It is typical of public real estate benchmarking reports to show only highly aggregated benchmarks based on buildings’ floor areas. They hardly provide disaggregated benchmarks for usage clusters. The aim of this study is to show the caveats from highly aggregated benchmarking without consideration of cluster-specific characteristics.
Based on the parameters of the German facility management association 812 standards, cleaning costs and costs for the surfaces of seven hospitals have been collected and allocated to specific room clusters. Using these basic data, a calculation and simulation conducted with the aim of simulating facilities that are comparable in the sum of costs yet feature varying sub-clusters as cost drivers. In particular, during this simulation, area ratios were varied randomly and the average cleaning costs per cluster were held constant for all hospitals. Therefore, the costs per square meter in the clusters of all simulated hospitals are identical and the full costs only depend on the area ratios.
The simulation shows that highly aggregated cleaning costs lead to large spans, and thus, to misinterpretations in the field of action. In the case, the aggregate benchmark ranges from 40.6 to 66.5 EUR/m², although, for all hospitals the same costs per square meter had been used. Thus, the bias results only from varying the share of area across the clusters. This finding is caused by a well-known statistical problem: the Simpson’s paradoxon, which currently receives little attention in real estate benchmarking.
The results show, that the regular benchmarking with high aggregated data, often used in practice, cannot be recommended. The author consider using a detailed benchmarking as meaningful and purposeful. To be able to make a detailed benchmarking, it is essential to identify and collect the influencing factors. Only if all important factors, in this case, the clusters will be regarded in the benchmarking, a reasonable benchmarking and useful interpretation can be given. Using a simple benchmarking to get a rough overview is refused steadfastly.
The study highlights that a comparison with public benchmarking reports (operation costs) must be taken with great caution. The author has quantified the bias from the aggregated benchmarking and have shown, that the Simpson’s paradox fully explains the consequences.
Emerald Publishing Limited
Copyright © 2019, Emerald Publishing Limited