https://orcid.org/0000-0002-1984-9616
![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
Obtaining geocoded electoral results at precinct level can be challenging in many countries. In this study, we present two validated methodologies developed to overcome these difficulties, and we build new geocoded electoral results for several Italian cities. Our dataset covers the last 20 years (1999-2022) and includes data for various types of elections, including national, regional, municipal and referendum. We provide an overview of some notable patterns in voting trends in major Italian cities. These include a high level of heterogeneity in voting within cities, an increase in spatial polarization of voting behaviour, and an increasing concentration of left-wing voters in central and wealthier areas of metropolitan cities. These trends may be influenced by a range of factors and can have significant implications for political representation and policy-making. Our dataset provides a valuable resource for understanding these trends and exploring their underlying causes.
Data availability statment
The data that support the findings of this study are openly available in a github repository that can be accessed at this link https://github.com/gabrielepinto/dati-sezioni-elettorali
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 For international dataset of elections see the CLEA project (Kollman et al. Citation2019). For Italy, see the Eligendo Project (Dell’Interno Citation2021).
2 Until a few years ago, precinct-level electoral data were also scarce outside Europe. However, there have been significant contributions to collecting this type of data in the United States (Baltz et al. Citation2022; Voting and Election Science Team Citation2020).
3 In Italy, electoral precincts correspond to ‘sezioni elettorali' or ‘electoral sections'. They are the smallest units into which electoral districts are divided and comprise between 500 and 1,200 voters.
4 We also provide:
1) A repository with the whole dataset and a Python package that can be used to reproduce the study and to replicate the methodology for other cities and countries (https://github.com/gabrielepinto/dati-sezioni-elettorali).
2) A user-friendly dashboard with interactive maps to explore the dataset (https://gabrielepinto-dashboard-sezioni-elettorali-appstreamlite-4t5mbj.streamlit.app/).
5 In the next section we will discuss circumstances that are specific to Italy. However, Willis et al report very similar experiences for the United States (Willis, Merivaki, and Ziogas Citation2021).
6 There are some contributions on this issue for the United States, mainly from a series of publications by Brian Amos and Michael McDonald. However, none of these specifically addresses the methodology for reconstructing precinct boundaries (Amos, McDonald, and Watkins Citation2017; Amos, Smith, and Claire Citation2017; Baltz et al. Citation2022; Voting and Election Science Team Citation2020). See also: https://openprecincts.org/about/.
7 The most important contribution on this type of data in Italy is from Keti, Monni and Tomassi, authors of the ‘mapparoma' project (https://www.mapparoma.info/). These authors provide electoral data (for the city of Rome only) for ‘urban zones' (‘zone urbanistiche’) (Lelo, Monni, and Tomassi Citation2021). In addition to expanding the coverage to other cities, we have also significantly increased the level of granularity of the data. This is 16 times greater (the city of Rome is divided into 155 ‘zone urbanistiche’ and 2,600 ‘sezioni elettorali'). Another important contribution that should be mentioned is the work by Corbetta and Piretti (Corbetta and Piretti Citation2009), who have created a historical atlas of Italian elections (1860 to 2008). However, their data are generally aggregated at municipality level.
8 For instance, see the obstacles to the collection of precinct-level data in the US (Willis, Merivaki, and Ziogas Citation2021).
9 According to a request for clarification (accesso civico), the Ministry of the Interior’s Internal and Local Affairs Department (Dipartimento affari interni e territoriali) is currently running an experimental program to check the feasibility of collecting precinct results from municipalities. No data have yet been made available by the Ministry. In the absence of a centralized archive and coordinated standards, collection of all precinct data requires an examination of 7,980 archives (almost the total number of Italian municipalities).
10 To clarify, each voter is assigned to a precinct (sezione), and can only vote in a specific corresponding polling station (seggio elettorale) that is located in a polling place (locale del seggio elettorale) representing multiple polling stations. The polling station is a specific room where only voters from a certain precinct can vote. The polling station is located in a building that constitutes the polling place, where there are multiple polling stations.
11 Article 34 of Presidential Decree No 223 of 20 March 1967.
12 Article 34 of Presidential Decree No 223 of 20 March 1967.
13 Article 40 of Presidential Decree No 570 of 16 May 1960; Article 48 of Presidential Decree No 361 of 30 March 1957; Article 51 of Presidential Decree No 361 of 30 March 1957.
14 Both the Presiding Officer and the poll clerks are appointed from among citizens enrolled in specific registers (‘albo degli scrutatori' and ‘albo dei presidenti di seggio').
