AN ERRORS-IN-VARIABLE MODEL OF THE EFFECT OF POVERTY ON URBAN HOMICIDE RATES

Inconsistent findings on the relationship between poverty and violent crime have led some authors to question the presence of a structural relationship. There is reason, however, to believe that many of the existing estimates are biased because measures of poverty contain errors which are confounded with disturbances in the estimated models. In this paper we specify and estimate a model which accommodates the measurement error and provides an instrumental variable estimate of the effects of poverty on homicide rates in the 49 largest cities in the U.S. Compared to similar OLS estimates, the instrumental variable estimates are much larger and fit a model in which poverty increases the homicide rate. The results are similar when homicides are divided into four types: family homicides, other primary homicides, robbery homicides, and other-felony homicides.