We argue that entrepreneurship research should use meta-analysis to integrate the findings of the field. A meta-analytical approach has several advantages as compared with narrative reviews: First, narrative reviews are likely to bias empirical evidence because they are limited by the information-processing capacities of the reviewers (Tett, Jackson, & Rothstein, 1991). This is often a downward bias leading to the conclusion of little positive knowledge in the field. For example, frequency counts of significant results ignore sampling errors of individual studies, reliability problems of instruments, range restrictions of samples, dichotomization of continuous variables, imperfect construct validity, and extraneous factors (Hunter & Schmidt, 2004). These issues usually result in a higher incidence of Type II errors (i.e., rejecting the hypothesis wrongly). Thus, narrative reviews are more likely to lead to the conclusion that there are no relationships between independent and dependent variables in entrepreneurship when in fact they are (Hunter & Schmidt, 1990; Tett et al., 1991). Second, meta-analysis accumulates studies based on a set of explicit decision rules and, therefore, is less biased by subjective perceptions of the reviewer than narrative reviews. Meta-analyses require judgments as well, e.g., when defining the area of the study or coding moderator variables. However, the decisions are public and open to criticism and replication by other scientists (Johnson & Eagly, 2000). Third, meta-analysis is based on many studies and, thus, avoids the influence of single studies. Fourth, meta-analysis controls for sampling error variance and, thus, controls for power deficits of individual studies (Hunter & Schmidt, 2004). For example, the Brockhaus and Nord (1979) study is frequently cited in the entrepreneurship literature for providing evidence that there is no relationship of personality characteristics with entrepreneurship. However, this study is based on a small sample of 31 business owners and therefore, has serious statistical power problems. Noteworthy, the effect sizes of small samples are less precise in estimating a population value than effect sizes of larger samples. Fifth, meta-analyses can correct many errors of individual studies (Hunter & Schmidt, 2004). Since meta-analyses estimate population correlations between given variables, it is important to correct for errors of studies (e.g., unreliability, range restriction, and sampling error) to achieve unbiased estimates. Sixth, meta-analysis allows an assessment of the magnitude of relationships and, thus, provides more precise and often comparable assessments of the validity of concepts. Thus, meta-analyses support the assessment of the practical significance of findings. Seventh, meta-analysis tests for variations in relationships across studies and, therefore, allows an assessment of the generalizeability of effects. If the size of reported relationships varies considerably between different studies, there will be context conditions that account for these variations. These context conditions are moderators that affect the size of relationships. The moderators may include study characteristics, method moderators, and theoretical moderators. Thus, meta-analyses also help to identify areas for new studies. Finally, meta-analysis techniques allow to test more than one independent and/or moderator variable by using methods based on regression analysis (Lipsey & Wilson, 2001). Using such procedures allows to estimate the independent contribution of variables on results, to control for methodological variables, and to test the interactions between moderator variables