It is widely accepted that maintaining a low crime rate is one of the government’s major responsibilities in order to ensure the attractiveness of cities. Furthermore, the cost of crime is considered to be quickly capitalized into housing markets serving as an early warning sign of neighbourhood transition. Various studies apply hedonic modelling to investigate to what extend different kinds of crime are capitalized into housing markets. This study follows a similar approach and applies a time fixed effect model on a small-scale panel data set of Hamburg city quarters.

During the past decade or so some new aspects have been added to this field of research as most hedonic models treat crime as an exogeneous variable. However, the literature has stressed the fact that crime should be considered rather as an endogenous variable. Relative to the total number of empirical studies on crime, only a few treat crime as endogenous applying valid instrumental variable approaches. Additionally, most studies are based on US data, manily due to data availability. Other studies that use data outside the US use rather large-scale data on city, state or country level.

This data set is collected from various public sources and applies familiar panel data models including an instrumental variable approach to account for the endogeneity of crime on a non-US data set. The land values of Hamburg are based on the sales-price collection of the governmental authorities and cover the entire city on a parcel level. The data is available for 6 consecutive years in total and can be filtered among others per dominant usage type of each parcel. The crime data is collected on a quarter level for the same periods and contains information about the crime level, the clearance rate and 9 different types of crime. Additionally, several socio-economic control variables on the city quarter level can be included.

The literature shows that in general changes of crime generate better results than levels of crime. Hence, we investigate a time fixed effect model regressing the changes of crime on the changes of land values for the respective periods. As instrumental variables, we use the density of public spaces and lags of the explanatory variable to account for the endogeneity. The preliminary results show a significant negative effect of crime on land values.