The first cross-disciplinary seminar of MEAFA took place today on "The Contemporaneously Endogenous Character of Accounting Variables".
I was very happy to present this paper to the Discipline of Econometrics and Business Statistics, given the very good turn up and the discussion which effectively covered all key econometric issues of the paper. I wish to thank Vasilis Sarafidis and Kandice Cherry for organising the event, with a special thanks to Kandice for indulging us with some delicious homemade caramelised slices.
This is a joint work with Stuart McLeay, which challenges prominent accounting research by treating a reconciliation of financial statements as a single matrix of endogenous information. Using a reconciliation of financial statements (balance sheets at time t-1 and t, and income and cash flow statements from time t-1 to t), the paper shows how all financial reporting variables are contemporaneously codetermined through the resolution of multiple accounting identities. Therefore, when endogenous accounting variables that are taken from the same reconciliation matrix are included on both sides of a regression equation, the Ordinary Least Squares estimator yields generally misspecified results, and the modelling requires instead a framework of simultaneous equations.
I was very pleased to see that the presentation stimulated a very lively discussion with some very useful comments and suggestions on how to proceed with the paper. Robert Bartels suggested ways of improving the debate over the issue of endogeneity, and Maurice Peat helped clarify the discussion with regard to the contemporaneous determination of accounting information and how this may cause misspecification under OLS estimation. Alan Woodland touched a key issue of the paper, that of the usefulness of the double-entry constraint as a parameter restriction, and Hajime Katayama was particularly concerned on the gain in efficiency when moving from estimating unrestricted systems of SUR to systems with parameter restrictions, such as in the application of this paper. Also, Vasilis Sarafidis pointed out the need for modelling firm-year datasets using panel methods and pointed out ways for estimating such models especially for situations where panels have a small time dimension, and Richard Gerlach explained the implications of a multinormal assumption in the estimation of the model. Finally, Remy Cottet referred to the reconciliation of financial statements as a classical classroom example for illustrating the source and consequences of endogeneity, and Dmytro Matsypura found very interesting how accounting is perfectly linearly deterministic at a specific point in time.
Overall, I feel privileged for having the opportunity to present this joint work in such a great audience and I sincerely thank everyone for their participation. You may view the abstract of the paper from the Discipline of Econometrics Seminar Series, and you may request a copy of the full paper by contacting MEAFA.