Abstract This paper proposes the use of street names as a source of geographic-specific cultural data for quantitative analysis in social sciences. Street names reflect the cumulative commemorative decisions of municipalities and, as such, can be used as proxies for their social and cultural characteristics, which is particularly useful given the scarcity of cultural data at the local level. The utility of street names as a data source is illustrated through the study of religiosity and local economic development. The street-name indicator of religiosity is shown to be strongly correlated with the cultural factor it is supposed to capture, and appears to be negatively related to economic development at the local level. Finally, the article explores a range of other potential empirical applications to important topics in social sciences. 1. Introduction This paper shows that street names are a very useful source of information for quantitative empirical analysis in social sciences. Street names reflect the commemorative decisions of each municipality over time, and as such can be understood as the city’s manifesto about its social, cultural and political values. In accordance with this view, this article proposes the use of the ‘city-text’ (i.e., city toponyms interpreted as a text) as a source of information to create socio-cultural indicators at the local level. Given the scarcity of data at the city level, street names broaden the possibilities of empirical analysis. For instance, they allow creating indicators of religiosity, national identity or male predominance at the local level—to mention just a few examples—which are generally unavailable from other sources such as surveys or censuses. The city-text as a data source for constructing indicators is particularly convenient given the current emphasis among empirical researchers on the causes and consequences of culture (e.g., Guiso et al., 2006; Alesina and Giuliano, 2015), and the advantages of working with city-level data, which help mitigate endogeneity problems.1 A well-established geographic literature maintains that ‘the names we choose symbolize a great deal about our culture and view of the world’ (Cooper et al., 2011, 460). The system of place names (or toponyms) configures the cityscape as a ‘“text” through which the commemorative priorities of a people can be read.’ (Rose-Redwood et al., 2010, 460). These priorities represent a set of cultural and social values, as well as a version of history and hegemonic narratives of former periods. In this sense, street names express something important for the community—such as a well-known person, event or abstraction—that deserves to be commemorated by an inscription in the cityscape (Augustins, 2004). The use of street names for commemorative purposes is a core feature of modern political culture, and they can be considered markers of political identity (Azaryahu, 1996). It has been argued that ‘street names socialize space and celebrate cultural identity [… and] are significant cultural indicators’ (Ferguson, 1988, 386). This article brings together quantitative empirical methods and insights from the city-text literature to make a twofold contribution. First, we propose the use of street names as a source of information to create indicators of the population’s attitudes and values. The rationale for this is that commemorative decisions in the form of street names arguably reflect the social reality and idiosyncratic characteristics of the local community. Second, we present an empirical application of this approach by analyzing the link between religiosity and local economic development. It is shown that an indicator of religiosity created from street names is highly correlated with the underlying factor it is supposed to measure, thus filling a gap in the literature on systematic evidence on the relationship between street names and local culture. In addition, we find that this measure of religiosity is negatively associated with economic outcomes at the municipal level, which is consistent with a negative view of the economic legacy of Catholicism (the main religion in Spain).2 Thus, this paper shows the utility of street names to study the interaction between culture and economic outcomes at the local level. This article is related to a strand of literature in geography that quantitatively analyzes town and business names to investigate patterns of regional identity (Ambinakudige, 2009; Cooper et al., 2011; Fuchs, 2015; Liesch et al., 2015; Weaver and Holtkamp, 2016). The key difference between this study and these previous works is its focus on street names. Compared with other toponyms, street names are very often the outcome of a political process of commemoration that explicitly aims to emphasize a salient social or cultural value in the local community. In addition, while toponyms like town names provide just one observation per municipality, the system of street names contains much more data capturing many cultural dimensions.3 These distinctive features are what provide street names with a special value added. The much richer information they contain allows creating indicators for measuring several dimensions of people’s cultural values, such as religiosity, political ideology, male predominance, etc., thereby constituting a versatile data source. We use the Spanish case as an example to illustrate how the city-text can be used to create socio-cultural indicators, allowing us to discuss conceptual issues and propose a methodology for the construction of street-name indicators. More specifically, we provide an application of this methodology by creating an indicator of religiosity at the municipal level using a dataset of more than 700,000 Spanish streets in 2001. Interestingly, there is a close relationship between this indicator and the population’s current religious attitudes. The correlation between our indicator of religiosity and a survey-based indicator of church attendance is above 50% at the provincial level, and highly statistically significant. In addition, municipal-level regressions show that the street-name indicator of religiosity is significantly associated with a lower incidence of separation and divorce and fewer de facto couples, with this result being robust to controlling for provincial-fixed effects and a set of relevant control variables. In turn, we use the street-name indicator of religiosity to analyze the relationship between this cultural factor and local economic development. We find that religiosity is negatively related to average gross income, as well as to other indicators of local development. Thus, this paper’s results indicate that street names reflect a society’s ingrained cultural characteristics, and can be used to study the relationship between culture and economic and social outcomes at the local level. We further propose other potential applications of street-name indicators in the last part of the paper. The use of street names as a data source broadens the possibilities of empirical analysis at the local level. Economics and other social sciences have recently placed more emphasis on this finer level of analysis (see, e.g., Trounstine, 2009). From an empirical point of view, the sample size increases dramatically when using municipalities (as does statistical power), and more importantly, endogeneity issues are better addressed at this level, as it allows the exploitation of within-regional variation. The utility of the city-text as a data source is particularly relevant for cultural variables, given the scarcity of data in this regard and the increasing interest in the role played by cultural factors in the economy (Scott, 2004; Guiso et al., 2006).4 Thus, researchers can use street names to create indicators capturing the specific cultural, social and political values of local communities. Moreover, to the extent that the city-text reflects the history of towns and cities, street names can also be used to analyze the legacy of historical factors and past events. Finally, the analysis of the city-text is not only interesting to social scientists from the point of view of its utility to create indicators. The process of street naming is important in itself, and may have consequences for the population. This has precisely been the focus of the critical literature on street names. The link between toponyms and the exercise of power has been extensively studied, particularly how political authorities use the former to promote particular views of history and national identity (Rose-Redwood et al., 2010). According to Azaryahu (1996), ‘the main merit of commemorative street names is that they introduce an authorized version of history into ordinary settings of everyday life’ (p. 312). Street names thus contribute to the (re)production of this narrative, becoming powerful instruments for legitimizing the sociopolitical order. Consider, for example, gender inequality and the historical marginalization of women in the public sphere. The overrepresentation of male names in the city-text not only reflects a historical and cultural fact, but also contributes to its perpetuation by making the presence of male names more ‘natural’ in daily life.5 Therefore, street names can also be used to explore the impact this symbolic infrastructure has on social behavior. The remainder of the paper is organized as follows. Section 2 provides a brief overview of the related literature. Section 3 discusses conceptual and methodological issues related to the use of street names to create socio-cultural indicators, while Section 4 illustrates an application to the study of religiosity and local economic development using Spanish street maps. Section 5 explores other potential applications of street-name indicators to relevant topics in social sciences. Finally, Section 6 puts forward some implications and concludes. 