Wikidata talk:WikiProject every politician/Political data model

From Wikidata
Jump to navigation Jump to search

structured data (Q26813700): next/previous[edit]

How about: 1) ? -> unicameralism (Q140247) (unicameral legislature (Q37002670)?)->bicameral legislature (Q189445)->tricameralism (Q747482)->tetracameralism (Q2475951)->etc.? --Fractaler (talk) 08:39, 23 November 2017 (UTC)

2) politic system? group of states? group of rulers? government of/by group of persons; country under group of rulers?:

anarchy (Q31895) -> monarchic system (Q22676587) -> oligarchy (Q79751): diarchy (Q936648) -> triumvirate (Q83204) -> tetrarchy (Q174450) -> pentarchy (Pentarchy (Q413340), Pentarchy (Q23863934), pentarchy (Q226106)) -> hexarchy -> heptarchy (Heptarchy (Q237945)) -> octarchy -> nonarchy -> decarchy (decarchy) -> ... Polyarchy (Q836853) --Fractaler (talk) 13:36, 13 December 2017 (UTC)

Members of a legislature[edit]

Hi! I have a question. What is more recommended: saying the parliamentary term (P2937) as a qualifier of a position held (P39) (like what we see in Marta Sibina (Q21293864)) or making a specific position (Q4164871) for a legislature (like Member of the 1st Scottish Parliament (Q30580541))? Thanks! --Davidpar (talk) 18:32, 1 December 2017 (UTC)

@Davidpar: I'm afraid there isn't really a simple answer to that! There are pros and cons of each, and there's no real consensus yet on which is better. What we would recommend however, is that even where a per-term item is used, a parliamentary term (P2937) qualifier is added anyway, so that queries don't need to know which model is in place — they can query for a subclass of the "main" membership item and use the qualifier, even if the item itself is more specific. (It's on my TODO list to fix this up for places not doing it yet.) At this point it's more important to get the information into Wikidata: we can easily migrate one approach to the other later. So, for now, I'd suggest simply using whichever is easiest for you and/or makes most sense to you. Is this in relation to a particular country/legislature? The answer might be simpler in relation to a specific context. --Oravrattas (talk) 18:56, 1 December 2017 (UTC)
@Oravrattas: Thanks for your complete answer. I'm starting the project for members of the Catalan Parliament and first of all I wanted to make sure I did it right. I think for now I'll add parliamentary term (P2937) qualifiers. --Davidpar (talk) 19:29, 1 December 2017 (UTC)

Ministerial position - add officeholder?[edit]

Can we add officeholder (P1308) to the preferable section of Ministerial position - per this example for Minister for Agriculture and Rural Development (Q32947389) ? --Tagishsimon (talk) 20:04, 11 December 2017 (UTC)

I think the general preference is for a position held (P39) on the item for the person, though I'm open to persuasion that we should recommend duplicating that information onto here as well. In general I'm a fan of having that information in both places, though I know that others disagree on that. --Oravrattas (talk) 21:53, 11 December 2017 (UTC)
Big-endians, the lot of them. I'm reminded that we found it a useful cross-check in Wikidata:WikiProject Heads of state and government. --Tagishsimon (talk) 00:23, 12 December 2017 (UTC)
Yes, I see it as being akin to double-entry bookkeeping, where having to express the same thing in two different ways makes it much more obvious when something is incorrect. Seeing them all in one place definitely makes it more obvious when something is missing, or where are there are incorrect overlaps etc (though we're also attempting to solve that with the PositionHolderHistory template: e.g. at Wikidata:WikiProject every politician/Estonia/officeholder/Q737115). Using officeholder (P1308) on the position itself also allows us to express much more clearly times when the position was empty (through a novalue claim). It does start getting rather unwieldy though if there have been hundreds of people holding a position. --Oravrattas (talk) 07:44, 12 December 2017 (UTC)
Can be unwieldy: agree. --Tagishsimon (talk) 14:07, 12 December 2017 (UTC)

Cabinet membership[edit]

Query on Cabinet_memberships, somewhat an echo of my previous question. We currently say:

On the individual cabinet, a has part (P527) list of each member in the form:

Is there worth in considering a different decomposition in which we specify the abstract positions that make up the abstract cabinet, and specify the officeholders thereof? In my little head, cabinets are composed of ministers; ministerial positions are held by politicians. So we would have:

I don't press this suggestion with any great conviction; but better to have these discussions early rather than late. --Tagishsimon (talk) 14:07, 12 December 2017 (UTC)

Yep, it's an interesting question: is the Cabinet made up of the positions, and the holder of that position is then in the cabinet as a result of that? Or is it made up of the people themselves, who might also happen to hold a certain position? I've actually tried going about this both ways: in Scotland we did the opposite of Wales (e.g. Second Salmond government (Q2577077)) to see what it would look like that way. I think my preference is for the 'people' rather than 'positions' version, as I think it's more of a reflection for what I think actually happens in practice — e.g. Cabinets often have members who don't actually hold a particular position: e.g. a "minister without portfolio". We could always have minister without portfolio (Q1937160) 'part of' the cabinet, with zero or more people holders at any given time, but that seems a bit more contrived that simply saying that someone is a member of the cabinet, and has a position held (P39) of minister without portfolio (Q1937160). I suspect that queries will be easier this way too, but that's more an instinct at this point rather than having tried extracting different types of data each way. I'm also not 100% sold on this though, so I'm definitely open to persuasion the other way (or an entirely different way!) --Oravrattas (talk) 16:07, 12 December 2017 (UTC)
@Tagishsimon:} after spending quite a while yesterday playing around with how to construct queries to get Cabinet memberships, I'm now more comfortable with solving this issue by simply removing the need to specify any has part (P527) on the cabinet :) — As long as each position knows that it's has part (P527) the Cabinet (or the Government, for junior roles), we can construct a list of members in a consistent manner — see for the example the Member List pages linked from Wikidata:WikiProject every politician/Wales/Cabinet. I'm sure that query can be made simpler (and adjusted to not think that current memberships end today), and I expect there are a few more edge cases that we'll need to cope with, but I think it's Good Enough For Now™, and simplifies our modelling requirements significantly. --Oravrattas (talk) 08:09, 18 December 2017 (UTC)
I like Wikidata:WikiProject every politician/Wales/Cabinet/Q27957659. part of (P361) on the position tickles the spot & yes, gets rid of Cabinet-Level_Positions. It'll be interesting to see if & how it breaks. I had a look at Ireland with a view to getting it to 4 stars, but I'm a bit too fried for that right now. In other news:
  • I noticed Government of Ireland (Q1672817) which conflates government and cabinet ... interim conclusion is that the model should give some more advice about the Government/Cabinet relationship.
    • Feel free to add some :) Though it's worth noting that the UK division here isn't a universal one — not all countries make that distinction. I don't know the Irish system well enough to know if it makes sense to split them there, though at the moment the only practical benefit to do so would be if we want to track positions that are part of the Government but not part of the Cabinet. --Oravrattas (talk) 13:52, 18 December 2017 (UTC)
  • I observe Parliamentary Terms - if using the instance of (P31) : legislative term (Q15238777) method, does not seem to point to the country or legislature of which is is a term. Although the 4-star prime objective is unaffected, this seems like an omission to me. I'm also concerned that if we want to point instances of terms at their legislatures (part of (P361) again?), then having two method - instance of (P31) : legislative term (Q15238777) versus via a country-specific item such as Parliamentary term in the Kingdom of Great Britain (Q21095053) means the part of (P361) is found either in the term item or in the country-specific item. Is that not bad news?
    • I'm not quite following what you mean by the part of (P361) here, but lots of legislative terms do point at the legislature they're part of, through of (P642) qualifiers:
    SELECT ?item ?itemLabel ?of ?ofLabel WHERE {
      ?item p:P31 [ps:P31 wd:Q15238777 ; pq:P642 ?of]
      SERVICE wikibase:label { bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en". }
    Try it!
    I think we can come up with a query that checks in either place easily enough. --Oravrattas (talk) 14:03, 18 December 2017 (UTC)
  • Do we want to tie instances of cabinets with instances of terms? See, for instance, w:en:Government of the 32nd Dáil which shows what we would term cabinets nested within the legislative term.
    • I'm not quite sure how we would do that. In countries with a Parliamentary system a new legislative term usually triggers a new cabinet, but the cabinet also usually outlives the legislative term. In other systems that's less clear-cut, and (for example) Presidential elections might be completely out of sync with legislative ones. I do agree it would generally be useful to know which of each overlap with which of the other, but I'm not sure if there's going to be anything better than just comparing dates. Suggestions welcome. --Oravrattas (talk) 14:07, 18 December 2017 (UTC)
I've not given the above much serious thought; these are just the impressions of the last days playing. --Tagishsimon (talk) 12:35, 18 December 2017 (UTC)
Half-formed thoughts. That's the phrase.
  • The model's prevailing assumption is that MPs are elected. But there are three classes of members: elected, appointed (e.g. Zambia - 9 by the president), and constitutional (e.g. Vice Predisident is a member in Malazi & in Zambia, iirc). We might want to made recommendations for qualifiers the three cases.
  • Cloudy thoughts about the distinction between the election dates and the date on which an officeholder assumes office - the structure is probably fine, but we might be more prescriptive in the model? Nice to have?
    • The current model assumes that the start time (P580) will be the day they actually assume office, which might be very different from the date on the elected in (P2715) item. (In the German Bundestag, for example, there was a month between them. In some cases that can be even longer.) +1 to making that clear in the documentation. --Oravrattas (talk) 14:10, 18 December 2017 (UTC)
  • *cough* acts of parliament statutory instruments *cough* --Tagishsimon (talk) 13:20, 18 December 2017 (UTC)

Legislators & their memberships[edit]

Next, at Legislators, we advise:

position held (P39) : a specific national position, eg Member of Parliament (Q16707842) *
preferably with a reference URL (P854) (ideally to the official site), and with modifiers (where known/appropriate):
parliamentary term (P2937), start time (P580), end time (P582), parliamentary group (P4100), electoral district (P768), elected in (P2715)

but the practise I see for the UK Parliament is membership of a numbered parliamentry term, such as Member of the 55th Parliament of the United Kingdom (Q35494253) seen in e.g. Alun Michael (Q19397).

I presume, for Alun, were we to follow the model advice, we would have multiple instances of Member of Parliament (Q16707842) each with a distinct parliamentary term (P2937)?

Might we consider amending the model description to reflect (at least) the UK practice and to make the suggestion for a pattern of term-specific membership items per Member of the 55th Parliament of the United Kingdom (Q35494253)? --Tagishsimon (talk) 18:17, 12 December 2017 (UTC)

There's some discussion on this further up this page at #Members_of_a_legislature. Changing the description to reflect this as a possibility is fine, though I wouldn't like to go as far as to recommend it by default. See Wikidata talk:EveryPolitician#too much specific items for legislature members (and further links from that) if you want more of the gory details! --Oravrattas (talk) 20:40, 12 December 2017 (UTC)
Thanks. No more questions for now ;) --Tagishsimon (talk) 17:09, 13 December 2017 (UTC)

Advice on modelling parliamentary group (P4100)[edit]

How should I go for modelling parliamentary group (P4100)? Should I create new parliamentary group for each legislative term (Q15238777) even though most of them stay the same? I have started writing script to import data about Saeima (Q822919). --Papuass (talk) 12:01, 16 February 2018 (UTC)

I think not. These things tend to span terms, as political parties do. Individuals will have one or more start dates for their membership of the P4100, and perhaps an end date. I see nothing pointing to terms in Wikidata:Property proposal/parliamentary group. And we could overlay term data, in a query, from the Parlamentry term qualifier of the membership of the legislature, so I see no good purpose in terming p.groups. --Tagishsimon (talk) 12:32, 16 February 2018 (UTC)
Thanks for the prompt reply. I saw no label (Q27950619), so I doubted. --Papuass (talk) 13:02, 16 February 2018 (UTC)

"applies to jurisdiction" for legislative terms[edit]

It is not clear why setting of applies to jurisdiction (P1001) is mandatory on legislative terms, since linking the legislative term to a legislature is mandatory, and the legislature already has applies to jurisdiction (P1001) as a mandatory field.--Tdombos (talk) 22:03, 13 March 2018 (UTC)

On the face of it, it's redundent in terms of normal form, but perhaps useful for practical query purposes, not least when there are two different possible routes from legislative term to the legislature - via a specific parliamentry term item, or via legislative term with the legislature as a qualifier? --Tagishsimon (talk) 12:15, 14 March 2018 (UTC)
@Tdombos, Tagishsimon: Yes, as the Query Service doesn't do any inference, queries need to be very explicit, which can sometimes be quite awkward, especially for beginners, so it's generally friendlier to "duplicate" information like this, especially as the data only needs to be written once, but might be queried millions of times. That said, there's perhaps a case for highlighting slightly differently data that a human should add, vs data bots could then derive and add to other places. (Though, of course, that would also require building such bots!) --Oravrattas (talk) 06:29, 15 March 2018 (UTC)

proportional system (electoral lists)[edit]

The current data model is guided towards plurality voting system (Q1211577) in which legislators are elected in single-member district (Q1052894), hence electoral district (P768) is mandatory. This model cannot accommodate electoral systems based on proportional representation (Q13365595) based on electoral list (Q618652) or Mixed electoral system (Q42417846) in which there are both single-member district (Q1052894) and electoral list (Q618652). A property would be needed to connect the legislator to the electoral list (Q618652) they were elected for (which might not be the same as parliamentary group (P4100) in case of electoral list (Q618652) with more than party). For mixed system there should also be a qualifier to set whether the legislator was elected in an single-member district (Q1052894) or via an electoral list (Q618652). There might be existing properties that can be used for this purpose, if yes, please point me to which ones should be used. If there are no existing properties that can be used, property proposals should be submitted. --Tdombos (talk) 14:25, 15 March 2018 (UTC)

@Tdombos: It's quite hard to reason about this in the abstract: do you have examples of some places that don't really work in the current model? (Note, btw, that even though the district is marked as compulsory, novalue is an acceptable value, where there is none, and helps differentiate between the case where we know that none applies, vs the data simply being missing). Recording the electoral list (Q618652) someone was part of seems more suitable on a candidacy in election (P3602), rather than a position held (P39), or am I misunderstanding? --Oravrattas (talk) 08:06, 16 March 2018 (UTC)
@Oravrattas: I like both of your suggestions (recording no value for electoral district (P768) when the person was elected via electoral list (Q618652); recording electoral list (Q618652) on candidacy in election (P3602)). What should be the property used for electoral list (Q618652) on candidacy in election (P3602)? Can you add a note on these to the data model? These decisions are not obvious, should be made explicit.--Tdombos (talk) 11:48, 16 March 2018 (UTC)
I'm not sure I really understand the nuances of different electoral systems well enough to be able to suggest a preferred model. There has been some talk about restoring WikiProject Elections (and some of the other discussions on that page talk about some of the complexity in this area); or perhaps other people here might have suggestions for what might be best. Do you have a specific example of something you'd like to record? That tends to make the discussion simpler. --Oravrattas (talk) 12:10, 16 March 2018 (UTC)

change of political group within legislative term[edit]

The current data model attaches parliamentary group (P4100) to legislator via position held (P39) as a qualifier. This means that no dates can be set for membership in parliamentary group (P4100). So how to handle situations when a legislator changes their parliamentary group (P4100) within one parliamentary term (P2937)? -- Tdombos (talk) 17:51, 15 April 2018 (UTC)

@Tdombos: The easiest way to do this is have a new position held (P39) term, using the same parliamentary term (P2937) qualifier, with the new parliamentary group (P4100); you'll also want to put dates on the two entries to reflect the change. See eg/ Michelle Thomson (Q20027903) as an example. Andrew Gray (talk) 18:40, 15 April 2018 (UTC)
And note the end cause (P1534). --Tagishsimon (talk) 19:38, 15 April 2018 (UTC)

Separate position items for each parliamentary term[edit]

This page (or more specifically, /Legislators) says that the value for position held (P39) should be a specific national position item, such as Member of Parliament (Q16707842), with a parliamentary term (P2937) qualifier like 57th United Kingdom Parliament (Q29974940). However, the United Kingdom currently doesn’t quite follow that model: it has sub-items for each legislative term, e. g. Member of the 56th Parliament of the United Kingdom (Q30524718), Member of the 57th Parliament of the United Kingdom (Q30524710), etc., which are subclasses of the national position item, and uses those for the position held (P39) main value (in addition to the parliamentary term (P2937) qualifiers, which are still present). At the Wikifying Westminster Workshop, someone (I’m not sure who, sorry… was it Andrew Gray?) mentioned that this model was better (at that point, the migration for UK items hadn’t been completed yet, so our queries had to deal with both possibilities), but we didn’t have time to discuss this in detail. But since we’re likely going to import all parliamentary terms of Germany soon (both a_ka_es and I have separately started working on this), I think now would be a good time to clear this up, so that we can use the superior data model for Germany as well – if it really is superior :) --Lucas Werkmeister (talk) 17:57, 13 June 2018 (UTC)

I've made two changes to the model documentation which I invite you to inspect & amend: Legislators diff & Legislative Positions diff. I favour the current UK model, but lack the energy to make an evidenced argument for its superiority beyond noting that it gives us a simple hook in the items useful for reporting, one that is arguably more efficient to use than date wrangling. --Tagishsimon (talk) 19:17, 13 June 2018 (UTC)
@Tagishsimon, Lucas Werkmeister: Looks good - thanks for the update. I think this is one of those cases where actual practice has gone a lot further than the documentation and we've forgotten to update it. In terms of which is better ... I think you can reasonably argue for either. I used to prefer the simple approach but Oravrattas (talkcontribslogs) convinced me to go with the complicated one :-)
The simple "Member of X" is a little easier to understand and write queries for, but some tools don't cope with it very well. "Member of Nth X" is a little more fiddly but the tools cope better and it is a little easier to do checks for mistakes - it also makes it easier to find people in a given term very easily as you don't have to query on start/end dates. I personally find it's also a bit easier to work with when you have a lot of properties - it's a little easier to find the right claim if they're differentiated like this.
I was a little worried that external users would find "member of Nth X" a bit weird, but they seem pretty relaxed about it and seem to think it's a comprehensible approach. Andrew Gray (talk) 12:07, 15 June 2018 (UTC)
@Lucas Werkmeister, Tagishsimon, Andrew Gray: There are a few connected, but slightly separate, issues in play here, which make this all a little complex.
  • Some of the benefits are around how to enter the data in bulk. Tools like QuickStatements, and from what I've seen so far, the new OpenRefine, don't cope with giving someone multiple statements that differ only by qualifiers: so if someone has been a member of the same legislature multiple times, they'll smash all of the subsequent memberships into the one statement. Splitting out individual "member of 10th Parliament", "member of 11th Parliament" etc, still doesn't get 100% of the way there, as these tools still do the wrong if someone has multiple memberships within a single term, but in most cases we've done this it gets us 95+% of the way there, and the edge cases can be tidied up by hand. This is a bit less of an issue now that we have User:PositionStatements Bot, but that's still not as straightforward for most people as QuickStatement or OpenRefine would be.
  • There are also benefits around reporting. Listeria, in particular, will produce much nicer reports for everyone who has been "member of the 10th Parliament", as it can then also drill into the qualifiers on that. In particular, it will also cope much better with people who have had that position more than once, whereas a hand-rolled SPARQL query will need to do lots of fiddly GROUP_CONCAT magic for those cases (as Listeria otherwise throws away subsequent rows with the same ?item).
In a world where these third-party tools were better suited to these sorts of issues, then I suspect we probably wouldn't have taken this modelling approach for the UK, so there's definitely a strong argument to say that these aren't good reasons for a potentially confusing approach. But it has paid off well so far, and there doesn't seem to be much evidence of people actually being confused by it. I don't think I would go as far as saying that it should be the preferred approach more generally, but I think it's an acceptable one, and it's used widely enough now that tools/queries working with this sort of data should always remember to add a P279* check.
One suggestion that Andrew made when we were looking at this before, which I think is definitely useful and important, is that even where something does use the "Member of Nth term" items, that should still make sure to have the parliamentary term (P2937) qualifier anyway, even if that seems slightly redundant, as it will allow us to simplify lots of queries significantly, and make it much simpler to to migrate to the "simpler" approach at a later point if/when preferred. --Oravrattas (talk) 12:29, 15 June 2018 (UTC)
I've further tweaked Legislators to reflect a continued requirement for P2937 where "member of ##th parliament" is used. --Tagishsimon (talk) 15:56, 15 June 2018 (UTC)

Handling redistricting[edit]

In the US most districts are changed every ten years, to reflect the population shifts measured by the US census. In some places, the towns included in a district of a certain name, such as the Vermont Rutland-3 state senate district, may be entirely different after redistricting. The current practice in Wikipedia for Vermont has been to include the year range in the name of the article, and consequently, generating a separate Wikidata item for each time period. Is this a best practice? See also the discussion at w:Wikipedia talk:WikiProject Elections and Referendums#Election district naming convention. Jc3s5h (talk) 14:11, 24 August 2018 (UTC)

It seems appropriate to do that here, since each instance is a new geo-shape. They can be linked via replaces (P1365) and replaced by (P1366). --Tagishsimon (talk) 14:18, 24 August 2018 (UTC)