Wikidata:WikiProject elections/Ontology proposal for Infobox election

With the creation of {{Infobox election}}, first in cawiki and, from 2020 summer, installed here in WD platform, the ontology for the "election items" has evolved trying to solve: the different kind of elections, the rational use of information in an infobox and the several levels of grouping or represent the information related with the "event of an election".

This is not the only aspect of an election; for instance, the information in a personal item changes with P39 property, or political party with P1410 are also a consequence of it. However, this has not been the main target of this topic.

This page gathers the design done until now and the decisions made in order to be discussed and tuned, if necessary.

Item levels

Elections understood as the act of voting and obtaining results, can be seen from several vectors that affect the items we need to structure the information. For example, a two-round election requires at least one item for each round, that is, for each "election moment"; legislative elections with some elected for each constituency, will require an item to collect the overall result that forms the chamber object of the election, but if we want to be able to express the result obtained for each constituency, we must have an item with candidates, results and elected at this level by each constituency. Another vector to consider is aggregations of results, often better known than baseline information, such as aggregate results by region or state. The following table presents a 4-level scheme of information structuring with a similar data model in order to harmonize the ontology and facilitate its use in searches or other types of subsequent exploitation.

Theoretical scheme

Each kind of items contain the whole specific information of its level. They are autonomous without depending on the existence of other level items of the same election.

1) Green & Blue levels should exist. Sometimes are the same (i.e. local elections)

2) Yellow levels are optional and can be multiple

3) Salmon level may exist in special double event. Infobox manages just a few cases

4) To avoid long list , only exists P361 (link to parent) between levels

Local

Presi- dentials

Legislatives

Mayor

Deputies

Parliament

Senate

A

"Upper aggregated level" (statistical)

?

x

G

"Groupped elections" (+1 election together)

maybe

maybe

maybe

maybe

P361 (↑)

P527 (↓) + P361 (↑)

O

"Operative level" (where the seats work)

same as C

same as C

x

x

x

P527 (↓)

"Two rounds parts"

P361 (↑)

I

"Intermediate aggregated level" (statistical)

no exists

x

x

x

P361 (↑)

C

"Constituency level" (where seats are won)

same as O

same as O

x

x

x

Description of the levels

Level O

This level is what we most commonly find, as it corresponds to the final result for which the elections have been held, either a direct result, such as direct suffrage to elect a mayor in a single vote, or the aggregation of seats from various constituencies to form a parliament. This level should exist for all elections. Ex: 2015 Catalan regional election (Q17082048) (second case in the example) is functionally the same as 2014 Indian general election (Q3587169) (third case in the example); both are parliamentary elections where those in Catalonia only have this item -without the results of the constituencies- and those in India have the complete sequence, as we will explain later. Although we should NOT use has part(s) (P527) to incorporate items of lower territorial (or jurisdiction) levels, there may be a P527 to contain any temporary subdivisions that may exist, as in the case of two-round elections.

Level C

This items collect the lowest level information in which a seat is generated. This description does not include lower levels that only provide votes but do not generate a seat, such as a neighborhood or a polling station. The information structure is similar to type "O" by adjusting in P541 the number of seats for the constituency. It must have the P361 to indicate the parent item of level "C", thus allowing to obtain a list of results of "O" by constituencies via SPARQL. In cases of singular direct election (mayor, president, ...) and unique constituency, a single item performs the functions of "C" and "O".

Level A & I

These are extra levels of aggregation that do not necessarily correspond to formal levels of the electoral process, but to usual statistical groupings in the social or cultural context in which elections occur. For example, there may be a "C" level with the result of a region in a state parliamentary election (eg: 2014 Indian general election in Andhra Pradesh (Q16874523), the second case in the example). An "A" level case is the total result of municipal councilors at the region or state level (eg: the second and third cases). All 3 examples have article in WP, as they correspond to "real events", although they only reflect information obtained in lower-level "electoral actions".

