Changes to the collection of short walk data in the NTS Glenn Goodman, DfT
A walk less than one mile but greater than 50 yards in distance. What is a short walk?
Timeline Develop weighting Short walks experiment Further experiment Public consultation Consultation response Cognitive testing Decision on method October 2015
Background: Household survey of personal travel First survey in Continuous since : n = 7k households, 16k people (63% resp. rate) Complex design – stratified two-stage random sample
Data collection
Travel week allocation Day 1 = Monday? Day 7 = Sunday? Respondents allocated start day – uniform spread across week and month
Data collection
The importance of the data Walking & cycling policy Transport forecasts and models
The decline of walking 234 trips 133 trips 58 trips 70 trips
Research outline Data are weighted to account for under- reporting – short walks only collected on day 7
Weighting for under-reporting
Research question Data are weighted to account for under- reporting – short walks only collected on day 7 Aims: To assess whether under-reporting of short walk trips exists If so, correct this under-reporting
Timeline Short walks experiment April – June 2013
Experiment outline Aim: To assess day 1 vs. day 7 Sample size: 1,000 Designed to detect a 5% point change £££
2013 short walks experiment
Under-reporting: adults
Under-reporting: children 11% 1.5%
Under-reporting: type of walk A B Short walk only trip AB Multi stage trip
Under-reporting: type of walk 1.1%
Timeline Short walks experiment April – June 2013 Public consultation July – Sept 2014
Consultation options 1 2 Day 7 Another experiment Produce weight 53 4 Day 1 Produce weight Break in series
Timeline Public consultation July – Sept 2014 Consultation response December 2014
Consultation response “…I would want to be very sure before making irreversible modifications to NTS protocols…the NTS has weaknesses and biases that are essentially stationary…turbulent biases are much worse…” 34 Day 1
Timeline Consultation response December 2014 Develop weighting Ongoing
The approach to weighting Short walk only trips by adults Logistic regression to calculate likelihoods Determine explanatory variables Produce weights
Developing back-series weight Age Economic status Car access
Example – car access 16% 7%
Developing back-series weight Age Economic status Car access Gender – not significant
Example – gender
Developing back-series weight Age Economic status Car access Gender Household region Settlement type (urban/rural) Mobility Income – not significant – excluded
Developing back-series weight Age + economic status 0-16 Full time Part-time Retired Other non-work Car access Main driver Other driver Non-driver No access
Logistic regression: Day 7
Logistic regression: Day 1
Relative likelihoods
Day 7 weighted Applying the weights Day 7 unweighted Day 1
Applying the weights
Historic likelihoods Age 0-16 Adult, full-time Adult, part-time Adult, other non-work Retired
Further work Cognitive interviews Comparisons with other studies London Travel Demand Survey Scottish Household Survey Further experiment
Timeline Develop weighting Short walks experiment Further experiment Public consultation Consultation response Cognitive testing Decision on method October 2015
Conclusion Experience of formal consultation process Highlights the value of ONS Methodology Advisory Service Lessons for other household surveys Better data on walking for the future!