Defining the Focus of this Work – the Construction of a Cathedral in Iberian Territories

To study the spatial autocorrelation of cathedral construction, it is important to define from the start the concept of cathedrals to understand the reason for this focus. Thus, the notion of a cathedral, or cathedral church, is associated with a church where the main figure of the diocese – the Bishop – is entitled to a chair (cathedra). Although the modern concept appeared only in the fourteenth century, in approximately 1340 (Usher Reference Usher2012), our study collected observations for each diocese of the Iberian Peninsula that included the year of the construction or expansion of the most relevant church. Thus, for generalization purposes, this study will primarily observe the religious buildings – as well as the ensuing improvements – of structures considered diocesan headquarters in the Iberian Peninsula since the twelfth century.

Studies that accurately detail the differences between the expressions of Catholic religious architecture, namely, the differences between chapels, churches, basilicas, Sés or cathedrals, are those by authors such as Icher (Reference Icher1998).

Money, Religiosity and Endogenous Resources – Economic Development in the Iberian Peninsula between the Twelfth and Fifteenth Centuries as a Catalyst for the Construction of Cathedrals

The analysis of the economic and social history of the middle ages in the Iberian Peninsula has been developing at a slow pace. The works of Virginia Rau, Damião Peres, Magalhães Godinho and Oliveira Marques created a significant impulse throughout the twentieth century. Subsequently, figures such as Balard et al. (Reference Balard, Genet and Rouche1994), Fourquin (Reference Fourquin1969) and Álvarez Palenzuela (Reference Álvarez Palenzuela2003) generated an equivalent development in a more recent period.

In invoking such works here, we do so with a sense of construction. In reality, we cannot disconnect the development of any cycle of religious construction from the cycle of surrounding economic development. Financing a religious temple has always been a complex task; this is even more so if the temple is categorized as a ‘cathedral’. Regardless of the religious sect in question, in the wake of the findings by Icher (Reference Icher1998), Fourquin (Reference Fourquin1969) and Fonseca (Reference Fonseca1984), the construction of religious temples combines the following levels of effort/cost.

• Efforts toward capital accumulation. Works for public use tend to be financially expensive, regardless of time and place. We restate here the thesis of Briscoe (Reference Briscoe1988) and Hutcheson (Reference Hutcheson1993), according to whom the most ‘grandiose’ works of any time (sometimes identified as ‘works of the regime’) occur in periods of capital accumulation in the surrounding communities that, in general, coincide with peak cycles in local and regional expansion. Thus, churches and cathedrals could only be raised when communities of believers perceived that there was an accumulation of capital. Although individual currencies have had their own cycles within each kingdom of the Iberian Peninsula, we follow the thesis of Valdés (Reference Valdés2014) that moments of greater economic development associated with the accumulation of resources were also periods with a greater dynamism in the circulation of registered money.

• Local or regional availability of materials and labour. If today the professionalization of the civil engineering sector makes it possible for teams to travel around the globe, in the times associated with the data analysed here, the scope of religious temples implied the use of available labour among locals from the region (Icher Reference Icher1998). Such essential availability did not prevent the use of architects from other spaces or even from other countries or the use of imported architectural materials/details. However, the close (regional) availability of labour and construction materials would make any construction project of this magnitude more viable by reducing associated costs.

• Religious efforts. These efforts of religious intentionality, according to Alashi (Reference Alashi2017), can be divided into three main directions: honour, worship and legacy. Thus, communities and leaders with the intention of building this type of religious space ‘feel’ the need to honour the divine (Lang et al. Reference Lang, Chan and Ragvald2005), to manifest the robustness of their faith and to leave this space for their heirs and for other believers (Alashi Reference Alashi2017).

We will now detail these main aspects of the economic history of the Iberian Peninsula between the twelfth and fifteenth centuries, explaining the construction of cathedrals.

From Spatial Autocorrelation to Contagion Analysis – Contagion, Mimicry or Cycles of Economic Expansion

From the above literature, we find three validated dimensions of local dynamics essential for the promotion of religious construction:

• the presence of vast resources (measured in monetary accumulation, following Olmos Reference Olmos2012);

• the existence of a growing population (Valdés Reference Valdés2014);

• finally, the presence of an expressive religiosity supported by local entrepreneurial leaders (Araújo Reference Araújo2011).

However, as Riccetti (Reference Riccetti2002) or Orme (Reference Orme2017) claim, in the cycle of expensive buildings, the following three factors become relevant:

• the presence of an architectural culture;

• the existence of a cycle of economic development extended to the entire surrounding space (in our case, the neighbouring dioceses);

• the influence of competition or complementarity in the aforementioned space (that is, of the surrounding dioceses).