2. Background and related literature The practice of commemorating national personalities and events in the cityscape is a modern phenomenon. Even though naming places after important figures has existed since Antiquity (e.g., Alexandria), its widespread use began toward the end of the eighteenth century in France. Traditionally, street names were vernacular, with a clear orientation purpose, and were associated with the local topography or history. During the French Revolution, however, the practice of using street names for political purposes became increasingly common, and other European countries subsequently followed the French lead. This practice was later included in the agenda of nation-building programs during the nineteenth century (Azaryahu, 1996). Today, commemorative street names are a familiar reality in many countries throughout the world. Street names are normal features of the cityscape, but they carry a deep symbolic meaning. The city-text is a repository of the commemorative decisions made by the local authorities with the power to name streets. It is the product of the prevalent and hegemonic values in a determined community at a given time. This rich information can be studied from two complementary perspectives: the analysis of place names and the analysis of place naming (Giraut and Houssay-Holzschuch, 2016). The former focuses on the name itself, studying its origin and meaning, while the latter is related to the political dimension of commemorating; that is, the power relationships associated with the process of choosing and using names for political purposes. Place name studies have investigated a variety of topics, including the reconstruction of settlement and migration histories (Arnold, 1977; Arge et al., 2005; Qian et al., 2016), the identification of past land uses (Conedera et al., 2007; Calvo-Iglesias et al., 2012) and regional and ethnocultural identity (Cooper et al., 2011; Fuchs, 2015; Weaver and Holtkamp, 2016). The crucial premise is that toponyms reflect a certain truth about social reality and history, and can therefore be used as a source of information to study different aspects of local communities and their interaction with time and space. Thus, place names can ‘be read as part of the official memory of a city or a nation, acting as evidence of social norms and values at a given time.’ (Hintermann and Pichler, 2015, 291). Weaver and Holtkamp (2016) state that ‘[p]lace names, or toponyms, are therefore traces of “cultural life” that reflect the thought processes of peoples in space–time’ (p. 208; see also Loy, 1989). Similarly, Fuchs (2015) argues the utility of toponyms to trace the geographic contour of socio-cultural features. More specifically, place names ‘provide historical and cultural information on settlement patterns and allow insights into the status, character, and mindset of settlers, eponyms (name givers), and local residents’, which help understand the culture that originated the names (Fuchs, 2015, 330). Within this branch of literature, the work most closely related to this article from a methodological point of view involves the identification of regional identity through traces such as town and business names (Ambinakudige, 2009; Cooper et al., 2011; Fuchs, 2015; Liesch et al., 2015; Weaver and Holtkamp, 2016). Typically, researchers create indicators of regional identity by detecting and counting how many towns or businesses within a certain geographic area contain a specific word that can be related to a certain ethnocultural trait. For instance, Fuchs (2015) studies Germanic toponyms in the American Midwest, and calculates the proportion of Germanic place names at the county level. Interestingly, he finds a highly statistically significant correlation between the proportion of place names and ancestry data. Weaver and Holtkamp (2016) use toponyms and business names to investigate the cultural geography of the Appalachian region in the USA. They find a highly significant relationship at the county subdivision level between Appalachian place names and business names. Although our research and these works share the use of a statistical analysis of toponyms to identify cultural patterns, this paper differs from the previous ones in many respects. We focus on street names (rather than on other toponyms, such as town names or business names), and our aim is not to study regional identity, but to propose a methodology to create socio-cultural indicators at the local level, and prove the validity of this approach by analyzing the correlation between street-name indicators and the factors they are supposed to measure. As mentioned above, street names are substantially different from other toponyms, and provide much richer information about local culture. Place names can also be considered from the perspective of the process of assigning names (i.e., place naming). While not the focus of this article, it is worth briefly introducing some work in this area, because there are potentially interesting applications to quantitative empirical analysis. The research question here involves the procedures and interests in the place-naming process (Giraut and Houssay-Holzschuch, 2016). Generally speaking, this literature focuses largely on street names rather than on other types of toponyms, and the central issue is the political dimension of the naming process. As stated by Rose-Redwood et al. (2010), ‘the naming process sheds light on power relations—how some social groups have the authority to name while others do not—and the selective way in which such relations reproduce the dominance of certain ideologies and identities over others’ (p. 462). This question has attracted considerable attention because street names are considered to have strong symbolic power. In the view of Azaryahu (1996), commemorative street names are a ‘powerful mechanism for the legitimation of the sociopolitical order’ (p. 311), whereby the source of their power lies in their naturalization of an official version of history by incorporating it into the common space of daily life. Thus, ‘the spatial configuration of historical street names defines a particular ‘city-text’ that pertains to a particular historical narrative’ (Azaryahu, 1996, 324). Although frequently taken for granted, street names contribute to the legitimization of the existing socio-political order and political authority (Tucci et al., 2011; Drozdzewski, 2014). This power of street naming systems is reflected in the frequent renaming of streets that accompanies major political change (see, e.g., Azaryahu, 1997; González Faraco and Murphy, 1997; Light, 2004; Palonen, 2008; Drozdzewski, 2014). Finally, there is a manifestly historical dimension in the analysis of street names. The current city-text is the outcome of a cumulative inscription process over a long period of time, whereby street-name indicators also reflect features of this long-term process.6 This historical dimension of the city-text is clearly manifested in the literature. In his analysis of Paris and Evora, Augustins (2004) observes that social groups are strongly attached to their place names, which makes them persistent. Similarly, Azaryahu (1996, 326) states that ‘[u]nless renamings were undertaken, a city-text would conserve notions and attitudes that prevailed in the different stages of its emergence’. Tucci et al. (2011) show, for the case of Milan, the coexistence of fragments from its long history inscribed into the city-text. Interestingly, Conedera et al. (2007) argue that many toponyms have very old origins (even pre-Roman in Europe’s case), and can therefore be considered historical records reflecting cultural heritage in a broad sense. 3. Street names as a data source for socio-cultural indicators 3.1. Necessary conditions of the data-generating process This section discusses certain conceptual and methodological issues related to the creation of socio-cultural indicators from street names. We begin by focusing on the necessary conditions of the naming decision process so that the resulting street names inform about the local community’s cultural and social values. In a nutshell, the basic argument underscoring this paper is as follows: the current system of street names is the result of a long process of commemorative decisions, and so the city-text reflects the commemorative priorities of the local community (Rose-Redwood et al., 2010). As these priorities are arguably the expression of social and cultural values, street names can be used as meaningful socio-cultural indicators. More formally, the suitability of street names as socio-cultural data depends on whether the following conditions are satisfied: Condition (1) Street names are the result of commemorative decisions. Condition (2) Naming (commemorative) decisions reflect the commemorative priorities of the local community. Condition (3) Commemorative priorities reflect people’s social and cultural values. It is evident that these three conditions cannot be taken for granted. To begin with, Condition (1) contains the precondition that streets are labeled with words (i.e., names). However, this is not necessarily the case, as streets may follow an alphanumeric system, which is common in US cities. Once this precondition has been satisfied, street names must be the result of commemorative decisions. This is not always the case either. Street names can be vernacular, as they were until the Modern Age, and only convey an orientation purpose. Vernacular—or autochthonous—names are those that have evolved spontaneously (naturally) over time, containing specific local references that only make sense for the local community (González Faraco and Murphy, 1997). In contrast, commemorative names are the result of a political process of naming to honor something important for those choosing the name. Condition (2) refers to how democratic and open the naming process is. The condition is met to the extent that the decision-making process is democratic and decentralized at the local level. If the authority to name is a prerogative of the local council, and it has been democratically elected, one expects that commemorative decisions will ultimately reflect the priorities of the local community. Completely different is the case in which the right to name belongs to the central government or to a non-democratically elected local ruler, as street names will then reflect their own particular commemorative priorities. It is worth emphasizing how the failure of this condition affects the suitability of street names as a proxy for socio-cultural values. Take, for instance, a town with a large percentage of streets with religious names. This may suggest a strong religious sentiment among the population. However, if the naming process is not democratic, this would simply reflect the religiosity of the authoritarian ruler. What’s more, a large percentage of religious names could even reflect a last resort propaganda effort to reverse a trend