Level G

Often an election day serves to group several elections: mayor + councilors, lower house and upper house, and so on. Even the name of grouped elections is more representative than the acts that make it up, such as: "General elections of <state name>". The item at this level collects the basic data common to all of them and links them using P527. It should not contain aggregate results, as the typology of elections that are part of it is usually heterogeneous.

Linking

To link different items of an election we will use part of (P361), it is, linking just to parent item. The reasons for limiting the use of the has part(s) (P527) are:

1. there are elections with hundreds of constituencies and it is ontologically incorrect, and
2. it is very likely that the lower level information is not 100% complete. For this reason, each level are autonomous in terms of information and it is not necessary to have to drill down.

There are two exceptions:

• The items of each of the votes that make up a two-round election
• Election items grouped within a "G" level item, as described above.

To navigate through the history of a given election, we can follow the follows (P155) & followed by (P156) and move along the same level. In the case of each round item in a two-round election, the first will have the P156 pointing to the second and no value in its P155; the item of the second round will be inverse, that is, the P155 will aim to the first round and the P156 will have no value.

Key properties by election type

The following table compares the properties related to the results of candidates and elected, in order to better understand the operation of the different classes of items and types of election.

Item level = O+C represents a unique item for both functions, as described above. The "case" column has links to the examples and the item that serves as a model.

P726: Can be filled in before election day. However, it can often represent a long number of entries that will not be elected and may even be entities that do not exist (nor will ever exist) in WD. Their qualifiers describe the circumstances of the candidate for this election: represents (P1268) (or member of political party (P102)), political coalition (P5832), party chief representative (P210) to fill with the lead candidate (Q42417559) when candidate is political party, and some others. After the vote, the votes received (P1111) are incorporated and, according to the rules of the type of voting, it will justify an entry for this candidate in the successful candidate (P991). Sometimes this upload is heavy and some editors choose to enter information after the election directly into the P991. The infobox knows how to handle this, although it doesn't seem advisable, as it can affect other forms of access.

P991: Must contain only candidates who, in accordance with the rules of the voting type, have been elected. When the type of voting is party-list representation, the qualifier number of seats in assembly (P1410) must be included. Sometimes some of the qualifiers (P1111, P1268, ..) of the P726 are replicated to facilitate some queries. They are NOT needed for the infobox, if they already exist on the P726.

P31: The value given is required, but not exclusive. It may have other values in addition.

P8196: It should be mandatory, although very few entries have it. The infobox has an algorithm to deduce it, but it can produce incorrect results if the data structure is not complete and correct.