We will now explore these dimensions.

Empirical Model

Based upon the theoretical discussion, our empirical model will explain the construction of cathedrals in the Iberian Peninsula between the twelfth and fifteenth centuries considering the following three assumptions.

• The principle of maintaining building upkeep, by which existing buildings tend to ask for renovations and improvements (although with decreasing marginal value, following Olmos Reference Olmos2012 or Riccetti Reference Riccetti2002);

• The principle of centre–periphery competition in regional economics, by which existing structures of capital located at a centre tend to attract investments from neighbouring areas (Orme Reference Orme2017);

• The mimetism effect, by which the abundance of inputs in a given diocese and in their neighbouring areas will motivate the construction of religious infrastructure (Araújo Reference Araújo2011).

Therefore, the likelihood of a given diocese constructing/expanding a cathedral is identified as y*. It is modelled as a function of neighbouring diocese activity (Neighbour) and of the existing structure of religious buildings (x). We follow the modelling established by Arima (Reference Arima2016).

Therefore, we will define y* _i = αNeighbour _i + x _i β + e _i . Assuming agents are value-maximizers (i.e., they will construct a cathedral if the construction is worth the effort), the probability of observing cathedral construction (y = 1) can be written as:

(1) $$\matrix{ {P\left( {y = 1{\rm{|}}x,Neighbour} \right) = P\left( {y* > 0{\rm{|}}x,Neighbour} \right)} \hfill \cr {\quad \quad \quad \quad \quad \quad \quad \quad \quad \quad = P\left( {e > - \alpha Neighbour - {\bf{x\beta }}{\rm{|}}x,Neighbour} \right)} \hfill \cr {\quad \quad \quad \quad \quad \quad \quad \quad \quad \quad = {\bf{\Phi }}\left( {{\bf{\alpha }}Neighbour + {\bf{x\beta }}} \right)} \hfill \cr } $$

(2) $${y_i} = \left\{ {\matrix{ {0, {\text {if}} \; y * < 0} \hfill \cr {1, {\text {if}} \; y * > 0} \hfill \cr } } \right.$$

where α is the vector of coefficients associated with Neighbour and β the vector of coefficients associated with the controls x. Following Arima (Reference Arima2016), we assume that the unobservable term e is normally distributed; then, the probability of a cathedral’s construction or expansion is given as P(y = 1|x, Neighbour) = P(y* > 0|x, Neighbour) = P(e > −αNeighbour − xβ|x, Neighbour) = Φ(αNeighbour + xβ). Φ(.) is the cumulative normal density function, which primarily leads to a nonspatial probit model.

We use the probit model with spatial dependency designed by Smith and LeSage (Reference Smith, LeSage, LeSage and Kelley2004). Following Arima (Reference Arima2016), this model is more appropriate for our observations in dioceses/religious areas because ‘[…] spatial dependencies are ascribed to occur between regions, not between individual observations. In their model implementation, SLS used county-level observations grouped by states (i.e., regions) and spatial dependencies were assigned between states instead of between individual counties.’

Therefore, following Smith and LeSage (Reference Smith, LeSage, LeSage and Kelley2004) and Arima (Reference Arima2016), we write our empirical model predicting the construction of a cathedral in a given century between the twelfth and fifteenth centuries as follows:

(3) $$y{^*_i} = \alpha Neighbo{r_{ij}} + {{\bf{x}}_{ij}}{\bf{\beta }} + \theta j + {e_{ij}}$$

with

$${\theta _j} = \rho \mathop \sum \limits_{k = 1}^m {w_{jk}}{\theta _k} + {\mu _j}$$

In equation (3), θ _j : j = 1, …, m is the effect of the neighbouring dioceses and may be defined as a spatial autoregressive process, where w _jk is the weight reflecting the degree of spatial proximity between dioceses j and k. ρ is the parameter that reflects the overall degree of spatial dependency between dioceses – we will give a particular focus to this parameter in further estimations. The idiosyncratic error (e _ij ) is assumed to be normally distributed, conditional on θ. μ _j is also assumed to be normally distributed.

Following Mourao and Vilela (Reference Mourao and Vilela2020), we can think of the previous models as binary-dependent variable models based on the usual spatial Durbin model:

(4) $${y_{it}} = \alpha + \rho {W_i}{y_{it}} + {X_{it}}\beta + {W_i}{X_{it}}\gamma + {e_{it}}$$

We also identify αNeighbour _ij in equation (3) as ${W_i}{X_{it}}\gamma $ in equation (4), and therefore, y* also depends upon the cumulative value of religious cathedrals in the neighbouring areas. Therefore, y* can be modelled as

$$y{^*_i} = {\mu _j} + {W_i}{X_{it}}\gamma \; + {{\bf{X}}_{ij}}{\bf{\beta }} + \;\rho \mathop \sum \limits_{k = 1}^m {w_{jk}}{\theta _k} + {e_{ij}}$$

Results and Discussion

Table 1 shows our results for the set of dioceses observed in the Iberian Peninsula.

