Table 1. Results of a Meta-analysis of Ernst and
Angst's (1983) Review of the Birth-
Order Literature, Using the Scoring Method. Townsend's (2000, Appendix A)
Uncorrected Results Are Listed to the Left; Corrected Outcomes are Listed to the Right).a
Personality Fully Partially Null Partially Fully p
Dimensionb __ Confirmingc ______ Confirming Results Opposed Opposed ___ N _ _ t-value d __ (2-sided)
Conscientious 9:20 2:3 25:29 0:4 0:0 36:56 3.80:5.00 <.001:<.001
Agreeableness 4:5 0:7 12:18 0:1 0:1 16:32 2.24:2.71 .041:<.001
Openness to 12:18 3:13 22:25 0:4 3:1 40:61 2.88:5.39 .006:<.001
Extraversion e 13:9 0:10 16:17 0:3 3:0 32:39 2.74:4.30 .010:<.001
Neuroticism 5:5 0:9 16:23 3:5 5:0 29:42 -0.45:2.55 .656:.017
a. For each personality dimension, "studies" are the unit of analysis but take into account multiple "findings" via a five-step scale for scoring individual "outcomes." The following hypotheses are tested in Table 1: Firstborns are expected to score higher in Conscientiousness and Neuroticism, and laterborns are expected to score higher in Agreeableness and Openness to Experience. For hypotheses relating to Extraversion, see note e. Owing to the meta-analytic amalgamation of related results, these 230 outcomes represent more than nine hundred individual birth-order findings derived from approximately 80,000 subjects. Townsend (2000, Appendix A) reports 18 "ambiguous" and 190 unambiguous results, for a total of 208 results. Townsend's 190 unambiguous results contain 23 redundant findings that fall under the same personality dimension as other previously counted findings, as well as 14 outcomes that are not assignable to any personality dimension. When scored according to the five-step scale employed here, which combines all results from each study into a single outcome on the same personality dimension, the number of truly codable outcomes in Townsend's analysis is reduced to 153. In the right-hand column of the table, Townsend's numerous errors and omissions have been corrected, based on the judgments of two investigators. These outcomes include findings from all controlled studies involving personality that listed in Part 1 of Ernst and Angst (1983:93-189), as well as five outcomes scored from Ernst and Angst's (1983:243-82) own results in Part 2 of their book. All outcomes listed here are controlled for sibship size (N=102), social class (N=32), or both variables (N=96).
b. Classifications of individual findings according to the Five Factor Model of personality have been made jointly by the author and Oliver John (University of California, Berkeley). We agreed on 88 percent of the classifications and, where possible, reconciled our differences in the remaining instances.
c. Classification of outcomes as "fully confirming," "partially confirming," "null," and so forth is based on a series of formal scoring rules that take into account the overall proportion of significant findings. For further information, see the text.
d. Values of t for the distribution around the null category have been calculated by giving full outcomes a weight of 2 (or -2) and partial outcomes a weight of 1 (or -1). Alternative methods of weighting yield similar results.
e. Proper predictions for Extraversion require that findings be classified by individual facets of the Five Factor Model. More specifically, firstborns are expected to score higher in assertiveness, whereas laterborns are expected to score higher on other facets of this personality dimension and on composite measures of Extraversion. Scored in this manner, there is a net gain of five fully confirming outcomes in Townsend's tallies, as corrected in the right-hand column of the table.
f. Statistical tests have not been performed on the overall meta-analytic totals because the pooled outcomes are not statistically independent for studies that contain findings relating to more than one personality dimension. For the 72 studies that report findings or multiple-finding "outcomes" relating to only a single personality dimension, which are independent of one another, the t-value for the entire sample is 4.77 (p<.001).