone ‘land line’ market share decreasing • Telephone screening technology improving • Modeling requirements/complexity is increasing • Policy questions being asked are ahead of tools • Surveys, in general, are saturating the area • Privacy, confidentiality, and identity theft are growing concerns
Essential HH Survey Goals • Appropriate capture and selection of HHs • representation of socio-economic markets • adequate capture of the ‘minority’ modes • Minimizing non-response • Minimizing under-reporting of travel • Maximizing Location Accuracy • Data that’s valid and ‘clean’
Survey Implementation Process • Planning • Survey Design • Assemble background data: Census STF1-4, CTPP, PUMS • Formulate survey approach, sampling procedures, • Design instrument(s) • Field Implementation • Pretesting, data collection • Data Preparation • Coding, cleaning, compiling • Data Analysis • Analyzing, Reporting, Using
Non-Response • The main concern: bias in the data • Response rate for 1994 HTS: 38% • 50 % Recruitment: unusable telephone numbers: fax machines, nonresident units, unoccupied units, etc. • 76% Retrieval: refusals, no telephone contact made, language problems, <50% HH members responded. • Who are non-responders? • Low income groups • Telephone ‘screeners’ • People who just are not home: high mobility groups! • People who are home, but do not travel – They don’t feel ‘applicable’ to a travel survey • Item non-response: income, age
How to Deal with Non-Response • HH non-reponse: • Ignore it (if sample size is sufficient without non-respondents) • Assumption: Non-respondents are similar to respondents (!) • Item non-response: • Impute values (Hot Decking) • Is a reasonable fix if the non-respondent population is different from respondent population • ‘Nearest-neighbor’ approach – uses like socio-economic and personal characteristics to ‘fill in’ item non-response and to adjust trip weights
Data Cleaninglogical, rational, and reasonable models require like data Question: How much time/effort is needed to clean data? Answer: How much time do you have? • 1994 HTS: 1.5 -2.0 person-years • What’s involved (cleaning HH/Trip/Person files): • one-way Frequencies – range/distribution • Cross-tabulations: logical /consistent/coherent • Trip-chaining: logical timing & sequence of trip itinerary • Address Matching: the big one • Validating against other data sources
Plans: 2007 Household Travel Survey • Project Director: Robert Griffiths, Technical Services Director, COG/TPB • Travel and Activity Survey – 10,000 HH • In-Vehicle GPS add–on – 250 households • Planned Survey Design • Address – based sample from USPS carrier route lists, as opposed to Telephone/Random-Digit-Dial(RDD)-based • Circumvents telephone-related issues cited above • Better control of uniform geographic capture, that is not ensured using RDD method • Differential sampling rates by area type • All households with deliverable mail address in sample, except those on ‘do-not-mail’ list (3%)
Sampling • 22 jurisdictions (modeled area) • Frame – mail carrier routes • Segmentation: • ‘Inner Ring‘ Jurisdictions • High density/mixed use areas (over-sampled to ensure capture of ‘minority’ modes (transit, ride share, walk, bicycle) • Low Density areas • ‘Outer Ring’ Jurisdictions
Data Collection • Initial mailing • Minimal household, person, vehicle characteristics asked • Follow-up telephone recruitment • Telephone/Internet travel-activity data retrieval (respondent’s option) • Real-time geocoding used
Pilot Test (in progress) • Assessing coverage of proposed mail route sample as opposed to RDD sample • Assessing the effect of financial incentives • Assessing interview method response rates • Testing conversion methods for non-respondents / non-response follow-up survey
Pilot Test … continued • Vehicle GPS add-on survey will be tested • will be used to assess under-reporting or over-reporting of trip making. • The assignment of “observed’ vehicle trips from the survey has historically resulted in an under-estimation of VMT. • Short non-work trips are typically under-reported, and so trip rates are usually increased to make up for the difference. • Is this the right thing to do? Other possible sources of error: • Commercial Vehicle trips not well reflected • External trips not well reflected • Error in Observed VMT
Schedule • Pilot Test Evaluation: Now • Main Survey: November 2006 – January 2008 • Survey will be collected throughout the 13 month period.
Conclusions on HH Surveys • Vital for formulating variable relationships in the work • But one piece of the data puzzle • Typically lag behind the questions being asked • Subject to problems relating to non-response, under-reporting, geographic coverage, modal coverage
Conclusions … • Modelers should be involved at the ‘front-end’ of survey design & development • Is the information obtained appropriate? • Are questions asked in the best way? • What are the limitations of the survey? • Sources of error abound, data is imperfect • Technology must continually be exploited to address issues
Read more about the Washington Household Travel Survey www.mwcog.org/hts