Raphael Nishimura Profile picture
May 12 40 tweets 10 min read Twitter logo Read on Twitter
Live tweeting Methods and Election Polling at #AAPOR here
First is Mike Witherly presenting The Effect of Random Ballot Order in the 2018 and 2022 City of Vancouver Municipal Elections Image
Vancouver municipal elections:
Rare instance of down ballot races
Viva Vancouver: "Random" ballots introduced in 2018 -- appear on the ballot by drawing lots
Reduce ABCD bias: alphabetical order would have a particular negative effect whose last name is Southeast Asian or Latino
Image
RQ:
Did randomized ballot matter?
Did candidate ethnicity matter?
In didn't matter in 2018, it had a very minor impact in 2022 Image
Image
Next is Jennifer Su (SSRS) talking about Emailing Registered Voters: Effectiveness and Sample Representation in Pre-Election State Polling Image
RDD vs. Hybrid Mixed Mode
CNN/SSRS polls:
Focus in MI, PA, WI
2020: RDD
2022: SSRS Opinion Panel + RBS (Aristotle)
Demographic distribution ImageImage
RBS design and contact protocols:
* 9 strata per state based on presence of email, phone number or both
* Undercoverage in RBS covered by inclusion SSRS Opion Panel (national probability-based panel recruited by ABS) Image
Demographic distribution: ImageImageImage
ImageImage
Phone still needed to reach certain groups
Weighted by 2020 recalled presidential vote Image
Next is @kwcollins (Survey 160) presenting Dynamic Response-Rate Adjusted Stratified Sampling for Election Surveys Image
Sampling inversely proportional to response rates, but they are not always stable in time
Proposal: Dynamic Response-Rate Adjusted Stratified Sampling (DRASS)
Adjust probabilities of selection according to the observed response rate in the stratum
Empirical test of this approach in NC, PA, AZ ImageImageImage
Next is @joywilke (BlueLabs) presenting Comparing Modes and Methodologies in Pre-Election Polling Image
Response rates can vary dramatically over the course of the campaign (presenting data from Mid-Sept and Early Nov 2022) Image
Cellphone was really reaching a different group Image
Significant difference across mode about how respondents answering being less motivated to vote by candidate party Image
And to the R-D margins on the elections estimates: ImageImage
Very large differences in costs between modes! Image
Next is Patrick Murray (Monmouth University) talking about The Media “Horse Race” Obsession: Can Polling Improve the Quality of Election Coverage? Image
Not a lot of research about media coverage of the election polls...
We have more tools to judge the quality of polls today.
Turnout models didn't capture voters that would typically only vote in federal elections Image
Polls were showing a close race, but media was not portraitying as such
Solution: present results in a different way, by level of support: Image
17% of media framed as a horse race "gap", which is not how the pollster originally presented the results
[This feels a lot like the media in Brazil showing the valid votes instead of total votes] Image
Next we have Donald Levy (Sienna College) and Spencer Kimball (Emerson College) presenting Comparing Modes and Methodologies in Pre-Election Polling Image
In the spirit of the conference, this is a good example of collaborations between two, presumably competitors, pollsters!
Two pollsters, Sienna and Emerson, polling during same time frame, same survey in NY
Differences in methodologies between the two polls:
Siena using a more traditional live-interviewer phone methodology
Emerson using a mixed-mode approach using IVR, online panel, and text to web Image
Some differences between the two polls: unweighted Emerson more Democratic than Siena, but much closer when weighted Image
Siena's main takeaway: remain as a phone shop 😅
Key to this relationship was trust -- sharing methodologies, data, etc.
@threadreaderapp compile unroll thread

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Raphael Nishimura

Raphael Nishimura Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @rnishimura

May 12
Live tweeting session Detection and Mitigation: Developing Monitoring Tools to Ensure Quality in Interviewer-Administered Data Collection at #AAPOR
First is Joe Murphy (RTI) presenting A Dashboard for Field Survey Data Quality Management Image
Why an *interactive* dashboard?
Different users, different needs
Gives user control to what they need
Read 19 tweets
May 12
Live tweeting the panel of Elections and Nonresponse now here at #AAPOR
First is Cameron McPhee (SSRS) presenting Underestimation or Overcorrection? an Evaluation of Weighting and Likely-Voter Identification in 2022 Pre-Election Polls Image
2022 Election Polls did really well, with maybe some under-estimation of Democrats Image
Read 40 tweets
May 11
Live tweeting the #AAPOR session The Panel on the Panel: Development and Testing of a Probability-Based, Nationally-Representative Survey Panel for Federal Use
First is Victoria Dounoucus (RTI) presenting Qualitative Work to Inform Contact Materials and Baseline Questions for the Ask U.S. Panel Pilot Image
Cognitive interview in Microsoft Teams for ~1 hour, with 30 interview (21 in English, 9 in Spanish)
Read 34 tweets
May 11
Live tweeting the "Gridded Population Surveys" at #AAPOR
First is Dana Thomson (U of Twente) giving an introduction to gridded population sampling Image
Gridded population datasets publicly available
Not all are equal Image
Read 43 tweets
May 10
Tweeting now session "Come In and Stay a While: Recruiting and Measuring Attrition" at #AAPOR
Kyle (SSRS) starts the session talking about chronic-nonrespondents in the SSRS probability panel Image
Defining chronic nonrespondents -- people who have not responded the survey after at least six invitation. Image
Read 40 tweets
May 10
Phillip Hastings (PRAMS) talks about maximum nonresponse bias in PRAMS. Seeing @rrandridge as one of the co-authors, I'm expecting to see some Pattern-Mixture Modelling and maybe even some SMUB or SMAB metrics. Image
Oh yeah, we have some Proxy Pattern-Mixture Models and some sensitivity analysisbright at the 3rd slide! My favorite approach to deal with Missing Not at Random!
For more about Proxy Pattern-Mixture Model, I recommend the excellent paper by Andrige and Little (2011): scb.se/contentassets/…
Read 10 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(