Abstract

Managers today have enormous opportunity to examine multiple market research reports. These reports are useful in creating marketing plans. Often, however, these multiple market research reports contain conflicting information. The question explored in this research is how a manager integrates conflicting quantitative information (projections) to form a forecast to be used in marketing planning.

The work builds from information integration theory (e.g., Anderson 1971) arguing that forecasts area weighted combination of the scale values of the projections. It is hypothesized that confidence in the projections and meaning of the projections (good/bad news) each impact the weights, and that the level of the new projection (lower vs. higher than the initial projection) impacts the scale values.

In terms of confidence, it is hypothesized that there will be more adjustment when there is low confidence in an initial projection and higher confidence in a new projection. This hypothesis is supported. In terms of meaning, it is hypothesized that similar to the literature on the negativity bias (e.g., Skowronski and Carlson 1989) and asymmetric loss functions (e.g., Weber 1994), there will be a consistent pattern of more adjustment toward projections meaning bad news than toward those meaning good news. And in terms of level, it is hypothesized that there will be more adjustment toward lower numbers than toward higher numbers due to Weber's Law.

Because these two latter hypotheses are sometimes inconsistent (e.g., lower number implies good news and higher number implies bad news), this research pitted the two against each other. The lower numbers effect was identified as the more powerful and consistent effect-swamping the effects of negative information. This finding is supported across a variety of conditions. It occurs in both managerial and non-managerial settings. And, it is not easily reduced via accountability, training to think about the meaning, or familiarity with the variables. The pattern of adjusting more toward lower numbers is identified as a bias (a consistent pattern) across seven experiments and in a meta-analysis.