Undisclosed tinkering in Excel behind economics paper
An economics paper has been retracted after it emerged the author used Excel’s autofill and even copied data from neighboring countries (e.g., UK for US, Netherlands for New Zealand) to patch missing values without disclosing this in the methodology. Commenters argue over whether the core problem lies in misused tools like Excel or in deliberate or negligent data falsification, noting that proper imputation methods and transparent reporting are standard expectations in serious research. The episode is cited alongside broader concerns about data handling, weak incentives, and low replication rates in parts of economics and the social sciences.
Excel “tinkering” vs misconduct
- Posters clarify that the issue is not an obscure Excel trick but crude data fabrication: copying values from adjacent rows/columns, even from different countries (e.g., UK for US, Netherlands for New Zealand).
- Some argue Excel’s autofill and UI “encourage” such behavior by hiding relationships and patterns; others insist the tool is neutral and the behavior is squarely the researcher’s responsibility.
- There’s disagreement on intent: some see deliberate falsification to complete a publishable dataset; others read the professor’s openness with the data as evidence of incompetence rather than fraud.
Imputation, econometrics, and standards
- Several commenters note that proper econometrics is explicit about missing data and uses partial identification or structured imputation; it does not silently fill gaps with other countries’ data.
- Some say the worst problem isn’t interpolation per se but borrowing values from similarly named or adjacent countries, which makes the dataset useless for inference.
- Others argue the core ethical breach is nondisclosure: if the imputation had been fully described, peer review would likely have rejected the work or treated conclusions as extremely weak.
Tools vs scientific diligence
- Multiple comments push back on framing this as an “Excel problem”: the same misconduct could occur with R, Python, or even pencil and paper.
- Counterpoint: better tooling and clearer UIs can reduce accidental errors and make questionable manipulations more visible, raising the floor of practice.
- Several note parallels with LLMs and “smart” fills: powerful but opaque pattern-filling tools require verification and can silently introduce spurious structure.
Social sciences, replication, and incentives
- One subthread generalizes to social sciences: anecdotes of tiny datasets inflated by aggressive imputation, weak quantitative skills, and unpublished “inconvenient” results.
- Others defend social science broadly, arguing that replication rates vary by subfield, incentives are misaligned across all sciences, and the solution is better incentives, not abandoning the field.
- Data from the replication crisis are cited: some psychology subfields replicate poorly; certain medical research replicates even worse.
Journal status and policy impact
- Commenters note the publishing journal is not a recognized economics journal and far from the discipline’s “top 5,” which are widely agreed upon and dominate tenure decisions.
- Historical Excel-related errors in influential macroeconomic work on austerity are recalled, with the observation that policymakers often cherry-pick studies that confirm prior ideological commitments.
University branding tangent
- A long side discussion covers legal protections and marketing around the terms “university,” “college,” “university of applied sciences,” etc., in several European countries, and how this can mislead international students about institutional status and degree value.