An Empirical Analysis of a Merger in the Australian Cigarette Industry

The cigarette market is of ongoing interest to competition authorities as it features a small number of rms selling di erentiated though similar products. However, the extent of market power and the response of competition authorities to mergers in this industry has not been recently analysed. In this paper we estimate a nested logit model of demand for cigarettes in Australia using a dataset obtained from a public database of internal tobacco industry documents and simulate a merger and the divestiture imposed by the Australian competition authorities. We use a nested logit as given current econometric technology it is more feasible for use by time and resource constrained competition authorities. We compare the predictions on the directions of price changes from the nested logit with the price changes following the divestiture. We successfully predict the direction of average price changes for the entrant receiving the divested brands but are less successful with the merged rm and the other incumbent. This may be because of a demand or cost shock to a major brand retained by the divested rm as it featured a large and unusual price change following the merger, compounded by the restrictiveness of the nested logit. However, our results suggest that for time and resource constrained competition authorities a structural econometric analysis using a nested logit may still be useful when having to quickly analyse mergers and divestitures. ∗This paper builds on the work done by Vivienne Pham for her honours thesis at La Trobe University. The authors wish to thank her co-supervisor, Damien Eldridge, for his guidance during this. Some of the work was done while David Prentice was a visitor at RSSS at the ANU and two visits at the University of Melbourne, and he is grateful for their hospitality and support. The authors would also like to thank Harry Clarke, Mick Coelli, Michelle Goeree, Philip McCalman, Harry Paarsch, Chris Skeels, Randal Watson, Jenny Williams and seminar participants at La Trobe University and the University of Melbourne for their useful comments. All results are preliminary and must not be cited without permission of the authors.