Type election	Item level	Case	Candidates		Identification		Vote to person / list	Elected singular	Notes
			P726	P991	P31 (specific)	P8196
direct (presidential)	O+C	Q1658720 ex. 1	Person P1111 + P1268 + P1352	Person	public election (Q40231) or subclasses	first-past-the-post voting (Q5557375)	Person	Yes	In this example, the list of candidates is ordered with qualifier ranking (P1352) according to results. It's not necessary for Infobox, because it sort by number of votes.
direct + Electoral college (presidential USA)	O+C	Q699872 ex. 1	Person P1111 + P1410 (suport E.college)	Person P1111	public election (Q40231) or subclasses	None	Person	Yes	In this example, P1410 does not represent the seats won (no seats are obtained), but the support from the candidate's delegates. The P1111 are the popular votes.
direct (local councillors)	O+C	Q61868584 ex. 3	Void. Uploaded in P991 post-election	Person P1111 + P102 + P1352 + P5832	public election (Q40231) or subclasses	None	Person	No	When the information of candidates has not been entered before the election day, from the point of view of the operation of the infobox, it is possible to choose to load, both the votes and the seats (when it touches), in the P991.
indirect (president elected by parlamentaries)	O+C	Q23558762 ex. 2	Person P1111 it uses P710 with list of voters !	Person P1111	public election (Q40231) or subclasses	indirect election (Q877353)	Person	Yes	In this example the nominal list of the 1239 parliamentarians who have voted in favor of the candidate has been added, when it would be more interesting to know the meaning of the vote (for, against or abstention) of each parliamentary group, as shown in Q97659738
limited multi-vot (constituency for senate) valid for Multiple non-transferable vote	C	Q75169736 ex. 1	Person P1111 + P1268	Person P1111 + P1268	public election (Q40231) or subclasses	limited voting (Q3786779)	Person	No	These two examples show the constituency level element (C) and that of the election chamber (O), in a case of nominal multi-vote. In the constituency, the sum of the votes of all the candidates is higher than the ballots cast (P1868) and therefore the proportion of each candidate describes the support they have had, but not the distribution, as the sum can be> 100%. At the next level (O), which corresponds to the whole chamber, it works with seats grouped by party to avoid long nominal lists. Total support is not obtained from the sum of the votes, as the figures add up to more than the votes distribution generating inconsistent ratios. For this reason, the ratio that the infobox shows in multi-vote elections is obtained from the distribution of seats between parties (P1410)
limited multi-vot (senate) valid for Multiple non-transferable vote	O	Q67795085 ex. 2	Party P1111 (aggregated)	Party P1410 (aggregated)	public election (Q40231) or subclasses	limited voting (Q3786779)	Person	No
limited multi-vot (senate within "general elections") valid for Multiple non-transferable vote	G	Q65121437 ex. 3	None	None	group of elections (Q76853179)	None	N/A	No	This is a "G" level item where there is no P726 or P991 information, but it is taken from the items listed on has part(s) (P527). In this case, one of them would be the Q67795085 of the immediately preceding case
two-rounds (French presidentials)	O global	Q7020999 ex. 3	Person P3831 + P5832 (+ P1111, if majority at first vote)	Person	public election (Q40231) or subclasses	two-round system (Q615255)	Person	Yes	In this example, the list of candidates is long because it originated before the primary election and the qualifier object has role (P3831) has been used to indicate their role. Elections in two rounds usually consists of 3 items: election (global) + first round + second round. However, if in the first round one candidate obtains an absolute majority, there will only be one item, as if it were an ordinary election. The recommended procedure is: create, before the elections, the global item with the P726 and thus be able to have information of the candidates. If in the first ballot one candidate obtains an absolute majority, it will only be necessary to load the qualifiers P1111 and P991 with the elected person name in the global item. On the other hand, if a second round is needed, we will do: A) create new first round item, B) move (or copy) the P726 from the global item to the new first round item, and C) then upload to that item the P1111 qualifiers of all candidates and the P991 of those who pass in the second round. D) then create the second round item by putting as P726, only the candidates who pass in the second round. When second round happens, just must to enter in the corresponding item, the P1111 for the finals candidates and P991 of final winner.
	O 1st. round	Q29836873 ex. 3	Person P1111	Person	voting round (Q24097670)	Optional: two-round system (Q615255)	Person	No
	O 2n. round	Q29837047 ex. 3	Person (only 2) P1111	Person	voting round (Q24097670)	Optional: two-round system (Q615255)	Person	Yes
party-list proportional representation (regional parliament)	O+C	Q17082048 ex. 2	Party P210 + P1111	Party P1410	public election (Q40231) or subclasses	party-list proportional representation (Q31764)	List	No
party-list proportional representation (Local elections)	O+C	Q20977896 ex. 1	Party P210 + P1111	Party P1410 Optional Person P1268	public election (Q40231) or subclasses	party-list proportional representation (Q31764)	List	No	Example of a municipal election. The basic level is the constituency and the elect will occupy the positions of the constituency. Therefore, only one item meets the "C" and "O" level. Because the choice type is party-list proportional representation, the information in P726 and P991 is by party. Should there be personal information of the elect ?. In this case, it is informed, although the infobox does not use it. It is a topic open to proposal and debate. The other two levels, "I" and "A" are aggregations used in the statistical, communicative and social field to understand the level of territorial presence of a party, although they do not respond to any level of political management.
	I	Q100707776 ex. 2	Party P1111	Party P1410	Summary data (Q100807399)	Optional: party-list proportional representation (Q31764)	List	No
	A	Q30279382 ex. 3	Party P1111	Party P1410	Summary data (Q100807399)	Optional: party-list proportional representation (Q31764)	List	No
party-list proportional representation with Parlamentary Group Constitution (Q101421918) (European Parliament case)	C	Q50393759 ex. 1	Party P210 + P1111	Party P1410	election to the European Parliament (Q1128324) or subclasses	party-list proportional representation (Q31764)	List	No	Example of a European Parliament election. The basic level (C) correspond to the states. Each state has a predetermines number of seats, which are assigned to a national parties following the state-rules election. The elected persons of all the states constitute the chamber of the Parliament attached to the Parliamentary Groups, not on behalf of their state. Therefore, the item of the Parliament (O) level is configured with the results of the national parties as in P726 (ie the input to the EP) and with the results of the Groups -after the reassignment- located in the P991.
	O	Q16999180 ex. 2 (limited to 1 state in this example until to have an agreement on ontology)	Party P1111 + P1410 + P4100	parliamentary group (Q848197) P1410	election to the European Parliament (Q1128324) + including (P1012) = Parlamentary Group Constitution (Q101421918)	Optional: party-list proportional representation (Q31764)	List	No

