arxiv: 2604.11760 · v1 · submitted 2026-04-13 · 💰 econ.GN · q-fin.EC

Recognition: no theorem link

Effects of interviewers on response to income and wealth items

Moslem Rashidi

Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:57 UTC · model grok-4.3

classification 💰 econ.GN q-fin.EC

keywords item nonresponseinterviewer effectsincome reportingwealth itemsSHARE surveyfinancial questionssurvey methodologymissing data

0 comments

The pith

Interviewers who expect higher income response rates receive more financial answers from respondents.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether interviewers' beliefs about how likely respondents are to answer income questions predict whether those respondents actually supply the data in a major European ageing survey. It connects interviewer survey responses on expected nonresponse to real item-level answers from over 41,000 people in twelve countries, while comparing several standard ways to handle missing covariates. In most model specifications the link holds: interviewers who anticipate better cooperation obtain fuller reports on income and assets. This finding matters because item nonresponse on financial topics remains a major source of bias in household surveys and any systematic way to reduce it could improve data quality on wealth, retirement, and inequality.

Core claim

Using linked interviewer and respondent records from SHARE Wave 6, the study shows that respondents interviewed by interviewers who expected higher response rates to income items were more likely to provide answers to four financial questions with high nonresponse. The association appears across complete-case analysis, multiple imputation, and a generalized missing-indicator approach with model averaging, although the latter does not produce clear improvements over simpler methods. The results indicate that interviewer expectations contain usable information for understanding and potentially reducing item nonresponse on sensitive financial topics.

What carries the argument

The observed association between interviewers' stated expectations of respondent willingness to report income (drawn from a separate interviewer survey) and the actual item response rates on income and wealth questions, estimated under alternative missing-data treatments.

If this is right

Interviewer expectations supply information that can be used to model and predict item nonresponse on sensitive financial questions.
Survey fieldwork could incorporate interviewer expectation measures when planning training or respondent assignment to reduce missing financial data.
Simpler methods for handling missing covariates perform at least as well as more elaborate model-averaging procedures in this setting.
The pattern holds across most specifications and multiple countries, suggesting it is not an artifact of one particular modeling choice.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the association stems from interviewer behavior rather than respondent selection, targeted training that raises expectations could improve data completeness on income and wealth items.
Future studies could test whether the same expectation-response link appears for other sensitive topics such as health or family details.
Randomized interviewer assignment experiments would help separate expectation effects from confounding assignment processes.

Load-bearing premise

That interviewers' expectations about response rates are not themselves shaped by unmeasured respondent traits or by the way interviewers are assigned to respondents.

What would settle it

A design in which interviewers are randomly reassigned to respondents and the statistical link between their prior expectations and observed response rates disappears.

Figures

Figures reproduced from arXiv: 2604.11760 by Moslem Rashidi.

read the original abstract

Item nonresponse to financial questions is a persistent source of survey error, especially in interviewer-administered surveys. We examine whether interviewers' expectations about respondents' willingness to report income are associated with actual item responses to income and asset questions in Wave 6 of the Survey of Health, Ageing and Retirement in Europe (SHARE). Using data from 41,934 respondents in 12 countries, linked to interviewer survey and roster information, we analyze responses to four financial items with substantial nonresponse. We compare three approaches to handling missing covariates: complete-case analysis, multiple imputation (fill-in methods), and a generalized missing-indicator framework with information-criterion-based model averaging. Across most specifications, respondents interviewed by interviewers with higher expected income response rates are more likely to provide financial information. However, model averaging does not yield clear gains over simpler approaches. The results suggest that interviewer expectations contain useful information for understanding and modeling item nonresponse to sensitive financial items, with potential implications for interviewer training and survey fieldwork design.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper links measured interviewer expectations to lower nonresponse on income and wealth items in SHARE but the association may reflect assignment processes rather than expectations themselves.

