The level of Aversion to Breaking Rules (ABR) is heterogeneous across nearby localities in many areas of the world and geographic sorting based on ABR may be a reason. In this paper we use an indicator of cheating that can be observed for migrants and remainers at the city/time level in Italy, to measure sorting based on ABR between the North and the South of the country and to assess its consequences for the local development of the affected areas.

Using Italian Census restricted data, we show that for Italians born between 1920 and 1970 there is evidence of a significant amount of cheating in the regis-tration of birth dates. Starting in early December of each year, the frequency of registered births per day declines substantially, while an abnormally large mass of registered birth is concentrated in the first five days of the following month of January. We also observe a significantly lower number of births on the 17th of each month.

We discuss possible motives for this tendency to lie in the registration of birthdates (e.g., delaying compulsory military service in the case of “January cheating” and superstition in the case of “17 cheating”), and we show that, irrespective of the motive for lying, a measure of this tendency has two major advantages for the analysis of sorting based on ABR.

1. It can be computed for small groups in the population at different points in time (generations during the 20th century) and specifically for migrants out of a given locality and remainers in the same locality.
2. It correlates closely across localities with other measures of cheating (like cheating on school test scores, ghost building, absenteeism), which instead cannot be typically observed for migrants and remainers within the same locality at different points in time.

Using a simple theoretical model, we define an agent as Propense to Breaking Rules (PBR) if she decides to cheat when the benefit of cheating is higher than deterrence. Vice versa, an (intrinsically) ABR agent is someone who never cheats whatever the levels of benefit and deterrence. Our model predicts that if two groups of agents have the same distribution of cheating benefits and face the same level of deterrence, the fraction of dishonest agents is higher in the group where more cheating is observed and where a change of deterrence (in whatever direction) induces a larger absolute change in observed cheating. Our identifying assumption in the empirical analysis is that within the same narrowly defined locality, the distribution of benefits and the level of deterrence are similar for migrants and remainers. In light of the model, using census data for the entire Italian population we first show that within narrowly defined localities, migrants from South to North are less likely to cheat on their birthdate than remainers in the South, while the opposite is observed for migrants from North to South versus remainers in the North.

We then exploit a pervasive institutional reform implemented by Fascism in 1926 to study how cheating on the date of birth reacts to changes in deterrence. There are three reasons why this reform is relevant in our context:

a. It created a new local administrative authority (the “Podest`a”) to increase control of the central state on the daily life of citizens.
b. It introduced measures to curb infant mortality, which included, crucially for our purpose, a registry of infant deaths occurring within 1 or 6 days from birth and an obstetrician service to help mothers giving birth at home.
c. It introduced monetary incentives to fertility and taxes on celibacy.

As a result of these changes, January birth date cheating suddenly declines in 1926, particularly in the South of Italy, to then go back to pre-1926 levels in 1942, when Fascism collapses. More importantly for our purposes, the reactions of January cheating to these changes in deterrence were smaller for migrants out of the South than for remainers in the South. We therefore conclude that Italy experienced sorting based on ABR between the North and the South and that the South suffered a drain of intrinsic ABR because of the internal migration movements of the 20th century.

Finally, we measure the ABR drain experienced by each locality, taking the difference between the intensity of January cheating for the population remain-ing in that locality in the 1991 census and the population born in that locality (which include remainers and people who emigrated out of the locality). We then show that localities experiencing a greater ABR drain, while controlling for the initial level of ABR, display lower productivity in recent years using the Ilzetsky and Simonelli (2019) measure of local vote counting productivity in national elections, a measure of individual productivity which holds technol-ogy and task constant. In a community affected by stronger ABR drain, vote counting is likely to take more time because of lack of trust between members of the team that count votes at each polling station.

In line with this hypothesis we find that our measure of ABR drain is sig-nificantly correlated with a lower speed of vote counting across localities. We also focus on a measure of value added per worker at firm level using Bureau van Dijk data. In local labor market affected by one standard deviation more ABR drain, firms have on average 0.5 percent lower valued added.