Frequent and/or Open questions

Some of the following points answer the Wikidata section: Wikidata:WikiProject_elections#Questions_to_solve.

Personal information. Person-oriented or party-oriented candidates

In this topic we must consider two vectors to make the analysis:

1. The first is whether it is an election where you vote, a person-oriented or party-oriented. If the vote is for one person, it makes sense to have the names of the candidates and elected officials on the election item. If the vote is party-oriented, it will be necessary to decide whether it is convenient to have information on the candidates and electors, which one and where it is located.
2. The second vector has to do with the number of people entries to have, in either of the two options described in the previous point. Proper design at this point should avoid the existence of long lists contrary to the ontology. Similarly, it is not advisable to incorporate data from potential candidates who will hardly be elected and thus avoid having to create items of people of whom we have little information and who would undoubtedly be eligible to be in WD.

Person-oriented case

It may be of singular result (only one elected person), such as presidents, mayors, etc. in a direct election, or also to occupy a limited number of seats among people on an open list, or among candidates nominated by a party, although the selection is per person. It is usually done with a “single non-transferable vote” or “limited voting” system, but not only.

• The information with the names of the candidates and elected officials must appear in the item of the constituency (C) which, let us remember, is where the seat is obtained.
• How many instances should there be? At a minimum, all elected must be in both P991 (elected) and P726, in which votes must be reported (P1111).
• Can only P991 be used and avoid duplicating the contents of P726, when the information is entered after the election ?. It is not correct, as the lack of homogeneity makes it difficult to treat. However, the infobox knows how to deal with it.
• Should there be instances for other candidates who have not won a seat? If we fill out the P726 before the election, we will probably be able to enter the information of other candidates who will ultimately not be elected. It would be advisable to be careful and avoid producing long lists to want to be exhaustive. Remember that the data of the "unelected" will only have a documentary value, but not operational, as they do not get any position. An alternative option could be to save the information of the "unelected candidates" using the P3602 in the person's item.

• For example in 2020 United States presidential election (Q22923830) we have the list of all the official candidates of each party and some of those who have been withdrawn as they lacked support (the end date of the election race is indicated). It is a small number, as the election is a unique seat and can start with 1 or 2 candidates per party. In the case of multiple seats like April 2019 Spanish Senate election in Barcelona (Q63825349) there were 23 candidates for 4 seats, the number of candidates could probably have been reduced by limiting it, for example, to those who had better options in the polls during the campaign.

• If no item has been created at the constituency level (C), can personal information be put on the item at the operational level (O)? The answer is NO, except in cases where the C and O levels are the same. Level O items that are the aggregation of multiple constituencies have too many seats to have the nominal list at that level.
• What personal information should be in the items above level C ?. No information. At these levels only the aggregate votes and seats per party should be listed.

