Optimizing the measurement scales of qualitative variables to improve bankruptcy prediction (2019)
Type of ContentTheses / Dissertations
Degree NameMaster of Commerce
PublisherUniversity of Canterbury
AuthorsBowatte, Piuminishow all
I extend Altman’s (1968) multivariate linear discriminant bankruptcy model by adding two variables, auditor-client familiarity (AFT) and senior management turnover (SMT). These two variables have not been considered in the bankruptcy modelling, albeit the literature suggests them to relate to firm performance. The addition of the two qualitative poses the question as to i) which measurement scale (nominal, ordinal or interval) should be used, and ii) how to determine the optimal scale distances such that the variables can be included into the regressions with the intention to improve their bankruptcy forecasting ability.
I collect financial and non-financial information for 70 UK construction companies to test above research questions. My final 7-variables model predicts 69-out-of-70 firms of my sample correctly into healthy and failed firms (as a comparison, the re-estimated Altman 5-financial ratios model predicts 56-out-of-70 firms correctly).
My thesis was motivated by the recent Carillion PLC collapse in the UK that attracted wider attention from the media and government. The case was portrayed as a surprise event, yet several models in my study indicate a strong decline in performance over 3 years prior to their collapse.