15 Some municipalities do offer an electoral precinct shapefile. The largest of these are Florence and Rimini.
16 For this reason, the procedure we have described here could be reproduced by aggregating voter list files (liste elettorali). The voter list file for Italian municipalities should be accessible for research purposes under Italian law. However, in practice, it is difficult to obtain these files in digital format. We tried to request the voter list file from the electoral office in Rome but without success.
17 Using the address shown in as an example, for the first observation, we submitted the query with the name of the street ‘Via Abano Terme, Roma, Italy' to the API for the geocoder. The API returned the location of the street with latitude and longitude. In some cases, a unique street corresponds to multiple precincts depending on the street number. For instance, in the example above (), ‘Via Abbiate Grasso' corresponds to multiple precincts: odd street numbers from 1 to 61 and even street numbers from 2 to 68 vote in precinct 2392, while all other residents of ‘Via Abbiate Grasso' vote in precinct 2391. In those cases, we submitted four different queries to the API: ‘Via Abbiate Grasso 1', ‘Via Abbiate Grasso 2', ‘Via Abbiate Grasso 68' and ‘Via Abbiate Grasso 61' for precinct 2392, and ‘Via Abbiate Grasso 70' and ‘Via Abbiate Grasso 63' for precinct 2391.
18 Clearly wrong geo-encoding occurs when 1) the coordinates received are outside the administrative boundaries of the municipality, or 2) the coordinates are extremely far from other points of the same precinct (we set the bar at three times larger than the mean distance between all other points of the same precinct). In Appendix 2 we have provided a comprehensive explanation of the process we have applied (Figure A-2, A-3 and A-4).
19 For example, in the Electoral Roadmap Dataset for the municipality of Rome, there are five precincts (with IDs from 2413 to 2417) that all correspond to only one unique address: Largo dell’Olgiata 15.
20 The address we geo-encoded is the ‘postal address' of a voter. This might not coincide with the voter’s residence if the street entrance is some distance from where the house is located.
21 In principle, both electoral precincts and census blocks are created from a dataset of addresses and street numbers. In theory, they should be reconcilable. However, the two procedures are carried out with clearly different objectives: census blocks are created for statistical purposes, while electoral precincts are only created to assign voters to polling stations in line with the limit imposed by existing legislation (as discussed in the first section of this paper). In a private conversation with municipal officers, those individuals confirmed to us that the two procedures are performed independently. The only exception is the boundaries of electoral constituencies (‘collegi elettorali'). In fact, when the new electoral law for parliamentary elections came into force in 2018, the National Institute of Statistics (ISTAT) had to design the new boundaries for the constituencies (Commissione Collegi Elettorali Citation2020). When the commission responsible needed to assign multiple constituencies to one municipality, it used statistical sub-units within the municipality. For instance, in the case of Rome, the commission used the boundaries of the ‘Zone Urbanistiche’, which represent an aggregation of census blocks. For this reason, the boundaries of electoral constituencies will match the boundaries of census blocks. However, in some cases, an electoral precinct could overlap different electoral constituencies. In such cases, the law envisages that the precinct be assigned to the constituency where the polling place is located (Article 3(2) of Legislative Decree No 189/2017).
22 Rimini, Florence and Genoa.
23 Available here: https://opendata.comune.rimini.it/dataset/sezioni-elettorali1.
We chose the municipality of Rimini (and not Florence or Genoa) because that municipality made available both the GIS shapefile and the Electoral Roadmap Dataset. We also had several personal contacts with the local office of the municipality who provided helpful explanations about the data. That was not the case for the municipality of Florence, which did not respond to our requests for clarification.
24 This is the clarification we received from the office: ‘The shapefile was drawn manually following the electoral roadmap, where the streets names and numbers belonging to each section are indicated. Because the borders do not pass through the center line, they are not absolutely precise and must be reviewed periodically, since new constructions in the map could be shown visually in another precinct' (translation from Italian to English by the author).
25 For example, the diameter of the circle that contains the entire area of the municipality.
26 In larger precincts the Centroid Distance metric is higher. The distribution of the metrics is right-skewed, with the mean larger than the median because of the presence of large extreme values.
28 We classified all parties competing in national elections into four main groups: left-wing, right-wing, Five-Star Movement (M5S) and centrist. In the ‘Appendix Table A-1', we have provided a full list of parties for each group.
29 The term ‘Partito delle ZTL' became popular after it was mentioned by the journalist Massimo Giannini on a political talk show (Di Martedì) on 13 March 2018 (just after the elections on 4 March). However, the term had already appeared in some twitter discussions from that time, and it seems that its coining can be attributed to the journalist Leonardo Panetta (Mediaset): https://twitter.com/marcobreso/status/972047046332035072.