A bottom up approach to estimate production-consumption matrices from a synthetic firm population generated by iterative proportional udating

Abstract

In order to model freight movements in a region or nationwide, data about regional production and consumption of goods is necessary. This type of data is typically obtainable on an aggregate level only. The data takes the form of production - consumption (PC) matrices with individual cells representing the amount of goods being produced and consumed at the origin and destination respectively. The cell values of the PC matrices are in essence, an aggregation of individual firm to firm relations in form of goods production, processing and consumption. Iterative Proportional Updating (IPU) is a mathematical technique, which generates synthetic populations using a representative sample of the desired population. The use of this technique in transportation domains is typically to generate synthetic individuals-households populations, but by using a descriptive enough sample of firms, IPU can generate a synthetic firm population preserving real life firm attributes and distribution. The quality of the synthetic firm population relies greatly on the quality of the sample used. A main advantage of using this approach is that with a small sample of real firms with well-defined desired attributes, modellers are able to regenerate a total firm population. This means if firm production volumes can me estimated, firm to firm goods transactions can be reproduced on a nationwide scale, opening the door for a bottom up microscopic modeling approach of demand and supply relations. Another advantage of the IPU approach used in this paper is that, using a synthetic firm population enables us to evaluate the effect of policy impacts such as land use and firm location choice policies. This approach is also in agreement with the current direction in modeling freight movement at a microscopic level, as opposed to the traditional four steps based macroscopic approach. In this paper we present how to obtain regional PC tables from an IPU generated firm population, from which PC matrices are easily calculated. We then compare our findings with traditional way to obtain the PC tables and analyse results.