Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Mindsets Revisited: Results of the Second Iteration of the Census Barriers, Attitudes, and
Motivators Survey
Monica J. Wroblewski, U.S. Census BureauFrederica R. Conrey, Booz Allen
Randal ZuWallack, ICF MacroRobynne Locke, ICF Macro
AAPOR Orlando, FL
May 17, 2012
Background CBAMS I Survey conducted in support of the 2010 Census
Integrated Communications Program (ICP) Fielded from July – August 2008 by ICF Macro
Measured attitudes towards and knowledge of the Census, potential motivators and barriers to participation, ranking of potential messages, and demographics
Revealed 5 distinct mindsets among the population that varied in knowledge of and attitudes toward the Census
Results contributed to the development of targeted 2010 Census messaging
CBAMS II - Methodology Fielded May – July 2011 Landline, cell phone, and in-person interviews Included HTC populations: American Indian Reservations, sites with
high Hispanic population density, sites with high Asian population density, and rural areas with high poverty
Conducted in English and Spanish for all modes and also in Vietnamese and Chinese for in-person surveys
$10 incentive for in-person interviews 4,071 completed interviews (combined response rate of 33.6%)*
2,004 landline (26% RR) 995 cell phone (16% RR) 1,071 in-person (64% RR)
*Response rates calculated using AAPOR RR3
CBAMS II Research Questions:
1. What is the best method for identifying Census mindsets;
2. Did mindsets change since the implementation of the 2010 Census ICP;
3. Can mindsets be categorized differently moving forward, and if yes, what are the new mindsets; and
4. What are the profiles of the new mindsets.
What is the best method for creating mindsets? Methods Tested
K-Means Q-Factor Latent Class Analysis (LCA)
Decision Criteria1. Segment Distribution and Size: Are the models producing actionable segments?2. Distinction of Segments: Clarity of profiles and distinguishing features3. Reclassification: Discriminant analysis to evaluate re-creation of segments
Used a total of 21 variables K-Means and Q-Factor Continuous or ordinal variables LCA Categorical variables
Tested 4, 5, and 6 mindset solutions for each method
Distribution of Mindsets for Four, Five, and Six Mindset Solutions Using Different Methods
Four Mindset Solutions Five Mindset Solutions Six Mindset Solutions Mindset K‐Means Q‐Factor LCA K‐Means Q‐Factor LCA K‐Means Q‐Factor LCA 1 36% 36% 36% 28% 28% 33% 23% 30% 24% 2 22% 29% 31% 26% 24% 25% 18% 22% 23% 3 22% 19% 18% 23% 19% 17% 16% 19% 18% 4 19% 16% 15% 14% 16% 13% 15% 16% 14% 5 9% 13% 12% 14% 11% 13% 6 13% 2% 9%
Range of Percent Intent to Respond for Four Mindset Solution Using Different Methods
Mindset K‐Means Q‐Factor LCA 1 76% 79% 88% 2 68% 65% 77% 3 56% 59% 62% 4 59% 59% 35% Overall 20% 20% 53%
Average range across all variables for 6 segment solution 23% for K‐Means 27% for Q‐Factor 37% for LCA
Error Rates in Classification for Four, Five, and Six Mindset Solutions Using Different Methods Four Mindset Solutions Five Mindset Solutions Six Mindset Solutions Mindset K‐Means Q‐Factor LCA K‐Means Q‐Factor LCA K‐Means Q‐Factor LCA 1 4% 29% 10% 4% 29% 11% 5% 33% 17% 2 10% 20% 10% 4% 20% 13% 9% 12% 19% 3 13% 8% 9% 12% 10% 12% 15% 26% 13% 4 10% 13% 2% 3% 14% 4% 3% 15% 2% 5 19% 17% 8% 10% 19% 14% 6 12% 4% 1% Overall 9% 18% 8% 8% 18% 10% 9% 18% 11%
Average error rate across solutions 9% for K‐Means 18% for Q‐Factor 10% for LCA
Summary: The Best Method for Creating Mindsets
LCA produces meaningful and actionable segment sizes K-Means tends to equalize segment distribution; Q-Factor produced a very small cluster
Groups identified using LCA had more distinct features than did groups identified using the other two approaches
K-Means and LCA had similar, low error rates, and Q-Factor analysis had consistently higher error rates
Recommendation: Use LCA to define CBAMS II mindsets
How are mindsets now different from mindsets created prior to the 2010 Census?
