Abstract
Making use of data from the Brazilian Census, this paper presents descriptive evidence on Brazilian slums while attempting to test the squeezing hypothesis from the squatting literature. A comparison of mean values for a host of household and neighborhood variables often shows wide, and usually predictable, differences in values between neighborhoods designated as slums by the Census and nonslum neighborhoods that lack this designation. The paper presents a variety of descriptive regressions making use of these rich data, while providing evidence consistent with the squeezing hypothesis in city-level regressions showing that a higher population share or land share in slums leads to higher formal rents.