The standard method for MBC is the serial cross-effects approach implemented in Sawtooth Software. Each item has its own model based on utilities from itself and a subset of the other items (cross-effects), creating a set of models {M}. The models in {M} are estimated independently and the respondent level utilities for a specific item will vary across each model. In contrast, we estimate all the {M} models simultaneously and with just one set of respondent utilities for each item, which is vastly more parsimonious and unified. For each specific model in {M} we size the impact of its cross effects using a nested logit.
