P2-2 Time series models of disaggregated retail price behaviour
PhD student: Christin Schulze-Bisping
Thesis Committee: Prof. Stephan von Cramon Taubadel, Prof. Tatyana Krivobokova, Prof. Bernhard Brümmer
In the analysis of Vertical Price Transmission (VPT) processes using aggregated retail prices we have found that information on price variation at the individual or market level is lost when aggregated prices are analysed. At best, those empirical results are only representative of average of individual behaviour. Hence, the results of price transmission studies depend on the spatial scale of the price data that is employed. Furthermore, disaggregated retail prices are mainly characterized by persistence and prevalence of the '9' as the last digit and cannot be modelled as unit root processes like aggregated prices.
Unit root processes are time series which are non-stationary due to a unit root but can be transformed to stationary processes by taking the first differences. Hence, the objective of my work is to develop and implement an adequate model for such disaggregated prices. Due to persistence and short time jumpy changes an unconditional hazard model has been considered as a first possible approach for modelling disaggregated retail prices. In general, in a hazard model price decisions are time- and state-dependent. Introduced in the macroeconomic theory time-dependent means that the probability of a nominal adjustment declines with the time elapsed since the last price change. Whereas state-dependency states that the probability of a nominal adjustment is highest when a store's price substantially diers from the average of other stores prices.
We will try to answer the question which features of the retailer (type, region, size; competitors) as well as the time elapsed since the last price change inuence the probability of a future price change. Moreover, in our case time-dependency is dened dierent than in the macroeconomic theory: in our model the probability of a change increases with the time elapsed since the last change. In other words, the longer prices have been constant, the more likely the next change, all other things being equal.
back to overview of all projects