To compare industrial accidents and communicate on their consequences and nature, quantitative measures have to be used. What is usually done is to put the values of the descriptors of an accident into a categorisation scheme and then compare them with an absolute system of reference (absolute ranking of accidents). In this paper, based on approaches developed within the European Union for accidental events causing major emissions, fires or explosions, a new method of scaling by means of the relative ranking of the values of the quantitative measures describing such events is presented. While the immediate objective of an absolute ranking type of scale is usually communicating the risk-related significance of accidents, the relative ranking approach is proposed to serve primarily as accident data analysis tool. After evaluating and discussing general characteristics of a successful accident gravity scale and reviewing on this basis a few currently available absolute ranking type of models, it is shown by way of example that the new relative ranking approach might help reducing some of the weak points related to subjective modelling assumptions still included in many existing approaches. The statistical methods used in this paper are standard, although the author has not seen elsewhere the unified treatment of the models given here, and their specific application to accident gravity scaling may be new. Further, although the reasoning in the paper deals with industrial accidents only, there is no reason why it should not also be applicable to disasters caused by natural hazards.; To compare industrial accidents and communicate on their consequences and nature, quantitative measures have to be used. What is usually done is to put the values of the descriptors of an accident into a categorization scheme and then compare them with an absolute system of reference (absolute ranking of accidents). In this paper, based on approaches developed within the European Union for accidental events causing major emissions, fires or explosions, a new method of scaling by means of the relative ranking of the values of the quantitative measures describing such events is presented. While the immediate objective of an absolute ranking type of scale is usually communicating the risk-related significance of accidents, the relative ranking approach is proposed to serve primarily as accident data analysis tool. After evaluating and discussing general characteristics of a successful accident gravity scale and reviewing on this basis a few currently available absolute ranking type of models, it is shown by way of example that the new relative ranking approach might help reducing some of the weak points related to subjective modelling assumptions still included in many existing approaches. The statistical methods used in this paper are standard, although the author has not seen elsewhere the unified treatment of the models given here, and their specific application to accident gravity scaling may be new. Further, although the reasoning in the paper deals with industrial accidents only, there is no reason why it should not also be applicable to disasters caused by natural hazards.