Table 1. Estimations of the spatial autocorrelation model (Iberian Peninsula, 1100–1500).

We preferred to divide the estimates in equation (3) by centuries to detail how the reaction of each diocese was different from century to century. In the first row, we have the estimated coefficients for the spatial autocorrelation parameter ρ. Thus, it is possible to observe that the dynamic capacity of neighbouring regions influenced the dynamics of religious construction in the dioceses of the Iberian Peninsula and had its period of greatest significance between the years 1200–1300 and 1300–1400. Between 1100 and 1200, we noticed that the dynamics of religious cathedral construction were mainly influenced by the existence of accumulated value in religious buildings (parameter β). The statistical significance found for all the estimated coefficients of parameter β (positive parameters) over the period reveals that the local dynamics of each diocese is the main factor affecting the construction or expansion of the cathedral. This perspective is convergent with the perspective of Icher (Reference Icher1998). On the other hand, we cannot fail to mention that the estimated value for ϒ reveals, for periods from 1200 onward, a sense of ‘spatial competition’ between neighbouring regions; the negative sign indicates that the greater the accumulated value of religious heritage in the surrounding dioceses, the expected likelihood of building or expanding a cathedral in a given diocese is less, for most spaces in the Iberian Peninsula. The values of log-likelihood as well as R ² suggest that the model estimated for the period between 1400 and 1500 has a higher quality than the models estimated for the remaining periods.

After discussing Table 1, we divide Iberian dioceses by considering them as Portuguese (if belonging to the kingdom of Portugal) or Spanish (if belonging to the kingdoms of Castile-Léon, Navarre or Aragon). We opted for this division to avoid generalist insights that would happen if we had not done this methodological step. In line with the arguments of diverse authors (Fonseca Reference Fonseca1984; Navascues Reference Navascues2010), given the autonomy of individual spaces – even in ecclesiastical terms – additional reasons arise for justifying this division.

Table 2 shows the estimated coefficients for equation (3) but only considers the data for the Portuguese dioceses. Some differences can already be found in light of the results shown in Table 1. First, the Portuguese dioceses only felt the effect of spatial correlation from 1300 onward. In particular, from 1400 onward, the dynamics of religious construction in the surrounding spaces stimulated religious construction in each diocese, with the curious aspect being that here, unlike the estimated parameter for dioceses across the Iberian Peninsula (Table 1), the estimated parameter for the Portuguese period between 1400 and 1500 is positive and significant. This evidence (ϒ positive and significant) indicates a complementary relationship between the religious capital accumulated in the neighbouring dioceses and the construction or expansion of the cathedral in each Portuguese diocese (Fonseca Reference Fonseca1984). It should be noted that this period in the Portuguese space is one of clear economic and demographic growth in most years as a result of the centrality that the country gained from the dynamics of sea voyages and Iberian discoveries. Once again, the estimated β values were significant and positive, indicating that dioceses with accumulated capital in religious constructions tended to carry out more modifications/expansions than spaces with less accumulation of religious capital.

Table 2. Estimations of the spatial autocorrelation model (Portugal, 1100–1499).

Table 3 shows the estimation of equation (3) only for Spanish dioceses in the period under analysis. The differences with the results achieved for the Portuguese dioceses are notable, especially in the values estimated for ϒ. These values were estimated to be significant and negative, which shows that, in Spanish dioceses, a spatial competition effect persisted between neighbouring dioceses. The more expanded a particular cathedral was, the lower the expected probability of the construction of religious buildings in neighbouring dioceses in the observed periods. The remaining parameters under observation – ρ and β – were in line with the results of the previous tables. Therefore, there were particularly positive and significant ρ values after 1200, showing that works in neighbouring dioceses led to significant development in each diocese in question. This result converged with our central hypothesis; that is, the influx of expressive inputs/resources in the various regions tended to produce more frequent ‘outputs’ in the set of analysed regions. These results also follow Olmos (Reference Olmos2012) and Valdés (Reference Valdés2014). Once again, dioceses with more highly valued accumulated capital tended to be more likely to have more religious construction works.

Table 3. Estimations of the spatial autocorrelation model (Spain, 1100–1499).