toward secularization in the population, thereby meaning the opposite to what could at a first glance suggest. It is therefore crucial for the naming process to be democratic. Regarding Condition (3), it is generally accepted in the literature that toponyms reflect the local culture, and that they can be used to study it (e.g., Augustins, 2004; Conedera et al., 2007; Weaver and Holtkamp, 2016). For example, communities characterized by a strong attachment to the nation are more likely to commemorate national symbols and heroes. Likewise, returning to the example in the previous paragraph, a Christian community with a strong religious sentiment arguably tends to name more places after Christ, saints, etc. Before using street names to construct socio-cultural indicators, it is important to examine whether the previous conditions are met. Condition (1) can be checked by using local institutional knowledge and by analyzing the nomenclature system used to name streets in a sample of municipalities within the area of study. Importantly, not all streets must meet this requisite. It suffices if a significant percentage of streets have commemorative names. With respect to the validation of Condition (2), it can be done by analyzing the formal process of street naming; that is, who has the prerogative (i.e., who the authority is) and how it is elected. Finally, Condition (3) can be validated by comparing indicators created from street names with other indicators capturing the underlying variable that the researcher seeks to measure.7 3.2. Methodological considerations Street-name indicators can be easily created through the following formula: Ind1 = SN, (1) where S refers to a variable counting the number of streets whose names contain a particular search text or keyword from a list of specific words or expressions, and N is the total number of streets in a municipality.8 The list of words or expressions for calculating S must be related to the socio-cultural factor that we want to measure. When creating a list of words, it is important to avoid potential biases by using methods that are not influenced by researcher subjectivity. A relevant methodological observation involves measurement errors. The main notion that this paper conveys is the use of street names to create indicators for quantitative analysis for a large cross-section of municipalities. This implies that the procedure for identifying and quantifying the search text should be as automatic as possible, so an individualized analysis of the streets for each city is unfeasible. Due to the virtually infinite number of possibilities in street names, it is impossible to take into account every potential name, and therefore measurement errors are always present. However, to the extent that these errors are not systematically related to a confounding factor, they will only produce an attenuation bias. The regression estimates when using street-name indicators can therefore be interpreted as a lower bound. Furthermore, these measurement errors imply that street-name indicators should be used preferably for comparative purposes across municipalities, but they are not as appropriate for making claims about absolute values for specific municipalities.9 Another methodological issue is that certain factors affecting the street-naming and renaming process may create a bias in the measure of street names. First, towns that have physically grown faster have many newer streets whose names are likely to reflect modern choices rather than those of the past. Conversely, cities that have experienced slow growth will probably have more traditional street names. While we acknowledge this source of potential bias, we consider that its impact is not severe. On the one hand, when naming new streets fast-growing cities will probably choose names that are consistent with their cultural values, which are supposedly reflected in their current street names. On the other hand, slow-growing cities may adapt their street names to their changing cultural and social values. Regarding this issue, if the street names of fast- versus slow-growing cities differ, it is more likely due to changes in cultural values in the former (as a consequence—for instance—of structural transformation, immigration, etc.) than to this source of bias. Related to the above point, a bias may also arise from a selective renaming process. For instance, capital cities with major redevelopments in the city center may have changed their street names more. In this sense, street names may also capture the historical urban structure, with cities with major redevelopments having more recent—and probably modern—street names. For the same reasons mentioned above, we consider that the impact of this potential bias is limited. In addition, it is a concern affecting only capitals or large cities, which are generally a minority in any country.10 Furthermore, the number of streets affected by redevelopments in large cities is—arguably—small compared to the total number of streets they have, suggesting that any selective renaming will have only a small impact on the street-name indicator.11 4. An application of street-name indicators: analyzing the effect of religiosity on economic development in Spanish municipalities This section illustrates the usefulness of street names as a source of geographic-specific cultural data with an empirical application to the study of religiosity and local economic development. We first conduct an exploratory text analysis of the main features of Spanish street names, then we construct an indicator of religiosity with street-name data and analyze its validity. Finally, we examine the relationship between our measure of religiosity and several proxies for local economic development. 4.1. Data preparation and initial text analysis Spain is a good case study because its streets largely satisfy the first two conditions in the previous section. Commemorative streets are very common in Spain, and it is the local council’s prerogative to name streets, with these councils being democratically elected since 1979. Consequently, municipalities have had enough time to change their street names whenever necessary to adapt them to the social and cultural values of the current population. Regarding data on street names, we use the 2001 Electoral Census Street Map provided by the Spanish Statistics Office (INE, 2016a). This dataset contains a list of Spanish streets in July 2001, indicating their name, type (avenue, street, square, etc.) and municipal code. There are a total of 730,082 streets in the dataset. Before using it to construct indicators, the data need to be prepared and refined: (i) repeated street names in the same municipality are removed, which reduces the number of streets to 701,429; (ii) the text is converted to lowercase, and accents are removed to streamline the text search process; (iii) a blank space is added at the beginning and end of each street name, which reduces potential confusions when identifying words12 and (iv) 83 streets with missing data are removed, whereby the total number of streets in the sample is 701,346. Figure 1 takes a first look at the data through a bar chart representing the forty most frequent words that appear in our dataset of Spanish street names.13 It clearly reflects the importance of religion, as the most frequent word is by far ‘san’ (i.e., ‘saint’). The second one (i.e., ‘cortijo’) can be considered vernacular because it refers to a type of farm that is widespread in Southern Spain. The third one is again a religious word, while the fourth refers to a male forename. A quick glance at the chart suffices to create a picture of some of the main characteristics of Spanish street names: predominant commemorative purposes, the importance of religion and the higher proportion of male names over female names. It is useful to repeat this visual analysis for each one of the Spanish regions in order to obtain a more accurate picture of regional differences and common features. Figure 1 View largeDownload slide The forty most frequent words in Spanish street names. Note: Bars indicate how many times each word appears in the street names. Figure 1 View largeDownload slide The forty most frequent words in Spanish street names. Note: Bars indicate how many times each word appears in the street names. The next step is to conduct an exploratory text analysis of the list of street names. It is important for several reasons: (i) to discover the most frequent types of names in the geographic area of study, (ii) to create lists of words capturing the socio-cultural dimensions we want to measure, while avoiding researcher bias and (iii) to check whether street names are used for commemorative purposes, and therefore whether Condition (1) above is met. We conduct the exploratory text analysis as follows. First, we extract the 200 most frequent words appearing in the street names in each one of Spain’s seventeen self-governing regions—Autonomous Communities—provided that the frequency of a word is above 10. By doing so for each region separately we ensure that we collect information for all of them, which is particularly important given the fact that some regions have vernacular languages. We compile the whole list of words and remove any repetitions, rendering a total of 1188 words. Second, we check both the meaning of each word and the function it generally performs in street names in order to classify the words into categories, such as ‘religious’, ‘surname’, ‘male forename’, ‘female forename’, ‘illustrious people’, etc. Accordingly, a word categorized—for example—as ‘religious’ allows us to identify whether a street has a religious name. Third, we distinguish between categories associated with commemorative street names and those associated with vernacular street names. Finally, we count how many streets contain each word, which allows us to estimate the relative importance of each category. Table 1 shows the results from this analysis. Panel A presents the list of categories associated with commemorative street names. There are 53 words with religious meaning (out of 1188 words analyzed), and 65,625 streets that contain any one of these religious words (Columns 1 and 2). There are 155 words corresponding to surnames and 46,297 streets containing any one of these surnames, etc. It is important to note that the total values reported at the bottom of Panels A and B