JPSM Distinguished Lecture 2019 Don Dillman: Web Push …...• Two-panel design: Internet First (invitation letter in first contact) and Internet Choice (questionnaire in first contact)

JPSM Distinguished Lecture 2019

Don Dillman: Web Push Surveys

NancyBates,SeniorResearcherforSurveyMethodology,U.S.CensusBureau

Discussant

1

Discussion Points

§ ApplicationsofWeb-Pushfor2020Census§ TailoredDesignfor2020Census§ Howcommunicationscampaigncanboostweb-push§ Modepreferencesamonghard-to-survey?§ Modechoiceexpectationsamongaudiencesegments§ ThingsIworryaboutwithWeb-pushin2020

2

2020 Census Design

§ FirsttimeeverU.S.DecennialCensuswillhaveinternetmoderesponse§ Threemodesinall:

§ Mail§  Internet§  Phone

§ TailoredDesign§  InternetFirst(Web-push)(78%ofhouseholds)§  Choice(internetormail)(22%ofhouseholds)§  Allhouseholdswilleventuallyreceiveamailform

§ Householdsthatdonotself-respondfollowedupinperson

3

•  Two-paneldesign:InternetFirst(invitationletterinfirstcontact)andInternetChoice(questionnaireinfirstcontact)

•  Newfor2020:fourInternetFirstcohortstobestdistributecallstoCensusQuestionnaireAssistance

•  InternetChoicepanelisinasinglecohort,mailedonthesamescheduleasInternetFirst,Cohort2

(*)Targetedonlytononrespondinghouseholds

Panel Cohort

Mailing1Letter(InternetFirst)orLetter+Questionnaire

(InternetChoice)

Mailing2Letter

Mailing3*Postcard

Mailing4*Letter+

Questionnaire

Mailing5*“It’snottoolate”Postcard

InternetFirst(Web-Push)

1 March12,2020 March16,2020 March26,2020 April8,2020 April20,2020

2 March13,2020 March17,2020 March27,2020 April9,2020 April20,2020

3 March19,2020 March23,2020 April2,2020 April15,2020 April27,2020

4 March20,2020 March24,2020 April3,2020 April16,2020 April27,2020

InternetChoice N/A March13,2020 March17,2020 March27,2020 April9,2020 April20,2020

2020 Census Web-push versus Choice

4

Criteria for Web-push versus Choice

TractsareassignedtochoiceiftheyrespondbymailmorethanbyinternetintheAmericanCommunitySurvey(ACS),andhaveatleastoneofthefollowing:§ Isalowrespondingtract,basedonACSself-responserates,§ Isatractwithhigherolderpopulation,basedonACSestimates,or§ Isatractwithlessinternetaccess,basedonFCCdata

§ Otherwise,tractsareassignedtoweb-push

5

Projected 2020 Census response by mode

Projectedself-responserates

ImplementationMethod Total Internet Phone Paper

InternetPush 69.0% 52.0% 5.2% 11.8%

InternetChoice 53.6% 14.6% 2.0% 37.0%Total

65.6% 43.8%

4.5%

17.4%

Projectedself-responseattheendofNRFU:

6

New contact methods to push to web

§ 2020CensuswillallowforNON-IDresponse§ Don’tneedmailpiecewithuniqueIDtorespondonline§ Advertisingcampaignopensupnewcontactmethods:

§  SeeURLonTV,print,orbillboardad§ HearURLonradioad§ Clickonadigitalad–godirectlytolandingpage§  InformedDelivery(viaemailfromUSPS)

§ PartnershipProgrammaterialswithURLplacedatevents,instorefronts,inlibraries,schools,etc.

7

NON-ID web option

§ 2020CensuswillallowforNON-IDresponse§ Don’tneedmailpiecewithuniqueIDtorespondonline§ Advertisingcampaignopensupnewcontactmethods:

§  SeeURLonTV,print,orbillboardad§ HearURLonradioad§ Clickonadigitalad–godirectlytolandingpage§  InformedDelivery(viaemailfromUSPS)

§ PartnershipProgrammaterialswithURLplacedatevents,instorefronts,inlibraries,schools,etc.

8

Example digital ad

9

Out-of-home example

10

Print example

11

USPS Informed delivery ™

12

13

14

How does Web-push versus Choice intersect with communications campaign?

15

16 Authorized Use Only

Note: U.S. population percentages do not add up to 100% due to tracts with no ACS mailout and, therefore, no tract segment assigned.

