View by Day
16:30 | Applied Micro Research Seminar
Max Planck Institute for Research on Collective Goods, Bonn, Germany
Authors:
Christian Traxler and Carsten Burhop
Extended Abstract:
Our study provides historical evidence on crime in Prussia. We collected a new data set
of convictions for more than 20 different types of crimes committed in the 37 Prussian
districts (Regierungsbezirke) between 1882 and 1912. In addition, we compiled data on beer
production from the official tax records (Brausteuer) for 14 Prussian provinces, an
administrative unit comprising several districts. At that time, transportation costs for beer
were substantial and market integration was weak. Beer brewers were mainly competing for
local markets. Hence, province level beer production serves as a good proxy for beer
consumption in a province.
A descriptive analysis of the crime data reveals that violent and property crime rates
have increased between the 1880s and the turn of the century. Subsequently, property crime
rates displayed a further increase whereas violent crime rates stagnated at a high level. The
increase in crime is mirrored in the beer data: per capital beer production, which accounted
for a major part of total alcohol consumption at that time, increased dramatically between
the 1880s and the turn of the century, modestly declining thereafter (in part, due to tax
reforms). The beer boom in the late 19th century was mainly due to the dispersion of new
technologies (cooling machines) and the success of bottom-fermented beers (Pils).
Within-differentiated, de-trended panel data indicate a strong correlation between beer
and crime rates suggesting a close association between the two. However, it would be naive
to make causal interpretations of this correlation. As is well understood in the literature, the
alcohol-crime link is hard to identify due to reversed causality, omitted variable problems
and measurement issues. We solve the identification problem by employing a novel
instrumental variables (IV) approach. Our IV strategy rests on the fact that beer production
in the 19th century was sensitive to grain inputs, which were in turn affected by weather
shocks. More specifically, we use one year lagged data on (regional) rainfall and temperature
as instruments. We first demonstrate that ‘bad’ weather during the past years sowing and
harvesting period for spring barley had a negative impact on barley yield, barley prices and –
ultimately – present year beer production.
Using exogenous variation in beer production induced by lagged weather conditions (and,
as an alternative IV, lagged barley yield) we estimate the impact of beer on different crime
rates. Our estimates suggest that a one percent increase of beer consumption leads to a 0.7
percent increase in violent crimes. Qualitatively and quantitatively, this result is very stable
with respect to different specifications and alternative IVs. The effect is mainly drive by the
strong effect of beer on simple and aggravated assault and battery. More severe violent
crimes (e.g., robbery and homicide), however, are not affected by beer consumption. Turning
to property crimes, we do not find any effect from beer consumption. However, once we
consider crime rates among specific parts of the population, we do find a significant effect of
beer consumption on property crimes – mainly petty larceny and theft – among young
criminals.
Our data allow us to run several robustness checks and plausibility tests. Making use of
other crime categories, for instance, we show that beer has a significant positive effect on
vandalism and property damage. In contrast, there is no effect on arson, bribery, infanticide,
etc. Moreover, we demonstrate that the overall effect of beer is driven by male criminals (and
beer drinkers).
The project makes two contributions to the literature: First, we present the first study
investigating the causal effect of beer consumption on crime using historical data. We show
that – in contrast to writings of 19th century scholars – beer consumption was not causal for
all types of crime, but only for some violent crimes. Our historical evidence might also be
informative to the returning debate on whether the prohibition in the US could have had any
beneficial effect on crime. Moreover, the evidence on the alcohol-crime link mainly comes
from studies on modern, industrialized countries. Our historical data might be more
informative to evaluate the crime impact of drinking in developing countries, which
nowadays host the most rapidly growing alcohol markets. It seems to be a relevant question
to analyze the relation between alcohol consumption and crime in these fast-changing
societies that transformed, as Prussia, from mainly agrarian into mainly industrial countries.
Second, we introduce a novel identification in the alcohol-crime literature. The few
studies that tackle the identification problem mainly used changes in alcohol taxation or
temporal and legal sales restrictions to break the simultaneity, we work with exogenous
weather changes which - in our case - affect input prices and alcohol output. The use of truly
exogenous changes in weather seems a promising new IV that might also be applicable in
studies using modern data.
