Abstract
My main objective in this paper is to evaluate how poverty was related to property crime and violent crime respectively in the Kingdom of Bavaria in the period 1835/36 – 60/61.
My point of departure is the book Statistik der Gerichtlichen Polizei im Königreiche Bayern und in einigen anderen Ländern, written by the German statistician Georg Mayr (1841 – 1925) and published in 1867. Mayr was employed at the Bavarian Statistical Bureau when the book was published. He was appointed Director of the Bureau two years later.
Mayr paid attention to the seven administrative regions Upper Bavaria, Lower Bavaria, Upper Palatinate, Lower Franconia, Middle Franconia, Upper Franconia and Swabia during the period 1835/36 – 60/61. I also limit my paper to the same seven administrative regions as Mayr did in his book.
What makes the analysis by Mayr particularly valuable is that he divided crime into different categories and looked at how poverty, then measured in terms of the rye prices, affected different types of crime. His main empirical findings were that higher rye prices tended to lead to more property crime (where theft was the dominant subcategory), but less violent crime (mainly consisting of injuries and rape).
I further consider the article Poverty and crime in 19th century Germany written by Mehlum, Miguel and Torvik and published in the Journal of Urban Economics in 2006. The authors use much of the same data as Mayr did in his book, but they make a twist in the analysis, using rainfall as an instrumental variable for the rye prices. They also give a different explanation of why higher rye prices tended to give less violent crime in the Kingdom of Bavaria at that time, namely that higher rye prices yielded higher beer prices and thereby less alcohol consumption, which again gave less violence. Mayr on the other hand simply argued that the different types of crime had different motives; whereas simple theft was driven by distress, the violent crime was a result of crudity and passion.
Although the beer prices may very well be important to correct for in the regression analysis, one cannot claim that higher rye prices led to higher beer prices for the period in question, due to the Reinheitsgebot, a food quality regulation from April 23rd 1516. This Bavarian Purity Law said that the only ingredients allowed to use in beer were barley, hop, and water.
After having found supplementary data, I calculate that the correlation coefficient for the barley and beer prices for the period 1835/36 – 60/61 is + 0.69 and thereby somewhat higher than the correlation coefficient for the rye and beer prices, which is + 0.64. That the former is larger than the latter is as expected, since beer was made with barley and not with rye. Interestingly, the correlation coefficient between the rye and barley prices is as high as + 0.95. So even if it was the barley prices that caused the higher beer prices and not the rye prices, at a first glance it still seems possible (since the rye prices and the barley prices were so highly correlated) that the reason why the rye prices and the violent crime were negatively related to each other was that beer was now more expensive, leading to lower alcohol consumption, and therefore people were less violent. I then exchange the rye prices with the beer prices in the data set which Mehlum et al use and I find that the variable for the beer prices is statistically significant in the regression for violent crime, even at the 1 % level. A 1 % change in the beer prices is associated with a - 0.9 % change in violent crime, so this effect is significant in every meaning of the word.
I further look into why the rye prices were so high in 1846/47 and 1853/54. When considering the workings of poverty in these years, one should both look at what caused the poverty and what possibilities the inhabitants had to escape the situation. Friedrich B.W Hermann, of whom Mayr was both a student and a successor to the seat at the Bavarian Statistical Bureau, wrote that after a period with constant grain prices, the reason why the prices fluctuated was the quality of the harvest; bad harvests led to high prices and good harvests to low prices, where less than average was produced in the former case and more than average in the latter case. The book by Hermann has been informative as to why the Kingdom of Bavaria experienced the years of high prices, but it has also shown that research on agricultural production was scarce in this period, and one should therefore be very careful with the utilization of these data. The data at hand are mainly average values and are also to some degree based on guesses.
I create a panel data set in an Excel-sheet and run two main regressions in STATA, one with violent crime and one with property crime as the dependent variable. The panel data set further contains variables on the rye prices, beggary, vagrancy, mortality, fertility, emigration and the adult male to female ratio. It differs from the data set created by Mehlum et al both by including different explanatory variables and also by including the rye prices for each region rather than using one single Bavarian rye price series. I expect the fluctuations for the rye prices, beggary, and vagrancy to be larger within the administrative regions than between them. The within estimator (the fixed effects estimator) should therefore be used. By plotting the dependent variable against the different explanatory variables separately, I also find a linear specification to be the most appropriate. I run two supplementary regressions in STATA, where I first consider beggary and then vagrancy to be the dependent variable. For all four equations I correct for heteroskedasticity and cluster by years, which Mehlum et al also correct for in their regression analysis. Clustering by years corrects for the possibility that the error terms are correlated between the regions within the same year. The main reason for expecting this is that the omitted variables may work in the same direction in all the seven regions.
In the regression where I consider property crime as the dependent variable, I find that the variable for the rye prices has a moderate positive effect, which is statistically significant even at the 1 % level. The elasticity of property crime with respect to the rye prices is 0.24, so that an increase in the rye prices of 1 % leads to an increase in property crime of approximately 0.24 %. Further, when the rye prices increase with a one standard deviation, there is an increase in property crime of an approximate 0.48 standard deviation. Mayr argued that theft was more common when it became more difficult to obtain food legally and that one naturally could conclude that property crime would increase with food scarcity. That these tendencies really were seen in Bavaria at the time is supported by my own empirical findings.
The variable for the rye prices further has a moderate negative effect on violent crime, also statistically significant at the 1 % level. An increase in the rye prices of 1 % leads to an increase in violent crime of approximately 0.20 %. When the rye prices increase with a one standard deviation, there is an increase in violent crime of an approximate 0.2 standard deviation. I will stress two possible explanations (of course there could be many more) of why there was a tendency for poverty to affect violent crime negatively in Bavaria in the period 1835/36 – 60/61. The first possible explanation is that in poor times the Bavarians were too exhausted by and too preoccupied with covering their most basic needs such as food and shelter to commit any violent crime. This is also how I believe Mayr (his explanation was somewhat unclear) saw the link between poverty and violent crime. The second explanation is the one which Mehlum et al present in their article. Since I find that the variable for the rye prices is statistically significant even at the 1 % level (and since the rye prices are so highly correlated with the barley prices and since barley was an important ingredient in the beer) I do not reject their way of reasoning, but I leave it out of the rest of my analysis.
The variable for the rye prices is statistically significant both in the regression where I consider beggary and in the regression where I consider vagrancy as the dependent variable, both times even at the 1 % level. Mayr assumed that the rye prices would show an even more intensive effect on these variables. By looking at the elasticities, this is supported by my empirical findings; an increase in the rye prices of 1 % leads to an increase in beggary of approximately 0.56 % and to an increase in vagrancy of approximately 0.51 %. I might have an endogeneity problem in both these regressions, though, due to the poorer health of beggars and vagrants, which might affect both fertility and mortality. Then I must exclude these variables, and since I already seem to have a severe omitted variable bias, not much explanatory power is left.
One has to be aware of the danger of the so-called “ecological fallacy” when interpreting the data; a logical fallacy inherent in making causal inference from group data to individual behaviors. This is exactly what I find to be the main problem with Mayr’s way of reasoning, although he certainly did point out that the statistician could only look into possible causal chains as long as these could be supported scientifically by connecting data on crime with data on factors that influence crime.
We may speak of tendencies on the aggregate level and then say what we find to be a possible relationship on the individual level. This is exactly what I have done in this paper. There are tendencies in the data, showing that in poor years, there was on the one hand less violent crime and on the other hand more property crime and more beggary/vagrancy in the Kingdom of Bavaria during the period 1835/36 – 60/61.