Discussion Papers

Antonia Niederprüm, Sonja Thiele

Forecast models for letter demand

No. 371 / December 2012

(Full version only available in German)


There is a broad range of different quantitative and qualitative methods for estimating letter volume development, as the analysis of international experiences with letter volume predictions in this study shows. Quantitative methods require both a longer time series and more detailed data – ideally quarterly data. These requirements can only be met by postal operators themselves in most cases but hardly by regulators or independent bodies. Quantitative methods do not allow to predict reversals of trend. Thus, they can only be applied where external conditions (e.g. economic development cycles) are steady. However, quantitative methods deliver good results for short term predictions up to three years.

The advantage of qualitative methods lies in the ability to predict a reversal of trend by relying not only on past data but by including experts’ opinions. In addition, they require only data of one single starting year as a data base. Due to these characteristics, qualitative methods can be applied by regulatory authorities for regulatory purposes, e.g. in price decisions.

In light of these finding we recommend a qualitative approach for predicting letter volume development in the German letter market. This approach should be based on a parametric model. In this study, we describe the development of such a model by six steps. First, the market has to be segmented. An appropriate segmentation should take into account the purpose of the prediction, data availability as well as rising complexity of the model in case of increasing number of segments. Second, the starting year of the prediction has to be determined. For the German letter market, there are several data sources which should be examined with respect to the services covered (e.g. by type of letter, weight and included postal operators) as well as on the available time series. Third, the horizon of the prediction has to be determined. Generally, uncertainty of the prediction results increases with the length of the prediction period. Fourth, appropriate indicators have to be chosen. Fifth the choice of the indicators as well as the modeling of their influence strongly impacts the results of the prediction. A thorough justification of the modeling based on extensive research and interviews with sector experts and market participants is especially important for qualitative predictions. As the sixth and last step, the indicators and their impact on the segments have to be parameterized.

The study finally compares letter volume developments in Germany and other European countries. This comparison shows the relevance of letter volume per capita for future volume losses. Due to the relatively low letter volume per capita in Germany, we expect only smaller volume losses in the German letter market compared to Nordic countries or the Netherlands.

Discussion Paper is available for download.

To top  |  Print