A new open-access paper supported by EMMELO: Correcting for Nonignorable Nonresponse Bias in Ordinal Observational Survey Data

A new open-access paper supported by EMMELO, "Correcting for Nonignorable Nonresponse Bias in Ordinal Observational Survey Data", has been published in Political Analysis. Written by Lukáš Lafférs, EMMELO's Jozef Michal Mintal and Ivan Sutóris, it addresses a growing problem in survey research. Standard weighting brings a sample into line with the population on demographics, but it cannot help when the people who decline to take part hold different views on the very question being asked. In that case, even a well-weighted survey can give a biased picture. This matters for research on contested topics such as gender attitudes and political extremism, where the people least willing to answer may be the ones whose views differ most.

The paper offers a practical correction for answers measured on ordinal scales, such as trust, satisfaction and policy attitudes. It uses information on how willing respondents were to cooperate to extrapolate from reluctant respondents to those who never answered. Applied to the 2024 American National Election Study, the correction shifts results markedly for life satisfaction, modestly for views of the economy, and hardly at all for the death penalty or trust in the media. Whether nonresponse bias matters depends on the question, and it can now be tested rather than assumed. The method runs in R on a standard laptop, and the code and a reproducible capsule are openly available.

Read the paper: https://doi.org/10.1017/pan.2026.10055