In the era of Brown v. Board, a series of experiments were presented in order to prove the harmful effects of segregation on black children. The design was thought up by Kenneth and Mamie Clark: essentially, they used multiple dolls of different races and asked students in segregated schools to pick which doll they preferred. What they showed was that black children were more likely to pick the white dolls over the black dolls (Clark and Clark, 1947). While this set of studies was not fundamental in the Brown v. Board case, they are probably the most well known part of it. This is despite the test being very bad.
To start, the main issue these studies faced was not including any results from integrated schools. This means there was no control, and therefore no way to prove these studies have any causation. This is also emphasized by Minow (2010). Guinier (2004) also noted that the Clarks failed to consider blacks in integrated schools could very well face greater stigma due to their increased contact with racist whites. This, already large, issue makes the other issues with the study all the more important. For example, the study only used children from the age of 3 to 7, which does not give enough time for people to develop and face significant life experiences which could easily change the gap. Furthermore, if the gap were to be genetic, there would be an issue of plasticity in the estimates (due to variable heritability over time). The application of these results to the Brown v. Board case is particularly interesting because of the author’s apparent concern with racial preference and its relationship with self-esteem. The integrationists appealed to ethnocentrism to prove the problem with segregated schools, contradicting their own belief in racial harmony.
This is not to say self esteem has no relation to one’s personal connection to their ethnicity. In fact, many studies support this being the case (eg. Jetten et al., 2015). But, we must avoid the composition fallacy. Take the two facts that ibuprofen treats headaches and cancer tends to cause headaches; we can not say, because this is true, ibuprofen treats cancer (this example from here). This is something the Clarks could have avoided simply by using a typical self esteem questionnaire for children. Because of the many large issues with this study, it is not very good. Despite this, it is, to this day, massively cited in psychology literature (Bergner, 2009).
More recently, another study attempted to show this gap and succeeded in doing so. Unfortunately, it also succeeded in having the same flaws, if not more, than the original study by the Clarks (Edney, 2006).
Instead of these bad studies, we should look to other longitudinal estimates of the racial gap in self esteem. Erol and Orth (2011) used data from the NLSY, including 8 assessments of self esteem and a sample of over 7,000 people. Over a 16 year period, blacks consistently had the highest self-esteem, whites shifted from having the second highest self-esteem to the lowest. The gap increased with age, due to black self esteem rising over time and white self esteem rising, plateauing at adulthood, and slightly decreasing at middle age. In general, the results support self esteem being low for all people in childhood, and generally increasing as they move into adulthood.
An even better study was done by Birndorf et al. (2005). They used a larger sample of 16,000 people from the National Education Longitudinal Study. They also compared the differences by sex. The results in this study are very clearly indicative of greater black self esteem in boys and girls from grade 8 to grade 12. The OR for boys was 1.38 vs. 1.00 and for girls was 2.06 vs. 1.00, blacks scoring higher in each. As we see, this is completely rid of effects of sociodemographics in the logistic regression analyses.
Some other evidence seems to contradict the doll tests even more, by using data from the segregation era. Literature reviews by Tashakkori (1993) and Tashakkori and Thompson (1991), for example, found that self esteem in blacks was overall greater than that in whites and the gap was actually smaller in integrated schools. Seligman (1987) reported on another set of doll tests done around the same time in Northern schools. What they found was that blacks preferred white dolls in the North, in fact, even more so than those in the South. This means that either integration could hurt black students as much as segregation or that the doll test is simply invalid. Both seem to be reasonable options. Finally, the Coleman Report suggested that blacks in the rural South were more likely to classify themselves as bright and able to learn than those in the North.
So, it seems that that old, small-sample, control-lacking study was wrong (who would’ve guessed?). This generally has no real-world implications, but it should be strongly advised psychologists do not continue to cite the doll tests as any significant sign of lower black self esteem. From what can be seen, the evidence suggests blacks today have greater self esteem than whites. Furthermore, there is little to no evidence that segregation hurt black self esteem, and some evidence to the exact opposite, but I find this detail has little importance.