Using multi-method data for more accurate research findings

The vast majority of personality research is based on a single method: self-reports. As about half of the variance in single-method trait scores is the result of systematic biases and random error, it is likely that most research findings are biased, even well-replicated ones. While some 'significant' associations may be entirely artefactual, most are likely to be underestimated, sometimes by as much as 50 %. Unfortunately, this is rarely discussed explicitly, let alone addressed empirically. After explaining the causes and extent of the problem, we argue that it can be effectively addressed by combining personality trait self-ratings with those of knowledgeable informants. To underscore the feasibility of multi-method research, we review recent large-scale studies that have combined self-reports and informant reports to provide more accurate answers to key questions in personality research, such as the heritability of traits and their association with important life outcomes. Since most associations are likely to be underestimated in typical single-method studies, multi-method studies will likely reveal higher correlations with commensurately stronger theoretical and practical implications. For example, single-method studies may have underestimated heritability by around a third and the predictability of life satisfaction from personality traits by around half. Personality psychologists have made great progress in incentivizing more reliable research; it is now time for the field to incentivize valid research, too.