In the past three decades, the meticulous approach adopted by historical seismologists has led to a remarkable improvement in the quality of the investigation, and it has enabled acquisition of information on the effects of earthquakes, often with a surprising amount of detail Stucchi The material available through these sources is particularly rich in the case of Italy Boschi The historical records for the period studied here refer to universal chronicles, monastic annals, ecclesiastical and liturgical sources, ancient literary sources, and coeval historiography Guidoboni In the period — and geographical area northern central Italy considered in this article, 28 earthquakes occurred; they hit cities included in our sample times obviously, a city can be struck by more than one seismic episode.
See a Problem?
They are reported in Table II , which indicates, for each earthquake, the year and the name of the city at the epicenter or the city nearest to the epicenter. For instance, in , two earthquakes occurred. The first had its epicenter in Arezzo and was registered in two cities included in our sample, Arezzo and Pistoia; hence, the number of cities in the corresponding cell is two although more than two cities among those not included in our sample were also hit.
The second had its epicenter in Cassino and was also registered by two cities, Cassino and Rome, neither of which is included in our sample; this is why the corresponding number of cities in Table II is equal to zero. The list shows all the earthquakes included in our data. Epicenter city is the city closest to the epicenter of the earthquake according to the geographical coordinates provided by the DBMI Year is the year in which the earthquake occurred. Registered quakes are the seismic events registered for our sample cities in the DBMI For each of the three augmenting criteria polygon , epicenter , circles ; see later discussion , the number of cities in our sample hit by a seismic event All and the number of cities hit before transition Identifying are reported.
Note that the numbers of registered earthquakes reported in column 1 may be larger than those in the next columns because the former include all earthquakes with D or F and without NR registered intensity in the original data set, whereas the latter only include earthquakes for which an intensity could be assigned, as explained in the text with greater detail. For a number of seismic events, intensity could not be registered in the original data set DBMI04 because of missing or inaccurate historical sources, these earthquakes are denoted by NR.
A possible concern about the data is that not necessarily all the seismic events that occurred in the Middle Ages could be originally recorded because of missing or inaccurate historical sources. Consider an earthquake with its epicenter in city j that occurred at time t. It may be that this earthquake was recorded in the DBMI04 for city j , but not for city i , which was also struck by the seismic episode, because the historical sources for that city were not handed down to the most recent periods. To take this possibility into account, we build an augmented data set according to the three following criteria.
For all the cities that in the sample period reported a seismic event with a valid registered intensity falling in either the D or F category in the DBMI04, we draw the outer convex polygon and impute an earthquake as occurring in city i at time t if this city was located within the polygon. The empty dots represent the cities hit by the earthquake in the DBMI Varese, Treviso, Venezia, Cesena, Genova, and Alessandria represent the vertexes of the outer polygon.
The full dots denote the cities that were assigned the earthquake because they were located within the polygon. The epicenter, Lanzise, is denoted by a star. After we augment the data set by means of this criterion, the total number of seismic episodes increases to of which hit cities before their transition to communes, if any, see Table II. Geographic distribution of the cities hit by the earthquake of Empty dots represent the cities that reported the earthquake in the DBMI04; full dots denote the cities that were assigned the earthquake by the polygon, epicenter , or circles augmenting criterion, respectively, in Panels A, B, and C.
The epicenter is denoted by a star. Consider an earthquake that occurred at time t with its epicenter in city j. Here we draw a circumference with city j at the center and with a radius equal to the distance between city j and the farthest city that reports a seismic event with a registered intensity in the DBMI We then assign a seismic episode to all the cities located within the circumference.
A graphical representation of this criterion is depicted in Figure I , Panel B, where again the empty dots indicate the cities that registered the episode in the DBMI04, the star denotes the epicenter, and the full dots are the cities imputed by the epicenter criterion. The total number of seismic events yielded by this criterion is of which hit cities before their transition to communes, if any. We then assign the earthquake to all the cities within the union of these circles.
Video Summary
Figure I , Panel C provides an illustration: The total number of episodes identified by this criterion is of which 99 hit cities before their transition to communes, if any. For all the seismic episodes included in the augmented data set according to one of these criteria, we also impute intensity and proceed as follows. Consider an earthquake with epicenter in city i ; we draw around city i a circumference with radius equal to the distance between city i and its closest city struck by that earthquake with intensity equal to D. All the cities included in this circumference are assigned intensity equal to D , whereas all the cities that lie outside this circumference but are still assigned an earthquake according to one of the three augmenting criteria report intensity equal to F.
