>Birth Order and Political Rebellion    



Frank J. Sulloway

University of California, Berkeley


This report is divided into five sections:

  1. An Introduction discussing technical details about two samples of political activists, whose political "radicalism" has been independently assessed by ten experts in international affairs.

  2. Results of a statistical analysis of individual differences in political radicalism, showing that laterborn political activists have tended to be more radical than firstborn activists, controlling sibship size and social class.

  3. Results of a statistical analysis of individual differences among political loyalists, showing that laterborns loyalists are judged as generally being more liberal than firstborn loyalists and are also more likely to have been revolutionaries at a previous stage in their political careers. 

  4. Combined statistical results for the two samples. 

  5. An Appendix containing expert ratings and biographical data for the two samples of political activists.


In Born to Rebel (Sulloway 1996) I analyzed a sample of 135 political activists previously studied by Rejai and Phillips (1979, 1983, 1988). In my assessment of birth-order differences in political behavior within this sample, I relied on Rejai and Phillips's own classification of these activists, via factor analysis, into nine political subgroups.  On a conservative/radical spectrum, these nine subgroups ranged from "defenders of the faith" (who generally sought to preserve important aspects of the status quo) to various kinds of radical revolutionaries, grouped by Rejai and Phillips under the categories of "agitators," "professional revolutionaries," and former revolutionaries-become-"elders."  In my own analysis of these data, I assigned defenders of the faith to the category of "conservative" revolutionaries, and I placed agitators, professional revolutionaries, and elders in a "radical" category.  In between, in a category of "moderate" revolutionaries, I placed the remaining five subgroups, which included "scholars," "aristocrats," "generals," "founding fathers," and "purists."

It would have been preferable if I had employed expert raters for this political sample in order to be assured of reliability in my interpretation of Rejai and Phillips's political classifications.  To rectify this omission, I have had Rejai and Phillips's sample of revolutionaries rated for radicalism by ten specialists in political history and international affairs. (For a list of these ten expert judges, see the Acknowledgments.)   Using a seven-step scale (extending from "very conservative" to "very radical"), these experts performed two different kinds of ratings.  To test my previous assessment of the differences in radicalism among Rejai and Phillips's nine political "types," I asked five of these experts to rate each of these political types.  The five experts' ratings were made after each judge had read Rejai and Phillips's (1983:104-111, 127-28) descriptions of their nine political types.  Each judge was also given a list of up to six activists receiving the highest factor scores for each political type.  For example, the description provided for Rejai and Phillips's "agitators" is as follows:

Agitators: This group, according to Rejai and Phillips, "virtually mirrors the popular stereotype of the Revolutionaries: young, lower class, socially deprived, with extensive record of radical activity" (Rejai and Phillips, 1983:110).

Examples: Charles Gagnon, Liu Shao-ch'i, Mao Tse-tung, Lin Piao, Yasir Arafat, and Fidel Castro.

In assessing the radicalism of these nine political groups, the mean interrater reliability for the five judges is .90 (Cronbach's alpha=.98), indicating that judges were able to reliably differentiate Rejai and Phillips's various political types on a conservative/radical spectrum.

Nine of my ten experts also rated Rejai and Phillips's sample of revolutionaries individually.  Each judge was given a full list of the 135 political figures in Rejai and Phillips's sample, presented in alphabetical order, and each judge recorded a rating next to every political activist on the list who was sufficiently known to them to permit an informed judgment.  An "informed judgment" was defined as one that the rater believed to have a margin of error of no more than one-step, in either direction, on the seven-step scale.  