read the letter

The main thing to know is that this paper reports an association in SHARE Wave 6: respondents paired with interviewers who expect higher income response rates are more likely to actually answer the financial questions. The link appears across most specifications on a sample of 41,934 people in 12 countries, using linked interviewer survey data and roster information. They test four items with high nonresponse and compare three missing-data strategies: complete-case, multiple imputation, and a missing-indicator approach with model averaging. The finding is incremental but uses real linked data rather than just inferring interviewer effects. The comparison of missing-data methods is a reasonable addition even though the averaging step does not clearly beat the simpler options. The work is honest about the empirical setup and sticks to what the data show without overclaiming causality. The soft spot is identification. Interviewer expectations could easily be shaped by the areas or respondent types they are routinely assigned to, and those same factors can affect willingness to report finances. The abstract and stress-test note do not show strong evidence that assignment is independent of those traits or that the authors controlled for within-interviewer variation. The missing-data checks address covariate gaps but not this selection issue, so the result stays associative. This paper is mainly for survey methodologists and SHARE users who care about item nonresponse in financial modules. It adds a modest data point on interviewer expectations but does not resolve the endogeneity problem. I would send it to peer review because the linked dataset and the question are worth referee input on identification and robustness, even if the current version needs work there.

Referee Report

2 major / 2 minor

Summary. The paper examines whether interviewers' expectations about respondents' willingness to report income (measured in a linked interviewer survey) are associated with actual item nonresponse to four financial questions in SHARE Wave 6. Using 41,934 respondents from 12 countries and three missing-covariate strategies (complete-case, multiple imputation, and missing-indicator model averaging), the authors report that higher expected response rates predict higher actual response rates across most specifications, with no clear advantage to model averaging.

Significance. If the reported association is not driven by selection into interviewer assignments, the result provides a practical signal for predicting and potentially mitigating item nonresponse on sensitive financial items, with direct implications for interviewer training and fieldwork protocols. The multi-method robustness check on a large linked dataset is a methodological strength, though the absence of gains from model averaging reduces the contribution on the missing-data side.

major comments (2)

[Empirical strategy / Results] The central association may be confounded by interviewer assignment processes. The manuscript does not report controls for area-level socioeconomic characteristics, prior interviewer experience, or roster-based matching that could jointly shape both pre-existing expectations (from the interviewer survey) and respondent item response; without such checks or within-interviewer variation, the link cannot be distinguished from selection (see abstract and the three missing-data approaches).
[Title and abstract] The title refers to 'Effects of interviewers' while the abstract and results describe associations only. No identification strategy (e.g., fixed effects, instrumental variables, or explicit exogeneity tests for expectations) is presented to support causal language, making the interpretation of the association as an 'effect' load-bearing for the paper's framing.

minor comments (2)

[Abstract] Clarify in the abstract whether the 'across most specifications' claim survives the full set of assignment-related covariates or is limited to the three missing-data variants.
[Results] The statement that model averaging 'does not yield clear gains' should be accompanied by a direct comparison table of coefficient magnitudes and standard errors across the three approaches.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below, indicating planned revisions to the manuscript.

read point-by-point responses

Referee: [Empirical strategy / Results] The central association may be confounded by interviewer assignment processes. The manuscript does not report controls for area-level socioeconomic characteristics, prior interviewer experience, or roster-based matching that could jointly shape both pre-existing expectations (from the interviewer survey) and respondent item response; without such checks or within-interviewer variation, the link cannot be distinguished from selection (see abstract and the three missing-data approaches).

Authors: We agree that interviewer assignment processes represent a plausible source of confounding that merits explicit attention. The linked dataset includes roster information used to connect interviewers and respondents, but the reported specifications do not include area-level socioeconomic covariates or measures of prior interviewer experience. In the revised manuscript we will add these controls (where available in the data) and present both the original and augmented results. We will also examine interviewer fixed-effects specifications to assess whether the association persists within interviewers. At the same time, we recognize that even with these additions the design remains observational and cannot fully separate expectations from selection; we will expand the discussion of this limitation and avoid any causal language. revision: partial
Referee: [Title and abstract] The title refers to 'Effects of interviewers' while the abstract and results describe associations only. No identification strategy (e.g., fixed effects, instrumental variables, or explicit exogeneity tests for expectations) is presented to support causal language, making the interpretation of the association as an 'effect' load-bearing for the paper's framing.

