African American Women’s Reports of Racism during Hurricane Katrina: Variation by Interviewer Race

Abstract

This study investigated the effects of interviewer race on low-income African American female hurricane survivors’ reports of racism during Hurricane Katrina and its aftermath (N = 41). Respondents were asked directly about the role of racism during the storm and evacuation by one of three interviewers (two White females and one African American female). Contrary to expectations, respondents were not significantly more likely to agree that racism played a role during the hurricane and its aftermath when with an African American interviewer compared to a White interviewer. However, when speaking to the White interviewers versus the African American interviewer, respondents were significantly more likely to use qualifying and contradictory statements and to make references to other races also being victims of the hurricane.

Despite the pervasiveness of racism in American society, mainstream psychology researchers have not fully considered its potential impact on research findings. In particular, there has been little attention given to the effect of interviewer race on data. For example, a PsycINFO database search (conducted in December 2010) using the search terms “interviewer race,” “race of the interviewer,” “interviewer ethnicity” and “ethnicity of the interviewer,” as well as identical searches using the terms “researcher,” “investigator” and “experimenter” in place of “interviewer,” resulted in only 51 journal articles. Of the 51 results, only eight were empirical studies drawing on qualitative interview data (as opposed to survey, test, or other data collected via interview methods), and only four of those eight were published in psychology journals (the remainder were published in interdisciplinary social science journals, primarily in Public Opinion Quarterly).

The lack of attention to the effects of interviewer race is a significant oversight, given that both quantitative and qualitative researchers frequently use interviewers to collect data. In this study we explored the impact of interviewer race (White vs. African American) by examining responses to questions about the role of racism during Hurricane Katrina among 41 low-income African American female hurricane survivors. We investigated whether interviewer race impacted respondents’ assessments of the role of racism during the hurricane and its aftermath, paying specific attention to the quality of the language used within their responses.

Racism and Hurricane Katrina

Hurricane Katrina struck the New Orleans area on August 29, 2005, and was the worst hurricane in recent United States history (Knabb, Rhome, & Brown, 2006). The hurricane and its aftermath had a major impact on vulnerable populations in New Orleans, particularly low-income African Americans. African American communities were damaged more than White communities (Logan, 2006), and African Americans affected by the storm have since reported higher rates of unemployment, psychological distress, and general life disruption than Whites (Elliot & Pais, 2006; White, Philpot, Wylie, & McGowen, 2007).

The disproportionate impact of Hurricane Katrina on low-income African Americans seems to be due, at least in part, to the interaction of racial and class injustices in pre-hurricane policies, as well as in the treatment of victims during the hurricane and its aftermath. For example, the city government failed to repair the levees surrounding the low-income African American community despite warnings of their fragility (Park & Miller, 2006). Evacuation policies relied on private means of transportation that were less available to African Americans (Lavelle & Feagin, 2006). Research has shown that African Americans were less likely than Whites to have an evacuation plan in place prior to the hurricane (Spence, Lachlan, & Griffin, 2007), and were less likely to have evacuated during the storm (Elliot & Pais, 2006).

The government has also been blamed for “bureaucratic failures” during the hurricane, wherein low-income African American citizens were put at heightened risk due to a strict adherence to rules and lack of use of discretion that would assist them (Molotch, 2006). Public administration scholar Christine Stivers (2006) has provided several examples of such failures, ranging from the slow transportation of food and water to the Louisiana Superdome, which served as a “shelter of last resort” to over 25,000 New Orleans residents (Brinkley, 2006), to the denial of small business loans to applicants from the hardest hit areas after the hurricane. Although race and class clearly interacted during the government’s response to the hurricane, many scholars, including Stivers, see racism as one of the major factors accounting for the increased risk of adversity among low-income African Americans during Hurricane Katrina and its aftermath.

