Saturday, May 08, 2010

The social construction of racism









Reader Linda Seebach wondered in the last post on the frequency of experiencing discrimination among blacks if there were some way to see if  "individuals' answers are consistently in one direction or another?"  This would assess the hypothesis that certain blacks tend to see racism all around in many forms, while others don't experience it in any of various possible ways, thus suggesting it is more in the mind of the victim than an objective reality.

In other words, if racist whites are all around, discrimination should tend to be experienced randomly by blacks. If racism, on the other hand, is found to be concentrated among a few victims, this suggests they are reading racism into events that others would interpret as benign.

Recall that most incidents that are labelled racist are ambiguous. Just look above at the list of questions that interviewers in the MIDUS Study asked respondents. If a waitress gives you bad service, it could be for many different reasons, only one of them being that she doesn't like members of your race. 

To test the competing hypotheses, I can calculate the correlations among the nine questions above. If the correlations are non-existent or low, this means that being a frequent victim of one type of mistreatment is unrelated to the frequency of another type. If, by contrast, the correlations are high, this suggests individuals tend to experience all types of mistreatment often, or none of them at all, which suggests it's much more about the victim and much less about others. 

It turns out that the correlations are overwhelmingly high. The average for all 36 correlations is .58, which means that a person's answer to one question predicts very strongly a similar answer to another question.

Another way to approach it is to calculate the alpha coefficient, which is a measure of the extent to which the questions tap an underlying factor, in this case the frequency of perceiving discrimination of all kinds. The coefficient is .92. This is a very high number and indicates that the questions are all tapping the same thing: a person's tendency to experience many interactions as racist. 

Many blacks are taught to see racists under every rock, so should we call the phenomenon we just observed in the data "the social construction of racism"?   

8 comments:

  1. Anonymous4:03 AM

    Aren't the questions rather correlated with each other though? If someone called you names or insulted you because of race, you are also more likely to think that you are treated with less respect and courtesy, also because of race. They all closely fall together.

    ReplyDelete
  2. Good point, but the results don't change if you focus on more discrete items. For example, thinking you are not smart, dishonest, and acting afraid of you are non-overlapping items. The average correlation among the three is .59, and the alpha coefficient is .81--high for only three items. Same results as for the full list.

    ReplyDelete
  3. "In other words, if racist whites are all around, discrimination should tend to be experienced randomly by blacks..."

    " The coefficient is .92. This is a very high number and indicates that the questions are all tapping the same thing: a person's tendency to experience many interactions as racist. "

    Even if the data rejects the above null hypothesis (the first quote), it does not mean that the conclusion is justified.

    I can immediately think of a few alternative hypotheses to explain the correlation in the data. The most obvious is that racial discrimination is real, but not all members of the target group are subject to it. In other words, racist whites are not all around, they are in specific places. Similarly, racial discrimination could be stronger among one generation than another, or stronger in one socio-economic class than another. So perhaps old southern blacks experience (or have experienced) much more discrimination than younger blacks who live in different parts of the country. Maybe poor blacks are often in subordinate positions where they are subject to the racist prejudices of others, whereas wealthier blacks have enough influence that racists would not dare to treat them like dirt.

    A second alternative is that discrimination is not targeted at race in itself, but instead is targeted at some cultural feature that is strongly associated with race (such as dialect or clothing styles). If some blacks show these cultural identifiers, they may be subject to discrimination while others who adopt "white" culture are not. It may not be literally the same as racial discrimination, but it is arguably just as bad and it is understandable that the target of this discrimination might not be able to tell the difference.

    BTW, I'm assuming that you are familiar with the idea of race as a social construct.

    ReplyDelete
  4. I just realized that your correlations referred to the aggregate data shown above, not data about how individual respondents answered the questions.

    That being the case, the correlation is meaningless. It can easily be explained by treating each category of discrimination as an event that occurs with its own small probability. Using that probability, we can derive an expected distribution based on the Poisson distribution. If two categories occur with the same probability, they will have the same distribution, and their correlation will be 100%, even if their is no correlation at the level of the individual respondents.

    ReplyDelete
  5. Ricketson: 1) the correlations are calculated based on the answers of the approx. 130 individuals--not aggregated data.

    2) To see such strong intercorrelations, white racists would have to be spatially concentrated indeed, and only some blacks would ever enter that circle, and they would have entered it often.

    3) No blacks were from the South. I read the codebook--they are all from Milwaukee.

    4) The Pearson correlation between age and a lifetime measure of all forms of reported discrimination is .01.

    5) The Pearson correlation between a measure of SES and the lifetime discrimination is .09--higher status blacks are slightly MORE likely to report discrimination. I think the cultural marker you speak of would be highly correlated with SES.

    ReplyDelete
  6. This data pattern is pretty powerful. I think this does show that seeing racism is more about subjective framing of events than reality. However, you only have 130 subjects to go on. I'd like to see this replicated with much larger representative samples. I doubt any researcher is motivated to do that though.

    ReplyDelete
  7. Thanks for clarifying the stats.

    ReplyDelete
  8. Ricketson: 1) the correlations are calculated based on the answers of the approx. 130 individuals--not aggregated data. 2) To see such strong intercorrelations, white racists would have to be spatially concentrated indeed, and only some blacks would ever enter that circle, and they would have entered it often. 3) No blacks were from the South. I read the codebook--they are all from Milwaukee. 4) The Pearson correlation between age and a lifetime measure of all forms of reported discrimination is .01. 5) The Pearson correlation between a measure of SES and the lifetime discrimination is .09--higher status blacks are slightly MORE likely to report discrimination. I think the cultural marker you speak of would be highly correlated with SES.

    ReplyDelete

Are gun owners mentally ill?

  Some anti-gun people think owning a gun is a sign of some kind of mental abnormality. According to General Social Survey data, gun owners ...