Sociology 498: Homelessness

Burt: Over the Edge

Overview

Martha Burt's book, her second on homelessness, includes one of the best collections of cross-city statistics on homelessness. Burt estimated homelessness rates in 147 cities as of 1989. Her data are much better than the HUD estimates that Honig and Filer used. She then collected other published data on these cities (most supplied by the Census) and asked what other characteristics of these cities were correlated with high homelessness rates. Her list of other characteristics included many of the usual suspects.

I have reproduced a table of correlations with homelessness. The table is adapted from her table 8.8 (pages 157-159). She had even more variables, many of which overlap, and so I have selected only some to simplify the table.

Units of analysis

Burt uses cities as her units to compare (as do Honig and Filer. Her criterion was all cities over 100,000 in population that were not suburbs of other large metropolitan cities. There were 147 such cities in 1986.

Measuring homelessness

Burt measures homelessness rates using a count of shelter beds in each city. For practical reasons, no street homeless are included in her measure. Her measure has some obvious weaknesses:

• If some cities have higher proportions of their homeless in shelters, they will score lower on her measure than on their true homelessness.
• If some cities have made a better effort at constructing sufficient shelters for their homeless, they will appear to have higher homelessness on the Burt measure than is their true ranking. Thus, any aspect of the city that would be correlated with shelter effort might have a correlation with Burt's measure, but not with the true homelessness rate.
• If shelter beds are not being used, Burt's measure will be an overcount of the homeless.

Despite these flaws, Burt's measure is useful. It was the only practical way to actually compare so many cities; trying to get even partially valid counts of the street homeless in so many cities would have been far too costly. We can have some confidence in her data because they seem to rank cities in much the same way as the subsequent counts made in 1990 by the U.S. Census (see the scatter plot of big cities). Both counts are imperfect. But if two imperfect counts show reasonably similar results, we can have more confidence that they are measuring something reliable about the cities. And cities vary so greatly in their homelessness rates, that even imperfect measures can distinguish between cities with high homelessness (e.g., Washington) and cities with low homelessness (e.g., Milwaukee).

Measuring city characteristics

Burt correlated her city measures of homelessness rates with a wide range of other characteristics of those cities. Measures of those characteristics came from a wide variety of sources such as the U.S. Census, HUD, and HHS. I have reproduced some of those correlations in a table that deserves close inspection. Some major hypotheses about the causes of homelessness are supported in this table; others are not.

Think about:

• What concept (i.e., our six causal theories) does each measure reflect?
• How well does each measure capture that concept?
• Are there other aspects of each of the seven causes that aren't captured by her measures (and are there any of the seven not addressed at all)?

see also

• the list of homelessness rates
• the plot of city homelessness rates as measured by Burt's 1989 study and the 1990 Census
• the plot of city homelessness rates by poverty rates
• the plot of city homelessness rates by vacancy rates
• the plot of city homelessness rates by % single