Analyzing and Predicting Forestry Accountancy Network Variables with Bayesian Belief Networks as Compared to Traditional Analyzing Methods

Abstract

The tremendous variability in physical conditions of forest enterprises as well as attitudinal aspects of their managers is seen as a major impediment to the understanding and optimization of forest management. For this reason, former studies using several methodological approaches-including meta analysis of econometric studies, binary choice models and stochastic frontier models-in many cases remained on a qualitative and more holistic level. This paper assesses the applicability of Bayesian Belief Networks (BBN) for the analysis of net income based on detailed 2006 economic data from the German federal accountancy network of forest enterprises larger than 200 ha. A network with one dependent (target) and 30 independent (explaining) variables was designed. The BBN has proven helpful for qualitative and to some extent quantitative analysis of economic data. It has become obvious that the completeness of populating the BBN model must be seen as a constraint. The speed of the calculations and the use of dependent probabilities can be seen as benefits of the BBN approach that reduce the risk of misinterpretation in comparison with traditional analysis methods such as the comparison of different strata. The visibility and presentability of the BBN approach facilitates its use in controlling and optimizing processes.

Forest Accountancy Data Networks as a Means for Investigating Small-Scale Forestry: A European Perspective

“…for the level of the individual enterprise as well as for a whole region, can be filled with the collected and analyzed data.'' Examples for such extensions to standardized statistics in terms of detailed studies are provided by Brandl (1998) and Hartebrodt et al (2010). Ultimately, there is the potential of utilizing accountancy networks as an established infrastructure for efficiently conducting specific surveys in regard to additional information and socio-economic variables such as attitudes, behavior or expectations, which are not covered by the standardized data frame (Sekot 2000).…”