authorship

New statement (23 Dec).

People

OT: Oyvind Thomassen, Seoul National University

HS: Howard Smith, Oxford

SS: Stephan Seiler, Stanford GSB

PS: Pasquale Schiraldi, LSE

Working papers

ST: Multi-Category Competition and Market Power: A Model of Supermarket Pricing (by OT and HS)

SSS: Measuring Market Power in the Supermarket Industry: A Model of Multi-Store and -Product Choice (by PS, SS and HS)

Summary

Seiler and Schiraldi demand to be made coauthors on a paper 2 months before it is submitted to (and later published in) AER, under threats of writing to editors disputing authorship and potentially preventing or delaying publication. Their reason: their other paper with one of the authors (who owned the data used in both papers) did not work out, even though Seiler and Schiraldi had taken over the core model from the the other paper years earlier.

Story

The working paper ST was published in the AER in August 2017.

The SSS project appears to have been initiated in 2008, the ST project in April 2010. Both used HS's UK Kantar scanner data, and the basic ideas / questions came from HS's earlier work on supermarkets (2004 REStud, 2005 JEMS).

SSS initially had a single (composite) good model with an AIDS utility form, extending HS's 2004 paper to allow consumers to visit two stores.

ST had a multiple (8) category (meat, fruit, dairy etc.) model with quadratic utility and nonnegativity constraints. The model and estimation framework was a natural extension of OT's ideas from his PhD dissertation given HS's idea about investigating cross-category pricing externalitites.

In the spring of 2011, SSS scrapped the single good / AIDS framework and took over the multi-category / quadratic utility framework of ST (without the knowledge of OT). See footnote (1).

About one month later, the first empirical results from ST were presented at a conference. See footnote (2).

All OT, SS and PS are all aware of HS's other project using the same data, at least from 2011. For several years, there is no talk about merging the two papers. I (OT) accepted that HS had this other project, and never had any reason to suspect bad faith on the part of the other coauthors, so I always assumed whichever paper was finished first would be submitted first, without any strategic considerations.

Around the summer of 2015 it seems that PS and SS are starting to worry. The two-step estimation method of SSS is proving problematic and no good results are forthcoming, although HS (who did the programming and estimation) appears to have worked a lot on this during 2014 and 2015. Meanwhile, ST (where coding and estimation is done by OT) has progressed very well.

Late October 2015 SS emails OT and demands that he and PS be made coauthors on ST. Footnote (3).

Initially OT resists, since SS and PS appear not to have contributed anything to ST: the data belonged to HS, the core model, now shared by both papers, came from ST, the economic questions were different.

Then PS threatens (footnote (4)) to email the editors of the top journals to create an authorship dispute about ST, preventing its publication. Unwilling to risk this, OT and HS agreed to include SS and PS as authors, with authors' names in reverse alphabetical order, to put OT and HS first in the author list.

More detail about how the core model came from ST

HS in email to SS 31 October 2015:

"I think its important here to note that Oyvind originally developed the quadratic category utility model with firm-category taste effects, with random utility in those taste effects (i.e. the model used in ST) separately. In our collaboration (ST) we wanted to study intra-store cross-category pricing externalities and their effects on market power relative to independent category sellers which do not internalize such effects). This is quite an interesting question to me and I have published a theory paper on the question. At this time SSS did not have categories, category-firm taste effects, or quadratic utility, and was not interested in the same questions.

When the question of categories came up after we had already developed our model and started presenting it (as a comment in a seminar) I let Pasquale know that I was already working on this with Oyvind, and I showed him the ST paper with the quadratic utility framework that Oyvind and I had developed at that stage. At that point Pasquale (and you presumably) wanted to adopt the ST framework into SSS, and asked me to agree to that, which I was wary about doing at the time, but I agreed, and as a result SSS dropped the model that it had used up to that point (which was based on an indirect utility function from an Almost Ideal model). At this point SSS imported the quadratic framework in ST, with the firm-category effects, the random draws at firm category level; SSS builds on this by adding the overlap variable and so on. So in essence I think its only fair to point out that firstly that the utility framework SSS originally used was at that point scrapped and replaced with the one in ST, and secondly that T had developed this earlier. Oyvind made I would say the main conceptual contribution to ST as well as the major time input and he is responsible for most of what is good about it including the quadratic specification with categories, firm-category taste effects, with random utility at that level. (He also contributed the idea for the estimation method used in ST, which is very clean, and different from that in SSS)."

(1) Version of SSS from 4 April 2011, just before their single good / AIDS model is scrapped for multi-category quadratic model. The email where HS sends this version to PS. Version of TS from 17 February 2011, with multiple categories / quadratic utility with nonnegativity constraints. Shared in email with PS on 8 April 2011.

(2) A version of TS with estimation results was presented at an INRA-IDEI conference in Toulouse 16-17 May 2011 and the EARIE conference in Stockholm 1-3 September 2011. Slides from Touluse 17 May 2011. After this, SSS adopts the multiple category / quadratic utility framework.

(3) First email (31 October 2015) from SS, demanding that SS and PS be included as coauthors on ST. OT's reply to SS from the same day, corroborated by HS in this email to SS, also 31 October 2015.

(4) After some discussion, OT and HS receive this threat from Schiraldi, that he will write to the editors of top journals to create an authorship dispute which will prevent the publication of our paper. This leads to SS and PS being taken on board as coauthors of ST, according to this agreement from 11 November 2015.