Could teachers’ racial stereotyping be a cause of black Caribbean boys’ under-achievement at school? That’s the question raised by a new paper by Simon Burgess and Ellen Greaves.
They compared two measures of children’s ability at age 11: their results on Key Stage 2 tests - which are externally marked - and teachers’ assessments of their ability.
In about three-quarters of cases, the two measures coincided, which is what you‘d expect. There were, though some cases where teachers’ assessments were higher than KS2 results, and some where they were lower. You’d expect this too, as there‘s a random element to exam results and teachers‘ opinions.
But here’s the thing - there are systematic racial differences in these deviations.
Take pupils who achieve level 4 - the expected level - in KS2 tests for English. Among white pupils, 12.4% got worse teachers’ assessments than their KS2 score, and 10.2% got better assessments. However, among black Caribbean pupils, 17.2% got worse teachers’ assessments and only 7.5% better. Among Pakistani pupils, 20.2% got worse assessments than KS2 scores, and only 6.2% better. For pupils entitled to free school meals, 19.1% were under-assessed by teachers and only 5.8% over-assessed.
There were similar, though smaller, patterns for maths tests.
The message here is depressing. Caribbean, Pakistani and poor kids are more likely than richer white ones to be regarded as poor students even if they are not.
Some teachers, then, seem sometimes* to be seeing what they expect to see. Because they expect Caribbean or poor kids to do badly, this is what they see. Consistent with this, teachers tend to over-assess Indian pupils’ maths abilities.
We shouldn’t, though, be too hard on them for doing this. As Dan Ariely has shown (pdf), seeing what we expect to see is a common trait. As he says:
There’s some experimental evidence that corroborates this, cited by Ariely. Psychologists at Harvard University got some Asian-American women to sit a maths test. Before the test, some were asked gender-related questions and some race-related questions. And they found that the women who answered the race questions performed better than those who answered the gender ones. This suggests that people can easily be influenced to live up or down to stereotypes. Women who were primed to think like women lived down to the stereotype (“girlies are bad at maths” ) whilst women promed to think of their race ( “Asians are good at sums”) lived up to it, even though the experimenters never mentioned these stereoypes explicitly.
Now, if highly intelligent adults are subject to stereotype effects when presented just once with quite weak cues, isn’t it likely that children will also be subject to them, when confronted continuously with stronger reminders?
Which raises a horrible possibility. Could it be that schools actually increase inequalities of educational achievement - between Caribbeans and whites or between rich and poor - through these stereotype effects? More generally speaking, could it be that the state - far from being a force for greater equality - is actually a force for inequality?
* These qualifications are important. We’re talking marginal statistical tendencies here, not universal facts.
** For my money, Predictably Irrational is by far the best of books of that type.
They compared two measures of children’s ability at age 11: their results on Key Stage 2 tests - which are externally marked - and teachers’ assessments of their ability.
In about three-quarters of cases, the two measures coincided, which is what you‘d expect. There were, though some cases where teachers’ assessments were higher than KS2 results, and some where they were lower. You’d expect this too, as there‘s a random element to exam results and teachers‘ opinions.
But here’s the thing - there are systematic racial differences in these deviations.
Take pupils who achieve level 4 - the expected level - in KS2 tests for English. Among white pupils, 12.4% got worse teachers’ assessments than their KS2 score, and 10.2% got better assessments. However, among black Caribbean pupils, 17.2% got worse teachers’ assessments and only 7.5% better. Among Pakistani pupils, 20.2% got worse assessments than KS2 scores, and only 6.2% better. For pupils entitled to free school meals, 19.1% were under-assessed by teachers and only 5.8% over-assessed.
There were similar, though smaller, patterns for maths tests.
The message here is depressing. Caribbean, Pakistani and poor kids are more likely than richer white ones to be regarded as poor students even if they are not.
Some teachers, then, seem sometimes* to be seeing what they expect to see. Because they expect Caribbean or poor kids to do badly, this is what they see. Consistent with this, teachers tend to over-assess Indian pupils’ maths abilities.
We shouldn’t, though, be too hard on them for doing this. As Dan Ariely has shown (pdf), seeing what we expect to see is a common trait. As he says:
When we believe beforehand that something will be good…it generally will be good - and when we think it will be bad, it will be bad (Predictably Irrational, p160-61)Even so, the effects of this might be very nasty. Burgess and Greaves fear that if teachers believe pupils are worse than they really are then they might put in less effort, or pupils who feel under-valued might not try so hard. The result could be that these stereotypes might be self-fulfilling.
There’s some experimental evidence that corroborates this, cited by Ariely. Psychologists at Harvard University got some Asian-American women to sit a maths test. Before the test, some were asked gender-related questions and some race-related questions. And they found that the women who answered the race questions performed better than those who answered the gender ones. This suggests that people can easily be influenced to live up or down to stereotypes. Women who were primed to think like women lived down to the stereotype (“girlies are bad at maths” ) whilst women promed to think of their race ( “Asians are good at sums”) lived up to it, even though the experimenters never mentioned these stereoypes explicitly.
Now, if highly intelligent adults are subject to stereotype effects when presented just once with quite weak cues, isn’t it likely that children will also be subject to them, when confronted continuously with stronger reminders?