However, despite evidence of racial discrimination in the government’s preparation for and response to the hurricane, poll data collected through telephone interviews suggests that not all Americans agree that racism played a part in what occurred (Page & Puente, 2005; Pew Research Center, 2005). For example, only approximately 60% of African American and 20% of White survey respondents agreed that the government’s response would have been faster if most of the victims were White (Page & Puente, 2005). Notably, these studies do not take into account participants’ perceptions of the interviewers’ race, which could have affected their responses to such questions. Furthermore, it is possible that the results would have differed in face-to-face interviews, wherein participants have more intimate contact with the interviewers, as well as much greater access to cues about the interviewers’ race.

Racism and Interviewer Effects

When considering participants’ responses to questions, especially those about racism, it is important to take into account the context of a given study. Literature drawn from sociology and political science shows that the dynamics between the researcher and respondent can affect the content of disclosures, particularly regarding racially explicit material (Anderson, Silver, & Abramson, 1988; Campell, 1981; Carr, 1971; Cotter, Cohen, & Coulter, 1982; Schaeffer, 1980; Singer, Frankel, & Glassman, 1983; Tucker, 1983). Using data from the mid-eighties, for example, Davis (1997a, 1997b) found that African Americans were less likely to endorse statements reflecting racial consciousness (e.g., that African Americans should form their own political party), and more likely to endorse statements reflecting a lack thereof (e.g., that African Americans are not oppressed by White people) when interviewed by a White interviewer than by an African American interviewer. In another study specifically designed to examine interviewer race effects, Krysan and Couper (2003) found that African American respondents gave more conservative responses to questions explicitly related to racial issues when interviewed by White interviewers. A study by Stangor, Van Allen, Swim, and Sechrist (2002) found that African American college students were more likely to regard discrimination as a cause of a failing grade when they made that judgment privately or with a member of their racial in-group. In contrast, when they expected to make the judgment out loud to a member outside of their racial group, they were less likely to report discrimination.

Interviewer race may also impact the language used in African Americans’ responses to questions about racism. Van den Berg, Wetherell, and Houtkoop-Steenstra (2003) have stated that open-ended interviews about sensitive and controversial topics such as racism and ethnicity frequently produce ambiguous and contradictory statements that are difficult to interpret. Supporting this view, Davis (1997a) found that African Americans were more likely to agree to mutually contradicting statements than close-ended questions (e.g., indicating that they support both the Democrat and Republican parties) when asked by a White versus an African American interviewer. In addition, in interpreting interviews about race in Australia, Buttny (2003) found that minority respondents were more likely to use a variety of speaking strategies to talk to about race with White than non-White interviewers. For example, they tended to quote out-group members (i.e., repeating what someone of another race said) and “the prototypical racial other” (i.e., using an individual’s actions or speech as representative of all members of the other racial group) more frequently when talking about race with White interviewers. Buttny postulates that these findings reflect the continuing stress of racism in respondents’ lives, and their avoidance of being held accountable for criticizing the majority group. Given the persistence of racism and dominance of Whites in the United States, it is possible that African Americans may also use these strategies here.

Davis (1997a) posits that variation in the content and quality of African Americans’ responses to interviewers’ race stems from several factors. First, differences in responses given to White versus African American interviewers may be indicative of cautiousness and distrust of Whites, and the general tension between the two racial groups. Others similarly suggest that not reporting discrimination to a member of an out-group could be due to fear of being negatively judged (e.g., Kaiser & Miller, 2001) or having an otherwise uncomfortable interpersonal interaction (Garcia, Reser, Amo, Redersdorff, & Branscombe, 2005). As a result, African Americans may give responses to White interviewers that they believe White interviewers want to hear, including opinions that distance themselves from other African Americans or from race-related issues. These strategies are considered adaptive considering current and historic racism. A potential implication of this perspective is that African Americans’ responses to African American interviewers might have greater validity due to respondents’ greater comfort and trust.