Rural Deltas and Urban Enclaves

43%

•  Low predicted rate of response, with the lowest percentage of that response coming online.

•  Found in rural parts of the southeastern United States, as well as concentrations in urban areas.

•  Low % college educated, low median household incomes, below-average levels of internet access, and majority non-Hispanic African American.

7% Of the U.S. Population Predicted

Self-Response

•  Low predicted rate of response, with a below-average percentage of that response coming online.

•  Found in California’s Central Valley and parts of New Mexico, Texas, Florida, as well as concentrations in urban areas.

•  High % foreign-born, low % college educated, and majority Hispanic.

14% 45%

Multicultural Mosaics

Predicted Self-Response

Of the U.S. Population

•  Below-average predicted rate of response, with below-average internet response.

•  Found in rural areas predominantly in the western United States, Appalachia, northern Maine, and Michigan’s Upper Peninsula.

•  High % owner-occupied housing and below-average levels of internet access.

5% 49%

Sparse Spaces

Predicted Self-Response

Of the U.S. Population

•  Below-average predicted rate of response, with a high percentage of that response coming online.

•  Found in communities around college campuses or military bases.

•  A majority 18-24, high % college educated, and high % renter-occupied housing.

2% 56%

Student and Military Communities

Predicted Self-Response

Of the U.S. Population

59%

•  Slightly below-average predicted rate of response, with a high percentage of that response coming online.

•  Found in densely populated metro centers.

•  High % college educated, above-average % foreign-born, high % 25-44 compared to the nation as a whole, and high median household incomes.

9%

Downtown Dynamic

Predicted Self-Response

Of the U.S. Population

60%

•  Slightly below-average predicted rate of response, with a below-average percentage of that response coming online.

•  Found in rural areas predominantly in the eastern United States, surrounding small towns and outside the suburbs of major cities.

•  High % owner-occupied housing, low % college educated, and below-average median household incomes.

16%

Country Roads

Predicted Self-Response

Of the U.S. Population

•  High predicted rate of response, with an above-average percentage of that response coming online.

•  Found in small towns and less densely populated areas surrounding urban centers.

•  Low diversity and a higher % 65 or older than the national average.

67% 21%

Main Street Middle

Predicted Self-Response

Of the U.S. Population

•  High predicted rate of response, with a high percentage of that response coming online.

•  Found in suburban neighborhoods of single-family homes.

•  High % college educated, high % married, and high median household incomes.

71%

Responsive Suburbia

Predicted Self-Response

24% Of the U.S. Population

Introduction to Tract Segments

17 Authorized Use Only

11% 70%

7% 8%

0% 0% 0%

Hispanic NH White NH Black NH Asian NH AIAN NH NHPI

NH Other

Race/Hispanic Origin Proportion of Segment

NH indicates Non-Hispanic.

[VALUE]

[VALUE] 6% 2%

Single-unit Multiunit (2-9)

Multiunit (10+) Mobile Home

Housing Type

[VALUE]

[VALUE]

Owner vs. Renter

Homeowner

Renter

6% 19%

8% 25%

29% 13%

<5 5-17

18-24 25-44 45-64

65+

Age

Group Quarters

1%

Moved in the Last Year

11%

Share of Self-Response Online

81%

Internet Access

90%

71%

Predicted Self-Response

14,669

Number of Tracts

24%ofU.S.Population

How will they respond?

Onaverage,thissegmentispredictedtorespondathighrates,withahighpercentageofthatresponsecomingonline.

Who are they?

Thetractsinthissegmenthavehighmedianhouseholdincomes,highpercentagesofmarriedhouseholds,andalargeproportionofcollege-educatedresidents.

Where are they?

Tractsinthissegmentcanbefoundinsuburbanneighborhoodswithsinglefamilyhomes.

Responsive Suburbia: Demographics

18 Authorized Use Only

[VALUE] [VALUE]

7% 14%

Single-unit Multiunit (2-9)

Multiunit (10+) Mobile Home

Housing Type

Age

5% 14%

7% 20%

31% 24%

<5 5-17

18-24 25-44 45-64

65+

9% 80%

5% 2% 2%

0% 0%

Hispanic NH White NH Black NH Asian NH AIAN NH NHPI

NH Other

Race/Hispanic Origin Proportion of Segment

Group Quarters

2%

Moved in the Last Year

12%

54%

Share of Self-Response Online

Internet Access

61%

49%

Predicted Self-Response

5,033

Number of Tracts

5%ofU.S.Population

Sparse Spaces: Demographics How will they respond?

