«Emergence of Endogenous Legal Institutions: Property Rights and Community Governance in the Italian Alps MARCO CASARI This article examines changes ...»
Casari Yet, it is well known from social choice theory that voting procedures are often characterized by instability and cycles among outcomes. How was this problem overcome in the case of the northern Italian communities? If we assume that there was homogeneity of interest among the villagers of a community, in the sense that either preferences were identical or highly correlated, then efficiency-enhancing policies should have had majority or supramajority support. For these reasons, the larger the group, the more difficult it would be to agree upon the provision of a charter.
Implication 4 (Population; Alternative to Implication 2) Communities with a large population are less likely to transition to a charter than small communities.
The communities that adopted a charter did so at different points in time, in some cases centuries apart (Figure 2). The patterns of geographical diffusion may be revealing about the underlying motive of adoption. We discuss two possible reasons, innovation by imitation and deterrence (Implication 5). A community of mostly illiterate peasants would find it hard to create a relatively sophisticated legal institution such as a charter from scratch, at least not without a pre-existing model.
Imitation of other communities seems more plausible than invention.
Imitation would be easier if the community was aware of the existence of this legal institution and, more importantly, there was social proximity with a working example of it.
An alternative reason for adopting a charter is as a defensive measure toward neighboring communities having adopted it first. A charter may have worked as a signaling device toward trespassers; when a community adopted it, it diverted violators from its resources to the resources of neighboring communities. Once the process of charter adoption got started, the other communities, especially the physical neighbors, felt an increasing pressure to adopt it as well.70 Implication 5 (Contagion) A community is more likely to adopt a charter when nearby communities have already adopted it.
information transmission. In the latter case, the increased surplus in the charter-establishing communities is, at least partially, offset by declines in surplus from increased outsider infringement in noncharter communities. In the latter case, it may also be that the aggregate costs of privateorder institutions are higher than the aggregate benefits. Once locked up into an all-charter situation, no community would have an incentive to switch back to informal cooperation.
the regressions (Table 2).71 The first charter was adopted in 1202 and the last one in 1795, right before Napoleon invaded northern Italy. By the end of the period, about 61 percent of the 231 communities coded in the dataset had a charter, which corresponds to 76 percent of the land.
Proxies for community remoteness are built using linear and altitude distances from major towns. Distances are measured in reference to 17 major towns that were headquarters of the decentralized government administration in the year 1810. In a mountain landscape, the length, as well as the steepness of a path could be related to how isolated a community is, although altitude and linear distances are correlated ( = 0.54). Community-level population data are from the 1810 census.72 A third dimension considered for remoteness is being on the border of the region analyzed.
Using the 1897 land survey data, one can cluster the land in a community into four categories. Category one includes vineyards, fruit gardens, and plow land (L1). Category two is meadows (L2), and category three comprises forests, alps, and grazing land (L3). Finally, category The dataset was built using both published and unpublished sources as detailed in the Appendix. The year of eventual adoption of a charter was recorded after accessing the original document, a transcription of it, or reliable news of the existence of the document. In 26 instances, the community charter regulated two or more villages. In that case the villages are consolidated into a single community. We believe that the reason for a joint charter was mostly to enjoy the protection of a stronger natural barrier such as a mountain range or a river. Moreover, seven communities were excluded because the charter date were unreliable. Trento was excluded as well because it was the major city in the region and had a unique peculiar statute.
Andreatta and Pace, Trentino. When break-downs at the village level were not available, we use the proportions from 1897 data (Consiglio provinciale d'agricoltura pel Tirolo, 1903).
Emergence 215 four is wasteland, lakes, and ponds (L4). We have chosen L1 as proxy for the land under private property and L3 as proxy for the land under common property. As mentioned, a quantitative analysis of the 1780 land registers of two communities suggests that L1 land was largely private property whereas L3 land was almost entirely common property.73 Meadows (L2) are excluded because of the high correlation between L3 and L2 (0.84) and because of the mixed property regime on this resource that emerges from land registers. To capture the possible nonlinear effect of L3, a dummy is created which is equal to one for communities with an endowment of L3 above the sample median. To control for fixed effects, 13 binary dummies were created for different areas of the region.
The logit regression presented in Table 1 estimates the likelihood that a community has already adopted a charter at any time before 1800.
Such likelihood significantly decreases with the remoteness of the community, measured as distance from the local town. This finding is in line with Implication 1. The other dimensions of remoteness, altitude difference from the administrative center and being at the regional border are not significant. Table 1 includes two specifications of the static model, (A) and (B), which yield similar results. Specification (B) drops some insignificant variables in order to reduce multicollinearity problems.
An important result is that the larger the community in terms of population, the more likely it is to adopt a charter. Implications 2 and 4 had opposite predictions regarding the impact of population. Implication 2 relied on the higher efficiency of the charter solution for larger communities and that is supported by the data. Implication 4 stressed the increasing difficulty to transition to a more efficient management regime as population size increases. The result suggests that if there were growing obstacles to the transition in larger communities, they were weaker in comparison to effect 2, at least in the long run. Finally, large amounts of commonly owned resources are positively correlated with the likelihood of adopting a charter. This finding is in line with Implication 3.
model. We use a discrete time version of the event history model. In particular, the time interval is five years; hence we have 120 intervals t between 1200 and 1800. The relevant observations for the analysis are those communities that have not yet adopted a charter at each point in time (risk set). A community that adopted a charter in a year that falls in time interval t contributes to the dataset with t observations. Those observations up to the time interval (t – 1) have a dependent variable set to zero; and the observation at the time interval t of adoption has a dependent variable set to one. For time intervals subsequent to t, no observations are included in the data set for that community. If a community never adopts a charter, it has 120 observations, one for each time interval. Otherwise, it has less.74 The actual dependent variable in the event history model is the hazard rate, P(t); given that a particular community has not yet adopted a charter (hence it is in the risk set), we model the probability that a charter will be adopted in the following time interval.
