Replicating and Extending Effects of “Achievement Gap” Discourse

Methods

We analyze data from two separate survey experiments: (a) the teacher sample from Quinn et al. (2019) and (b) a second U.S. sample drawn from Amazon MTurk (n = 500). (See online Appendix A for sample descriptive statistics with comparisons to nationally representative data, randomization balance, and disaggregated results. ¹ ) In each experiment, respondents were randomly assigned to one of two versions of our main survey item. One version used the term racial achievement gap, while the other used racial inequality in educational outcomes. ² The main item read:

As you may know, there is [a racial achievement gap/racial inequality in educational outcomes] between Black and White students in the U.S. Thinking about all of the important issues facing the country today, how much of a priority do you think it is to [close the racial achievement gap/end racial inequality in educational outcomes] between Black and White students?

Response options were “not a priority,” “low priority,” “medium priority,” “high priority,” or “essential” (adapted from Valant & Newark [2016]). For the MTurk sample, we used this item with five additional items to create an index (Cronbach’s α=.97; see online Appendix B).

The teacher sample also completed an implicit association test (IAT; Greenwald et al., 1998) measuring respondents’ automatic associations between race (Black/White) and academic competence (see online Appendix C). We use the IAT as a moderator to test whether the effect of AG language on priority ratings differs depending on teachers’ implicit stereotypes.

After answering the gap/inequality item, respondents in the MTurk sample answered 10 stereotype items (based on the General Social Survey) in which they rated racialized groups (Black/White) on five bipolar traits (hardworking/lazy; intelligent/unintelligent; competent/incompetent; capable/incapable; confident/unconfident). The differences in respondents’ average trait ratings of racialized groups compose our stereotype index (Cronbach’s α=.78; see Table 1 note and online Appendix D).

OPEN IN VIEWER

Table 1

Effects of Achievement Gap Language on Priority Ratings for Racial Disparities and on Explicit Racial Stereotypes (MTurk Sample)

	Ach. Gap	Inequality
	Mean(SD)	Mean(SD)	Diff.(SE)	P
Priority outcome
Priority index	3.30 (1.18)	3.62 (1.22)	−0.32** (0.11)	0.003
High priority/essential (1 = Y, 0 = N)	0.44 (0.50)	0.57 (0.50)	−0.13** (0.04)	0.003
Priority (single item)	3.26 (1.20)	3.59 (1.26)	−0.33** (0.11)	0.003
Stereotype outcome
Stereotype index (White mean index – Black mean index)	0.48 (1.09)	0.30 (0.87)	0.18* (0.09)	0.039
Black mean index	4.77 (1.12)	4.92 (1.08)	−0.15 (0.10)	0.14
White mean index	5.25 (0.86)	5.22 (0.87)	0.04 (0.08)	0.628
N	250	250

Note. “Ach. Gap” = respondents randomly assigned to the “Achievement gap” version of priority items; “Inequality” = respondents randomly assigned to the “inequality in educational outcomes” version of items. “Diff.” = Ach Gap – Inequality mean difference; standard error of the mean difference is in parentheses in the “Diff.” column. p = p-value for t-test of equal means across the two conditions. Priority (single item): “As you may know, there is [a racial achievement gap/racial inequality in educational outcomes] between Black and White students in the U.S. Thinking about all of the important issues facing the country today, how much of a priority do you think it is to [close the racial achievement gap/end racial inequality in educational outcomes] between Black and White students?” Responses on 5-point scale: 1 = not a priority; 2 = low priority; 3 = medium priority; 4 = high priority; 5 = essential. “High priority/essential” = 0/1 variable indicating that respondent answered “high priority” or “essential” on priority item. Priority index = mean on priority item and five similar items (see online Appendix B for scale detail and online Appendix F for additional robustness checks). Stereotype items are composed of 10 items: respondents rated Black Americans and White Americans on five bipolar traits (hardworking/lazy; intelligent/unintelligent; competent/incompetent; capable/incapable; confident/unconfident), each with a 7-point scale in which 7 = the respondent believes that “almost all” of the given racialized group tends to exhibit the positive pole of the trait and 1 = “almost all” exhibit the negative pole. Black (White) mean index = mean score respondents gave across five traits for Black (White) Americans; Stereotype index (White mean – Black mean) = White – Black difference in mean score across five traits. See online Appendix D for scale detail and online Appendix F for robustness checks.

*

p < 0.05. ^**p < 0.01. ^***p < 0.001.

Results

In Figure 1, we present results from a fitted logistic regression model predicting whether teachers rated closing the gap/ending inequality as “high priority” or “essential.” We find that teachers’ implicit racial stereotypes moderate the effect of AG language on priority ratings (see online Appendix E for estimates and robustness checks). On the x-axis, positively signed IAT scores represent teachers’ automatic association of White students as being more competent than Black students, negatively signed scores represent the reverse, and zero represents neutrality. As seen by the vertical distance between the two fitted curves, the negative effect of AG language is largest among teachers holding strong implicit anti-Black/pro-White stereotypes. Teachers with automatic associations of Black students as being more competent than White students (negatively signed IAT scores) give high priority regardless of framing condition. The negative effect of AG language on priority levels is statistically significant for respondents with IAT scores above .45. Said differently, higher levels of implicit racial bias predict lower prioritization of racial equity when AG language is used, but “inequality” language neutralizes that negative relationship (IAT does not significantly predict priority level in the “inequality” condition).

OPEN IN VIEWER

Figure 1.

Teachers’ implicit racial stereotypes (IAT score) moderate the effect of achievement gap language on the priority ratings teachers give to racial disparities in education (n = 675)

In Table 1, we show the effects of AG language in the MTurk sample. First, we replicate the finding from the teacher sample in Quinn et al. (2019): AG language lowers the extent to which respondents prioritize racial equity (as measured by the original priority item as well as by the priority index). Descriptively, this negative main effect is larger in the MTurk sample (ES = –.26 SD vs. –.11 SD for teachers).

Consistent with the theory that AG-framing activates deficit mindsets, we also find that AG language increased explicit anti-Black/pro-White stereotypes (see Table 1 and its note for details). On the stereotype index, positively signed values indicate anti-Black/pro-White stereotypes (0 = neutrality). As seen, respondents in both conditions expressed significant anti-Black/pro-White stereotypes. The AG language, however, increased the magnitude of stereotyping by .19 SD (see online Appendix F for robustness checks).

Discussion

We present further evidence that the language used to discuss racial equity in education matters. First, we improve the external validity of Quinn et al. (2019) by replicating the negative effect of AG language in a sample of non-teachers. Importantly, we show evidence consistent with the hypothesis that AG language primes deficit thinking. We show that the effect from Quinn et al. (2019) was driven by teachers who held stronger implicit stereotypes of Black students being less competent than White students. Furthermore, in the new non-teacher sample, AG language increased explicit anti-Black/pro-White stereotypes compared to “inequality” language.

Scholars and advocates sometimes use language associated with deficit models in service of advancing an anti-deficit agenda; as such, it is important to take account of context and to recognize that the use of a term does not automatically ground a discourse in a deficit framework (Patton Davis & Museus, 2019). At the same time, phrases can call to mind specific frames with which they are often associated (Lakoff, 2004). The findings in the present study suggest that in the absence of an explicitly anti-deficit framework, the language of “racial achievement gaps” may carry more negative connotations than the language of “racial inequality”. Combined with recent experimental evidence that a TV news story reporting on racial achievement gaps magnified viewers’ racial stereotypes (Quinn, 2020), the present findings indicate that care should be taken in how racial disparities are framed in efforts to advance educational equity.

Supplemental Material