1. Best Practices Met Council Household Travel Survey (HTS) 2010-2012 May 8. 2013

2. Metropolitan Council Household Travel Survey: Design, Implementation, and Lessons Learned 14th TRB Planning Applications Conference 5-9 May 2013 Columbus, OH Authors: Jason Minser, Abt SRBI Laurie Wargelin, Abt SRBI Jonathan Ehrlich, Metropolitan Council Anurag Komanduri, Cambridge Systematics Kimon Proussaloglou, Cambridge Systematics Cemal Ayvalik, Cambridge Systematics. The HTS was Part of a Total Travel Behavior Inventory with Cambridge Systematics as Prime

3. Household Travel Survey (HTS) Overview 19-counties in Minnesota (16) and Wisconsin (3) Address-based sampling approach – three tiered stratification plan Comprehensive inventory of... 12,000 households’ 24-hour travel for persons 6 years of age and older Subsample of 250 GPS households Special subgroups MnPass users University students

4. Why Address-Based Sampling (ABS)? Includes cell-phone only households Supplemental cell phone data collection is expensive with low response rates – not geographically based With ABS geographically based sample can be pulled and weighted by census tract Duel frame sampling (ABS/RDD) is difficult to weight Weighting is not the solution to unrepresentative samples – only builds in the bias. 4

5. ABS Allows Targeting by HH Type As random, proportional data collection proceeds to 60-75% of recruitments, results are monitored and examined on a weekly basis Since ABS is census tract/block based – at this point - oversamples can be ordered for tracts/blocks with high incidence rates for under-represented variables of interest— i.e. those with H2R household characteristics Examples (Zero-vehicle HHs, 4+ person HHS) Implementation can be targeted by geographic area This approach is called Responsive Interviewing Design (Groves and Herringa ) 5

6. Highly Stratified ABS Sampling Plan ABS improves the ability to define geographic strata 7 distinct regions sampled at differential rates Stratified by household size and total number of vehicles

7. Imbedded Design for Hard – to – Reach Populations Clearly define populations But not every group can be hard to reach or treated as such Identify appropriate levers; not all levers work for each group Use post-collection techniques sparingly (e.g., weighting) Address Based Sampling Cell onlyGeography Multi-Methods PhoneWebMail Design Sponsor Legitimacy IncentivesReminders

8. Design Qualities of 2010-11 Minnesota HTS Advance letters tailored to region and sponsorship emphasized Recruitment by phone (matched) and web (unmatched) Personalized activity-based travel diary Reminder calls made evening before scheduled travel day (weekdays only) Contingent incentive offered in follow-up for specific sub-groups ($20) Follow-up by phone, web, and mail

9. GPS Subsample 250 Total Completed Households Data collection consistent with main study 7-day travel period One day diary Diary date randomly assigned within 7-day period Pilot study conducted Incentive experiment conducted Full GPS/diary comparison conducted by PlanTrans, Inc. Matched ABS Unmatched Phone Recruit Web Recruit Travel Packet Deployment 7-day Travel Period Travel Packet Returned Diary and GPS Device Validation Completed HH Incomplete HH

10. Study Compliance and Participation Thirteen percent of households recruited by web Parity between web and mail-back of retrievals Phone becoming less of a retrieval method Method consistent with multi-method scenario Recruitment PhoneWebTotal Retrieval MethodN%N%N% Phone1,56315.1462.61,60913.3 Web4,29241.51,21168.15,50345.5 Mail-back/Multiple Methods4,46943.352229.34,99141.2 Total Completed10,3241001,77910012,103100 Total Households Recruited22,98687.33,35612.726,342100

11. Design Strategies to Overcome Low Responding Markets Tailored strategies are crucial throughout the process to overcome low responding markets Sampling Recruitment Retrieval Customized Letters Phone and Web Targeted Sampling Customized Letters Phone and Web Targeted Sampling Unmatched/ Cell only 4+ HH 0-vehicle Low Income University 4+ HH 0-vehicle Low Income University Non- Responders Incentives Phone or Web Reminder Letters Targeted Sample Incentives Phone or Web Reminder Letters Targeted Sample Phone, Web, Mail Reminder Calls Reassignment Incentives Phone, Web, Mail Reminder Calls Reassignment Incentives Sampling geographic strata requires close and continual monitoring

12. Recruitment at 60% Complete 12

13. Retrieval at 30% Complete 13

14. Strategies for Corrections at this Point Incentives for under-represented cells going forward: $20 for 0-Vehicle and 4+ Person Households. Added incentives for all hhs. with incomes <$25,000. Additional incentives: -- Reassignment: $20 offered to recruited hhs. in under- represented cells--that did not complete. Assigned a new travel day. Refusal Conversion -- $25 to University Oversample -- $10 to GPS households that had not returned units. Increased reminder calls to under-represented cells Introduced targeted sample by census tract geography for under-represented cells 14

15. Results of Corrective Actions on Final Retrieval 15

16. Geographic Representativeness Regional Classifications CompletedACS 2006 - 2008Diff N%N% Minneapolis1,78914.8158,38911.8-3.0 St. Paul1,1229.3108,7088.1+1.2 Core West – Suburban Minneapolis 3,31327.4345,26225.7+1.7 Core East – Suburban St. Paul2,92224.1213,20015.9+8.2 Core South – MPO Core1,63813.5278,35220.7-7.2 Ring Counties, Minnesota8416.9176,19113.1-6.2 Ring Counties, Wisconsin4783.963,9774.8-0.9 Total12,1031001,344,079100-6.2

17. Demographic Representativeness HH Characteristics CompletedACS 2006 - 2008Diff HH SizeN%N% 13,71730.7377,97328.1+1.4 24,80439.7445,44339.70 3+3,58229.5520,66338.8-9.3 Total12,1031001,344,079100-7.9 Number of VehiclesN%N% 04834.091,5866.8-2.8 13,80631.4413,35730.8+1.4 2+7,81464.6839,13662.5+2.1 Total12,1031001,344,079100+0.7

18. Best Practices and Lessons Learned

19. Best Practices for a Representative HTS ABS allows inclusion of cell-only households and improves geographic targeting by key variables Establish legitimacy of sponsor – advance letters/postcards/website Continual monitoring as random data collection proceeds as designed. Continual interval data review/editing and reporting -- so that hhs. not meeting quality standards can be replaced as data collection proceeds Implementation of corrective actions at approximately the two- thirds recruitment point Results in a representative sample that can be weighted with bias minimized. 19

20. Contact Information Laurie Wargelin Abt SRBI l.wargelin@srbi.com