A paper by A. Terracciano and a raft of coauthors in Science has this abstract:
Most people hold beliefs about personality characteristics typical of members of their own and others' cultures. These perceptions of national character may be generalizations from personal experience, stereotypes with a "kernel of truth," or inaccurate stereotypes. We obtained national character ratings of 3989 people from 49 cultures and compared them with the average personality scores of culture members assessed by observer ratings and self-reports. National character ratings were reliable but did not converge with assessed traits. Perceptions of national character thus appear to be unfounded stereotypes that may serve the function of maintaining a national identity.
Upon reading the introduction, the point of the paper actually seems to be very minor. They cite a number of earlier works to the effect that human populations actually do differ in the frequencies of different personality types (e.g. extrovert/introvert, etc.). And they claim that the mix of personality types may correlate with "culture-level variables such as individualism-collectivism".
So what they have examined is whether the actual mix of personality traits in a country matches the stereotypes people hold about their own country itself. If these do in fact match, then the "aggregate personality data" from a population ought to indicate the same values as reported by people from the population themselves. The study claims that this correspondence is true for gender stereotypes:
For example, gender stereotypes depicting women as warm and men as assertive are widely held around the world (15). Crosscultural studies using both self-reports and observer ratings have shown that women in fact score higher on measures of warmth, whereas men score higher on measures of assertiveness (10, 16). Assessed gender differences are small but are largely consistent with gender stereotypes (17, 18), so those views appear to have a basis in the characteristics of individuals.
In contrast, the study found that stereotypes that people hold about their "national character" don't correspond to the actual mix of personalities in their populations.
Perceptions of national character are not generalizations about personality traits based on accumulated observations of the people with whom one lives; instead, they appear to be social constructions that may serve different functions altogether. ... Whatever their origins, stereotypes may be perpetuated by information-processing biases in attention/perception, encoding, and integration of information (2, 30). They become cultural phenomena, transmitted through media, hearsay, education, history, and jokes.
It's hard to understand exactly what this means without looking at the data. For example, the data indicate that Canadians rank their national character as relatively nonassertive, while Americans rank our character as fairly highly assertive. But in reality, both countries' personality "aggregate" scores are middling for assertiveness. In fact, what I get from looking at these values is that countries tend to have key traits that they rank highly as part of their character, but the personalities in these countries tend to be much less extreme. So Americans appear to think other Americans are not very trustworthy or straightforward (except, interestingly in the four-state sample, in Illinois) but we actually are. Canadians tend to think their compatriots are not anxious or hostile, but they actually are.
Come to think of it, this study appears a ripe source for its own stereotypes. But the comparison of Canada and the U.S. is interesting, because it shows a clear similiarity between Canadians and Americans in their observer-ranked personality aggregates, but a huge difference in the way people in the two countries perceive their compatriots. Even so, the result is not that people are all alike; it is that people define their cultures for cultural reasons rather than purely psychological observations.
Terracciano A. et al. 2005. National character does not reflect mean personality trait levels in 49 cultures. Science 310:96-100. Full text (subscription)