Comparison based upon retrospective fit of segments with variables common to CBAMS I and II (26 variables) using LCA
LCA model selection statistics suggest 4 mindsets for CBAMS I plus the Unacquainted
4 mindsets generally have characteristics that align with Leading Edge, Head Nodders, Insulated and Cynical Fifth
Group sizes changed
Additional large mindsets emerge in CBAMS II when adding a fifth and sixth segment, but not for CBAMS I
Implications Increase in Census affinity Cynical Fifth equivalent is less knowledgeable, suggesting migration
of most knowledgeable members to Leading Edge equivalent Increased knowledge of Census uses among Insulated equivalent Unacquainted down from 7 to 3% suggesting an overall increase in
Census awareness
Comparison of Mindset Sizes
CBAMS ILCA
CBAMS IILCA
Leading Edge equivalent 20% 35%*
Head Nodders equivalent 20% 19%
Insulated equivalent 23% 23%
Cynical Fifth equivalent 30% 20%*
Unacquainted 7% 3%*
* Significant at p < 0.05
What are the CBAMS II Mindsets?
Results from previous question suggested there are now more than 5 segments
Explored 5, 6, 7, and 8 segment solutions with expanded set of variables Common to CBAMS I: affinity, knowledge, beliefs New to CBAMS II: trust in government/privacy, important issues,
paperwork, motivators
Recommendation: 7 segments
A. Gov’t Perception
B. Knowledge,Internet Preferences
C. Gov’t Trust, Paperwork Attitudes, andInternet Preferences
D. Knowledge,Paperwork Attitudes, and Internet Preferences
A
BDC
What are the final CBAMS II mindsets? 3 High Affinity Groups/Census is Important: Government-Minded Compliant and Caring Dutiful
1 Indifferent Group: Local-Minded
3 Low Affinity/Census is not Important Uninformed Cynical Suspicious
Government Minded (19%) Attitudes Completing the Census is a civic responsibility
Knowledge High knowledge of what the Census is and is not
used for Motivators Political Representation; Roads and Highways;
Fire and Police Services Demographics Educated, Affluent, Married, White, Mostly born in
the US, and Mostly speak English only
Compliant and Caring (15%) Attitudes Completing the Census is a civic responsibility Thinks there is personal benefit to completing the Census
Knowledge High knowledge of what the Census is and is not used for
Motivators Schools; Daycare; Health Care; Elder Care; Job Training
Demographics Female, Married, but otherwise Demographically Diverse
Dutiful (14%) Attitudes Completing the Census is a civic responsibility Responsibility to let the government know what the
community needs Knowledge Knowledgeable about Census uses Misinformed about what the Census is not used for
Motivators Political Representation; Schools; Fire and Police
Services; Healthcare Demographically Diverse
Local-Minded (12%)
Attitudes Believe government has their best interests in mind
Knowledge Misinformed about what the Census is not used for
Motivators Schools; Daycare; Health Care; Elder Care; Job Training
Demographics Female, Less Educated, Lower Income, Non-White,
and Many Immigrants
Uninformed (16%) Attitudes Does not believe the Census can personally
benefit or harm Knowledge Knowledgeable about what the Census is not used for Misinformed about Census uses
Motivators Hospitals and Elder Care
Demographics Less Educated, Lower Income, and Many do not speak
English in the home
Cynical (10%) Attitudes Concerned about personal information and privacy Low trust in the government
Knowledge Knowledgeable about what the Census is not used for
Motivators Political Representation; Roads and Highways; Fire
and Police Services Demographics Male, Older, Married, White, Mostly born in the US,
and Most speak English only
Suspicious (14%) Attitudes Low intent to reply Thinks the Census can personally harm
Knowledge Misinformed about Census uses
Motivators Schools and Healthcare
Demographics Young, Single, Mobile, Less Educated, and Ethnically
Diverse
Mindsets in Action Developing an easy to use segmentation tool Recruit for Focus Groups, etc. Append to Pre-Existing Surveys
Coordinating outreach efforts with Census partners
Future Iterations of CBAMS Monitor Mindsets Use results to aid in the design of the 2020 Census
communications campaign
Conclusions Identified a new method to classify mindsets: LCA
Found that, with respect to CBAMS I mindsets, CBAMS II mindsets are more knowledgeable about the Census with higher Census affinity
Seven final CBAMS II mindsets