A British polling outfit surveyed Iraqis about deaths in their household and came up with a tally of
1.2 million dead Iraqis as the central point of their estimate (obviously, with a big number like that, the confidence interval includes a wide range). Kevin Drum
points out skeptically that this survey seems to indicate that car bombings are being massively underestimated in Iraq -- out of 20 or so a day, only 2-3 are getting reported. I agree with him that this seems wrong, but it also seems implausible to me that families would be massively overreporting deaths.
This article available online at:
http://www.theatlantic.com/politics/archive/2007/09/surveying-iraqs-mortality/46253/