The mean interrater reliability by this second method of assessment is .77, yielding an effective reliability for pooled ratings of .91 (Cronbach's alpha). Altogether, 89 political activists were rated by my experts on an individual basis, and an additional 42 activists were rated as members of one or another of Rejai and Phillip's nine political types. The remaining 4 individuals, for whom Rejai and Phillips's factor analytic data do not permit "type" classifications, were assigned radicalism ratings based on the mean of the sample.1

Based on these numerous expert ratings as a measure of political radicalism, the correlation between birth order and support for radical political causes lies between .22 and .34, depending on how birth order and political radicalism are coded and on which controls are included in the analyses. Controlled for sibship size and social class, laterborns were significantly more likely than firstborns to support radical political changes (r=.34, N=70, p<.005--see Table 1).2

Table 1. Effect sizes for birth order and political radicalism, based on ratings by ten expert judges, using Rejai and Phillips's (1983) sample of 135 revolutionaries

Independent variable(s)     Dependent
Birth-Order   Correlation     N       t-value        p<
Birth order (coded 
dichotomously: firstborn/
  Ratings by     ..revolutionary
        .22  74      1.90     .07
Birth order (coded dichotomously: firstborn/laterborn)   Individual ratingsb         .33  74      2.93     .005
Birth order (linear contrast: first/middle/last)c   Individual ratings         .29  74      2.60     .02
Birth order (linear contrast: first/middle/last--controlled for sibship size and social class)d   Individual ratings         .34  70      2.97     .005


a. Here and elsewhere, 7 only children are included with firstborns. Omission of these 7 politicians does not appreciably alter the overall statistical results (see note d).  

b.  Individual ratings are available for 66 of the 74 politicians (89 percent) on whom birth-order information is available. In 7 of the 8 instances in which individual ratings are not available, radicalism scores have been based on "type" ratings by five experts, with the one remaining radicalism score being based on the sample mean.

c. Six laterborns are known, ambiguously, as either middle or youngest children. In tests based on the linear contrast, these 6 laterborns have been assigned the mean birth-order value for these two categories. Omitting these 6 politicians from the sample, the correlation becomes .32 (N=68, p<.01).  

d. Omitting the 7 only children from this controlled sample, the partial correlation becomes .28 (N=63, p<.04).  Two members of the controlled sample are laterborns known ambiguously as either middle or youngest children (coded with the mean birth-order value for these two categories).  Omitting these 2 politicians from the sample, the partial correlation becomes .33 (N=68, p<.01). Without the control for sibship size, the partial correlation is .34. (N=70, p<.005)  Without the control for social class, the partial correlation becomes .33 (N=70, p<.01).  Without either control, the correlation becomes .33 (N=70, p<.01).

Because my experts rated Rejai and Phillip's 135 revolutionaries individually as well as by "revolutionary type," it is possible to test my claim (Sulloway 1996:296) that Rejai and Phillips's "defenders of the faith" and members of the IRA in Northern Ireland are "conservative" political activists relative to other members of their sample. Although Townsend (2000) has strongly criticized this assertion, he did not subject his claim to any empirical test. In actual fact, the 11 members of these two political groups received significantly lower radicalism ratings than the 124 other political figures included in Rejai and Phillips's study (r=-.57, N=135, p<.0001).


Rejai and Phillips (1988) also provide biographical information on a group of 50 "loyalists." Like Townsend (2000), I initially considered using this sample as a control group until I examined it more closely and discovered what Townsend, remarkably, fails to mention about it. Rejai and Phillips's loyalists are not, as might first seem, a sample of opponents of political revolutions. Rather, Rejai and Phillips asked experts to nominate as "loyalists" those prominent politicians who were in key political positions when revolutions began. Some of these people turn out to have been strongly opposed to revolutionary change, but others were not. Still others are former revolutionaries who happened to be in power when a new challenge to authority arose. If a revolution broke out in Cuba today, Fidel Castro would qualify as a "loyalist" according to Rejai and Phillips's classification scheme, even though Castro is also included in their sample of revolutionaries. At least a dozen members of Rejai and Phillips's sample of loyalists turn out to be former revolutionaries or liberal reformers, a point that Rejai and Phillips fully acknowledge but that Townsend fails to disclose.3