This estimation is carried out with the following logit regression model
where P(t) is the number of charters adopted in time interval t divided by the number of communities in the risk set in the same time interval t.
There are three classes of explanatory variables: Time trend, a(t), century fixed effects; Time-invariant variables, x1, remoteness proxies, private property size, common property size, and area fixed effects; and Time-varying variables, x2(t), community population, contagion variables, and specific historic event dummies.
In addition to the time-invariant variables x1 of the static model, the explanatory variables a(t) and x2 were added to the dynamic model.
Population data at the community level are not available from primary sources for each five-year interval. The 1810 data were taken as the main reference and have been scaled proportionally over time using the Italian population data from Athos Bellettini’s work.75 In some ways, it is unsatisfactory because it ignores internal migrations, differential growth within the region, and differences in population trends between This methodology can handle two issues present in the data set, censoring and time-varying explanatory variables (Allison, Event History). Censoring occurs because the period considered is finite and the event of a charter adoption does not occur for all the units. Time-varying explanatory variables such as if a neighboring community has already adopted a charter could be relevant before the community itself adopts the charter but not after that event.
Bellettini, La popolazione italiana.
Casari Italy and Trentino. A possible ambiguity in the interpretation of the results arises if charter adoption, because of its higher efficiency, had an impact on population size.76 To control for fixed effects, both area and century dummies were employed. As in the static model, 13 binary dummies were included for different areas of the region. In addition, five century dummies were created to control for variations in the trend of the baseline hazard function over time (no dummy for 1700–1800). In addition, dummies for important historical events are also included to control for population shocks such as that from the Black Death (1350–1400), or institutional shocks that might have changed the propensity of the Prince to grant charters, such as the Peasant War (1525–1535) or the Council of Trento (1545–1565); the Italian crisis of the first half of the seventeenth century (1600–1650) is probably a mixed case. As it turns out, none of these latter historical dummies show a significant impact.
Another class of regressors concern contagion effect, i.e., the impact of the previous charter adoption by another set of communities.77 Three different reference sets of communities are considered: the whole region, administrative neighbors, or geographical neighbors. Contagion proxy 1 records the number of charters already adopted in the whole region up to the previous time interval. Contagion proxy 2 is built in two steps. First, the region was divided into 89 nonoverlapping and exhaustive clusters of communities. This partition is taken from the 1800 administrative districting of Trentino. The regressor is a dummy variable indicating whether there is at least one community in the cluster that has adopted a charter at or before time interval (t – 1). Contagion proxy 3 considers the set of communities that shares borderlines with the community itself. Once the set of neighbors is identified, we build a dummy variable indicating whether at least one neighbor has adopted a charter in the previous time interval. This last variable captures the impact of a charter as a signaling device. The three proxies are highly correlated.78 Table 3 presents four different specifications of the event history model. The first column is the most general, whereas the other three are If population is assumed to be endogenous then both static and dynamic regressions are biased toward finding a positive coefficient for “population.” On the positive side the proxy built for the dynamic model eliminates most of the endogeneity in the timing of population change.
Even with exact five-year interval data on village populations, the interpretation of the causality in the charter-population relationship would be ambiguous.
Consider a model of contagion, where the dynamic is governed by a logistic function:
Dx(t) / dt = rx(t)((1 – x(t)) / K), where r, K 0 and x(t) is the number of communities who have already adopted a charter. The cumulative number of charters, x(t), is an S-shaped function of time. The regressor is the number of communities with a charter in the previous time interval as the variable to be explained is P(t) = (dx / dt) / (1 – x) = rx / K.
= 0.47 for (1) with (3); = 0.63 for (1) with (2); = 0.45 for (3) with (2).
Emergence 219 limited to significant regressors and just one contagion proxy to avoid multicollinearity problems. All the results from the dynamic model confirm the findings of the static model with regard to remoteness, population, and common property size.
Although the signs are correct for all contagion proxy coefficients, the only significant impact comes from the regional proxy 1. Even when proxy 2 or proxy 3 is the only contagion proxy in the model, it is not significant at a 10 percent level. In other words the existence of other communities with a charter significantly raised the likelihood that a community would adopt a charter itself. The influence was not local but there was a general adoption trend in the region. Neither the adoption by physical neighbors (proxy 3) nor by the immediate administrative group around the community had a significant impact (proxy 2). These results have important consequence for evaluating the motivation behind a charter adoption. Setting up a new charter was not an effort to offset the better protection strategy of immediate neighbors; the interpretation of charter adoption as a zero-sum game is not corroborated by the data. The significance of the general proxy 3 is ambiguous. The variable could capture general factors missing from the model such as commodity prices that influenced charter adoption by all communities.
Still, it is compatible with an interpretation of the charters as legal innovation that spreads the more rapidly the lower the costs of information about its content and procedure of adoption.
DISCUSSION AND CONCLUSIONS