Authors: We accept the referee's observation. The title uses 'Effects,' which can imply causality, while the analysis and abstract are framed in terms of associations. We will change the title to 'Associations between Interviewers' Expectations and Response to Income and Wealth Items' (or a close variant) and ensure the abstract, results, and discussion consistently employ associative language. No identification strategy for causal effects is available in the data, and we will make this explicit. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical association on external survey data

full rationale

The paper reports an observational study linking interviewer expectations (from a separate interviewer survey) to item nonresponse in SHARE Wave 6 data. It employs standard regression specifications, three missing-data treatments (complete-case, multiple imputation, missing-indicator with model averaging), and reports associations across specifications. No derivations, first-principles predictions, self-definitional constructs, or fitted parameters renamed as out-of-sample predictions appear. The central result is a statistical association tested on independent data sources; it does not reduce to its inputs by construction. Self-citations, if present, are not load-bearing for any uniqueness claim or ansatz. This is the normal case of a non-circular empirical paper.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard survey-statistical assumptions about the measurement of interviewer expectations and the handling of missing covariates; no free parameters, invented entities, or ad-hoc axioms are described in the abstract.

axioms (1)

domain assumption Interviewer expectations measured in the separate interviewer survey are valid proxies for the expectations that influence respondent behavior during the interview.
This assumption is required to interpret the linked data as evidence of an expectation-response link.

pith-pipeline@v0.9.0 · 5462 in / 1238 out tokens · 62572 ms · 2026-05-10T15:57:41.084444+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

6 extracted references

[1]

and Burnham, K.P

Anderson, D.R. and Burnham, K.P. (2002). A voiding pitfalls when using information- theoretic methods. The Journal of wildlife management 66, 912–918. Banks, J., Muriel, A., and Smith, J. (2011). Attrition and he alth in ageing studies: Evidence from ELSA and HRS. Longitudinal and Life Course Studies , 2: 1–29. Bergmann, M., Franzese, F., and Schrank, F. ...

2002
[2]

and Scherpenzeel, A., (2 017)

Bergmann, M., Kneip, T., De Luca, G. and Scherpenzeel, A., (2 017). Survey participation in the survey of health, ageing and retirement in Europe (SHA RE), Wave 1-6. Munich: Munich Center for the Economics of Aging . Berk, M. L., and Bernstein, A. B. (1988). Interviewer charac teristics and performance on a complex health survey. Social Science Research, ...

1988
[3]

and Augustin, N.H

Buckland, S.T., Burnham, K.P. and Augustin, N.H. (1997). Mo del selection: an integral part of inference. Biometrics 53, 603–618. Cunha, C., Matos, A.D., Voss, G. and Machado, C.,

1997
[4]

In Challenges and Trends in Organizational Management and Industry

Inter viewer characteristics and nonresponse survey outcomes: A Portuguese case study. In Challenges and Trends in Organizational Management and Industry. Springer , Cham. 95–111 Dardanoni, V., Modica, S., and Peracchi, F. (2011). Regress ion with imputed covariates: A generalized missing-indicator approach. Journal of Econometrics , 162: 362–368. 22 Dard...

2011
[5]

Dardanoni, V., De Luca, G., Modica, S., and Peracchi, F. (201 5). Model averaging estima- tion of generalized linear models with imputed covariates. Journal of Econometrics , 184: 452–463. De Luca, G., and Peracchi, F. (2012). Estimating Engel curve s under unit and item non- response. Journal of Applied Econometrics , 27: 1076–1099. Durrant, G. B., Grove...

2012
[6]

and Kohnke-Aguirre, L

Singer, E. and Kohnke-Aguirre, L. (1979). Interviewer expe ctation eﬀects: A replication and extension. Public Opinion Quarterly , 43: 245–260. Sudman, S., Bradburn, N.M., Blair, E.D., and Stocking, C. (1 977). Modest expectations: The eﬀects of interviewers’ prior expectations on response s. Sociological Methods & Research, 6: 171–182. Tourangeau, R., an...

1979