ABSTRACT: In the 1980s several estimates were made of the size of the homeless population nationwide. An examination of these estimates shows that claims by advocacy groups that 2–3 million persons were homeless on any night were unjustified. Instead, social scientific studies agree that the number is probably about 300,000–500,000. This discrepancy tests the ability of the media to distinguish between good social science research and mere guesstimates. An examination of US magazines and newspapers found that, as a group, the media were more likely to cite high estimates than low estimates and this changed only slightly over time. Two factors seem to explain why journalists cannot distinguish between good and bad estimates: media bias and the process whereby information is gathered.