The Poisson Extension of the Unrelated Question Model - Improving Surveys with Time-Constrained Questions on Sensitive Topics

Sep 17, 2026

The paper introduces the Poisson extension of the unrelated question model (UQMP), combining the Poisson technique with the unrelated question model to estimate how often a sensitive behavior occurs while preserving participant anonymity. It contrasts UQMP performance with the standard Poisson model using direct questions in a survey on drinking and driving. Although prevalence estimates differ markedly, both models indicate high rates among German traffic participants who generally engage in the behavior. The study discusses how online respondents may skim instructions, affecting result quality, and argues the UQMP or similar approaches remain valuable.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Survey prevalence and time-constrained questioning

From yes/no prevalence to frequency rates

Participant effort and ambiguity in prevalence estimates

Survey Research Methods (2024) Vol. 18, No. 1, pp. 21-38 [https://doi.org/10.18148/srm/2024.v18i1.8252](https://doi.org/10.18148/srm/2024.v18i1.8252)[ ](https://doi.org/10.18148/srm/2024.v18i1.8252)European Survey Research Association ©2024 Author(s) CC BY 4.0 The Poisson Extension of the Unrelated Question Model: Improving Surveys with Time-Constrained Questions on Sensitive Topics Benedikt Iberl1 · Anesa Aljovic1 · Rolf Ulrich1 · Fabiola Reiber2 1Eberhard Karls University of Tübingen 2University of Mannheim The Poisson model (Iberl & Ulrich, 2023) is a new survey technique that enables the estimation of how frequently a certain behavior occurs, while employing easy-to-answer yes/no-questions that refer to a speciﬁc time frame (e.g.,“Did you participate in gambling during the last 12 months?”) . In this paper, this model is combined with the unrelated question model (UQM) by Greenberg et al. (1969). The UQM is another survey technique that guarantees complete and objective anonymity to participants in order to achieve more valid survey results when asking sensitive questions (e.g., about drug use) . The resulting Poisson extension of the UQM (UQMP) is expected to yield valid estimations for how many participants engage in a researched sensitive behavior, and how regularly they do so. The performance of the UQMP was compared to the performance of the standard Poisson model, employing direct questions, in a survey on drinking and driving. While prevalence estimates differ greatly between the UQMP and the standard Poisson model, the results of both models indicate a high rate of drinking and driving among those German trafﬁc participants who generally engage in this behavior. The different prevalence estimates could be due to the fact that some participants in online studies read instructions superﬁcially, lowering the quality of results; we discuss possible causes for these problems and why the UQMP or similar approaches can be valuable nonetheless. Keywords: Randomized Response Technique; Unrelated Question Model; survey research; time-constrained questions; prevalence curve; Poisson process 1 Introduction In survey research, one is often interested in obtaining prevalence estimates describing a certain target behavior. Prevalence estimates can be useful in many politically or socially important ﬁelds, such as for the assessment of public opinion or to evaluate the frequency of criminal or risky behavior like drug abuse. Oftentimes, these prevalence estimates are produced by posing yes/no questions that refer to a particular time frame, such as “Did you gamble in the past 12 months”(e.g., Andrie et al., 2019; Atzendorf et al., 2019; Beck et al., 2021; Birkel et al., 2022; Burr et al., 1989; Ferrante et al., 2012; Han et al., 2015; Isolauri & Laippala, 1995; Linton et al., 1998; McCabe et al., 2006 ; McKetin et al., 2006 ; amaz et al., 2014; Sawyer et al., Corresponding author: Benedikt Iberl, Eberhard Karls University of Tübingen, Tübingen, Germany (Email: [benedikt-jonas.iberl@uni-tuebingen.de](benedikt-jonas.iberl@uni-tuebingen.de)) 2018; Virudachalam et al., 2014) . In this paper, we will call such questions time-constrained yes/no questions. However, one might not only be interested in whether the concerning behavior has occurred within a certain timeframe, but also how often the behavior is shown. So, besides the information about whether someone was gambling in the last year, a researcher might be interested in the rate of this behavior (i.e., the average frequency of the concerning behavior per time unit) . To measure this rate, one could simply ask participants how often they have engaged in the behavior in question within a certain time frame (e.g.,“How often did you gamble in the past 12 months?”) . This kind of questioning technique is also widely used in prevalence research (e.g., Cullen et al., 2018; Miller et al., 2020; Molinaro et al., 2018; Seitz et al., 2020; Soga et al., 2021)

FAQs

What does the Poisson extension of the unrelated question model (UQMP) aim to estimate?

UQMP is designed to estimate both how many participants engage in a sensitive behavior and how regularly they do so, while using anonymity-protecting questioning approaches.

How is UQMP evaluated in the paper?

Its performance is compared with the standard Poisson model using direct questions in a survey on drinking and driving, focusing on differences in prevalence estimates.

Why might UQMP and the standard Poisson model yield different prevalence estimates?

The paper discusses that online study participants may read instructions superficially, which can reduce result quality and lead to different prevalence estimates.

Sep 17, 2026

Uploaded by River Wang