However, as Bobo and Fox (2003) note, researchers should not assume that African American’s assessments of discrimination with an African American represent their “true” opinion. For example, Davis (1997a) speculates that, with African American interviewers, African Americans may feel pressure to report a heightened sense of racial solidarity or support for causes that are related to African Americans or persons of color and contrary to the dominant White culture. According to Davis, these responses may be influenced by a desire not to be labeled a “sell-out” (p. 312) and may be encouraged by subtle verbal and non-verbal cues of African American interviewers. Such pressure could likewise account for variation in African Americans’ responses to African American and White interviewers; however, here it would not necessarily imply that their responses to African American interviewers have greater validity than their responses to White interviewers. More generally, racial deference toward both African American and White interviewers may also be influenced by social desirability. That is, interviewees may avoid giving certain responses to both the White and African American interviewers in order to avoid directly offending them.

In this study, we investigated whether low-income African American female hurricane survivors’ assessments of racism during Hurricane Katrina and its aftermath differed by race of the interviewer. Based on previous research, we hypothesized that respondents would be more likely to assert that racism occurred with an African American interviewer compared to a White interviewer. We also expected greater ambiguity in the language they used to talk about race and racism with White compared to African American interviewers (e.g., referencing others’ points of view, giving contradictory answers, etc.).

Method

Procedure and Respondents

Respondents were drawn from a larger study of low-income parents who had been enrolled in three community colleges in the city of New Orleans in 2004-2005 prior to Hurricane Katrina. The purpose of the larger study (ongoing) is to examine whether performance-based scholarships affect academic achievement, health, and well-being among these students (Richburg-Hayes et al., 2009). To be eligible for the study, students had to be between the ages of 18 and 34, be parents of at least one dependent child under the age of 19, have a household income under 200 percent of the federal poverty level, and have a high school diploma or equivalent. The initial sample contained 1,019 participants, 492 of whom had completed one-year follow-up survey batteries prior to Hurricane Katrina.

Following the hurricane, between March 2006 and January 2007, 402 of these 492 participants (81.7%) were relocated and completed a comprehensive survey containing the same items as the one-year follow-up, plus additional items assessing their experiences during and after Hurricane Katrina. From these 402 participants, 57 were selected to complete comprehensive, semi-structured qualitative follow-up interviews, conducted between August 2006 and March 2007. These 57 respondents were selected because they had lived in either Orleans or Jefferson Parish prior to the hurricane and had suffered some damage to their pre-hurricane home. Respondents were also selected to contain a comparable number of respondents who had returned to New Orleans after the storm and respondents who had relocated elsewhere (e.g., to Baton Rouge, Houston, or Dallas). Since the overall sample was predominantly female (e.g., 92.5% of the 1,019 participants who completed the baseline survey), the investigators decided to include only women in the interview sample.

Two of the 57 respondents did not self-identify as African American and were dropped from our analysis. Of these 55 African American respondents, 41 were either directly asked the question, “Do you think race played a part in what happened with Hurricane Katrina?” or spontaneously discussed this issue during the course of their interviews. We dropped the remaining 14 respondents from our analysis because they were not asked this question directly and did not discuss race spontaneously, or because their responses were inaudible due to technical difficulties during the interviews. Two White female interviewers interviewed 31 (75.6%) of these 41 respondents, and one African American female interviewer interviewed the remaining 10 (24.4%) respondents. These 41 respondents’ average age was 23.37 (SD = 3.38). All were parents, with 53.7% reporting one child, 17.1% reporting two children, and the remainder reporting three or more children prior to Hurricane Katrina. The large majority (78.0%) was living without a spouse or partner at the time of the hurricane. The mean pre-hurricane monthly personal income for the sample was $541.23 (SD = $575.72; Median = $300.00; Range: $0.00 to $2000.00).

It is important to note that no respondents were asked directly about how they thought the interviewers’ race affected their responses to the questions. Also, the qualitative study was not initially conducted with the purpose of examining interviewer race effects, and so interviewer characteristics other than race were not held constant (with the exception of gender, since all interviewers, like all respondents, were females). While we are conscious of the limitations that such an indirect and non-experimental research design presents for the interpretation of our results, it is equally important to note that this situation simulates the “real world” of much survey and interview data that are collected in mainstream psychological and other social science research, in which interviewer race effects are not directly controlled, but may nonetheless operate. Interviews typically lasted one to two hours and covered a range of topics including respondents’ evacuation experiences, childhood and family of origin histories, education and work histories, partnering and parenting histories, and their expectations and hopes for the future.

