Modelling negatives in stochastic reserving models

Incremental loss data triangles regularly contain negative values, which may be due to salvage, subrogation and/or conservative case reserve estimates. Assumptions made by many stochastic claims reserving models are not always satisfied when negative values occur in incremental loss data triangles. This paper takes advantage of a mixture model [Kunkler, M., 2004. Modelling zeros in stochastic reserving models. Insurance: Mathematics and Economics 34, 23–35] to combine both a reflected and a regular generalised linear model to extend many previous stochastic claims reserving models for incremental loss data triangles that contain both positive and negative values.