NAMs and the Flynn effect: In the last post on the Flynn effect, I used GSS data to show that the mean Wordsum scores of people born in the United States of Mexican and Puerto Rican descent have risen over the past two decades. I list below the mean scores for a larger number of ethnic groups over the past four decades.
Mean Wordsum scores
This decade 5.53*
This decade 5.38*
This decade 6.95
This decade 6.40
This decade 6.59
This decade 6.47
This decade 6.23
This decade 7.03
This decade 7.47
* Significant change between first and last period, .05 level, two-tail test.
The American Indian sample size is very low after you remove blacks and whites who say their main ethnicity is Native American, so I had to limit the comparison to two periods. The difference between the two would very likely be statistically significant if the sample were larger.
There is a clear pattern here: none of the white groups have seen significant improvements since the early 1970s when the vocabulary test was first given, but all of the non-white groups have improved. I suggested before that fewer NAMs have good English language models, but the increased hours of television viewing might have helped compensate in the last couple of decades. Kids watch several hours of TV each day, and they often watch peers that they want to imitate. Characters on television usually speak well. (Of course, it doesn't help if they are speaking another language well). White kids often have good models anyway, so people on television with good vocabularies are redundant.
Anyway, that is one explanation, but others ideas to account for the pattern are appreciated.
(TV can't be THAT powerful, or generations wouldn't continue to have strong accents.)