Data Analysis Plan

The research team first developed general descriptive codes based on prior research on disasters and covering a broad range of topics (e.g., education, work, relationships) (e.g., Gibbs, 1989; Norris et al., 2002). The team coded the original 57 interviews with these descriptive codes using Atlas.ti, a computer software program that allows for the qualitative analysis of large bodies of textual, graphical, audio, and video data. To generate a coding scheme for the current study, three independent raters reviewed the results of the descriptive codes related to discrimination, racial/ethnic relations, social class, and the government’s response to Hurricane Katrina. Each rater then generated a list of content codes related to respondents’ perceptions of racism, and the three raters decided on a final list of detailed codes through discussion and mutual consensus.

Because we aimed to explore interviewer race effects both qualitatively and quantitatively, we focused our analyses mainly around responses to the standard interview question that directly tapped into 41 of the respondents’ perceptions of racism (“Do you think race played a part in what happened with Hurricane Katrina?”). Two researchers coded for the presence or absence of each content category in respondents’ responses to that question (or in their spontaneous discussion of the role of racism during the hurricane and its aftermath if they were not asked the question directly, as mentioned earlier). A third researcher collected and compared the two other researchers’ analyses, compiling, cleaning, and adding her own analysis of the detailed codes along the way. Finally, the three authors reviewed each other’s analyses, discussing places of consensus and resolving remaining discrepancies. Due to the small size of our sample, as well as the exploratory nature of the study, we set the cutoff criteria for statistical significance of the quantitative analyses at p = .10 (Bernard, 2000; Cascio & Zedeck, 1983).

Coding Schema

Perceived racism

Based on the 41 respondents’ responses to the standard interview question about race, three mutually exclusive Perceived Racism categories were created. The first category designated respondents who asserted that racism played a part during the hurricane and its aftermath (“Yes”), and the second designated those who asserted that it was not a factor (“No”). The third category designated respondents who said that they were unsure about whether racism played a role during the hurricane (“Maybe/Unsure”). We hypothesized that respondents would be more likely to assert that racism had played a role with the African American interviewer, and more likely to assert that it had not, or that that they were unsure, with a White interviewer. To facilitate our quantitative analyses, a dummy-coded variable was created for each Perceived Racism response category.

Quality of responses

In addition to coding for respondents’ overall perceptions of racism, we also coded such responses for various features. Quotes that illustrate each feature that was coded are listed in Table 1. Each of the Quality of Responses codes was also quantified as a dummy variable. That is, for each, respondents were given a 0 if they did not use that type of statement, and a 1 if they did.

Table 1.

Qualitative Content Codes, and Exemplary Quotes

Content Code	Example
Qualifying Statement	RESPONDENT: [Yes, I think race played a part] because at the time they was only picking up White people. WHITE INTERVIEWER: They were? RESPONDENT: No offense.
Contradictory Statement	WHITE INTERVIEWER: Do you think that race played a role in what happened with the evacuation or with rebuilding? RESPONDENT: With the evacuation, I want to say no, but it seemed like it did. It seemed like it did. But I’m going to say no. I want to say no, I’m going to say. I want to say, no, but the way -- the way they [did] things, I think it did.
References to Others’ Opinions	RESPONDENT: Really, I don’t think that race played a factor in it. No. I know a lot of people say that. But I don’t know, I guess some people feel like you have to blame someone.
Racism Doesn’t Bother Me	RESPONDENT: I thought about it but I don’t try to think so much into it.
Class is a Bigger Issue	RESPONDENT: To a certain degree, I would say. I don’t think it was the sole factor for the reasons why things went the way they did. I think to a certain degree. I think that economical status probably played a big part in it. Because some people feel that every-body in New Orleans was in poverty. That nobody had [anything].
Other Races Were Victims	RESPONDENT: I’m not going to say it was a color thing because there were some white folks that [were] left behind. It was just unfair to New Orleans period.

