Improving Minority Population Sampling with BISG Probabilities: Evidence from a Survey of Jewish Americans
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.
Tom: I'm Tom.
Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.
Tom: We're looking at the paper: "Improving Minority Population Sampling with BISG Probabilities".
Jane: Here's the gist of the paper. The paper proposes Bayesian Improved Surname Geocoding (BISG) to enhance the efficiency of sampling geographically dispersed minority populations, specifically focusing on a national survey of Jewish Americans.
Tom: BISG generates individual-level probabilities of minority group membership based on names and residential addresses.
Jane: These probabilities are then incorporated into a stratified Poisson probability sampling design, aiming to improve upon standard disproportionate stratified sampling methods.
Tom: The methodology involves leveraging BISG probabilities within a stratified Poisson sampling framework, allowing for the optimal allocation of stratum-level target sample sizes using these probabilities.
Paper summary: Tom: So to wrap up this discussion on "Improving Minority Population Sampling with BISG Probabilities: Evidence from a Survey of Jewish Americans", the authors are proposing using BISG probabilities within a stratified Poisson probability sampling design to efficiently sample dispersed minority populations.
Jane: The main implication is that even when you can't individually identify people in your frame, you can use this name and location-based probability method to get results that are comparable to expensive surveys like the Pew Research Center survey.
Meng: It makes the process way cheaper for organizations trying to understand these specific communities. It turns a much more costly data collection effort into something much more manageable.
Lu: The work shows that this approach is robust, meaning it stays valid even if those initial BISG probabilities are biased, provided you handle non-response correctly in your final analysis.
Tom: The paper suggests that for sampling the American Jewish population, BISG provides a way to get high success rates, which is what matters when you're trying to gather reliable data on a specific group.
Jane: It’s about using sophisticated statistical modeling to overcome real logistical hurdles in minority population research.
Lalam: The final point is that the BISG method, with filtering applied, achieves a sample size of forty-nine thousand people and produces results comparable to established methods like Pew's survey, which is what they claim holds up under scrutiny on page one of this work >
Conclusion: Tom: So, we're wrapping up on this paper about improving how we sample minority groups using something called BISG probabilities for Jewish Americans.
Jane: It really boils down to taking names and where people live to get a better idea of their group membership, which you can then use in a sampling plan.
Lu: The core idea is building these probabilities using Bayesian hierarchical models that look at data from Jewish obituary records, specifically linking surnames and first names to geographic locations.
Meng: From an engineering standpoint, it’s smart because it lets us allocate our survey resources more efficiently based on those calculated probabilities instead of just picking random people.
Tom: The authors show that this method gets results that are surprisingly close to what a much bigger, more expensive survey like Pew Research Center found.
Jane: They point out they're talking about saving a huge amount of money—about twenty-three dollars per person compared to two hundred twenty-eight for a full interview.
Lu: And the success rate for their proposed Poisson sampling method with filtering is estimated above fifty-eight point six percent for a target sample size of fifty thousand people, which is quite high compared to simple random sampling.
Meng: That's good because we’re dealing with real data at scale, and that success rate suggests the model isn't just theoretical noise.
Tom: The paper validates a predictive model estimating Jewish identity based on name and geography for registered voters, which is a big step for targeted research.
Jane: It shows that even if you can't directly ask everyone in the group, this probability approach gives you estimates that align well with established benchmarks.
Lu: They did find some interesting correlations between state distributions estimated by AJPP and the actual obituary locations, though they noted California and New York were a bit under-represented while Florida was over-represented.
Meng: So it’s not perfect, but those kinds of localized biases are exactly what we need to fix when we deploy these systems in production.
Tom: The title itself says "Improving Minority Population Sampling with BISG Probabilities," which perfectly captures the mechanics and the goal of making sampling more precise.
Jane: And the implication is that for researchers, especially those studying smaller, dispersed communities, this can be a very cost-effective way to get high-quality data.
Lu: It opens up a new way to integrate name and place data into larger survey designs without needing expensive individual interviews for every single person.
Tom: This moves us from just counting what we find in existing records to building models that predict group membership across a whole population structure.
Jane: If this works consistently, it suggests that we can build more accurate demographic pictures of specific groups using the information already available in public, record-based datasets.