Party-oriented case

In the case of elections with voting on party lists, the personal information of the members of the list is not essential to the election items. Often the P726s in these items contain the name of the party’s candidacy leader in the P210 qualifier, it's correct, but not mandatory. Instances of P991 can also be found with elected officials and party instances mixed. This structure is a bug and should be fixed. The information of the elected persons must appear in the P39 of the item of the elected person that should have the qualifier P2715 pointing towards the item of the elections. If you want to have the list of elected people without having to do a SPARQL of the P39, it would be more reasonable, although equally redundant, to do it alternatively:

• create a list item, such as "elect list to ...." with all elect in P527 or P710
• locate the list of members within the item of the legislature of the chamber. This solution could be equally inadequate in chambers with hundreds of deputies
• locate the list of names in the item of the parliamentary group, in case there is an item for the legislature.
• If none of the previous options are available, we really need to have personal names in the election item and the number of elected are only a few, they can be incorporated as a P991 qualifier of the party’s P991 instance.

 To be decided

Item of lower level than the constituency

Can we have items with information below the constituency level (C)? Yes they can exist, although it will NOT have the P991, as the elected seats appear at the constituency level. Therefore, it may be interesting to have this information to be able to analyze participation ratios, evolution of the vote between parties, etc. but without being able to determine which representative corresponds to each subdivision.

It already exists one example (as I knew now): Q20105870. To identify among other election item, we'll use:

• instance of (P31) = electoral result (Q19571328)

+ of (P642) = organizational subdivision (Q9261468) ..... to be free to apply to any kind of combination

+ applies to part (P518) = <election in place of year> .... the item of the referenced elections

Concatenating indirect elections

Indirect suffrage elections are in fact the concatenation of two elections: the popular vote to the members of a corporation (parliament, senate, city council, ..) who are granted the election of the executive office that will preside over the state, the region, the municipality, etc. This second election does not usually have its own article in the WPs, as it appears as one more section of the electoral process. Now, from WD’s point of view, they are two events directly linked, but separated in time and with different participants. The proposed solution is:

• create an item for the election among elected or voting of investiture with a structure equal to that of a direct suffrage, usually with a single candidate, where the census of voters is the number of deputies, with votes distribution, abstentions and votes collected by the candidate (qualif.P1111).
• in addition, and unlike a popular election, number of support votes (P8683), number of negative votes (P8682) and number of abstentions (P5043) may be indicated for each party and round (if the rules provide for a minimum and a second round), as shown in Presidential investiture in Catalonia, 2016 (Q97659738).
• Link the item of the popular vote with this second, using the P1536 in the first pointing to the second and the P1478 in the opposite direction.

When sufficient information is not available to construct the second item, basic information can be incorporated into P1536 with:

• value = appointment (Q1318578) or investiture (Q167407)
• of (P642) = name of the elect to invest
• replaces (P1365) = predecessor name

and the infobox will present it in the same way.

This treatment of P1536 also applies to direct suffrage to link elections to the usual formal act of handover with the previous incumbent, as can be seen in 2017 French presidential election (Q7020999).

Elections with hetereogeneous rules for constituence

The elections to the European Parliament are the sum of the results obtained in the elections held in the member states of the European Union. These elections are held on different days and each state represents a constituency in relation to the configuration of the chamber. That is, from the point of view of the levels described above, state elections represent level C and the item of constitution of Parliament represents level O. The rules of operation of state elections are not homogeneous. This fact is not a problem for the constitution of the parliament, as each state has a certain quota of seats, but not the way the deputies are selected.

Therefore, the voting system (P8196) of the items corresponding to the states, may be different from that of the constitution of the Parliament.