First, we coded for Qualifying Statements, which were operationalized as statements that seemed to minimize the impact of assertions of racism made to the interviewer (e.g., “no offense,” “I don’t want to say this, but…”). These statements usually occurred at the beginning of respondents’ responses as prefaces to an affirmative answer that racism did play a role during the hurricane.

Next, we coded for Contradictory Statements. In these statements, respondents asserted that racism played a role during the hurricane, but later went on to rescind or alter their responses during the interview. These responses stuck out as being difficult to code regarding whether or not respondents perceived racism to have played a role either during or after Hurricane Katrina.

Our next code, References to Others’ Opinions, describes responses in which respondents alluded to thoughts, ideas, and opinions about racism, or personal experiences of racism, of people other than themselves. These responses referred to either specific people in their lives (e.g., family members, friends) or to a more amorphous group of others (e.g., “some people”). This type of response is similar to quoting out-group members or “the prototypical racial other,” as described earlier (Buttny, 2003).

The code Racism Doesn’t Bother Me designated responses in which respondents said that racism did play a role during the hurricane, or that they have heard others say that it played a role, but that they, personally, were not affected or bothered by racism. These respondents stated that racism does not play a major role in shaping their personal experiences, or that it is something they rarely consider in their everyday lives.

The code Class is a Bigger Issue designated responses by respondents who emphasized the importance of socioeconomic status in shaping what happened during the hurricane. In these responses, respondents seemed to downplay or negate the role of racism during the hurricane, while at the same time identifying that class discrimination, which they felt took precedence, had occurred.

Our last code, Other Races Were Victims, designated responses by respondents who pointed out that groups of people other than African Americans were also affected by the hurricane to be evidence that racism did not play a role, or at least that racism cannot fully explain what happened during the hurricane and its aftermath.

A commonality among all of these Quality of Responses codes is that they generally downplay the presence and role of racism during and after the hurricane, whether respondents asserted that racism had occurred or not. Similarly, several of these Quality of Responses codes serve to distance the respondents from their evaluations that racism shaped the government’s response to and effects of the hurricane. We therefore hypothesized that the Quality of Response codes would appear more frequently with the White interviewers compared to the African American interviewer.

In sum, Perceived Racism categories characterize respondents’ full responses, whereas Quality of Responses categories were applied to specific statements or utterances within such full responses. For this reason, although our codes for Perceived Racism and Quality of Responses may sometimes overlap (e.g., a Quality of Response code that includes a Contradictory Statement might be more likely to be coded as “Maybe/Unsure” under Perceived Racism), they do not entirely overlap.

Results

Perceived Racism

Among the 41 respondents, 19 (46.3%) reported that racism played a role during or after Hurricane Katrina, 13 (31.7%) replied that racism might have played a role or that they were unsure, and 9 (22.0%) did not perceive racism to have played a role. The breakdown of respondents’ reports of perceived racism is presented in Table 2. According to a 3-X-2 chi-square analysis, respondents’ perceptions of racism did not significantly differ by race of the interviewer, χ2 (2, N = 41) = 1.19, p = .55.

Table 2.

Perceptions of Racism for Full Sample and by Interviewer Race

	Full Sample (N = 41)	White Interviewer (n = 31)	African American Interviewer (n = 10)
Yes	46.3% (19)	45.2% (14)	50.0% (5)
No	22.0% (9)	25.8% (9)	10.0% (1)
Maybe/Unsure	31.7% (13)	29.0% (9)	40.0% (4)

Note. A 3-X-2 chi-squared analysis detected no significant differences in perceptions of racism by interviewer race, χ2 (2, N = 41) = 1.19, p = .55, Cramer’s V (measure of effect size) = .17.