Which raises a horrible possibility. Could it be that schools actually increase inequalities of educational achievement - between Caribbeans and whites or between rich and poor - through these stereotype effects? More generally speaking, could it be that the state - far from being a force for greater equality - is actually a force for inequality?
* These qualifications are important. We’re talking marginal statistical tendencies here, not universal facts.
** For my money, Predictably Irrational is by far the best of books of that type.
While it's easy to beat people with the racism stick, there are other possibilities. A teacher's prediction could be influenced by e.g. the behaviour of the pupils in class - badly behaved pupils get more negative predictions than well behaved ones.
Certainly, this could explain the free-school meals bias. Many students from those families are socialized to a different set of standards from those the school expects, and so behave worse in class than their more prosperous classmates. It would have been nice if the study had corrected for the teacher's assessment of classroom behaviour.
Posted by: William | October 29, 2009 at 04:12 PM
"However, among black Caribbean pupils, 17.2% got worse teachers’ assessments and only 7.5% better."
It's worth spelling out that this means that 75 % of black Carribean pupils are correctly assessed by their teachers (which is surprisingly high to my mind, although I suppose tests have got much more standardised and it is easier to assess therefore).
This means that in a school year of, say 120 kids with (say) 40 each of black carribean, asian and white typoes, we might expect that 30 of the black kids will be correctly assessed, 7 will do better than expected and 3 will do worse. This compares with 31 of the white kids correctly assessed, 5 doing better than expected and 4 doing worse. You make the point that these are marginal effects but it is worth spelling out because it suggests that teacher racism cannot be very active, even if it is there at a deep level (probably down to differences in speech patteerns than differences in colour, I suspect: research has shown that 'speaking black' has this effect on prospective employers, while looking black doesn't). In fact, looked at that way teachers seem to be doing quite well in difficult circumstances.
By the way, isn't there a contradiction in your speculation that this research might indicate that lowered expectations migh lower acheivement, because if it it did, this research would show an achievement to expectation one-to-one match.
Posted by: John Meredith | October 29, 2009 at 04:47 PM
Agree with the comments above - but isn't it good that we have the external tests against which we can make these comparisons? Also, presumably, the teachers making the assessments got to compare their results with the objectively marked ones and maybe reflected on the differences and learned from them. Not like when I was a lad. The objectively, externally marked 11+ was, I suspect, a rubber-stamping of teacher assessment carried out much earlier. Anecdotal reasoning here http://a419.blogspot.com/2009/07/grammar-schools.html
(includes a hat-tip to Chris :) ).
Pete
Posted by: Pete B | October 29, 2009 at 08:06 PM
There is nothing necessarily irrational about expecting a whatever kid to do worse, IF in fact the whatever category tends to do worse. By comparing kids who score 4, you may be simply selecting the surprise outperformers in a given group.
The prior probability of a given score assigned by a teacher should take into account all available information, including group performance characteristics.
This would only be irrational if this group bias was overweighted relative to information about that individual's performance, and I think that this is what you and the paper are suggesting is happening here. It may be so, however the results looked at do not particularly support this contention.
In any class of two groups where one group markedly outperforms the other, the optimum predictive strategy for individual pupils will take into account group membership as an additional predictive factor, the weight of which will depend on the quantity and predictive value of information about the individuals class performance.
Of course this is literally prejudice, but then the teachers were asked to pre-judge the pupils.
Posted by: Andrew Sales | October 29, 2009 at 09:28 PM
@ John Meredith -
You cannot infer those proportions. These were pupils who scored 4. Not all black kids or white kids.
If one group does much worse than the others then only a few of the very best will score well (relative to the other group). These few will likely turn out to have had too pessimistic expectations by their teachers if the teachers were applying an optimal predictive strategy (i.e. using all the information) in the context of imperfect individual information.
In other words, amongst NBA basket ball players, the four foot high ones are going to have surprised a few of their teachers with their subsequent success. At least relative to the seven footers. This doesn't mean that the teachers were irrational to guess that the short ones wouldn't end up basketball pros, even given a bit of information about their individual performances.
Posted by: Andrew Sales | October 29, 2009 at 09:37 PM
@ John Meredith -
You cannot infer those proportions for black or white kids. The results are given for kids who scored 4, not all kids.
In other words if you look at NBA basketball players, you might find the four foot high ones surprised their teachers with their subsequent success. At least relative to their seven foot peers. This doesn't mean that the teachers were irrational in guessing that the four footers had a slim chance. This is true even though the teachers might have had some information about individual performance.
Posted by: Andrew Sales | October 29, 2009 at 09:40 PM
"could it be that the state - far from being a force for greater equality - is actually a force for inequality ?"
What's the state got do do with it ? Would you expect different results if all schools were private ?
Posted by: Laban | October 30, 2009 at 09:35 AM
It is so much easier to blame students deficiencies on the teacher's expectations (which simply requires brainwashing the teachers to fully embrace political correctness) then the amount of homework they do (which would require getting the little darings to actually do some work).
Posted by: Michael E Piston | October 30, 2009 at 03:06 PM