Americans are taught to judge people individually, and in situations where you can get to know someone, this makes sense. But the truth is that you can predict aggregates better than individuals.
Using General Social Survey data, I calculated the correlation between a measure of IQ (WORDSUM) and income (REALINC) for almost 17,000 people born in the U.S. It turned out to be .28. If you square that number, you get .08 which is called R-squared. It is interpreted as the proportion of variation in income that can be explained by your IQ. In other words, if I know one thing about a person--his IQ square--I am not going to be able to predict his income level with any accuracy at all.
But the situation changes dramatically if I calculate mean IQs and mean incomes for the 29 ethnic groups which have at least 30 respondents in each group. Now the correlation jumps all the way up to .77. If we square that, we get .59, which means that 59 percent of the variation in mean income is explained by the variation in mean IQ scores. So if I've got a random group of, say, Americans of Russian descent, chances are their average IQ is high, and I can make a pretty good bet that the group will earn an above-average income as well.
This is why Richard Lynn and Tatu Vanhanen's approach is so effective in IQ and the Wealth of Nations. You grab a random guy in Japan, he might be smart or dumb; he might be rich or poor. But tell me the mean IQ of the country is 106, and I'm putting my money on it being a wealthy place.
HBD-ers are criticized for focusing on groups, but reality is most predictable at that level, and being scientific is being concerned with prediction.