Quality of Responses

As seen in Table 3, respondents differed in the language with which they spoke about racism with African American versus White interviewers. To test for significant differences between respondents interviewed by African American and White interviewers, 2-X-2 chi-square analyses were conducted for each Quality of Responses code. Given that some of the cells contained less than five respondents, we employed Fischer’s exact significance test (Agresti, 1992). Analyses demonstrated significant differences in the Qualifying Statements (χ2 (1, N = 41) = 9.37, p < .01), Contradictory Statements (χ2 (1, N = 41) = 3.21, p < .10), and Other Races Were Victims (χ2 (1, N = 41) = 2.89, p < .10) codes, such that each was more likely to be applied for respondents interviewed by a White versus an African American interviewer. Differences regarding the other qualities of responses were not significant.

Table 3.

Qualitative Contents for Full Sample and By Interviewer Race

	Full Sample (N = 41)	White Interviewer (n = 31)	African American Interviewer (n = 10)	χ2	Cramer’s V
Qualifying Statement	41.5% (17)	54.8% (17)	0.0% (0)	9.37***	.48
Contradictory Statement	19.5% (8)	25.8% (8)	0.0% (0)	3.21*	.28
References to Others’ Opinions	24.4% (10)	29.0% (9)	10.0% (1)	1.46	.19
Racism Doesn’t Bother Me	12.2% (5)	16.1% (5)	0.0% (0)	1.84	.21
Class is a Bigger Issue	51.2% (21)	54.8% (17)	40.0% (4)	.67	.13
Other Races Were Victims	31.7% (13)	38.7% (12)	10.0% (1)	2.89*	.27

^*Note. = p ≤ .10

^**= p ≤ .05

^***= p ≤ .01

We selected the following excerpts from interview transcripts as illustrative of the three categories with significant differences. First, regarding Qualifying Statements, an African American mother of five, who had evacuated before the storm hit and who was still living in Houston at the time of her interview, qualified her spontaneous response about race to a White interviewer (with italics added for clarity and emphasis):

WHITE IV: Do you think this would have happened if it would have been another city?

RESPONDENT: No.

WHITE IV: Why not?

RESPONDENT: I don’t want to say race plays a part, but I’m going to say it. I think it played a part. The majority [of people in New Orleans] are African Americans.

By contrast, the following African American mother of two, who had evacuated before the storm hit and who had relocated back to New Orleans by the time of her interview, gave a much more direct response to an African American interviewer when explaining this sentiment:

AFR. AMER. IV: Do you think race has played a part in Hurricane Katrina?

RESPONDENT: Yes, yes, yes.

AFR. AMER. IV: How?

RESPONDENT: If you notice they let a breach come back through the levee back there, in the 9th ward. That’s only be cause it was predominaly Black people. If you look at it, all the White people’s area was saved.

Second, regarding Contradictory Statements, an African American mother of four, who had not evacuated until after the storm hit and who had relocated back to New Orleans by the time of her interview, offered a contradictory response to a White interviewer. Her original response was an emphatic “yes”; however, she quickly altered her response to a “no”, explaining that other races were hurricane victims, too:

WHITE IV: Do you think that race was a part of …?

RESPONDENT: Yes.

WHITE IV: How so?

RESPONDENT: Well no, I’m going to change that. Because there were all kinds out there. There were Caucasian, there were African American, there was Spanish, there was Vietnamese. There [were] all kinds. So, no. We were all out there and we came as one to help each other during that time.

By contrast, the following African American mother of one, who had evacuated before the storm hit and who was still living in Houston at the time of her interview, did not budge from her original “yes” response when answering the same question posed by an African American interviewer:

AFR. AMER. IV: How do you feel about the way the government responded to Hurricane Katrina?

RESPONDENT: I thought that they were racist, and I still think they racist at this point. I mean they didn’t put FEMA people on the phone to make sure [they help you]. They harass the hell out of you.

