Published October 23, 2026 | Version 1.0

Beyond the Standard R-Indicator: Optimism Trap or Efficiency Trap in Targeted Survey Design Using Paradata?

  • 1. ISTAT

Description

The transition to self-completion modes in European social surveys challenges representativeness when nonresponse is Not Missing At Random (NMAR) and driven by a latent attitudinal factor. Using a Monte Carlo simulation calibrated to the Italian Consumer Confidence Survey (ICC) — a univariate attitudinal survey in which the latent factor driving nonresponse coincides with the construct being measured — this paper evaluates a paradata-derived auxiliary variable, response latency, for targeting NMAR nonresponse, and proposes an augmented R-indicator for the diagnostic monitoring of latent selectivity. Three results emerge. First, once the right-censoring of response latency is properly modelled, latency-based targeting is statistically indistinguishable from simple demographic targeting: the apparent advantage of latency in earlier evidence was an artefact of using information unavailable at the moment of the prioritisation decision. Second, a stepwise two-phase design, implemented as a first approximation to a genuine Adaptive Survey Design, performs substantially worse than the single-phase baseline; the failure is structural rather than parametric or modelling-based. Third, the standard R-indicator is blind to latent selectivity and, under the monotonic NMAR mechanism modelled here, can improve while bias grows. The augmented R-indicator R* proposed in this paper incorporates paradata-derived proxies and captures nearly all the diagnostic power of the theoretical lower bound; the empirical hierarchy R(X, Z) ≤ R(ρ) ≤ R* ≤ R(X) is confirmed for γ ≥ 1.0, with deviations at lower NMAR intensities attributable to finite-sample effects. R* is strategy-dependent, and simple demographic targeting remains preferable.

Preprint deposited to establish intellectual priority. DOI: 10.5281/zenodo.23663248

Files

paperpreprint.pdf

Files (783.2 kB)

Name Size
md5:d17f22fb778577814f92d3a6fe17a9bc
783.2 kB Preview Download

Additional details

Related works

Cites
Publication: 0322-788X (ISSN)
Is supplement to
Conference paper: 3059-2143 (ISSN)

Dates

Available
2026-10-26

References

  • Lima, R. (2026). Design-Based Vs Model-Based Sampling for Consumer Confidence: A Monte Carlo Comparison of ISTAT vs. Adaptive Design. In: Martella, F., Arima, S., Marino, M.F., Mollica, C. (eds) Statistical Science: From Theory to Applied Research III. SIS-FENStatS 2026 2026. Italian Statistical Society Series on Advances in Statistics. Springer, Cham. https://doi.org/10.1007/978-3-032-30881-8_88 Download citation .RIS.ENW.BIB DOI https://doi.org/10.1007/978-3-032-30881-8_88 Published 17 July 2026 Publisher Name Springer, Cham
  • Lima, R. (2014) Italian Methods and Practices Regarding the LFS: Case of Response Rate and Response Burden, Statistika vol. 94 (3) 2014