Beware of Amazing Statistics

August 05, 2008

Sometimes a new statistic will come out that shows amazing progress or decline in a certain area. These numbers may seem too good to be true, which often is a sign that there is some "fuzzy math" going on. For instance, numbers recently came out claiming that homelessness dropped 30 percent in just two years (from 2005 to 2007). Some have pointed to the finding as proof that the Bush administration's policies are working. Richard Land, head of the Southern Baptist Convention's Ethics & Religious Liberty Commission, heralded the numbers and used them to praise the Bush administration and attack the media for not reporting it more.

The number, however, sounded too amazing to be true. If there was such a large and quick decline in homelessness, would not more people have noticed even before the report was released? As it turns out, the way homeless people were counted changed. Thus, some and maybe even most of the change could be because of a different statistical approach. In the 2005 count, researchers walked around talking to homeless people, but were also allowed to made counts of other apparently homeless people they saw. In 2007, however, the researchers were not allowed to count someone unless they talked with that person. Thus, if a researcher saw three homeless people hanging out together under a bridge and one of them refused to talk to the researcher, then only two were included in the 2007 number but all three were counted in 2005. Since the method of counting was so different, the numbers cannot actually be compared. It may be that the Bush administration's policies have reduced homelessness (although probably not by 30 percent), but that cannot be proven since the methodology was changed. Thus, it is not wise to trumpet such numbers or claim media bias against those who ignored the useless statistical comparison.