do not correspond to the sum of values in Columns (2), (3) and (4). This is because the same street name may contain—for example—both a religious word and a male forename (e.g., ‘San Juan’), so it is counted separately in each individual category, but when adding up all the categories the street is counted only once. For this reason, the sum of the number of streets reported for each category in Column 2 is larger than the total.14 Table 1 Initial text analysis of Spanish street names Number of words Number of streets containing words included in each category Percentage of streets (%) Over the respective panel Over the total (1) (2) (3) (4) Panel A: Commemorative words Categories Religious 53 65,625 29.5 16.1 Surname 155 46,297 20.8 11.3 Male forename 115 91,809 41.3 22.5 Female forename 22 14,733 6.6 3.6 Illustrious people 71 20,424 9.2 5.0 Professions 16 14,588 6.6 3.6 Crown and aristocracy 11 8,281 3.7 2.0 Regions and countries 85 25,025 11.3 6.1 Other commemorative 24 11,282 5.1 2.8 Total 552 222,124 100.0 54.4 Panel B: Vernacular words Categories Urban geography 158 62,245 33.4 15.2 Rural geography 139 67,599 36.3 16.5 Geography (generic) 114 10,766 5.8 2.6 Nature 41 18,551 9.9 4.5 Other vernacular 131 48,557 26.0 11.9 Total 583 186,461 100.0 45.6 Total commemorative and vernacular 1,135 408,585 100.0 Panel C: Non-classified 53 8,289 Number of words Number of streets containing words included in each category Percentage of streets (%) Over the respective panel Over the total (1) (2) (3) (4) Panel A: Commemorative words Categories Religious 53 65,625 29.5 16.1 Surname 155 46,297 20.8 11.3 Male forename 115 91,809 41.3 22.5 Female forename 22 14,733 6.6 3.6 Illustrious people 71 20,424 9.2 5.0 Professions 16 14,588 6.6 3.6 Crown and aristocracy 11 8,281 3.7 2.0 Regions and countries 85 25,025 11.3 6.1 Other commemorative 24 11,282 5.1 2.8 Total 552 222,124 100.0 54.4 Panel B: Vernacular words Categories Urban geography 158 62,245 33.4 15.2 Rural geography 139 67,599 36.3 16.5 Geography (generic) 114 10,766 5.8 2.6 Nature 41 18,551 9.9 4.5 Other vernacular 131 48,557 26.0 11.9 Total 583 186,461 100.0 45.6 Total commemorative and vernacular 1,135 408,585 100.0 Panel C: Non-classified 53 8,289 Notes: This analysis is conducted using the 200 most frequent words in the street names of each Autonomous Community, provided that the frequency of a word is at least 10. Totals in Columns (2), (3) and (4) do not correspond to the sum of values corresponding to all categories. This is because the same street name can contain words from several categories, so it is counted separately in each row, but when adding up all categories the street is counted only once. When the same street name contains both a commemorative and a vernacular word, it is considered that the commemorative character prevails and, consequently, the street is considered commemorative. The category ‘geography’ refers to broad geographic features that are not specifically related to the urban or rural world. Panel C contains polysemic words, conjunctions, prepositions and single letters. Regarding the relative importance of each category, religiosity is a highly significant cultural factor, accounting for around 30% of the total number of commemorative streets and 16% of the total number of streets (Panels A and B). It is also interesting to look at male and female names. The ratio of male-to-female forenames is as high as 6.2 (i.e., 91,809/14,733). However, this does not accurately capture the overrepresentation of men in street names. When we consider that the words included in categories such as ‘professions’ and ‘illustrious people’—among others—almost always refer to male individuals, the male/female ratio increases to 7.6. Yet this ratio is higher (9.3) when we exclude streets related to religious figures, since their proportion is higher among streets with female names. It is also important to note that according to this initial analysis the number of commemorative streets is larger than the number of vernacular streets. This indicates that commemorative streets are a widespread phenomenon in Spain, and consequently Condition (1) above holds. Finally, at the bottom of Table 1 it is reported that 408,585 streets contain any one of the 1,135 words classified as commemorative or vernacular. This indicates that by analyzing the 200 most frequent words in the street names of each region we have been able to account for almost 60% of all Spanish streets (i.e., 408,585/701,346), which means that the exploratory text analysis conducted is effective in capturing the characteristics of the Spanish city-text, and can therefore be used as a guide for the construction of socio-cultural indicators. 4.2. A street-name indicator of religiosity Religious attitudes are a key cultural factor that has attracted considerable interest from social scientists since Weber’s seminal work on the cultural origins of capitalism (Weber, 1976 [1904–1905]; see e.g., Barro and McCleary, 2003; Guiso et al., 2003; Becker and Woessmann, 2009; Cantoni, 2015). However, indicators measuring religiosity are generally not available at the city level, particularly for a large sample of municipalities. In this section, we use street names to construct an indicator of religious attitudes at the local level that can be used as a proxy for this cultural factor. Cities and towns differ in the prominence they give to religion in the cityscape. If we assume that municipalities with a higher percentage of streets named after religious figures (or associated with religious events or symbols) give more importance to religion than municipalities with a lower percentage, then street names can be used to create an indicator of religiosity. The indicator is defined as follows: Relig = RN×100, (2) where the numerator, R, captures the number of streets containing a word from a list of religious terms (e.g., ‘San’ [Saint], ‘Virgen’ [Virgin], etc.), while the denominator, N, refers to the total number of streets in each municipality. We take the list of religious words from Table 1 (i.e., 53 words) to avoid potential researcher bias, which could easily be introduced if we created a list of religious words that we ‘think’ are most likely to be used in street names—as the resulting list would reflect words that are familiar to us. Naturally enough, this methodology produces measurement errors in the numerator of Formula (2). It is virtually impossible to create a comprehensive list of all potential religious words that can be used in street names. However, to the extent that errors are not systematic, Relig is a valid indicator to compare this cultural factor across municipalities. Bearing in mind the previous caveats, there are 65,625 religious streets. The proportion of religious streets is on average 11.77% across municipalities. Figure 2 shows the distribution of the variable along with the kernel density estimation. It is relatively close to a normal distribution, albeit skewed. To avoid an excessive influence of observations with very high values in religiosity, we exclude 53 observations with values higher than 50% in the municipality-level analysis conducted below. Panel A in Figure 3 shows the geographic distribution of this indicator aggregated at the provincial level to facilitate its visualization. Higher values of religiosity are found along Spain’s northern Meseta and on the western reaches of the Pyrenees. Figure 2 View largeDownload slide Distribution of the street-name indicator of religiosity. Note: Epanechnikov kernel function used; bandwidth equal to 2.5. Figure 2 View largeDownload slide Distribution of the street-name indicator of religiosity. Note: Epanechnikov kernel function used; bandwidth equal to 2.5. Figure 3 View largeDownload slide (A) Street-name and (B) survey-based indicators of religiosity. Note: A darker color indicates more religiosity. Figure 3 View largeDownload slide (A) Street-name and (B) survey-based indicators of religiosity. Note: A darker color indicates more religiosity. Is the street-name indicator of religiosity really capturing the intensity of the population’s religious sentiment? It is not easy to analyze the validity of this indicator—and consequently whether Condition (3) above is met—due to a lack of data, particularly at the local level. One possibility is to aggregate street-name data at the provincial level in order to compare it with a survey-based indicator of religiosity, which can be constructed at this level. This indicator measures the percentage of people attending religious services at least once a month (CIS, 2005), and is depicted in Panel B in Figure 3. Interestingly, high values in the street-name and survey-based indicators of religiosity tend to overlap. There is indeed a strong and statistically highly significant correlation between both variables (ρ = 0.52, p-value < 0.001). This suggests that the street-name indicator accurately captures the population’s religious attitudes. Table 2 reports further evidence on the strength of this relationship, which is robust to the inclusion of: (i) GDP per capita; (ii) latitude, which controls for the North–South gradient in Spain; (iii) the percentage of landless workers in 1797 as a proxy for historical inequality, which has been a very important factor in Spain’s development (Oto-Peralías and Romero-Ávila, 2016) and (iv) four dummy variables measuring whether the region has a second official language besides Spanish, which control for the cultural and linguistic particularities of these territories. Remarkably, the coefficient on the street-name indicator remains positive and highly statistically significant in all cases.15 The strength of the relationship between this indicator and the measure of church attendance is high. In the most saturated model (column 6), a one-percentage-point increase in religious streets raises church attendance by 1.9%.16 The partial R2 of the street-name indicator in this model is 18.4%, which also indicates a high explanatory power. Table 2 Street-name indicator of religiosity: assessing validity at the province level (I) Dependent variable is percentage of people attending the church every month (1) (2) (3) (4) (5) (6) Street-name indicator of religiosity 2.124*** 2.186*** 2.007*** 2.103*** 2.253*** 1.856*** (0.545) (0.553) (0.534) (0.521) (0.614) (0.588) GDP pc 2001 −0.5 −0.301 (0.374) (0.476) Latitude 0.551 0.064 (0.341) (0.426) Historical