By considering "career radicalism" scores for the 42 loyalists who have been individually rated by my experts, we can now examine my assertion that Rejai and Phillips's loyalist sample is distinctly heterogeneous in its political views, making it inappropriate for use as a "control group." Compared with the 31 loyalists who experienced no prior revolutionary activity, the 11 politicians specifically identified by Rejai and Phillips either as revolutionaries-turned-loyalists or as former liberals were rated substantially higher on an overall scale of political radicalism (r=.65, N=42, p<.0001). Moreover, loyalists who were former liberals or radicals are more likely to be laterborn than are those loyalists who maintained a conservative outlook throughout their political careers (r=.20, N=46, one-tailed p<.10). In short, any attempt to compare Rejai and Phillips's (1988) loyalists, en bloc, with Rejai and Phillips's (1983) revolutionaries is methodologically inappropriate because the two samples are both distinctly heterogeneous groups that intermingle conservatives with liberals and thereby potentially obscure any inherent birth-order effects.


"Career" ratings for political allegiances allow us to incorporate Rejai and Phillips's sample of loyalists into an overall analysis of political radicalism. Because Rejai and Phillips's loyalists and revolutionaries are drawn from very different events and countries, the most appropriate way to analyze these data is to compute separate effect sizes for birth order and radicalism for each of the two samples and then to combine these effect sizes via transformations from r to zr and back again to a mean-weighted r. Based on this method of combining z-scores, birth order correlates significantly with political radicalism across Rejai and Phillips's two samples, with correlations ranging from .20 to .27, depending on how birth order and radicalism are coded. More important, birth order is significantly associated with political radicalism after one controls sibship size and social class (r=.27, N=115, p<.01).Townsend (2000), who arrives at a very different answer, does so because he naively combines political apples and oranges, and also fails to conduct appropriate statistical tests on any of his data. 5


The Appendix that follows lists 135 political "revolutionaries" and 50 "loyalists" from Rejai and Phillips (1979, 1983, 1988). Missing data are indicated by a "99." Codes for birth order (assessed as a functional construct--note 2) denote firstborns ("FB"), only children ("OC"), middle children ("MC"), youngest children ("YC"), and laterborns whose exact status as middle children or youngest children is unknown ("LB").  Overall radicalism ratings are an average of the ratings provided by ten expert raters. Whenever available, individual ratings have been employed in the statistical analyses (N=131). Ratings by political type, based on Rejai and Phillip's (1979, 1983, 1988) descriptions of their nine political types (as operationalized by five experts), have been used in the remaining instances (N=54).  Social class, which has been included in various regression models as a control variable, was assessed on a five-step scale from 1 (aristocracy) to 5 (manual labor), based on father's occupation. 


Click here to access these biographical data (Table 2)



I thank Lincoln Moses and Lynn Gale at the Center for Advanced Study in the Behavioral Sciences for valuable statistical advice in connection with this article. I am grateful to the following experts for providing ratings on politicians: László Bruszt, William Deverell, Norman Naimark, Françoise Pascals, Gérard Roland, Howard Rosenthal, Hilda Sabato, James Scott, Piotr Sztompka, and Björn Wittrock. This essay was prepared while the author was Fredrick Redlich Fellow at the Center for Advanced Study in the Behavioral Sciences. I am grateful for financial assistance furnished by the Center's Foundations Fund for Research in Psychiatry and by the National Science Foundation (Grant SBR-9022192).  Portions of this report have appeared in Sulloway (2000) and are used here by permission of the Association for Politics and the Life Sciences.


1. For each politician, the "type" rating was based on the highest factor loading received by this person in Rejai and Phillips's (1983) multivariate analysis.

