Place classification - town vs suburb, remote rural areas, how does a CDP relate to place

Let me be more succinct.

In Maine we decided that boundaries of unincorporated towns should be tagged with the same admin_level but a different border_type as incorporated towns, because other than the incorporation status, they were more alike than they were different, and therefore we treated them as a collective group. This distinction was made easy due to things like signposted borders and a space-filling geography where the unincorporated towns have a similar size.

Are these unincorporated towns in Nevada an equivalent case to Maine or is there some aspect about them that makes them different?

Or maybe it’s New England’s towns that are the odd ones out. The Midwest’s townships and Wisconsin’s towns are all admin_level=7. These words are all terribly overloaded, so I wouldn’t actually read too much into them.

I wasn’t under the impression that Maine’s townships are even relevant to this discussion. The parallel I was actually trying to make is that, like New York City’s community districts, Nevada’s unincorporated towns have legally defined boundaries, but their governing boards are only advisory. Thus the question of whether their boundaries are administrative in nature. (I’m unsure if an unincorporated town’s boundaries are signposted on the spot; that may be a key difference.)

You brought up Nevada in the necropost - it seemed to me that the Maine case was a clear parallel to that situation. I would also discount Nevada’s merely-advisory governments as governments in the same way we agreed that Maine did not have governments in their townships. And, like in Maine, I would consider city/town boundaries as a collective group even if some are government-less.

In Maine, townships have no government, while towns and cities do. In Nevada, unincorporated towns have no non-advisory government, while cities do. So by this reasoning, unincorporated towns would have administrative boundaries despite their advisory governments, not because of them.

1 Like

I would say rather that the collective group of town/townships/city boundaries are administrative regardless of the fact that some portion lack governments.

When the whole of them lack governments, that feels like a different class of feature and hence a different case.

That’s fair. I think the hard part is deciding what’s in the class and what isn’t. Nevada isn’t the only state with something along these lines. In some sense, California’s county service areas serve the same purpose as Nevada’s unincorporated towns, but a CSA’s board is the county board of supervisors ex officio, rather than a separate advisory body. Does the lack of even direct advisory representation affect the district’s prominence enough to result in a change in tagging? Or are the similar levels of service a red herring because one is a place and the other is not? Is this where we can finally draw the line between administrative and something else?

1 Like

It’s getting murky in these waters. Any political scientists tuned in?

This needs to be spelled out more, but so far, to steal a @stevea phrase, I find myself nodding in agreement.

In partial answer to Zeke’s “if we put such a high value on equivalency…” (we don’t): each “stack of numbers” are values in a hierarchy as “steps on a ladder.” With each “deeper” (higher) integer subset value in that stack, that stack of numbers become more and more unique. So-called “equivalency” (only approximating, not true) only increases somewhat as you go up in the hierarchy and find other members that are “this and that with you” at the level shared. OSM does it everywhere like this. Somewhere “up at” 1 or 0, we denote all humans on Earth.

Each stack of numbers is unique. Around here (2 in this particular dialog context / linguistic register) 4s are 50+ mostly equivalences, except for half a dozen “uniques” at the federal level 2. Each of those 50 equivalences (amongst each other, in a Union at the federal level) are sturdy towers of “how we do it around here.” (At 4, the level of a state). Territories, Commonwealths, a District, those half dozen have their own methods to do things at their own 4 level, because that how things fit together into this particular hierarchy (at this level, 2 where these substacks fit together). So, even at our 4, there isn’t “full” equivalence, but it is explained (in wiki table and footnotes) as constructed as it is (the data in OSM are). And briefly, again, here.

What seems going on now is the explosive realization that there are blisteringly-large (relatively speaking) distinct columns of numbers which exist. And a 10 here isn’t a 10 anywhere else, because that’s not how it works “down here” (where stretching to achieve “equivalence” is pretty hopeless, due to the deep level, except perhaps for a neighboring / bordering place where your stacks are pretty similar if not identical). Down at 9 and 10, the explosive growth of possibilities that are different than “your” (local, from another perspective in the hierarchy) ken of it is not only likely different, it is almost automatically, mathematically, probabilistically different.

Sure, you can infer similarities, but only in a “similar fashion” like brothers and sisters.

It’s fascinating. Glad to be discussing.

I listen.

A place=*, a boundary=administrative, a border_type=*, these all denote different things. Sometimes our Census Bureau is saying something and it often makes sense to say it like that (with a boundary=census, for example). These can and do affect one another in large scope given the cross-products across “stacks of number hierarchies.” Not always, but in “could go in a particular direction” in an almost infinitely complex set of methodologies.

I’m not sure whether I agree or disagree here. Probably parts both. I assess that we’ve made an effort, pretty successfully, to match up the steps on the admin_level ladder, horizontally. The lower the admin_level, the firmer the matching, as the constructs tend to be the same.

Where we are trying to decide which admin_level to assign to a boundary, and we have the wiggle room of plus or minus one (as we tend to because we tend to skip numbers to leave space) – how do we decide where to place it?

I maintain that the way we do it is to look for similarity in parallel hierarchies. It’s not perfect, but from a data consumer perspective – the more similarity we are able to achieve, the more useful the data.

So far at least, the principle of lower numbers enclosing higher numbers (with some exception where overlaps occur) seems to generally hold.

I could see, for example with NYC, if we felt equivalence was more important than enclosing hierarchy, we could have chosen the scheme of:

New York State - 4
New York County (same geography as Manhattan) - 6
New York City - 8
Manhattan (same geography as New York County) - 9

That would have preserved equivalence between counties in other states (which are 6) and municipalities in other states (which are usually 8). But, it would have “broken” in the fact that a city (level 8) encloses several counties (level 6). So, our decision to make NYC a special case at level 5 was a choice to break equivalence in favor of hierarchical consistency.

There are certainly multiple ways to skin this cat and I don’t see anything wrong with this choice. We didn’t break equivalence willy nilly, we did so because we had competing objectives and had to make a choice.

The fact that NYC is an outlier is not terribly a problem since (1) it’s well-documented and (2) there’s only one case of it that data consumers need to worry about.

Key concept: everything is an outlier. Consistency in structure is partly imagined / fictional and a fragile consensual hallucination. Yet there are (sibling, cousin…) similarities.

We do OK, even as we keep explaining its nuances to ourselves. This isn’t exactly easy.

Edit: NYC as an outlier isn’t unique at 5, as so are Tennessee’s Divisions.

admin_level is, by definition a hierarchy. We don’t “honor equivalence” and the objectives are not competing. The ONLY choice was to stay hierarchically consistent in the, uh, hierarchy.

It is, using the definition of:

any system of persons or things ranked one above another.

However, the world is complicated and in cases where it is not clear how things rank with respect to each other, we are forced to decide. That’s what I mean by competing objectives.

Which is why

breaks down when the hierarchy is not clear.

We have chosen to assign this rank on the basis of nesting, in other words, if one thing contains multiple other things, the enclosing thing is the higher rank.

However, not all boundaries nest so neatly. Boundary geeks know that there are very many towns in the United States that overlap multiple counties. Therefore, we cannot use the principle of nesting here. Instead, what we have decided to do in these cases is to rank the whole class of counties as higher than the whole class of cities and towns. This is a reasonable choice because counties mostly do tend to nest multiple cities/towns, and are much larger divisions of the state on the whole than cities and towns are.

The state of New York is one of those states where municipalities can overlap multiple counties:

Now of these (and of all the examples in the US), only NYC fully overlaps more than one county.

So what ranks higher - city or county? We had a choice in logic we could have applied here:

  1. Treat NYC as a member of the whole class of cities and towns in NY and rank the whole class of cities and towns as below the whole class of counties in NY. This results in the contradiction that a lower admin_level is spatially contained within a higher admin_level,
  2. Treat NYC as a region of New York that is the lone member of a solitary class of boundary, and rank that above counties, which are themselves ranked above the whole class of other cities and towns. This results in the contradiction that a city is ranked above a county in the hierarchy.
  3. Treat NYC like a consolidated city-county and rank it as a county, but bump it up one level because it consolidates more than one county.

Logic #2 is the logic behind ranking the districts of Tennessee - those form three subdivisions of the state that are neatly divided further into counties. No special case there, and if indeed they belong in the admin_level hierarchy, 5 seems a plainly obvious spot to place it.

Logic #3 is what we actually did and is pretty defensible.

I’d like to pick into the logic a bit because I think it’s interesting. We tag consolidated city-counties at admin_level=6 (example - Denver) because they act as both city and county, and county is more important. I recall past conversations where it’s been suggested that these might also get a second boundary relation, having the same geometry, to represent the city, but tagged at a higher admin_level values. I don’t believe this was ever done, broadly.

It is curious that in NYC, we have done this of sorts:

(when you operate a site based on a big database of boundaries, you notice when weird things like this show up)

The Brooklyn boundary has been tagged at level 7 since 2019, so I was surprised just now that this little oddity hadn’t yet found its way into the extensive footnotes on the United States admin level wiki page. I’m sure this divergence will be resolved one way or another soon enough.

Anyways, if this 7/6 scheme is dubbed a correct tagging scheme, then it opens the question of whether other consolidated city-counties should also have both the “county” and the “consolidation” mapped.

So ranked hierarchy, yes. But you still have to make some determinations on what criteria you use for the ranking, and what to do when the complexities of the real work challenge those criteria.

Lots to say about that, I’ll start with Brooklyn being 7. IMO, Brooklyn “should” match other NYC boroughs, like Kings County, at 6. I’d consider this an oversight that others (here, in OSM, familiar with admin_level…) could get behind and say “yeah, let’s retag that 7 as a 6, to match its siblings.” They are in the same, exact “city,” after all. But what do I know?, last time I was in Brooklyn was 2007. If it has been mistagged for five years, let’s fix it.

But should “city” as NYC calls itself, be found to be equivalent to other cities, say, outside of New York (state)? (Let’s leave out INSIDE of New York state for now). No, not really or only superficially on a “cousins can be alike, but they are not really the same” basis.

These are hazy, heads-nod agreements to similarity and structure-that-fits-consistently-within-a-bigger-picture, but maybe we aren’t shaking hands, hugging and kissing each other about everything. “Kissing cousins?”

Still and all, it is somewhat amazing that OSM (and the People of our country and Earth) do this the way that we do, with as little rancor and disagreement as we have experienced over the decades we do (and have been) doing this. I’ll take a hazily-focused, what I’ve called a “slightly runny watercolor” picture of things. I’ll also (happily!) take improvements to our view(s) of things.

Still nodding my head, and pretty amazed at this discussion, it’s great!

Thought experiment.

As a counter-factual, if a long time ago, suppose I tagged Rhode Island as admin_level=3, its counties as admin_level=4, and its towns as admin_level=5. Suppose I did this long before such things were more standard and somehow nobody noticed.

And suppose someone came along and just now noticed that RI had this number scheme that didn’t look like the rest of the country and wanted to change it to the 4/6/8 scheme that is the most common amongst states. Suppose then that I (stubbornly) argued that “it’s just a hierarchy, the ranking order is correct, and it doesn’t matter which values are used. Who cares that I’ve not skipped numbers? They’ve been like this a long time!”

Would you be okay with this tagging situation? Are you saying there is absolutely no value in equivalence horizontally, as long as vertical ordering is preserved?

I say “there is a history to how things have evolved.” I say “horizontal and vertical (both) become more strict the longer they have been established.”

Saying “Rhode Island got 3, so…” is tantamount to starting with a false proposition, which logic discourages as a strong foundation, as starting from a falsehood means you can prove anything is true.

We have the present, we have what is established, we have the ability to change (what might widely be agreed to have entered erroneously, and should be corrected, by wide consensus), though, only if “not too sclerotic” (as it has become long-established).

There is an aspect to doing this which reminds me of “Instantly Funk” (a childhood toy from the 1970s which I think was colored cornstarch, to which water was added): “it’s liquid, it’s solid…”.

I am not sure you are quite understanding the currently-tagged-situation-in-the-map as I re-read this. There are five boroughs each tagged admin_level=7, and there are five totally separate county boundary relations tagged admin_level=6. Each borough boundary is coterminous with a county boundary. So there are a total of ten mapped boundary relations for the five entities.

So if you are saying that we should instead have ten boundary relations within NYC tagged admin_level=6, then no, that would be a bad situation for data consumers.

My apologies. I succumb to a bad habit of mine and one I attempt to remedy but don’t always “chase from both directions” (as correctness should chase bad data): the wiki states “Borough” at 6 in NYC and “empty” at 7. I mistakenly looked here (wiki) rather than at map data (definitive) at what was really going on and I missed actual map data for our wiki shorthand at denoting it. So you (and the map data) are correct with 6 and 7 in NYC as they are now.

I’ll correct the wiki(s). Wiki chases map chases wiki chases map chases wiki…

Edit: Wiki edits completed.

My understanding is that although each borough is conterminous with a county, they are still at least somewhat separate entities. The borough is a division of the New York City government while the county is a division of the New York State government. Seems like the counties mostly just handle judicial matters. While it seems a bit silly to have 10 boundaries for 5 areas, this does seem to reflect reality if my understanding is correct.

2 Likes