inequality −0.137*** −0.185* (0.049) (0.1) Catalan −8.986* −7.075 (5.025) (6.279) Valencian −10.67*** −11.363*** (3.35) (3.236) Basque −0.51 −6.67 (2.742) (4.714) Galician 4.987 −4.02 (4.303) (5.596) R2 0.27 0.30 0.30 0.34 0.40 0.46 Observations 50 50 50 50 50 50 Dependent variable is percentage of people attending the church every month (1) (2) (3) (4) (5) (6) Street-name indicator of religiosity 2.124*** 2.186*** 2.007*** 2.103*** 2.253*** 1.856*** (0.545) (0.553) (0.534) (0.521) (0.614) (0.588) GDP pc 2001 −0.5 −0.301 (0.374) (0.476) Latitude 0.551 0.064 (0.341) (0.426) Historical inequality −0.137*** −0.185* (0.049) (0.1) Catalan −8.986* −7.075 (5.025) (6.279) Valencian −10.67*** −11.363*** (3.35) (3.236) Basque −0.51 −6.67 (2.742) (4.714) Galician 4.987 −4.02 (4.303) (5.596) R2 0.27 0.30 0.30 0.34 0.40 0.46 Observations 50 50 50 50 50 50 Notes: Variables descriptions are provided in Online Appendix Table A1. The estimations include a constant term, which is omitted for space considerations. Robust standard errors are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively. It has been argued in previous sections that street names, as marks of cultural heritage, also capture the history of the population. Therefore, we would expect a positive association between religiosity in the past and the percentage of religious streets today. Following Oto-Peralías and Romero-Ávila (2016), we measure religiosity in the past as the percentage of population that was a member of the clergy (both secular and regular) at the end of the eighteenth century. Table 3 shows that this indicator of religiosity in the past is strongly and positively associated with our street-name indicator, both at bivariate level and when including control variables.17 All these results suggest that the street-name indicator actually captures the population’s religious attitudes, and that in turn it is partially determined by a long historical process. Table 3 Street-name indicator of religiosity: assessing validity at the province level (II) Dependent variable is street-name indicator of religiosity (1) (2) (3) (4) (5) (6) Religiosity in the eighteenth century 2.802*** 2.784*** 2.74*** 3.238*** 2.162** 1.947* (0.748) (0.718) (0.773) (0.835) (1.039) (1.104) GDP pc 2001 0.013 0.087 (0.107) (0.214) Latitude 0.136 0.166 (0.085) (0.138) Historical inequality −0.026 −0.033 (0.016) (0.035) Catalan −1.717** −2.404 (0.753) (1.748) Valencian 1.208 0.689 (1.322) (1.298) Basque −2.168* −4.609** (1.179) (1.806) Galician −2.283** −4.01** (1.021) (1.979) R2 0.21 0.21 0.23 0.24 0.32 0.46 Observations 50 50 50 50 50 50 Dependent variable is street-name indicator of religiosity (1) (2) (3) (4) (5) (6) Religiosity in the eighteenth century 2.802*** 2.784*** 2.74*** 3.238*** 2.162** 1.947* (0.748) (0.718) (0.773) (0.835) (1.039) (1.104) GDP pc 2001 0.013 0.087 (0.107) (0.214) Latitude 0.136 0.166 (0.085) (0.138) Historical inequality −0.026 −0.033 (0.016) (0.035) Catalan −1.717** −2.404 (0.753) (1.748) Valencian 1.208 0.689 (1.322) (1.298) Basque −2.168* −4.609** (1.179) (1.806) Galician −2.283** −4.01** (1.021) (1.979) R2 0.21 0.21 0.23 0.24 0.32 0.46 Observations 50 50 50 50 50 50 Notes: Variables descriptions are provided in Online Appendix Table A1. The estimations include a constant term, which is omitted for space considerations. Robust standard errors are in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively. Although the previous evidence seems convincing, it is still possible to argue that some unobservable province-level characteristics are driving the results. To address this concern, we test for the existence of a relationship between the street-name indicator and religious attitudes at the municipal level. This approach allows us to include province dummies to exclusively exploit within-province variation. However, a disadvantage here is the lack of specific indicators on religious attitudes, so it is necessary to use proxies. In this regard, the ratio of separated and divorced people to married ones is a suitable proxy, as the doctrine of the Catholic Church (by far the predominant religion in Spain) is against separation and divorce (Lehrer, 2004). Column 1 in Table 4 shows a highly significant and negative bivariate relationship between the previous ratio and the street-name indicator of religiosity.18 Column 2 includes province-fixed effects to control for all idiosyncratic province-level characteristics. The size of the coefficient decreases, but it remains highly statistically significant. Thus, within provinces, municipalities with a higher percentage of religious streets tend to have a lower incidence of separation and divorce. Columns 3–5 control for other relevant factors such as population, altitude, geographic location and demographic and socio-economic indicators. Notably, the coefficient remains negative, similar in size and highly statistically significant. Column 6 adds the annualized population growth rate over the period 1900–2001. This helps control for the possibility that cities that have grown faster, and consequently have a higher percentage of new streets, may have chosen more modern names for these streets, thereby reducing the percentage of religious streets. Again, the coefficient on the variable of interest is not affected by this additional control.19 Table 4 Street-name indicator of religiosity: assessing validity at the municipality level Dependent variable is ratio separated and divorced to married people (1) (2) (3) (4) (5) (6) (7) Street-name indicator of religiosity −0.0465*** −0.0187*** −0.0178*** −0.0174*** −0.0145*** −0.014*** −0.034*** (0.01) (0.006) (0.005) (0.005) (0.005) (0.005) (0.008) Population in 2001 0.0146*** 0.014*** 0.0112*** 0.0104*** 0.006*** (0.002) (0.002) (0.002) (0.002) (0.001) Population in 2001 squared −0.000005*** −0.000005*** −0.000004*** −0.000003*** −0.000002*** (0.000) (0.000) (0.000) (0.000) (0.000) Altitude −0.0005* −0.0003 −0.0002 −0.0009** (0.000) (0.000) (0.000) (0.000) Quadratic polynomial in latitude and longitude [0.0250] [0.2612] [0.2388] [0.5357] Average age −0.0208 −0.0094 0.0447 (0.013) (0.016) (0.03) Education 0.0224*** 0.0224*** 0.0568*** (0.006) (0.006) (0.01) Average socioeconomic condition 2.6311*** 2.6199*** 0.6777 (0.514) (0.515) (0.833) Annualized 1900–2001 population growth rate 11.7311 44.9654*** (8.036) (10.581) Province-fixed effects Yes Yes Yes Yes Yes Yes R2 0.02 0.32 0.33 0.33 0.36 0.36 0.66 Observations 8023 8023 8023 8023 8023 8022 3128 Dependent variable is ratio separated and divorced to married people (1) (2) (3) (4) (5) (6) (7) Street-name indicator of religiosity −0.0465*** −0.0187*** −0.0178*** −0.0174*** −0.0145*** −0.014*** −0.034*** (0.01) (0.006) (0.005) (0.005) (0.005) (0.005) (0.008) Population in 2001 0.0146*** 0.014*** 0.0112*** 0.0104*** 0.006*** (0.002) (0.002) (0.002) (0.002) (0.001) Population in 2001 squared −0.000005*** −0.000005*** −0.000004*** −0.000003*** −0.000002*** (0.000) (0.000) (0.000) (0.000) (0.000) Altitude −0.0005* −0.0003 −0.0002 −0.0009** (0.000) (0.000) (0.000) (0.000) Quadratic polynomial in latitude and longitude [0.0250] [0.2612] [0.2388] [0.5357] Average age −0.0208 −0.0094 0.0447 (0.013) (0.016) (0.03) Education 0.0224*** 0.0224*** 0.0568*** (0.006) (0.006) (0.01) Average socioeconomic condition 2.6311*** 2.6199*** 0.6777 (0.514) (0.515) (0.833) Annualized 1900–2001 population growth rate 11.7311 44.9654*** (8.036) (10.581) Province-fixed effects Yes Yes Yes Yes Yes Yes R2 0.02 0.32 0.33 0.33 0.36 0.36 0.66 Observations 8023 8023 8023 8023 8023 8022 3128 Notes: Variables descriptions are provided in Online Appendix Table A1. The estimations include a constant term, which is omitted for space considerations. Robust standard errors clustered at the province level are in parentheses. The p-values of the joint significance test for the quadratic polynomial in latitude and longitude are in brackets. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively. Column 7 restricts the sample to municipalities with more than 1,000 inhabitants. This is a way of mitigating the impact of measurement errors, as errors in identifying and quantifying religious streets have more severe consequences in small municipalities, where both the numerator and denominator in Formula (2) are smaller. It is reassuring that the effect of religiosity is now stronger, which is to be expected, as measurement errors create an attenuation bias. As a final check, we replicate the previous analysis using as dependent variable the percentage of de facto couples, which is also ‘anti-religious’ behavior, as reflected by the expression ‘living in sin’. The results are fully consistent with the previous ones.20 To sum up, the analysis of street names at the provincial and municipal level reveals the indicator of religiosity’s validity for capturing the underlying cultural variable it is supposed to measure, thereby supporting Conditions (1)–(3) for the case of Spanish street names. However, it is worth noting certain limitations of this econometric exercise. One is the intrinsic difficulty of measuring religiosity, particularly at the local level, which makes it difficult to evaluate how well the street-name indicator is capturing this cultural factor. Another is the ever-present issue of omitted variables, which cannot be completely ruled out in an observational study like this—although it can be argued that the endogeneity problem is less pervasive in this case because our purpose is not to document a causal effect but a correlation between several measures of religiosity. 