Presented to:

Presented by:

Transportation leadership you can trust.

Second Day Response Rates:

Implications for CMAP’s Travel Tracker Survey

13th TRB Planning Applications Conference, Reno

Cemal Ayvalik and Eric Petersen

May 12, 2011

2

IntroductionMany MPOs implement 2-day travel surveys

Benefits: Statistical insights at a lower cost than surveying twice as many households.

Risks: Significant day-to-day variation in individual travel:– major activities at the tour level,– number of stops and duration, – scheduling of activities, – mode shifts – and route choice.

Respondent fatigue likely to bias the results.

3

Measuring Fatigue

Stop activities may be omitted on second day.

Work tours may be less affected than other travel purposes.

Issues possible in full chain of activities or omissions in maintenance or leisure activities on the second day.

Markers that might raise flags or minimize concern. – The respondent actually indicates whether the data is being read off

hard copies of a travel diary or reported from memory.

– One respondent providing information on all household members.

Clues for response fatigue.

4

Correcting for Response FatigueDepending on the severity of the problem, there are a variety of options:– Discarding the entire record;– Discarding the second day and treating the first day as if

it were one-day data;– Reweighting either the second day or both days;– Adjusting the existing data with respect to VMT or

activity duration;– Synthesizing missing information.

However, modelers must know extent of the problem before making any adjustments or corrections.

5

Case study: Chicago Metropolitan Area

In the 2008Travel Tracker household survey conducted for the Chicago Metropolitan Agency for Planning (CMAP):

– More than 14,300 households, 10,000+ in 6-County Area.

– About 30 percent of households in 6-County area were asked about their travel activities over 2 days (over 3,100 households).

– The rest responded via one-day travel diaries.

Is respondent fatigue an issue that leads to degradation in the data outweighing the advantage of 2-day surveys?

6

Household Level Observations

Household Trip Reporting from the 2-Day Survey

Fewer Trips in the Second Day 1,349 43.4%

More Trips in the Second Day 1,122 36.1%

Same Number of Trips 640 20.6%

3,111 100.0%

7

Household Level Observations

Household Trip Rates by Survey Duration and Day of the Survey

Survey Duration Day N Mean Std Dev

1-Day 1 7,652 9.147 6.933

2-Day 1 3,263 8.844 6.713

2-Day 2 3,196 8.625 6.666

8

Person Level Observations

Person Trip Reporting from the 2-Day Survey

Fewer Trips in the Second Day 2,175 34.9%

More Trips in the Second Day 1,991 31.9%

Same Number of Trips 2,067 33.2%

6,233 100.0%

9

Person Trip Rates by Survey Duration and Day of the Survey

Survey Duration Day N Mean Std Dev

1-Day 1 16,033 4.366 2.602

2-Day 1 6,778 4.257 2.510

2-Day 2 6,588 4.184 2.444

Person Level Observations

10

Tour Level Observations

Diaries from Typical Travel DayWeekdaysHome-Based complete toursAdults, Age 16 or OverWork vs. Non-Work Tours

11

Tour Level Observations

Number of Tours by Survey Type and Day of the Survey

Survey Duration Day Tour Type Number of

Tours Percent Mean Std Dev

1-Day 1Work 7,046 34.4% 0.52 0.54

Non-Work 13,421 65.6% 0.99 0.99

2-Day

1Work 2,056 38.0% 0.58 0.55

Non-Work 3,357 62.0% 0.95 1.00

2Work 2,014 36.2% 0.54 0.55

Non-Work 3,557 63.8% 0.95 0.97

12

Tour Level Observations

Tour Lengths by Survey Type and Day of the Survey

Survey Duration Day Tour Type Number of

Tours Percent Mean Std Dev

1-Day 1Work 7,046 34.4% 1.59 1.98

Non-Work 13,421 65.6% 1.89 1.79

2-Day

1Work 2,056 38.0% 1.75 2.01

Non-Work 3,357 62.0% 1.73 1.70

2Work 2,014 36.2% 1.64 1.99

Non-Work 3,557 63.8% 1.81 1.79

13

Testable Hypotheses

Hypothesis #1: Will respondent fatigue decrease the tour count and number of stops by tour type reported in 2-day surveys ?

Hypothesis #2: Will respondent fatigue decrease the tour count and number of stops by tour type reported on the second day?

Hypothesis #3: Will respondents reading from travel diaries suffer less fatigue?

Hypothesis #4: Will respondent fatigue increase as the number of household members increases?

14

Tour Level Comparisons

Reduction in number of tours and stops in tours is considered as an indicator of response fatigue.

Tours in the first day of 2-Day survey and the 1-Day survey are compared first.

Differences between the days in 2-Day survey were analyzed.

Completion of a log, household size and survey type are used as explanatory variables.

Number of tours and number of stops in a tour for work and non-work tours were dependent variables.

15

Tour Level Comparisons

Work Tours Non-Work Tours

Tours Stops Tours Stops

Survey Duration More tours in 2-Day

More stops in 2-Day

Fewer tours in 2-Day

Shorter tours in 2-Day

Complete Logs Shorter tours from memory

Fewer tours by memory

Shorter tours by memory

Interaction Dampenedin 2-Day

1-Day vs. 2-Day Surveys

Survey Duration More tours in 2-Day

More stops in 2-Day

Fewer toursin 2-Day

Shorter tours in 2-Day

Household Size More tours for Size 3+

More stops as size increases

More Tours except for Size 3+

Shorter tours for Size 3+

Interaction Inconclusive Interaction Higher reduction

for Size 3+

16

Tour Level ObservationsFirst vs. Second Day in 2-Day Survey

Work Tours Non-Work ToursTours Stops Tours Stops

Survey Days Fewer work tours in Day 2

Complete Logs Fewer Tours from memory

Shorter Tours from memory

Interaction

Survey Days Fewer work tours in Day 2

Household Size More tours for Size 3+

More stops for Size 3+

More tours with size (Size 3+)

Shorter tours for Size 3+

Interaction

17

Summary of ResultsTwo different types of fatigue can be evaluated:– Fatigue across survey types (mix of single and multi day diaries).

– Fatigue across survey days within multi-day surveys.

Based on trip comparisons:

Less travel is reported by 2-day survey.

Less travel is reported on second day of 2-day survey.

Mandatory travel similar on first and second day.

Fewer non-mandatory stops on the second day.

18

Summary of ResultsBased on tours:

H1:

Equivalent number of tours reported by survey type.

Fewer non-mandatory tours in the 2-day survey.

Complexity of mandatory tours seems to be increasing in 2-day survey. – Are shorter non-mandatory tours condensed into longer

mandatory tours in reporting or is it due to day-to-day variation?

H2:

No major differences between first and second day.

More mandatory tours reported in first day.

Summary of Results

H3: Shorter tours were observed from respondents who did not fill out a travel log. – The reason why they did not fill a diary is unknown. May be fatigue or

the travel activity was actually short enough to recite from the memory.

H4: Larger households - higher tour count and more stopsThree-person households had unique patterns– More work tours – Primarily due to differences in member relationships and composition.

19

Next StepsDay-to-day variation.Unclear whether this is non-random.Use GPS data to establish a degree of day-to-day variation.Looking at the variation by different segments including:– household life cycles, – time of day, – activity patterns (linked activities, tours by complexity),– geography, and– data retrieval methods.

Matched-Pair Design to control for socioeconomics.Model tour types and lengths using 1-Day survey data. Cross-validate using 1-Day and 2-Day survey data.

20