Department of Statistics, Harvard University · Department of Political Science, Tufts University · Department of Government, Harvard University
stat.AP, stat.ME
Submitted: 2026-05-06
Updated: 2026-11-05
Code: https://github.com/kchaz/BISG_sampling_public
Project page: https://www.pewresearch.org/religion/2021/05/11/jewish-americans-in-2020
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 99/100
The gist: The paper proposes Bayesian Improved Surname Geocoding (BISG) to enhance the efficiency of sampling geographically dispersed minority populations…
Key concepts
- Bayesian Improved Surname Geocoding (BISG)
- A statistical method that uses Bayesian hierarchical models to estimate the probability of an individual belonging to a minority group. It combines information from geo-located surnames in obituary data with first name details to create a more accurate membership prediction.
- Stratified Poisson Probability Sampling
- A sampling design where the target sample size for each geographic stratum is determined using the calculated BISG probabilities. This method optimizes how many people to sample from different areas to ensure representation while maintaining statistical efficiency.
- Efficiency Gains
- The benefit of using BISG is that it allows researchers to achieve results comparable to expensive, detailed surveys much more cheaply. The method demonstrated substantial gains, yielding estimates closely aligned with those from a costly survey for a fraction of the expense.
Terminology
Summary
The paper proposes Bayesian Improved Surname Geocoding (BISG) to enhance the efficiency of sampling geographically dispersed minority populations, specifically focusing on a national survey of Jewish Americans. p. 0 Although stratified sampling can
offer efficiency gains, these gains are typically modest unless the minority population is highly
concentrated within a small number of strata. In this paper, we propose using Bayesian Improved Surname Geocoding (BISG) to enhance the efficiency of minority population sampling. In this paper, we propose using Bayesian Improved Surname Geocoding (BISG) to enhance the efficiency of minority population sampling. BISG generates individual-level probabilities of minority group membership based on names
and residential addresses. Our
approach leverages Bayesian Improved Surname Geocoding (BISG), a widely used technique for
estimating the probability that an individual belongs to a minority group based on their surname
and geographic location (e. We show that these BISG probabilities can be incorporated
into a stratified Poisson probability sampling design, yielding substantial efficiency gains for sampling minority populations.Where the paper says it
Where the paper says it
Where the paper says it
The methodology involves leveraging BISG probabilities within a stratified Poisson sampling framework, allowing for the optimal allocation of stratum-level target sample sizes using these probabilities. p. 7 We then propose an alternative Poisson sampling method based on BISG
probabilities. We show how to optimally allocate stratum-level target sample sizes using BISG
probabilities. We incorporate these probabilities into a stratified Poisson probability sampling design. Applying the proposed approach to a national survey of Jewish Americans,
we find that our estimates closely align with those from a large-scale Pew Research Center survey of the same population, which relied on a substantially more expensive sampling strategy
involving geographic stratification and screening.Where the paper says it
Where the paper says it
The study utilizes Bayesian hierarchical models to estimate BISG probabilities, incorporating both geo-located surnames available in Jewish obituary data and first name data extracted from obituary texts. p. 1 Standard approaches to
BISG probability estimation may be used here when applicable, but in Section 2. 3, we present a
method for estimating the BISG probabilities when geo-located surnames are available as in our
obituary data. We show that these BISG probabilities can be incorporated
into a stratified Poisson probability sampling design, yielding substantial efficiency gains for sampling minority populations. Importantly, the proposed methodology remains valid even if BISG
probabilities are biased. We first introduce the standard disproportionate stratified
sampling and then propose our Poisson sampling method based on BISG probabilities. The standard disproportionate stratified sampling method for rare population sampling chooses
target stratum sample sizes ng· that sum to a fixed target size n and minimize the approximate
variance of the stratified estimator for some outcome Y .Where the paper says it
Where the paper says it
Where the paper says it
The data used includes information from deceased individuals across 109 Jewish funeral homes in 30 states, with some locations having very high observation counts (e. p. 4 This provides information about which names are rare or common in
at least a broad subset of Jewish Americans. After manually assembling a list of 109 Jewish funeral homes across the United States (see
Appendix B for details), we contracted with PBI Research Services, a commercial provider that
compiles obituary records. Because our obituary data are aggregated at the state level, we use the AJPP’s
