Methods to Deal with Non-Working Matched Phone Numbers in an - - PowerPoint PPT Presentation

Methods to Deal with Non-Working “Matched” Phone Numbers in an Address Based Sample Survey

Anna Fleeman & Tiffany Henderson

67th Annual Conference of the American Association of Public Opinion Research

Background

• Traditional landline RDD has rapidly declined
• ver last few years
• CPO HHs, zero-banks, VoIP assignment
• Cell RDD often cost-prohibitive
• Federal law mandates hand-dialing and dispositioning
• Use of address based sample increasing
• Nearly full coverage of all US households (CDSF)
• Well suited for screening surveys and small

geographic areas

Addresses Selected from CDSF Phone Append Process

Background

Phone Match No Phone Match Treated akin to RDD sample Mail survey to

• btain phone #

Issue

• What to do when matched phone numbers are

dispositioned as non-working?

• Cannot remove from response rate calculations

as with RDD sample

• Sampling unit = address
• To resolve issue:
• Send mail survey/screener to address
• Request contact phone number(s)

treat as no phone match

Research Question

• Can non-working dispositions be used as a

proxy for vacant / ineligible addresses?

• Much cheaper and quicker to rely on phone
• No mailing of surveys, waiting for returns to

come in, scanning of returned surveys

• Lead to calculating an adjustment factor for

ABS response rates

Methodology

• 85,000+ addresses selected from ABS frame
• Concentrated in Northern Alabama and Joplin, MO
• Sample Target: HHs with 12-17 year olds
• September 2011 – March 2012
• 51% phone match rate
• Up to 10 attempts to all to matched phone #s

Methodology

• 9,477 matched phone #s dispositioned as

non-working

• Treated as no phone match records
• Sent mail survey/screener to address
• Requested contact phone number(s)
• Invitation letter and postage-paid return envelope
• $5 promised incentive
• Out-going envelope addressed to name, if available,

and current resident:

• Sam Smith or Current Resident
• Current Resident

Methodology

Results

Business 4% Disconnect / Not In Service 93% Fax/ Modem 3%

Type of Non-Working Matched Phone Numbers

n= 9,477

Results

Undeliverable 12% No Return 84% Returned Surveys 4%

Mail Survey Outcome

n= 9,477

Results

Different Address / Different Name 4% Same Address 85% Different Address / Same Last Name 11%

Address Comparison of Returned Surveys

n= 333

Results

Deceased 1% Vacant 3% No Such Number/ Street 1% Not Deliverable as Addressed 81% No Mail Receptacle 2% Unclaimed/ Rfsd 1% Addressee Not Known at Address 11%

Undeliverable Mail Types

n= 1,115

Conclusions

• Non-working phone dispositions cannot be

used as proxy for vacant/ineligible addresses

• The real story…
• USPS undeliverable codes not uniformly used
• USPS inconsistent forwarding rules
• “…or Current Resident” ≠ forwarded or ANK status

Next Steps

• Better understand USPS undeliverable codes
• Push for standardization
• Need a better solution for “…or current

resident / occupant”

• Mail in ABS study should not be forwarded
• Address = sample unit
• For our survey research field, these are key as

use of ABS increases

Thank You

Questions / Comments

a.fleeman@srbi.com t.henderson@srbi.com