2. For this sample, birth order has been coded functionally (how the politicians were raised rather than born). For instance, the first child to survive infancy, although not a biological firstborn, has been counted as a functional firstborn.  Rejai and Phillips (1983, 1988), who also code functional birth order, agree closely with my own assessments of this variable (r=.93, N=40). I have also documented birth-order information for 34 politicians for whom this information is partially or completely missing in Rejai and Phillips's tables.  In various statistical models based on the data in Table 1, appropriately adjusted values for birth order have been employed in two instances in which individuals spent part of childhood in more than one sibling position, thus reflecting functional birth order.  These adjustments, using weighted averages, reflect the proportion of time spent in each sibling position up to the age of twenty.  Vo Nguyen Giap lost an elder sister when he was about eight.  His birth rank in his sibship of four children was second or third, so he either remained a middle child or became a functional firstborn.  After the age of eight, the mean value for these two ordinal positions has been used for Giap.  Nelson Mandela was his mother's eldest child but a middle child and youngest son of his father, who had multiple wives.  The mean value for these two ordinal positions has been used for Mandela.  Overall, the correlation between biological and functional birth order in this sample is .96 (N=74), and these minor differences in how birth order is coded for Giap and Mandela do not affect the conclusions reached in my analyses. Models that employ the EM algorithm to deal with missing data yield almost identical parameter estimates to those presented in Table 1. In noting that middleborns are underrepresented in their sample of revolutionaries, Rejai and Phillips (1983) appear to confuse the relationship between birth order and eminence with the relationship between birth order and radicalism. Middleborns are generally underrepresented in samples of people who are famous enough to become part of the historical record, but this statistical fact says nothing about birth-order trends in radicalism, for which Rejai and Phillips do not provide a formal measure.

3. For identification of the nine former revolutionaries-turned-loyalists, and the three loyalists said to be former liberals, see Rejai and Phillips (1988:40-41, 127-28). One loyalist, Albert Luthuli, was actually nominated as a revolutionary by one of Rejai and Phillips's historical advisors (1983:9).

4. In seven instances in which loyalists were not individually rated by my experts, they have been assigned career radicalism scores based on their membership in one of two relevant loyalist subgroups, as rated by five experts. The two relevant subgroups are "loyalists" and "loyalists who were former revolutionaries" or former "liberals" (Rejai and Phillips 1988:40-41, 127-28). For Rejai and Phillips's two political samples (loyalists and revolutionaries combined), the correlation between individual and "type" ratings for radicalism is .82 (N=131, based on ratings by ten experts).  In Rejai and Phillip's (1988) sample of loyalists there is one individual who grew up in two different sibling positions.  Maurepas lost an elder brother when he was six, causing him to become a functional firstborn (Table 2). A mean-weighted average for his birth order, before and after the age of six, has been used in the statistical models (see note 2). 

5. Combining samples drawn from different populations can, under some circumstances, obscure significant effects present within each population, a phenomenon known as Simpson's paradox, which is operative in this case. The method of combining effect sizes, employed here, is recommended to prevent this problem (Rosenthal 1987:219).



Rejai, M., and K. Phillips (1979). Leaders of Revolution. Beverly Hills and London: Sage Publications.

Rejai, M., and K. Phillips (1983). World Revolutionary Leaders. New Brunswick, N. J.: Rutgers University Press.

Rejai, M., and K. Phillips (1988). Loyalists and Revolutionaries: Political Leaders Compared. New Brunswick, N. J.: Rutgers University Press.

Rosenthal, R. (1987).  Judgment Studies: Design, Analysis, and Meta-Analysis.  Cambridge: Cambridge University Press.

Sulloway, F. J. (1996). Born to Rebel: Birth Order, Family Dynamics, and Creative Lives. New York: Pantheon.

Sulloway, F. J. (2000).  Born to Rebel and its critics.  Politics and the Life Sciences, 19:181-202.

Townsend, F. (2000). Birth order and rebelliousness: Reconstructing the research in Born to RebelPolitics and the Life Sciences, 19:135-156.



©2002 Frank J. Sulloway, Ph.D.