Abstract:
Several behavioral models assume that choice over multi-attribute goods is systematically affected by the ranges of attribute values. Two recurring principles in this literature are contrast, whereby attributes with larger ranges attract attention and are therefore overweighted, and normalization, whereby attributes with larger ranges are underweighted as fixed differences appear smaller against a larger range. These principles lead to divergent predictions, and yet, both contrast-based and normalization-based models have found strong empirical support, albeit in different contexts and with different experimental designs. The question remains: when does one effect emerge over the other? We experimentally test a unifying explanation: normalization dominates in simple choices, while contrast dominates in complex choices. We conduct an experiment with real-effort tasks in which we manipulate attribute ranges in both simple and complex choices. We find that, indeed, contrast dominates as the number of attributes increases. We also find that contrast emerges with cognitive load induced by time pressure.