Onaverage,thissegmentispredictedtorespondatabelow-averagerate,withabelow-averagepercentageofthatresponsecomingonline.

Who are they?

Thetractsinthissegmenthavealargeproportionofowner-occupiedhousingunits,below-averagelevelsofinternetaccess,andahigherpercentageofresidents45orolderthanthenationalaverage.

Where are they?

TractsinthissegmentcanbefoundinruralareaspredominantlyinthewesternUnitedStates,Appalachia,northernMaine,andMichigan’sUpperPeninsula.

[VALUE]

[VALUE]

Owner vs. Renter

Homeowner

Renter

NH indicates Non-Hispanic.

19 Authorized Use Only

9% 18%

69% 1% 1% 0% 0%

Hispanic NH White NH Black NH Asian NH AIAN NH NHPI

NH Other

Race/Hispanic Origin Proportion of Segment

[VALUE] [VALUE]

13%

3%

Single-unit Multiunit (2-9)

Multiunit (10+) Mobile Home

Housing Type

[VALUE] [VALUE]

Owner vs. Renter

Homeowner

Renter

Age

7% 19%

11% 26% 25%

12%

<5 5-17

18-24 25-44 45-64

65+

Group Quarters

2%

Moved in the Last Year

16%

Share of Self-Response Online

35%

Internet Access

60%

43%

Predicted Self-Response

6,362

Number of Tracts

7%ofU.S.Population

Rural Deltas and Urban Enclaves: Demographics How will they respond?

Onaverage,thissegmentispredictedtorespondatalowrate,withthelowestpercentageofthatresponsecomingonline.

Who are they?

Thetractsinthissegmentincludeareaswithlowmedianhouseholdincomes,below-averagelevelsofinternetaccess,andlowpercentagesofcollege-educatedresidents.Thissegmenthasamajorityofresidentswhoarenon-HispanicAfricanAmerican.

Where are they?

TractsinthissegmentcanbefoundinruralpartsofthesoutheasternUnitedStates,includingMississippi,Alabama,Georgia,andSouthCarolina,withadditionalconcentrationsinurbanareas.

NH indicates Non-Hispanic.

Tailored Design in 2020

Census2020tailoring:§ Targetedmailingbi-lingualmaterials§ TargetedInternetFirst§ TargetedChoice§ Targetedadvertising(messagingandbuys)

But…§ Notusingmultiplecontactstrategies(e-mail,text,phone)§ Noincentives§ Nocustomizedlandingpages§ Nocustomizedmailmaterials

20

TAILOREDINTERVENTION

CityCommons

CollegeTowns

SouthwesternFamilies

Messaging §  Quick–10min/10Qs§  Singleparents§  Schoolsandservicesfor

families/children§  Schoollunches,afterschool

programs,TANF/WIC§  Jobtrainingprograms

§  Quick-10minutes/10Qs§  Mobileoptimization§  Bikelanes§  Dogparks§  Environment§  Celebrities§  Education

§  Families/multi-generational

§  Childrenarefocus§  Communitycenters§  Confidentiality

Incentive Cash AmazonGiftCard Cash

In-Languagematerials No No Spanish

Customizedlandingpage Singleparents YoungpeopleDiversefaces

SpanishFamilies

Modetargeting Choice Web Choice

Textreminder Yes Yes No

USPSInformedDelivery No Yes No

21

What worries me about Web-push for 2020 Census? § Publicperceptionthatyouhavetorespondonline?§ Somehard-to-surveypopulationsprefermail?

§  HispanicsinCensusBarriersAttitudesMotivatorSurvey§  (37%statedpreferenceformail;33%internet;29%nopreference)

§  Stakeholderqualitativeresearch(recentimmigrantspreferredmail)§  Yet…weproject56%ofself-responsefromMulticulturalMosiacsegmentwillbeinternet?

§  Onesegmentprojectedtohavemajorityofresponsesbymail§  RuralDeltasandUrbanEnclaves

§ Webinterfaceusabilityissues?§  NON-IDoption§ Mobileoptimization

22

To conclude: Web-push has arrived for the 2020 Census § PushversusChoicehasbeentestedleadingupto2020withsuccess§ Weexpectmajorityofself-responsestocomefromweb(!)§ RespondentscanuseSmartphones(w/oID)toparticipate§ But…needastrongcommunicationstrategytoreassurethosewhoprefermail

23

Thankyou!

24