Table II shows for each of the three augmenting criteria explained here the total number of cities in our sample hit by an earthquake columns 2 , 4 , and 6 and the number of cities in our sample hit before transition columns 3 , 5 , and 7 , which will be used for the identification of the relevant parameters see later discussion. In some cases we report earthquakes that were not registered in the cities of our sample but were still detected by our augmenting criteria for instance, the earthquake that originated in Sansepolcro [not included in our sample] in was not originally registered in our sample of cities but was assigned to Arezzo according to the circles criterion.
In what follows we refer to the polygon criterion as our preferred augmenting methodology. Under the assumption of common trends, we expect the differential effect between episcopal and non—episcopal see cities to be the result of the overlap between political and religious authorities in the former group of cities.
This observation is at the core of our differences-in-differences strategy. Time fixed effects absorb any potential event contemporaneous with the earthquake for all the cities e. Time fixed effects ensure that identification is obtained conditional on shocks common to cities with and without earthquakes in each year.
Solidarity (Polish trade union) - Wikipedia
Finally, the city-specific time trends account for possible slow-moving social and cultural variables e. In fact, in our data, both probabilities are positive functions of time see Tables I and II. Importantly, these trends may vary from city to city because, for instance, cities experiencing increases in education and civic capital accumulation steadily enhance their ability to register earthquakes. The omission of one or both sets of fixed effects or of the city-specific trends would lead to a potential bias in the coefficients of interest.
Their inclusion allows us to exploit the random nature of the timing of the seismic episodes that is the central feature of our design. We explore the dynamic treatment effects by including lagged earthquake variables because we are interested in not only the contemporaneous effect but also its duration. As we will show, this expectation is confirmed by the data. Given the large number of fixed effects included, model 1 and its modifications are estimated adopting a linear probability model LPM. The limitations of this approach and alternative functional forms are discussed in Section IV.
This method requires a threshold distance after which the dependence disappears. This is justified by the fact that, for earthquakes registered by multiple cities in the original data set, the average distance between the epicenter and the farthest city with a registered intensity is about km. The three criteria described in Section III. D polygon , epicenter , and circles assign possibly missing seismic events to cities by starting from recorded earthquakes respectively, the vertexes of the polygons, the epicenter, the center of the circles.
Hence, our augmenting procedures necessarily cannot take into account earthquakes that were registered in no city according to the DBMI04 because the relative documents were not handed down to the historical sources. This may be important in the presence of pretrends. For instance, if transitions in episcopal cities, unlike in non—episcopal ones, happen during a period of political turmoil or in the presence of other circumstances that negatively influence the probability of registering an earthquake, we would erroneously attribute to a seismic episode a negative impact on the probability of becoming commune, when no true differential effect exists.
This, in turn, would imply a negative differential effect of the seismic event between the two groups of cities. If these circumstances last for some years after the transition, we would also observe negative differential effects of the leads of the earthquake variable on the transition.
Thus, finding no negative differential effects of the earthquake on past transitions is consistent with the absence of pretrends. To verify the presence of possible pretrends and provide preliminary evidence on the differential effects of earthquakes on the transition probability, we estimate model 1 including both lags and leads in the earthquake variable. Since the outcome is an absorbing state, lead and lag effects must be estimated separately. In estimating the coefficients on the leads, we leave the outcome variable equal to 1 if a transition occurred in city i and at time t and set it equal to 0 in the following periods before t , time is not defined and the city drops.
On the left, we report coefficients on the leads as they capture a relation between the occurrence of an earthquake in a given period and past transitions. On the right, we measure the contemporaneous coefficient and coefficients on the lags, reflecting the dynamic treatment effects, as they reveal a relation between an earthquake at time t and future values in the outcome variable. Year fixed effects, city fixed effects, and city time trends always included.
Three empirical patterns are apparent in Figure II. These results suggest that the occurrence of an earthquake slows the transition to communal institutions only for episcopal see cities. The effect appears to last longer when we consider the circles augmenting criterion Figure II , Panel C and to last for a shorter time when we adopt the other two criteria, polygon and epicenter respectively, Figure II , Panels A and B. If anything, the effects are virtually zero with point estimates slightly above zero. Third, the large confidence intervals associated with a number of coefficients suggest that they are quite imprecisely estimated.
It is likely that exploiting the annual variability in the earthquake variable generates noise in the estimation. In fact, the latter finding advises us to adopt a model that aggregates the effects of the earthquakes across consecutive years that we report in the following sections. The p -values computed from a test of joint significance of the lagged differential effects grouped by five-year periods are also reported.