state-level estimates in our analysis. These data show that the Jewish population in the United States is highly geographically
concentrated: approximately half of all Jewish Americans reside in just four states—New York,
California, Florida, and New Jersey—while many states have Jewish population shares below 1%. Because our obituary data are aggregated at the state level, we use the AJPP’s
state-level estimates in our analysis. These data show that the Jewish population in the United States is highly geographically
concentrated: approximately half of all Jewish Americans reside in just four states—New York,
California, Florida, and New Jersey—while many states have Jewish population shares below 1%. As a result, stratified sampling based on geography
alone offers only limited gains in efficiently identifying Jewish respondents. Our approach combines existing estimates of the geographic distribution of the minority group
with overall name data from the U. L2 is a leading national non-partisan firm and the oldest
organization in the United States that supplies voter data and related technology to candidates,
political parties, pollsters, and consultants for use in campaigns. L2 combines public state-level
lists of registered voters with other commercial data and modeling to give insights on voters. We first introduce the standard disproportionate stratified
sampling and then propose our Poisson sampling method based on BISG probabilities. The standard disproportionate stratified sampling method for rare population sampling chooses
target stratum sample sizes ng· that sum to a fixed target size n and minimize the approximate
variance of the stratified estimator for some outcome Y .Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Sample from geographically dispersed minority populations efficiently by using Bayesian Improved Surname Geocoding (BISG) to generate individual-level probabilities of minority group membership based on names and residential addresses. This allows for a stratified Poisson probability sampling design, yielding
substantial efficiency gains for sampling minority populations.
-
Estimate the probability that an individual belongs to a minority group by combining multiple data sources, including name distribution through obituaries from Jewish funeral homes, geographic distribution estimates from sources like Brandeis's American Jewish Population Project (AJPP), and baseline name distributions from the U.S. voter file.
-
Implement a stratified Poisson sampling allocation formula that optimally targets stratum-level sample sizes by using the probability P(R = 1 G = g, I = 1) derived from BISG probabilities, which is
larger when P(R = 1 S = s, G = g) is larger for many surnames within a given stratum.
-
Develop a Bayesian hierarchical model with MCMC sampling to estimate the conditional posterior mean of P(S = s G = g, R = 1), using shrinkage towards the overall surname distribution to mitigate high variance in small counts. This allows for
valid downstream analysis even when estimated BISG probabilities have error.
-
Design survey non-response mitigation strategies by employing post-stratification or raking methods initialized with Inverse Probability Weighting (IPW) weights, which are then adjusted to match demographic distributions from the gold standard Pew survey to ensure
remarkably close alignment between our survey and Pew.
Abstract
Sampling geographically dispersed minority populations poses substantial challenges when individual group membership cannot be directly observed. Although stratified sampling can offer efficiency gains, these gains are typically modest unless the minority population is highly concentrated within a small number of strata. In this paper, we propose using Bayesian Improved Surname Geocoding (BISG) to enhance the efficiency of minority population sampling. BISG generates individual-level probabilities of minority group membership based on names and residential addresses. We incorporate these probabilities into a stratified Poisson probability sampling design. Applying the proposed approach to a national survey of Jewish Americans, we find that our estimates closely align with those from a large-scale Pew Research Center survey of the same population, which relied on a substantially more expensive sampling strategy involving geographic stratification and screening. At a fraction of the cost, our survey reproduces nearly identical patterns observed by Pew, including estimates of religious denomination prevalence and participation in specific religious activities.
Sources
- Estimating Racial Disparities When Race is Not Observed
- Using Embedding Models to Improve Probabilistic Race Prediction
Related papers
- Unraveling time-varying causal effects of multiple exposures: integrating Functional Data Analysis with Multivariable Mendelian Randomization
- Strategic Causal Policy Learning with Manipulable Eligibility
- BayesJudge: Uncertainty-Aware Bayesian Meta-Evaluation of Human and LLM Judgments
- Automating structural reliability analysis with a multi-agent large language model framework
- When Do Sports Leagues Stop Being Competitive? Bayesian Detection of Competitive-Balance Eras from Match-Level Data
- A Bayesian Approach for Earthquake Impact Modelling