Multi-state. European Parliament case

Despite the comment in the previous section about the "non-affectation of the heterogeneity of voting systems", there are other aspects that make it different from classic parlamientary elections:

1. The state results correspond to parties and coalitions that win seats.
2. The parliament is structured in Groups where the deputies adhere according to their national party of origin

Therefore, between the constituency level item (C), to the chamber level item (O) there is a reassignment of the deputies, as they do not represent their states, but their parliamentary group and national party.

• For state-level items (C), when they exist, the structure would be as already described: P726 with the candidate parties and their votes in P1111; in P991 the political parties with their seats in P1410.
• For the parliamentary level (O), which must show the reallocation in groups and the final design of the chamber, we propose two options to structure it following a similar idea where the P726 summarizes the result of all the states (by parties) and P991 showing the new group structure and its components:

Option A

• Place in P726 the national parties that have obtained seats indicating the votes obtained (qual. P1111) and -exceptionally- the seats obtained (qual. P1410)
• Place in P991 the parliamentary groups with the total number of seats they represent in qual. P1410. Incorporate the qualifier P527 or P1012 with the national parties that make it up.

This option allows you to have the data broken down as they have been issued (by party) and how many seats each parliamentary group has and what makes it up. However, the P726 will have more than 150 entries that will be repeated as a qualifier for the P991

Option B

• Similar to option A, the P726 will contain the national parties that have obtained seats, with results in qualifiers: P1111 + P1410 obtained at constituency level (C), and -in this option- the P4100 indicating the Parliamentary Group where they will be attached.
• Place in P991 the parliamentary groups with the total number of seats they represent in qual. P1410. In this case, the qualifier P527 proposed in option A is not necesary as the party-group relationship can be get from P726.

This option has also long list of instances with national parties (aprx.150), but it is not redundant in P991, as happen in option A.

• See example option B: 2016 United States presidential election (Q699872)

In either option, it should be identified that this electoral element contains in P991 a reassignment structure in groups instead of national parties, as it has in P726. To do this, we will add to the P31 that contains the value "parliamentary elections xxxxx", the qualifier including (P1012) = Parlamentary Group Constitution (Q101421918).  To be decided

Electoral college

Some elections combine direct suffrage in some constituencies, with indirect suffrage in others where voters elect a delegate who, with their vote in support of the candidate, represents the general support of the constituency. To represent this type of "qualified vote" conducted by elected delegates, number of seats in assembly (P1410) will be used with the number of delegates as a qualifier P726. Recall that the P1410 is normally used in the P991 in party-oriented elections. Therefore, this definition does not present any possible confusion.

• See example: 2016 United States presidential election (Q699872)

Use of Participant (P710)

There are 450 elections with participant (P710) property. It may correspond to:

• Party name for a USA primary elections where the scope are: year (P585) + the part (P710) + the state (P1001). The name and results of candidates should be in P726+P991
• List of voters in president election for Germany.
• Candidates instead using P726. As Danish Landsting, 2018 United States House of Representatives elections (Q28404971),Indonesian legislative election, European Parliament election in Finland, and some others.

 To be defined / decided

Use of other rare properties

It is necessary to analyze the complete list of properties that use claim P31/P279 = public election (Q40231) in order to identify basic and common properties to be included in the ontology and to detect errors or misuses that need to be fixed.

Use of Ranking (P1352)

Some items have the ranking (P1352) as a qualifier in P726 or P991 to indicate final position among candidates after elections.

Infobox doesn't use it to order nor to know who is the winner. However, could be usefull for use in SPARQL as a faster way to make a selection.

We should consider whether to incorporate this property into the data model, or simply consider it optional.

 To be defined / decided

Properties common to many projects

• official website (P856), different from (P1889), organizer (P664), has list (P2354), facet of (P1269), topic's main category (P910), Commons category (P373)

Identifiers

• > 100 uses: Freebase ID (P646), UOL Eleições ID (P5892), Democracy Club election ID (P6649), Placar UOL Eleições ID (P6672), Tagesschau election ID (P4247)
• < 100 uses: Encyclopædia Britannica Online ID (P1417), BBC Things ID (P1617), Library of Congress authority ID (P244)