RT Article
T1 Police, crime and the problem of weak instruments: revisiting the ‘‘More Police, Less Crime’’ thesis
JF Journal of quantitative criminology
VO 32
IS 1
SP 133
OP 158
A1 Kovandzic, Tomislav
LA English
YR 2016
UL https://krimdok.uni-tuebingen.de/Record/1559166959
AB Objectives A key question in the general deterrence literature has been the extent to which the police reduce crime. Definitive answers to this statement, however, are difficult to come by because while more police may reduce crime, higher crime rates may also increase police levels, by triggering the hiring of more police. One way to help overcome this problem is through the use of instrumental variables (IV). Levitt, for example, has employed instrumental variables regression procedures, using mayoral and gubernatorial election cycles and firefighter hiring as instruments for police strength, to address the potential endogeneity of police levels in structural equations of crime due to simultaneity bias. Methods We assess the validity and reliability of the instruments used by Levitt for police hiring using recently-developed specification tests for instruments. We apply these tests to both Levitt’s original panel dataset of 59 US cities covering the period 1970–1992 and an extended version of the panel with data through 2008. Results Results indicate that election cycles and firefighter hiring are “weak instruments”—weak predictors of police growth that, if used as instruments in an IV estimation, are prone to result in an unreliable estimate of the impact of police levels on crime. Conclusions Levitt’s preferred instruments for police levels—mayoral and gubernatorial election cycles and firefighter hiring—are weak instruments by current econometric standards and thus cannot be used to address the potential endogeneity of police in crime equations.
K1 instrumental variables
K1 Endogeneity
K1 Crime
K1 Police
K1 Instrumental variables
DO 10.1007/s10940-015-9257-6