= A New Paradigm for Polling =

'''A New Paradigm for Polling''' (DOI: https://doi.org/10.1162/99608f92.9898eede) was written by Michael A. Bailey in 2023. It was published in the ''Harvard Data Science Review'' (issue 5, no. 3).

The author builds upon [[StatisticalParadisesAndParadoxesInBigData|Meng (2018)]]. They argue that nonresponse in polling has exacerbated to the point that random sampling is a defunct practice.

They instead argue for a paradigm of '''random contact''' surveys, wherein a representative sample much smaller than the true population is contacted. Probability sampling goes out the door, such that the only relevant sampling fraction (''f=n/N'') is the response rate among those contacted (''p,,r,,''). The motivation is that nonresponse leads to an unrepresentative analytic sample, and the nonresponse bias accumulates with the population size, leading to a problem that cannot be solved by collecting more data. Working with a representative nonprobability sample, especially one has been assembled from highly responsive individuals, will lead to smaller actual error.

Assuming that the set of individuals contacted is representative, in that {{attachment:eq.svg}} holds, the Meng equation can be reformulated as:

{{attachment:meng.svg}}

The author also calls for experimental research to estimate ''ρ,,R,Y,,''.



----
CategoryRicottone CategoryReadingNotes