4.3. Are more religious municipalities relatively poorer? This section uses the street-name indicator of religiosity to investigate whether more religious municipalities are relatively less developed. Previous work in the literature on religion and economic development generally finds a negative effect associated with Catholicism, which is the religious denomination captured by the indicator (e.g., La Porta et al., 1997; Inglehart, 1999; Stulz and Williamson, 2003; Arruñada, 2010).21 For reasons of data availability, the impact of religion has mainly been analyzed at the country or regional level, but barely at the city level. The street-name indicator allows us to conduct the analysis at this finer level. Our main indicator of local economic development is the average gross income of taxpayers in each municipality in 2013, which is available for municipalities of more than 1000 inhabitants. We also use other development indicators from census data with more geographic coverage, namely, average socioeconomic condition, average number of vehicles per household and the importance of the service sector. Columns 1, 3, 5 and 7 in Table 5 report a negative (within-province) relationship between religiosity and these indicators. Columns 2, 4, 6 and 8 add demographic and geographic controls such as total population in 2001 (both linear and squared terms), average age of the population, altitude and the quadratic polynomial in latitude and longitude. Notably, the coefficient on religiosity declines slightly but remains negative and almost always statistically significant.22 The rest of the columns analyze the effect of religiosity on proximate determinants of development, such as human capital and entrepreneurship. Religiosity also appears negatively related to both factors, although its relationship with entrepreneurship is more robust. This might suggest that the moral values conveyed through Catholicism are not drivers of an entrepreneurial culture. Although all this evidence is suggestive, it should be noted that its character is correlational. The main objective of this section has been to illustrate the usefulness of street names as a data source for analyzing the interaction between cultural factors and social and economic outcomes. Table 5 Religiosity and local economic development Dependent variable → Average gross income Average socioeconomic condition Average number of vehicles per household Service sector share (% of total employment) Education Non-agricultural entrepreneurs (% of total employment) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Street-name indicator of religiosity −81.8757*** −32.535* −0.0004** −0.0002 −0.0017*** −0.0008* −0.0818*** −0.0565** −0.063*** −0.0002 −0.0329*** −0.0263** (21.191) (16.607) (0.000) (0.000) (0.001) (0.000) (0.029) (0.025) (0.022) (0.016) (0.011) (0.012) Population in 2001 20.282*** 0.0002** −0.0011*** 0.0984*** 0.0464*** −0.0198*** (4.447) (0.000) (0.000) (0.013) (0.011) (0.003) Population in 2001 squared −0.006*** −0.0000001* 0.0000004*** −0.00003*** −0.00002*** 0.000007*** (0.002) (0.000) (0.000) (0.000) (0.000) (0.000) Average age −491.3685*** −0.0029*** −0.0233*** −0.432*** −1.423*** −0.1737*** (49.425) (0.001) (0.001) (0.081) (0.053) (0.041) Altitude −0.6552 0.0001*** 0.0000 0.0008 0.0021 0.003*** (0.599) (0.000) (0.000) (0.002) (0.001) (0.001) Quadratic polynomial in latitude and longitude [0.1186] [0.0001] [0.0132] [0.6240] [0.0003] [0.2956] Province-fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R2 0.51 0.63 0.56 0.58 0.46 0.63 0.20 0.26 0.35 0.55 0.06 0.09 Observations 2933 2933 8023 8023 8023 8023 8023 8023 8023 8023 8023 8023 Dependent variable → Average gross income Average socioeconomic condition Average number of vehicles per household Service sector share (% of total employment) Education Non-agricultural entrepreneurs (% of total employment) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Street-name indicator of religiosity −81.8757*** −32.535* −0.0004** −0.0002 −0.0017*** −0.0008* −0.0818*** −0.0565** −0.063*** −0.0002 −0.0329*** −0.0263** (21.191) (16.607) (0.000) (0.000) (0.001) (0.000) (0.029) (0.025) (0.022) (0.016) (0.011) (0.012) Population in 2001 20.282*** 0.0002** −0.0011*** 0.0984*** 0.0464*** −0.0198*** (4.447) (0.000) (0.000) (0.013) (0.011) (0.003) Population in 2001 squared −0.006*** −0.0000001* 0.0000004*** −0.00003*** −0.00002*** 0.000007*** (0.002) (0.000) (0.000) (0.000) (0.000) (0.000) Average age −491.3685*** −0.0029*** −0.0233*** −0.432*** −1.423*** −0.1737*** (49.425) (0.001) (0.001) (0.081) (0.053) (0.041) Altitude −0.6552 0.0001*** 0.0000 0.0008 0.0021 0.003*** (0.599) (0.000) (0.000) (0.002) (0.001) (0.001) Quadratic polynomial in latitude and longitude [0.1186] [0.0001] [0.0132] [0.6240] [0.0003] [0.2956] Province-fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R2 0.51 0.63 0.56 0.58 0.46 0.63 0.20 0.26 0.35 0.55 0.06 0.09 Observations 2933 2933 8023 8023 8023 8023 8023 8023 8023 8023 8023 8023 Notes: Variables descriptions are provided in Online Appendix Table A1. The estimations include a constant term, which is omitted for space considerations. Robust standard errors clustered at the province level are in parentheses. The p-values of the joint significance test for the quadratic polynomial in latitude and longitude are in brackets. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively. 5. Other potential applications: some examples This section briefly illustrates the versatility of street names to investigate other relevant topics that have attracted considerable interest from social scientists in recent years, including gender roles, national identity and transitional justice. 5.1. Gender roles Cultural values regarding gender roles constitute a central topic for social and economic development (e.g., Duflo 2012; World Bank, 2012). However, indicators measuring this factor are generally unavailable at the city level, particularly for a large sample of municipalities. Street names allow the construction of indicators of male predominance that can be used as a proxy for attitudes about the role of women in society. An indicator of male predominance can be defined as the percentage of streets with male names over the total number of streets with male and female names. An important issue here is to collect a list of male and female names. We use the 1000 most common forenames in both genders in Spain (INE, 2016b), and then complement the list with the names collected in Table 1.23 Panel A in Figure 4 shows the geographic distribution of the indicator’s values. At the national level, the percentage of streets named after men over the total named after men and women is 83.1%. If streets named after religious figures are removed, the percentage increases to 86.8%, which is a value close to that found in other studies that focus on specific cities.24 There are notable differences both between and within provinces, as given by the fact that the set of province dummies explains only 6% of the variation in the street-name indicator in a regression with municipal-level data. This suggests that an analysis at the local level could be fruitful to know more about the effect gender role attitudes have on socio-economic outcomes. Figure 4 View largeDownload slide Examples of other applications. (A) Male predominance. (B) Spanish identity. (C) Francoist streets. Note: A darker color indicates a higher value. Figure 4 View largeDownload slide Examples of other applications. (A) Male predominance. (B) Spanish identity. (C) Francoist streets. Note: A darker color indicates a higher value. 5.2. National identity Political identity and, more particularly, national identity are key political phenomena. National identity is one of the main factors of personal attachments (Rodrik, 2013), and can foster civic involvement (Huddy and Khatib, 2007). Local communities commemorate symbols, dates and events related to the nation they are part of in order to proclaim and emphasize their sense of belonging and identity. Therefore, street names can be used to study the intensity of national political identity. To illustrate this, we construct a street-name indicator measuring the intensity of Spanish identity at the municipal level by calculating the percentage of streets that contain words related to ‘Spanishness’. Panel B in Figure 4 shows the geographic distribution of values. According to this indicator, areas in central Spain tend to show a stronger Spanish identity, as does the region of Aragon (specifically the province of Zaragoza) in the north. By focusing on the local level, the study of the causes and consequences of national identity could help to provide new insights and evidence. 5.3. Memory, transitional justice and law enforcement An important topic in the literature on street names has been the use that political regimes make of the public space to commemorate their martyrs, heroes and glorious events in order to promote particular notions of national identity and history (Rose-Redwood et al., 2010). In this regard, the study of historical memory and attitudes toward transitional justice is an interesting application of street-name indicators. Concerning the Spanish case, in 2001 there were more than 2000 streets commemorating figures or events related to Franco’s dictatorship, and more than 1000 municipalities had at least one such street. Panel C in Figure 4 shows the geographical distribution of values of the percentage of Francoist streets. The presence of Francoist symbols raises the question of why so many municipalities still retain them in the twenty-first century. In other words, what factors explain the lack of support for simple measures of transitional justice such as the removal of Francoist symbols from the streets?25 An initial answer to this question would arguably be that the population in municipalities with Francoist street names has a more pronounced right-wing ideology. Indeed, we find a strong and highly significant correlation between the vote share of right-wing parties and Francoist streets.26 In addition, it would be interesting to study the impact of Spain’s 2007 Law of Historical Memory, according to which government entities must remove Francoist symbols from the public space. As there are still today at least 1237 streets with Francoist names,27 one can investigate the factors affecting the enforcement of this law, which is a relevant question for scholars interested in law enforcement in general. 