They suggest that the differential effect of an earthquake in the first five years following the event is robust to the inclusion of an increasing number of lags. Moreover, as one can see, the effect of an earthquake lasts at most 20 years i. Consequently, we perform the empirical analysis including lags up to 20 years. The estimated coefficients in this model represent the average effect of an earthquake over each five-year period. Columns 1 - 2 - 3 of Table III show our main results from regression 2 on adopting, respectively, the polygon , the epicenter , and the circles criterion.
Estimation by OLS of model 2. Columns 1 - 2 - 3 refer to the data set augmented by exploiting only information from the earthquakes which report a valid intensity in the DBMI04; columns 4 - 5 - 6 refer to the data set augmented by also employing the seismic events for which intensity was not registered in the DBMI04; columns 7 - 8 - 9 refer to the data set that only considers destructive earthquakes with intensity between 6 and 10 physical damage to people or objects.
Our results suggest that the occurrence of an earthquake has always a negative and statistically significant differential effect between episcopal and non—episcopal see cities in the first five years following the event. Similar conclusions are drawn when we consider the epicenter and the circles criteria columns 2 and 3.
Columns 4 - 5 - 6 of Table III report our regression output on employing an extended data set that also exploits information from earthquakes with unreported intensity in the augmenting procedure described in Section III. D respectively for the three criteria, polygon , epicenter , and circles.
Our conclusions are not altered in any relevant way. In columns 7 - 8 - 9 , we show corresponding results obtained on a more restrictive data set that only comprises destructive earthquakes, i. Again results are consistent with those presented in columns 1 - 2 - 3 , but suggest a somewhat larger effect.
- Japan-Bashing: Anti-Japanism since the 1980s (Routledge Contemporary Japan Series).
- The Third Alternative: Christian Self-Government?
- Freckles, the Irish Papillon.
Estimation by OLS of model 3. Columns 1 - 4 - 7 report the effects of destructive earthquakes with intensity between 6 and 10 physical damage to people or objects ; columns 2 - 5 - 8 report the effects of seismic events with intensity between 2 and 5 no physical damage.
The subscripts 0, -5, , refer to the five-year lag respectively: To check the robustness of our results, we implement a placebo test in the spirit of Chetty, Looney, and Kroft adopting our preferred augmenting criterion polygon. In our sample of years and cities, 28 earthquakes took place in northern central Italy, which according to the polygon criterion generated city-episodes the total number of seismic events reported in Table II.
Although randomly assigned, the placebo earthquakes reflect the true time and space clustering of the real data. The time clustering is obtained every 50 years: We repeat this procedure 10, times employing alternative numbers of replications do not affect our results in any significant way and save the estimated coefficients. D, and iterating 10, times. The purpose of this test is to check how many times these randomly generated placebo point estimates happen to be smaller or too close to our true point estimate.
If in our main results we were erroneously rejecting the null hypothesis that our coefficient of interest is equal to 0 i. As can be seen from Figure IV , the point estimates generated in the falsification test are almost always to the right of meaning larger in value than the true estimated coefficients. This does not obtain in only 1. As is clear from Table II , a large fraction of the seismic events considered in our exercise are generated by the earthquake that struck northern-central Italy in In order to investigate how this or other particular episodes contribute to generating our previous results, we re-estimate model 2 by isolating the effect of each single earthquake captured by a dummy variable that equals 1 if earthquake x struck city i at time t , and 0 otherwise from the effect of all the other seismic events captured by a dummy variable that equals 1 if an earthquake different from x struck city i at time t , and 0 otherwise.
Adopting our preferred imputation criterion, we perform 20 different regressions 20 is the number of events that, according to the polygon criterion, hit at least one city before transition to a commune, if any: In Table V , we report our coefficients of interest, namely the differential effects of earthquakes excluding x on the transition probability of Episcopal see cities in the first five years.
Panel A displays results obtained on adopting the data set that only exploits information from earthquakes with a registered intensity in the DBMI When we exclude the earthquake which struck northern-central Italy in this earthquake hit 92 out of cities , the point estimate on the quake variable in the first five-year interval for the Episcopal cities remains negative To show how sensitive our estimated standard errors are to the thresholds of km employed in Conley's correction, we compute the spatially corrected standard errors obtained on adopting thresholds of and km estimation output is shown in Online Appendix Tables A.
Finally, we investigate whether our findings are robust after we drop cities one-by-one this set of regressions is available upon request. No compelling evidence suggests that the results change on the exclusion of some cities in any significant way.
