If these data were gathered through government records, the figures would have to be weighted to account for guns the government didn't know about, from old hand-me-down shotguns to smuggled fully autos. Even in New Jersey, where I had purchased 8 handguns over the years (and had to obtain a separate permit to purchase each one), my local police department had records for only 2 of them, and I had (through a dealer) legally sold one of those to a friend in Pennsylvania. So available information would be inaccurate even from New Jersey, a state that is obsessive about gun paperwork.
Usually information of this kind is collected through surveys. Of course, if the survey is to be truly scientific, it has to be conducted carefully, and demographers, statisticians, and other experts have to weight the results intelligently. But it's hard enough to get accurate information on what people eat and drink, or how they manage their household budgets, and researchers have studied those areas for many years. Trying to get a true picture of gun ownership would, I think, be extremely difficult.
And trying to measure the ownership of illegal firearms (from unregistered and thus technically illegal guns in Illinois to sawed-off shotguns regularly used in Bronx robberies to the mountain of illegal guns in the possession of organized crime) would be like trying to gauge the underground (cash) economy: it's vast, and nobody really knows its extent.