6. Conclusions The wealth of information contained in street names has largely been ignored by quantitative empirical research. Street names are the result of historical processes, and reflect the population’s commemorative priorities (Rose-Redwood et al., 2010). They can be interpreted as a text from which we can read their social and cultural values. According to this view, this paper proposes the use of the city-text as a data source for constructing socio-cultural indicators for quantitative empirical analysis in social sciences. Given the scarcity of data at the city level, street names broaden the possibilities of empirical analysis. This is particularly convenient given the current emphasis on the causes and consequences of culture, and the advantages of working with city-level data. After discussing conceptual and methodological issues, we apply this paper’s premise to the study of religiosity and local economic development. First, we construct a street-name indicator of religiosity. Empirical evidence at the provincial and municipal level shows that this indicator can capture the population’s religious attitudes, which suggests that street names actually reflect the local community’s social and cultural values. Second, we show that this measure of religiosity is negatively correlated with several proxies for local economic development, which is consistent with previous work arguing that Catholicism (the main religion in Spain) may lead to worse economic outcomes. We further illustrate the versatility of street-name indicators in a variety of relevant topics, such as gender roles, national identity and transitional justice. The use of street names as a data source to create cultural indicators at the local level opens many possibilities for empirical analysis, including the study of the potential consequences on individuals’ behavior. However, researchers should be careful when applying this article’s idea. It is important to familiarize oneself with the main features of street names in the geographic area of study, to check whether the conditions discussed above are likely to be met and to be aware of the implications of measurement errors. Supplementary material Supplementary data for this paper are available at Journal of Economic Geography online. Footnotes 1 The use of city-level data helps to mitigate endogeneity problem by allowing the inclusion of regional (e.g., provincial) fixed effects. This makes the empirical analysis more convincing by exclusively exploiting variation within relatively small and homogeneous geographic areas. In addition, given the much larger sample size, it is possible to include more control variables, which also help address the omitted variable bias. 2 See, for instance, Putnam (1993) and Landes (1998). 3 Business names also provide more than one observation per municipality, but they do not reflect the community’s commemorative priorities. Instead, their purpose is commercial, and hence their utility to create cultural indicators is much more limited than street names. Indeed, their use in the literature has been confined to the study of regional identity. 4 There are a growing number of empirical studies on the interaction between culture and economics. See, for instance, Duranton et al. (2009), Falck et al. (2012), Maseland (2013), Nathan and Lee (2013), Bauernschuster and Falck (2015), Tubadji and Nijkamp (2015), Guiso et al. (2016), Spigel (2017), and Oto-Peralías and Romero-Ávila (2017). 5 As a reaction to this, there are social movements promoting a more egalitarian presence of women in toponyms. An example is the Italian project ‘Toponomastica femminile’ (for more details, see www.toponomasticafemminile.com). 6 Street names—even in the case of new inscriptions—are also the result of historical processes of cultural creation. 7 An analysis of the correlation between street-name indicators and the underlying socio-cultural factor helps validate Conditions (1)–(3) simultaneously. The reason is that a strong relationship in this regard means that ‘street names’ (first part of Condition 1) reflect ‘people’s social and cultural values’ (last part of Condition 3). 8 In the literature that uses business names to study regional and ethnocultural identity, it is also common to standardize by the total number of cases (Reed, 1976; and more recently, Ambinakudige, 2009; Cooper and Knotts, 2010; Cooper et al., 2011). 9 Measurement errors have more severe consequences for small municipalities than for large ones. The denominator, N, in Formula (1) is smaller, and so a variation in S due to an error will produce a larger variation in the indicator than for large municipalities. 10 In Spain, the country used here for illustrative purposes, around 1.5% of municipalities have more than 50,000 inhabitants, and only 0.7% more than 100,000. Similarly, in England and Wales only 1.3% of towns have more than 100,000 inhabitants (Office for National Statistics, 2013). Part of the difference between both countries is that the units of observation for Spain are municipalities, while for England and Wales they are built-up areas. 11 In regression analysis, both potential biases can be mitigated by controlling for city population growth during the previous decades and for dummy variables capturing whether the municipality is a capital or a large city. This is done in Section 4.2 when assessing the validity of the street-name indicator of religiosity. 12 To illustrate this point, consider that we are interested in counting the number of streets containing the forename ‘Juan’ (i.e., John in English) and suppose there is a street containing the name ‘Juana’ (i.e., feminine of Juan). If we simply search for ‘Juan’, we count the street name ‘Juana’ as having the search text. By contrast, if—as explained above—we add a blank space both to the street name and to the search text, now ‘ Juan ’ is not contained in ‘ Juana ’. 13 Articles and other very frequently used words (i.e., prepositions) are excluded. 14 Words are classified into categories according to their meaning and, importantly, according to the primary function they actually perform in street names. We have therefore revised the list of street names containing each word to check whether the classification is valid. When a word is impossible to classify because—for instance—50% of the times it works as vernacular and the other 50% as commemorative, we include it in Panel C (non-classified), which also contains conjunctions and single letters. 15 We have also checked that controlling for education and socioeconomic condition does not affect the coefficient of interest. These results, along with the sources, descriptive statistics and correlation tables of all the variables used in the article are available in the Supplementary Material (Online Appendix Tables A1–A8). 16 In other words, a one-standard-deviation increase in religious streets increases the percentage of people attending religious services every month by 0.46 standard deviation. 17 It is worth noting a similar relationship between religiosity in the past and in the present when the survey-based measure of religiosity is used as dependent variable. The standardized beta coefficients in the most saturated model (column 6) are very similar in both cases, being about 0.32–0.37 (see Online Appendix Table A9). This provides confirming evidence that religiosity is a persistent cultural factor that is well reflected in the street names. 18 As indicated above, we exclude a few observations with values higher than 50% in the indicator of religiosity to avoid these outliers exerting an excessive influence on the coefficient. 19 We have also checked that the results are robust to controlling for dummy variables capturing whether the municipality is a capital city or has more than 100,000 inhabitants, as they are more likely to have redeveloped their city center (see Online Appendix Table A10). 20 As an aside, one may wonder whether the street-name indicator of religiosity measures this cultural factor more accurately than the ratio of separated and divorced people to married ones and the percentage of de facto couples. This appears to be the case. The correlation at the provincial level of the survey-based indicator of religiosity (i.e., church attendance) with the street-name indicator (ρ = 0.52) is higher than with separated and divorced people (ρ = −0.41) and de facto couples (ρ = −0.44). When church attendance is regressed on these three variables, only the street-name indicator is statistically significant. These additional results are presented in Online Appendix Tables A11 and A12. 21 The negative legacy of Catholicism has been attributed to factors such as the hierarchical structure of the Catholic Church that undermines horizontal bonds and trust (Putnam, 1993), a culture of intolerance (Landes, 1998) or a historical hostility to lending on interest (Stulz and Williamson, 2003). 22 It is worth noting that the control set in Table 5 is different than in Table 4. This is because here we are trying to analyze the effect of religiosity on development, and it is important to avoid the inclusion of ‘bad controls’, that is, variables that can themselves be considered outcomes (Angrist and Pischke, 2009, Ch. 3). Indeed, total population in 2001 and the average age of the population could be considered bad controls, biasing the true value of the coefficient on religiosity. Online Appendix Table A13 replicates Table 5, but including only exogenous controls and total population in 1900. The effect is stronger in this case. 23 See the Supplementary Material for details about the construction of this indicator and the others used in this section. 24 For example, the percentage of streets commemorating a woman in Vienna is about 10% of all streets commemorating a person (see Hintermann and Pichler, 2015, and references therein). 25 See Aguilar et al. (2011) for a survey-based study of the determinants of attitudes toward transitional justice at the national level in Spain. 26 This result is reported in the Supplementary Material (Online Appendix 11, Tables A14 and A15), and provides additional support to this paper’s argument that street names reflect people’s values. 27 This figure corresponds to the January 2016 Electoral Census Street Map (INE, 2016a), and has been calculated using the same list of words as for 2001. These 1237 Francoist streets are distributed across 688 municipalities. Acknowledgements I am particularly indebted to the editor and referees of this journal for many valuable comments which led to a substantial improvement of the original manuscript. Financial support by the Carnegie Trust for the Universities of Scotland [grant no. 70433] is gratefully acknowledged. References Aguilar P., Balcells L., Cebolla-Boado H. ( 2011) Determinants of attitudes toward transitional justice: an empirical analysis of the Spanish case. Comparative Political Studies , 44: 1397– 1430. Google Scholar CrossRef Search ADS Alesina A., Giuliano P. ( 2015) Culture and institutions. Journal of Economic Literature , 53: 898– 944. Google Scholar CrossRef Search ADS Ambinakudige S. ( 2009) Revisiting “the south” and “dixie” delineating vernacular regions using GIS. Southeastern Geographer , 49: 240– 250. Google Scholar CrossRef Search ADS Angrist J. D., Pischke J.