Third, regarding Other Races Were Victims, an exemplary quote came from an interview with an African American mother of one, who had evacuated before the storm hit and who had relocated back to New Orleans by the time of her interview. Even though this respondent explained to a White interviewer that victims of the hurricane and evacuation were majority African Americans, she simultaneously downplayed the importance of racism by emphasizing that other races were also affected by the storm. In her responses to follow-up questions, she emphasized her “open-mindedness” and disapproval of other African Americans who focus too strongly on the role racism during the hurricane:

WHITE IV: Do you think that race played any role in what happened? Like, with the way in which the government responded?

RESPONDENT: No, because there [were] all kinds of races that [were] stuck out, so I don’t feel as though it was race at all. I mean, it’s the majority of Black people, but…

WHITE IV: What about what Nagin said about how Black people are not coming back to the city, aren’t getting the same help to come back?

RESPONDENT: I thought it was funny when he said that…

WHITE IV: You don’t think it’s true?

RESPONDENT: No.

WHITE IV: Why do you think he said that?

RESPONDENT: I don’t know. That’s why I say his mouth gets him in trouble. I don’t know why. I was like, “Is he serious?” Everybody has their different things, but to me, I’m open-minded. I feel as though it’s not a race thing. It’s majority Black, but, I mean, you never know who else [was] stuck out. You’re just looking at yourself for now…. But they have a lot of people in the same situation that [are] not Black.

By contrast, the following African American mother of one, who had also evacuated before the storm hit and relocated back to New Orleans by the time of her interview, connected the “majority Black” hurricane victims to the role of race more clearly and directly with the African American interviewer:

AFR. AMER. IV: Do you think race has played a part in what has happened in Hurricane Katrina?

RESPONDENT: Yes.

AFR. AMER. IV: Tell me [what makes you say that].

RESPONDENT: Because the majority of all the people that died in the hurricane, they weren’t White, they were Black people. And it’s not no secret that the majority of the reason why a lot of the people couldn’t get out of the city is because they were Black and unemployed, living in projects, no cars, no means of transportation, no way to even get out to go anywhere.

In this particular case, differences by interviewer race translated into different overall responses to the question of whether race played a role during the hurricane and its aftermath (“No” to a White interviewer and “Yes” to an African American interviewer). However, as we have shown regarding Qualifying Statements and Contradictory Statements, differences in the quality of respondents’ statements by interviewer race sometimes emerged even when respondents’ overall perceptions of racism did not differ by race of the interviewer.

Discussion

The purpose of this study was to explore the impact of interviewer race on 41 low-income African American female hurricane survivors’ assessments of racism during Hurricane Katrina and its aftermath. We predicted that respondents would be significantly less likely to assert that racism played a part during and after the disaster with the White interviewers than with the African American interviewer. This hypothesis was not supported, however. Following a high-profile national disaster, respondents were equally likely to report that racism played a role during or after the hurricane to the African American and White interviewers. This might have been due in part to the self-protective nature of attributions to causes other than racism (Schmitt & Branscombe, 2002). Future research should therefore account for additional variables that could impact such disclosures. For example, according to the Rejection-Identification Model (Branscombe, Schmitt, & Harvey, 1999), stronger in-group identification lessens the negative psychological impact of attributions to discrimination. Along these lines, past research indicates that individuals who identify less with their social groups report different perceptions of the prevalence of discrimination depending on whether they are talking to individuals in their in-group versus out-group (Postmes, Branscombe, Spears, & Young, 1999).

In addition, we sought to explore differences in the quality of the language these respondents used when speaking with the White versus African American interviewers. Based on previous literature, we predicted that African American respondents interviewed by White interviewers would be more likely to use strategies that create distance between themselves and their responses regarding the sensitive topic of racism, resulting in answers that were less direct and more ambiguous. Our results provided some support for this hypothesis and suggest that even though respondents were not necessarily more likely to report that racism played a role during or after the hurricane to the African American than White interviewers, they were still more likely to use a variety of strategies that softened the potential impact their responses with the White interviewers compared to the African American interviewer. When discussing the role of racism, they were significantly more likely to use qualifying statements, to contradict previous statements, and to assert that members of other racial groups were also victims of the hurricane.

These statements, which were used more frequently with White interviewers, may have been employed to avoid offending or rupturing a connection with the interviewers. The African American respondents likely did not know where the White interviewers stood on the topic of racism, and it is possible that they felt more uncomfortable or reluctant to disclose their opinions on the subject to them, resulting in a greater level of contradictory statements made about racism. It is further possible that qualifying responses and allusions to other races being victims of the storm were used to minimize any potential offense to the White interviewer. In contrast, the weaker usage of such strategies in interviews with the African American interviewer may indicate greater comfort with a same-race interviewer, or a heightened sense of racial solidarity and willingness to voice perceptions of racism during the hurricane (Davis, 1997a). Future research could explore further why African Americans may be likely to employ such strategies more frequently with White than African American interviewers, both within and outside of disaster settings such as Hurricane Katrina.

More broadly, our results demonstrate the importance of paying attention to both interviewer and respondent characteristics when undertaking studies that involve contact between researchers and respondents, especially in psychological studies involving questions about sensitive topics such as racial discrimination. To the extent that interviewer characteristics in particular affect the quality of the data provided to researchers by respondents, we can make sure to collect data on them in the first place (as the present study did by documenting the three interviewers’ as well as all of the respondents’ racial backgrounds), to better control for them in our analyses, and to continue to investigate the reasons behind their impact.

The results of the study also have implications for clinicians working with clients of other races, particularly White clinicians working with African American clients. As has been noted elsewhere (Hays, 2001; Sue & Sue, 1999), clinicians should strive to maintain an awareness of how their own racial backgrounds affect their relationships with clients, what information clients will disclose, and especially how they will disclose and discuss sensitive topics. Both academic research and clinical interviews are active processes within which interviewer and clinician characteristics influence the type of, and the manner in which, information is shared by participants and clients. This is consistent with feminist scholars’ assertion that the power and authority imbued in the researcher’s role, personal position, and insider status significantly influence the co-construction of the interview (Hesse-Biber, 2007).

Although the results of this study offer preliminary data on how interviewer race affects discussions of racism, it is not without limitations. First, it explores only differences in African Americans’ responses with African American and non-Latino White interviewers after a major disaster. Interviewer effects are likely to occur across other racial dyads, depending on the sociocultural and historical context in which interviews are situated, as well as in non-disaster situations. Furthermore, all respondents and interviewers were female; the benefit of this is that gender was held constant across all participants in our study, but the manner in which differential responses manifest themselves could also be different among men and, or between same-sex and cross-sex interview pairs, especially surrounding sensitive issues regarding gender. In addition, we cannot address whether additional interviewer characteristics (such as research skills, comfort in soliciting responses pertaining to race, age, socioeconomic and educational background, and propensity to focus on certain topics over others) may have impacted the results, since they were not held constant in the larger project. Future studies can therefore expand upon the current study by systematically including a larger sample of interviewers and systematically varying their race, class, and other background characteristics. They can also ask respondents directly about how they perceive the interviewer’s race and identity characteristics, which would provide an interesting additional perspective from respondents’ points of view even in situations where interviewer characteristics cannot be fully controlled.

Finally, our small sample size limits the statistical power of our quantitative analyses, and the study was focused only on racism during a major disaster situation, which limits the potential generalizability of our results to non-disaster situations. Future research should continue to explore the impact of interviewer characteristics in non-racism and non-disaster situations. Nonetheless, the results reiterate the importance of paying attention to the roles of both interviewer and respondent race in psychological research, particularly when such research involves face-to-face surveys or interviews, focuses explicitly on sensitive issues such as racism, and occurs during disaster situations. To take responses at face value would be to overlook the interactional nature of such research, the historical and current effects of racism in the United States, and the tension and discomfort that discussions about racism often evoke. As we have shown, even in an instance where overall responses do not suggest that interviewer race effects, a more thorough analysis of the language used in responses can suggest otherwise.