Panel A refers to the dataset augmented exploiting only information from the earthquakes which report a valid intensity in the DBMI04; Panel B refers to the dataset augmented also employing the seismic events for which intensity was not registered in the DBMI Conley's standard errors corrected for spatial dependence with threshold distance of km are reported in round brackets; standard errors clustered at the city level are in square brackets.
One potential concern with using a LPM is that this estimation method could provide an imprecise approximation of the marginal effects, especially when there is a mass of zeroes in the dependent variable, as in our design. We consider them in turn. A conditional logic model in our case suffers from two substantial limitations. First, this method leads us to discard all the information from cities that never transited to communes and does not reach convergence when year fixed effects and city-specific time trends are included.
Nonetheless, a logic model conditional on city fixed effects allows estimation of the parameters of interest and may prove useful for verifying whether the implied negative and statistically significant effect of earthquakes on the transition probability for episcopal cities is confirmed in a nonlinear model.
In fact, reassuringly, the conditional logit model yields results qualitatively similar to those provided by the LPM for episcopal see cities results are in Table A. A duration model would not be appropriate in our context either. Indeed, a Cox model with no time-varying covariates would ignore the panel structure of the data, whereas a Cox model with time-varying covariates i.
Moreover, the Cox model incurs the same estimation problems as a standard logit whenever the independent variable perfectly predicts the outcomes see note This makes estimation of the dynamics of the effects of earthquakes not feasible when the outcome variable transition is always equal to 0 in correspondence to the earthquake variable, or its lags, equal to 1. Finally, a duration model with or without time varying covariates would make it difficult to distinguish between short- and long-run effects of possibly multiple earthquakes on the transition probability.
The Third Alternative: Christian Self Government
To conclude, the LPM seems the most appropriate model in our design to exploit the random nature of the timing of earthquakes. This is especially the case when considering the more serious limitations of alternative models conditional logit or Cox model. Providing empirical test of this hypothesis is no simple exercise, given the absence of systematic data in the medieval period and the impediments to appraising religious beliefs directly.
Yet an indirect measure of the intensity of the religious feelings in a given city in a certain period could be given by the number of constructions and ornamentations of religious buildings. Medieval history furnishes several cases in which people reacted to the dread and consternation caused by a seismic event with an increase in votive offerings and enrichments of religious buildings. These were powerful means to please God and to seek reconciliation after a crisis. One example is provided by the history of Bergamo.
The year was crucial for the monastery of Santo Sepolcro in Astino located near the city of Bergamo: Further examples are suggested by the history of Verona. If democracy based on national identity is the system of governance most consistent with Christianity, the EU stands contrary to a Christian vision for Europe and from a Christian perspective the only solution is to progressively dismantle it. Re-empowering the nations of Europe, their people and leaders, to make decisions for themselves and as to how they will be governed is part of what it means for them to be fully human, a precondition for the deep spiritual renewal that needs to take place.
Account Options
This is not something that can be driven, top-down, from a secular institution; it is the role of the family and churches from the bottom up. The UK is fully capable of charting its own course in the world.
It led the way in the defeat of slavery and industrialised genocide. It remains a permanent member of the UN Security Council. Its contributions to world knowledge, progress, and culture are outstanding. All this has, of course, taken place in a nation underpinned by strong Christian beliefs and traditions, the influence of which still remain. The solution to the Brexit negotiations from a Christian perspective is simple. The EU should let the UK go with its blessing, rather like the father of the lost son who gave his younger son his inheritance early so that he could pursue his dreams, rather than seeking to make him stay or to punish him to frighten others.
The UK is not, of course, a lost son; the point is simply that this is the best way to create relationships based on love, not fear. Let my people go, so that they may worship me. Influential voices are now demanding yet deeper European integration. French President Emmanuel Macron has called for EU identity cards, a shared defence budget, a European military intervention force and an EU defence force.
Martin Schulz, the German SPD leader and former European Parliament president, has called for a new constitutional treaty to create a United States of Europe, with those countries that refuse automatically losing EU membership. Encouragingly, Angela Merkel and others appear sceptical. There is no doubt that these people are well-intentioned. But how will this power be exercised by those who may come after them? You are commenting using your Twitter account. You are commenting using your Facebook account. Notify me of new comments via email.
Notify me of new posts via email. This site uses Akismet to reduce spam. Learn how your comment data is processed. Comment Cancel reply Enter your comment here Fill in your details below or click an icon to log in: Email required Address never made public.
Subscribe to Faith and Politics by email. Post was not sent - check your email addresses! Sorry, your blog cannot share posts by email.