-S. ( 2009) Mostly Harmless Econometrics: An Empiricist’s Companion . Princeton: Princeton University Press. Arge S. V., Sveinbjarnardóttir G., Edwards K. J., Buckland P. C. ( 2005) Viking and medieval settlement in the Faroes: people, place and environment. Human Ecology , 33: 597– 620. Google Scholar CrossRef Search ADS Arnold C. L. ( 1977) Early Anglo-Saxon settlement patterns in southern England. Journal of Historical Geography , 3: 309– 315. Google Scholar CrossRef Search ADS Arruñada B. ( 2010) Protestants and Catholics: similar work ethic, different social ethic. The Economic Journal , 120: 890– 918. Google Scholar CrossRef Search ADS Augustins G. ( 2004) Naming, dedicating: street names and tradition. History and Anthropology , 15: 289– 299. Google Scholar CrossRef Search ADS Azaryahu M. ( 1996) The power of commemorative street names. Environment Planning D: Society and Space , 14: 311– 330. Google Scholar CrossRef Search ADS Azaryahu M. ( 1997) German reunification and the politics of street names: the case of East Berlin. Political Geography , 16: 479– 493. Google Scholar CrossRef Search ADS Barro R., McCleary R. M. ( 2003) Religion and economic growth across countries. American Sociological Review , 68: 760– 781. Google Scholar CrossRef Search ADS Bauernschuster S., Falck O. ( 2015) Culture, spatial diffusion of ideas and their long-lasting imprints—evidence from Froebel’s kindergarten movement. Journal of Economic Geography , 15: 601– 630. Google Scholar CrossRef Search ADS Becker S. O., Woessmann L. ( 2009) Was Weber wrong? A human capital theory of protestant economic history. Quarterly Journal of Economics , 124: 531– 596. Google Scholar CrossRef Search ADS Calvo-Iglesias M. S., Díaz-Varela R. A., Méndez-Martínez G., Fra-Paleo U. ( 2012) Using place names for mapping the distribution of vanishing historical landscape features: the Agras field system in northwest Spain. Landscape Research , 37: 501– 517. Google Scholar CrossRef Search ADS Cantoni D. ( 2015) The economic effects of the protestant reformation: testing the Weber hypothesis in the German lands. Journal of the European Economic Association , 13: 561– 598. Google Scholar CrossRef Search ADS CIS. ( 2005) Barómetro Autonómico (I). N° Estudio–2610. Centro de Investigaciones Sociológicas. Available online at: http://www.cis.es/. Conedera M., Vassere S., Neff C., Meurer M., Krebs P. ( 2007) Using toponymy to reconstruct past land use: a case study of ‘brüsáda’ (burn) in southern Switzerland. Journal of Historical Geography , 33: 729– 748. Google Scholar CrossRef Search ADS Cooper C. A., Knotts H. G. ( 2010) Declining Dixie: regional identification in the modern American south. Social Forces , 88: 1083– 1101. Google Scholar CrossRef Search ADS Cooper C. A., Knotts H. G., Elders K. L. ( 2011) A geography of Appalachian identity. Southeastern Geographer , 51: 457– 472. Google Scholar CrossRef Search ADS Drozdzewski D. ( 2014) Using history in the streetscape to affirm geopolitics of memory. Political Geography , 42: 66– 78. Google Scholar CrossRef Search ADS Duflo E. ( 2012) Women empowerment and economic development. Journal of Economic Literature , 50: 1051– 1079. Google Scholar CrossRef Search ADS Duranton G., Rodríguez-Pose A., Sandall R. ( 2009) Family types and the persistence of regional disparities in Europe. Economic Geography , 85: 23– 47. Google Scholar CrossRef Search ADS Falck O., Heblich S., Lameli A., Südekum J. ( 2012) Dialects, cultural identity, and economic exchange. Journal of Urban Economics , 72: 225– 239. Google Scholar CrossRef Search ADS Ferguson P. P. ( 1988) Reading city streets. The French Review , 61: 386– 397. Fuchs S. ( 2015) An integrated approach to Germanic place names in the American midwest. The Professional Geographer , 67: 330– 341. Google Scholar CrossRef Search ADS Giraut F., Houssay-Holzschuch M. ( 2016) Place naming as Dispositif: toward a theoretical framework. Geopolitics , 21: 1– 21. Google Scholar CrossRef Search ADS González Faraco, J. C., Murphy M. D. ( 1997) Street names and political regimes in an Andalusian town. Ethnology , 36: 123– 148. Google Scholar CrossRef Search ADS Guiso L., Sapienza P., Zingales L. ( 2003) People’s opium? Religion and economic attitudes. Journal of Monetary Economics , 50: 225– 282. Google Scholar CrossRef Search ADS Guiso L., Sapienza P., Zingales L. ( 2006) Does culture affect economic outcomes? Journal of Economic Perspectives , 20: 23– 48. Google Scholar CrossRef Search ADS Guiso L., Sapienza P., Zingales L. ( 2016) Long-term persistence. Journal of the European Economic Association , 14: 1401– 1436. Google Scholar CrossRef Search ADS Hintermann C., Pichler H. ( 2015) Gendered spaces in the city: critical topography in geography education. Journal for Geographic Information Science , 1: 287– 298. Huddy L., Khatib N. ( 2007) American patriotism, national identity, and political involvement. American Journal of Political Science , 51: 63– 77. Google Scholar CrossRef Search ADS INE. ( 2016a) Electoral Census Street Map. Instituto Nacional de Estadística. Available online at: http://www.ine.es/. INE. ( 2016b) Most frequent names and surnames (Birth Statistics). Instituto Nacional de Estadística, Available online at: http://www.ine.es/. Inglehart R. ( 1999) Trust, well-being and democracy. In Warren Mark (ed.) Democracy and Trust , pp. 88– 120. New York (NY): Cambridge University Press. Google Scholar CrossRef Search ADS La Porta R., Lopez-de-Silanes F., Shleifer A., Vishny R. W. ( 1997) Trust in large organizations. American Economic Review , 87: 333– 338. Landes D. ( 1998) The Wealth and Poverty of Nations . London, UK: Little, Brown, and Company. Lehrer E. L. ( 2004) The role of religion in union formation: an economic perspective. Population Research and Policy Review , 23: 161. Google Scholar CrossRef Search ADS Liesch M., Dunklee L. M., Legg R. J., Feig A. D., Krause A. J. ( 2015) Use of business-naming practices to delineate vernacular regions: a Michigan example. Journal of Geography , 114: 188– 196. Google Scholar CrossRef Search ADS Light D. ( 2004) Street names in Bucharest, 1990–1997: exploring the modern historical geographies of post-socialist change. Journal of Historical Geography , 30: 154– 172. Google Scholar CrossRef Search ADS Loy W. G. ( 1989) Geographic names in geography. Yearbook of the Association of Pacific Coast , 51: 7– 24. Google Scholar CrossRef Search ADS Maseland R. ( 2013) Parasitical cultures? The cultural origins of institutions and development. Journal of Economic Growth , 18: 109– 136. Google Scholar CrossRef Search ADS Nathan M., Lee N. ( 2013) Cultural diversity, innovation, and entrepreneurship: firm-level evidence from London. Economic Geography , 89: 367– 394. Google Scholar CrossRef Search ADS Office for National Statistics. ( 2013) 2011 Census: Characteristics of Built-Up Areas. Available at http://www.ons.gov.uk/ons/rel/census/2011-census/key-statistics-for-built-up-areas-in-england-and-wales/index.html. Oto-Peralías D., Romero-Ávila D., ( 2016) The economic consequences of the Spanish Reconquest: the long-term effects of Medieval conquest and colonization. Journal of Economic Growth , 21: 409– 464. doi:10.1007/s10887-016-9132-9. Google Scholar CrossRef Search ADS Oto-Peralías D., Romero-Ávila D. ( 2017) The consequences of persistent inequality on social capital: a municipal-level analysis of blood donation data. Economics Letters , 151: 53– 57. Google Scholar CrossRef Search ADS Palonen E. ( 2008) The city-text in post-communist Budapest: street names, memorials, and the politics of commemoration. GeoJournal , 73: 219– 230. Google Scholar CrossRef Search ADS Putnam R. ( 1993) Making Democracy Work. Civic Traditions in Modern Italy . Princeton: Princeton University Press. Qian S., Kang M., Wang M. ( 2016) An analysis of spatial patterns of toponyms in Guangdong, China. Journal of Cultural Geography , 33: 161– 180. Google Scholar CrossRef Search ADS Reed J. S. ( 1976) The heart of Dixie: an essay in folk geography. Social Forces , 54: 925– 939. Google Scholar CrossRef Search ADS Rodrik D. ( 2013) Roepke Lecture in Economic Geography—who needs the nation-state? Economic Geography , 89: 1– 19. Google Scholar CrossRef Search ADS Rose-Redwood R., Alderman D., Azaryahu M. ( 2010) Geographies of toponymic inscription: new directions in critical place-name studies. Progress in Human Geography , 34: 453– 470. Google Scholar CrossRef Search ADS Scott A. J. ( 2004) A perspective of economic geography. Journal of Economic Geography , 4: 479– 499. Google Scholar CrossRef Search ADS Spigel B. ( 2017) Bourdieu, culture, and the economic geography of practice: entrepreneurial mentorship in Ottawa and Waterloo, Canada. Journal of Economic Geography , 17: 287– 310. Stulz R. M., Williamson R. ( 2003) Culture, openness, and finance. Journal of Financial Economics , 70: 313– 349. Google Scholar CrossRef Search ADS Trounstine J. ( 2009) All politics is local: the reemergence of the study of city politics. Perspectives on Politics , 7: 611– 618. Google Scholar CrossRef Search ADS Tubadji A., Nijkamp P. ( 2015) Cultural gravity effects among migrants: a comparative analysis of the EU15. Economic Geography , 91: 343– 380. Google Scholar CrossRef Search ADS Tucci M., Ronza R. W., Giordano A. ( 2011) Fragments from many pasts: layering the toponymic tapestry of Milan. Journal of Historical Geography , 37: 370– 384. Weaver R., Holtkamp C. ( 2016) Determinants of Appalachian identity: using vernacular traces to study cultural geographies of an American region. Annals of the American Association of Geographers , 106: 203– 221. Google Scholar CrossRef Search ADS Weber M. ( 1976 [1904–1905]) The Protestant Ethic and the Spirit of Capitalism . London: Allen and Unwin. World Bank ( 2012) World Development Report 2012: Gender Equality and Development . Washington, DC: The World Bank. © The Author (2017). Published by Oxford University Press. All rights reserved. For Permissions, please email: firstname.lastname@example.org
Journal of Economic Geography – Oxford University Press
Published: Jan 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.
All for just $49/month
Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.
Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.
It’s easy to organize your research with our built-in tools.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera