For the UN System Data Commons platform team. Prepared 16 Sep 2026 by the NYC Voluntary Local Review team (Builders' Day participants). This is not a defect report — nothing here is broken. It is an observation about what the graph can and cannot represent, found while building a city↔UN indicator crosswalk, and it bears directly on the question we most want to ask you: is there a sanctioned path for a city to contribute a series?

We should be precise about ownership. The SDG indicator framework is set by the UN Statistical Commission and the IAEG-SDGs, not by this platform. What follows is not a request to change it. It is a measurement of a gap between what the framework asks and what cities actually publish, offered because it bounds what municipal data could ever be ingested here.

The finding in one line

Zero of the 519 named SDG base indicators mention elections, voting or turnout. Twenty-three cities publish 469 datasets that do.

How the zero was established

We enumerated the SDG goal framework by walking ->relevantVariable from the seventeen goal roots: 689 base indicators. Of those, 519 return a name; the other 170 return none (see the second section below). Searching all 519 names for election | electoral | vote | voter | voting | turnout | ballot | referendum | suffrage | polling | candidate returns nothing.

We also checked the 170 unnamed DCIDs by mnemonic. Two look electoral and neither is:

Neither returns a name from get_variable_metadata, so we could not confirm either from the graph itself and are inferring from the series-code convention. If either is in fact electoral, this finding weakens and we would want to know.

What the framework does measure nearby, and why none of it covers the gap:

So the framework counts seats, measures who occupies them, and asks people how they feel about it. Nothing measures the conduct of an election: turnout, results, polling-station distribution, ballot accessibility, electoral register coverage.

What the cities publish

Scanning 17,162 datasets across 48 municipal portals for electoral vocabulary in the title, in six languages:

• 469 datasets, across 23 cities and 23 portals.
• 64% of them fall in the bottom quartile of their own catalog by similarity to any SDG indicator — measurably further from the framework than the average municipal dataset (mean affinity 0.332 against 0.407 for all datasets).

Milan alone is 66% of the datasets, so the dataset count is not evidence of breadth. The city count is: excluding Milan entirely leaves 159 datasets across 22 cities, which is the number this finding actually rests on.

The most telling detail

For each electoral dataset we recorded which SDG indicator the matcher ranked closest. The six most common answers across all 469:

For thirty-five municipal election datasets, the closest concept in the entire SDG framework is municipal waste collected. That is not a failure of the matcher so much as a description of the space it is searching.

Corroboration

The same category surfaced independently in two runs that share no cities and no embedding model:

Two disjoint sets of cities, two different models, one category.

Limits, stated

• This is a vocabulary result, not a conceptual proof. A dataset far from every indicator means no indicator's text is near its text. We hold this to be evidence of a framework gap only because it recurs across 23 independent cities; a single city's filing habit would prove nothing.
• Our matcher's precision is roughly half on hand-read candidate lists, and one positive control in nine fell into the tail in the English run (NYC's Housing Maintenance Code Violations, a hand-verified match for 11.1.1, at the 21st percentile). The tail contains real matches.
• The English cluster mixes two things: municipal elections and council roll-call votes (Winnipeg's Council Voting Data, Calgary's Council and Committee Votes). Both are democratic process; only the first is an election.
• Full method and every caveat: inverse crosswalk and the non-English run.

A second item, which is yours

170 of the 689 enumerated SDG base indicators return no name. get_variable_metadata answers for them but the name field is absent, and they return no observations either — so from outside there is no way to tell what DI_ILL_OUT, EN_BITR_REP_DV or SE_ACS_ELECT measure.

It is a quarter of the enumerated framework, and it has a direct cost for anyone building on the graph: our first matching run fell back to the DCID mnemonic as query text, so a quarter of "the framework" was represented by strings like DI ILL OUT, against which nothing can match. That inflated our gap measurements until we excluded them. We now match against the 519 named indicators and say so on the page.

A name on those 170 would fix it. It is also what stopped us confirming the two electoral-looking mnemonics above.

Sample: AG_FPA_COMM, DI_ILL_IN, DI_ILL_OUT, EN_ATM_CO2MVA, EN_BITR_REP, EN_BITR_REP_DV, EN_HAZ_TREATV, EN_LKW_QLTRB, EN_LKW_QLTRST, EN_LND_SLUM, EN_MAR_BEALIT_BV, EN_MAR_CHLANM. The full list is in the inverse crosswalk artifact.

Why we are raising it

Our project is a live Voluntary Local Review workbench: NYC series mapped to UN indicator DCIDs with a comparability grade on every mapping. Everything we have built runs city → UN, which can only ever surface what the framework already asks about. Running it backwards is how this appeared.

If city-level contribution is on the roadmap, then the categories cities actually publish — and electoral administration is among the largest and most consistent of them — are the ones that will have nowhere to land. We would be glad to discuss it at Builders' Day on 22 September.

Reproducing

git clone https://github.com/sarapis/undatacommons-nyc
python3 probe/fetch_municipal.py
python3 probe/inverse.py                    # English catalogs
python3 probe/inverse.py --language non-en  # the other five languages